diff --git "a/checkpoint-7605/trainer_state.json" "b/checkpoint-7605/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-7605/trainer_state.json"
@@ -0,0 +1,56616 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.3042,
+  "global_step": 7605,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 2e-06,
+      "loss": 3.1659,
+      "theoretical_loss": 3.484796891144814,
+      "tokens_seen": 1661665280
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4e-06,
+      "loss": 3.0486,
+      "theoretical_loss": 3.484785010063602,
+      "tokens_seen": 1661730816
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6e-06,
+      "loss": 2.5634,
+      "theoretical_loss": 3.4847731295821447,
+      "tokens_seen": 1661796352
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8e-06,
+      "loss": 3.0242,
+      "theoretical_loss": 3.48476124970039,
+      "tokens_seen": 1661861888
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1e-05,
+      "loss": 2.685,
+      "theoretical_loss": 3.4847493704182817,
+      "tokens_seen": 1661927424
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2e-05,
+      "loss": 2.7646,
+      "theoretical_loss": 3.4847374917357676,
+      "tokens_seen": 1661992960
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4e-05,
+      "loss": 2.8002,
+      "theoretical_loss": 3.4847256136527935,
+      "tokens_seen": 1662058496
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6e-05,
+      "loss": 3.0059,
+      "theoretical_loss": 3.4847137361693052,
+      "tokens_seen": 1662124032
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8e-05,
+      "loss": 3.0361,
+      "theoretical_loss": 3.4847018592852494,
+      "tokens_seen": 1662189568
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2e-05,
+      "loss": 2.79,
+      "theoretical_loss": 3.484689983000571,
+      "tokens_seen": 1662255104
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2e-05,
+      "loss": 2.7581,
+      "theoretical_loss": 3.4846781073152173,
+      "tokens_seen": 1662320640
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4e-05,
+      "loss": 2.6647,
+      "theoretical_loss": 3.484666232229134,
+      "tokens_seen": 1662386176
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6e-05,
+      "loss": 2.8627,
+      "theoretical_loss": 3.484654357742267,
+      "tokens_seen": 1662451712
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8e-05,
+      "loss": 2.8585,
+      "theoretical_loss": 3.4846424838545627,
+      "tokens_seen": 1662517248
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3e-05,
+      "loss": 2.7554,
+      "theoretical_loss": 3.4846306105659677,
+      "tokens_seen": 1662582784
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2e-05,
+      "loss": 2.5777,
+      "theoretical_loss": 3.484618737876427,
+      "tokens_seen": 1662648320
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.4000000000000007e-05,
+      "loss": 2.5762,
+      "theoretical_loss": 3.4846068657858877,
+      "tokens_seen": 1662713856
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.6e-05,
+      "loss": 2.6419,
+      "theoretical_loss": 3.484594994294296,
+      "tokens_seen": 1662779392
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.8e-05,
+      "loss": 2.6613,
+      "theoretical_loss": 3.4845831234015976,
+      "tokens_seen": 1662844928
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4e-05,
+      "loss": 2.6489,
+      "theoretical_loss": 3.484571253107739,
+      "tokens_seen": 1662910464
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.2000000000000004e-05,
+      "loss": 2.8326,
+      "theoretical_loss": 3.484559383412666,
+      "tokens_seen": 1662976000
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.4e-05,
+      "loss": 2.771,
+      "theoretical_loss": 3.4845475143163256,
+      "tokens_seen": 1663041536
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.6e-05,
+      "loss": 2.8005,
+      "theoretical_loss": 3.4845356458186627,
+      "tokens_seen": 1663107072
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.8e-05,
+      "loss": 2.6649,
+      "theoretical_loss": 3.484523777919625,
+      "tokens_seen": 1663172608
+    },
+    {
+      "epoch": 0.0,
+      "objective/train/docs_used": 918485,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.4342758655548096,
+      "objective/train/theoretical_loss": 3.4845119106191578,
+      "objective/train/tokens_used": 22097376,
+      "theoretical_loss": 3.4845119106191578,
+      "tokens_seen": 1663238144
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5e-05,
+      "loss": 2.6636,
+      "theoretical_loss": 3.4845119106191578,
