{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.39718790960003175,
  "global_step": 20000,
  "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
    }
  ],
  "max_steps": 50354,
  "num_train_epochs": 9223372036854775807,
  "total_flos": 6.6891364171776e+17,
  "trial_name": null,
  "trial_params": null
}