+      "tokens_seen": 1663238144
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.2e-05,
+      "loss": 2.8955,
+      "theoretical_loss": 3.4845000439172074,
+      "tokens_seen": 1663303680
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.4e-05,
+      "loss": 2.5046,
+      "theoretical_loss": 3.48448817781372,
+      "tokens_seen": 1663369216
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.6e-05,
+      "loss": 2.709,
+      "theoretical_loss": 3.484476312308643,
+      "tokens_seen": 1663434752
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.800000000000001e-05,
+      "loss": 2.4005,
+      "theoretical_loss": 3.484464447401921,
+      "tokens_seen": 1663500288
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6e-05,
+      "loss": 2.5362,
+      "theoretical_loss": 3.4844525830935007,
+      "tokens_seen": 1663565824
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.2e-05,
+      "loss": 2.5606,
+      "theoretical_loss": 3.4844407193833282,
+      "tokens_seen": 1663631360
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.4e-05,
+      "loss": 2.6669,
+      "theoretical_loss": 3.484428856271351,
+      "tokens_seen": 1663696896
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.6e-05,
+      "loss": 2.626,
+      "theoretical_loss": 3.484416993757514,
+      "tokens_seen": 1663762432
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 2.5635,
+      "theoretical_loss": 3.484405131841765,
+      "tokens_seen": 1663827968
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.000000000000001e-05,
+      "loss": 2.454,
+      "theoretical_loss": 3.4843932705240483,
+      "tokens_seen": 1663893504
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.2e-05,
+      "loss": 2.5559,
+      "theoretical_loss": 3.4843814098043113,
+      "tokens_seen": 1663959040
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.4e-05,
+      "loss": 2.1608,
+      "theoretical_loss": 3.4843695496825005,
+      "tokens_seen": 1664024576
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.6e-05,
+      "loss": 2.6816,
+      "theoretical_loss": 3.4843576901585616,
+      "tokens_seen": 1664090112
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.8e-05,
+      "loss": 2.5955,
+      "theoretical_loss": 3.4843458312324413,
+      "tokens_seen": 1664155648
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8e-05,
+      "loss": 2.7015,
+      "theoretical_loss": 3.484333972904086,
+      "tokens_seen": 1664221184
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.2e-05,
+      "loss": 2.5895,
+      "theoretical_loss": 3.484322115173442,
+      "tokens_seen": 1664286720
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.400000000000001e-05,
+      "loss": 2.6653,
+      "theoretical_loss": 3.4843102580404555,
+      "tokens_seen": 1664352256
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.599999999999999e-05,
+      "loss": 2.4374,
+      "theoretical_loss": 3.4842984015050726,
+      "tokens_seen": 1664417792
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.8e-05,
+      "loss": 2.6655,
+      "theoretical_loss": 3.4842865455672403,
+      "tokens_seen": 1664483328
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.999999999999999e-05,
+      "loss": 2.1725,
+      "theoretical_loss": 3.484274690226904,
+      "tokens_seen": 1664548864
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.2e-05,
+      "loss": 2.6074,
+      "theoretical_loss": 3.484262835484011,
+      "tokens_seen": 1664614400
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.400000000000001e-05,
+      "loss": 2.2801,
+      "theoretical_loss": 3.484250981338507,
+      "tokens_seen": 1664679936
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.6e-05,
+      "loss": 2.6196,
+      "theoretical_loss": 3.484239127790339,
+      "tokens_seen": 1664745472
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.800000000000001e-05,
+      "loss": 2.349,
+      "theoretical_loss": 3.484227274839453,
+      "tokens_seen": 1664811008
+    },
+    {
+      "epoch": 0.0,
+      "objective/train/docs_used": 919051,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5782582759857178,
+      "objective/train/theoretical_loss": 3.4842154224857955,
+      "objective/train/tokens_used": 23735776,
+      "theoretical_loss": 3.4842154224857955,
+      "tokens_seen": 1664876544
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001,
+      "loss": 2.5304,
+      "theoretical_loss": 3.4842154224857955,
+      "tokens_seen": 1664876544
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000102,
+      "loss": 2.5036,
+      "theoretical_loss": 3.484203570729313,
+      "tokens_seen": 1664942080
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000104,
+      "loss": 2.4537,
+      "theoretical_loss": 3.4841917195699517,
+      "tokens_seen": 1665007616
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000106,
+      "loss": 2.2072,
+      "theoretical_loss": 3.484179869007658,
+      "tokens_seen": 1665073152
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000108,
+      "loss": 2.5087,
+      "theoretical_loss": 3.4841680190423787,
+      "tokens_seen": 1665138688
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011,
+      "loss": 2.2714,
+      "theoretical_loss": 3.4841561696740597,
+      "tokens_seen": 1665204224
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000112,
+      "loss": 2.4838,
+      "theoretical_loss": 3.4841443209026477,
+      "tokens_seen": 1665269760
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000114,
+      "loss": 2.6179,
+      "theoretical_loss": 3.4841324727280893,
+      "tokens_seen": 1665335296
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011600000000000001,
+      "loss": 2.5005,
+      "theoretical_loss": 3.4841206251503305,
+      "tokens_seen": 1665400832
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000118,
+      "loss": 2.562,
+      "theoretical_loss": 3.484108778169318,
+      "tokens_seen": 1665466368
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012,
+      "loss": 2.5914,
+      "theoretical_loss": 3.4840969317849986,
+      "tokens_seen": 1665531904
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000122,
+      "loss": 2.3786,
+      "theoretical_loss": 3.484085085997318,
+      "tokens_seen": 1665597440
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000124,
+      "loss": 2.3953,
+      "theoretical_loss": 3.4840732408062234,
+      "tokens_seen": 1665662976
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000126,
+      "loss": 2.5667,
+      "theoretical_loss": 3.484061396211661,
+      "tokens_seen": 1665728512
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000128,
+      "loss": 2.5867,
+      "theoretical_loss": 3.4840495522135777,
+      "tokens_seen": 1665794048
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013000000000000002,
+      "loss": 2.4648,
+      "theoretical_loss": 3.4840377088119188,
+      "tokens_seen": 1665859584
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000132,
+      "loss": 2.3746,
+      "theoretical_loss": 3.484025866006632,
+      "tokens_seen": 1665925120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000134,
+      "loss": 2.4735,
+      "theoretical_loss": 3.4840140237976636,
+      "tokens_seen": 1665990656
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013600000000000003,
+      "loss": 2.3041,
+      "theoretical_loss": 3.48400218218496,
+      "tokens_seen": 1666056192
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013800000000000002,
+      "loss": 2.5027,
+      "theoretical_loss": 3.483990341168467,
+      "tokens_seen": 1666121728
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014000000000000001,
+      "loss": 2.3187,
+      "theoretical_loss": 3.4839785007481323,
+      "tokens_seen": 1666187264
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014199999999999998,
+      "loss": 2.6289,
+      "theoretical_loss": 3.483966660923902,
+      "tokens_seen": 1666252800
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000144,
+      "loss": 2.5713,
+      "theoretical_loss": 3.483954821695723,
+      "tokens_seen": 1666318336
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000146,
+      "loss": 2.5538,
+      "theoretical_loss": 3.4839429830635407,
+      "tokens_seen": 1666383872
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000148,
+      "loss": 2.4974,
+      "theoretical_loss": 3.4839311450273023,
+      "tokens_seen": 1666449408
+    },
+    {
+      "epoch": 0.0,
+      "objective/train/docs_used": 920018,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.4347381591796875,
+      "objective/train/theoretical_loss": 3.4839193075869543,
+      "objective/train/tokens_used": 25374176,
+      "theoretical_loss": 3.4839193075869543,
+      "tokens_seen": 1666514944
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00015,
+      "loss": 2.4437,
+      "theoretical_loss": 3.4839193075869543,
+      "tokens_seen": 1666514944
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000152,
+      "loss": 2.3852,
+      "theoretical_loss": 3.4839074707424444,
+      "tokens_seen": 1666580480
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000154,
+      "loss": 2.5793,
+      "theoretical_loss": 3.4838956344937175,
+      "tokens_seen": 1666646016
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000156,
+      "loss": 2.4539,
+      "theoretical_loss": 3.483883798840721,
+      "tokens_seen": 1666711552
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000158,
+      "loss": 2.466,
+      "theoretical_loss": 3.4838719637834012,
+      "tokens_seen": 1666777088
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00016,
+      "loss": 2.4158,
+      "theoretical_loss": 3.4838601293217053,
+      "tokens_seen": 1666842624
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000162,
+      "loss": 2.3978,
+      "theoretical_loss": 3.483848295455579,
+      "tokens_seen": 1666908160
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000164,
+      "loss": 2.3159,
+      "theoretical_loss": 3.4838364621849705,
+      "tokens_seen": 1666973696
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00016600000000000002,
+      "loss": 2.5254,
+      "theoretical_loss": 3.483824629509824,
+      "tokens_seen": 1667039232
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00016800000000000002,
+      "loss": 2.2786,
+      "theoretical_loss": 3.483812797430088,
+      "tokens_seen": 1667104768
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00017,
+      "loss": 2.4052,
+      "theoretical_loss": 3.4838009659457088,
+      "tokens_seen": 1667170304
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00017199999999999998,
+      "loss": 2.5102,
+      "theoretical_loss": 3.4837891350566323,
+      "tokens_seen": 1667235840
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000174,
+      "loss": 2.4416,
+      "theoretical_loss": 3.4837773047628064,
+      "tokens_seen": 1667301376
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000176,
+      "loss": 2.2819,
+      "theoretical_loss": 3.4837654750641764,
+      "tokens_seen": 1667366912
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000178,
+      "loss": 2.4608,
+      "theoretical_loss": 3.4837536459606904,
+      "tokens_seen": 1667432448
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 2.367,
+      "theoretical_loss": 3.4837418174522936,
+      "tokens_seen": 1667497984
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000182,
+      "loss": 2.4133,
+      "theoretical_loss": 3.4837299895389333,
+      "tokens_seen": 1667563520
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000184,
+      "loss": 2.515,
+      "theoretical_loss": 3.4837181622205566,
+      "tokens_seen": 1667629056
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000186,
+      "loss": 2.4365,
+      "theoretical_loss": 3.48370633549711,
+      "tokens_seen": 1667694592
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00018800000000000002,
+      "loss": 2.4476,
+      "theoretical_loss": 3.483694509368539,
+      "tokens_seen": 1667760128
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00019,
+      "loss": 2.6584,
+      "theoretical_loss": 3.4836826838347923,
+      "tokens_seen": 1667825664
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000192,
+      "loss": 2.2917,
+      "theoretical_loss": 3.483670858895815,
+      "tokens_seen": 1667891200
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000194,
+      "loss": 2.5954,
+      "theoretical_loss": 3.483659034551555,
+      "tokens_seen": 1667956736
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00019600000000000002,
+      "loss": 2.4191,
+      "theoretical_loss": 3.4836472108019585,
+      "tokens_seen": 1668022272
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00019800000000000002,
+      "loss": 2.4538,
+      "theoretical_loss": 3.483635387646972,
+      "tokens_seen": 1668087808
+    },
+    {
+      "epoch": 0.0,
+      "objective/train/docs_used": 921237,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6976654529571533,
+      "objective/train/theoretical_loss": 3.483623565086542,
+      "objective/train/tokens_used": 27012576,
+      "theoretical_loss": 3.483623565086542,
+      "tokens_seen": 1668153344
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002,
+      "loss": 2.6155,
+      "theoretical_loss": 3.483623565086542,
+      "tokens_seen": 1668153344
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000202,
+      "loss": 2.5577,
+      "theoretical_loss": 3.4836117431206164,
+      "tokens_seen": 1668218880
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000204,
+      "loss": 2.397,
+      "theoretical_loss": 3.483599921749141,
+      "tokens_seen": 1668284416
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000206,
+      "loss": 2.5303,
+      "theoretical_loss": 3.483588100972063,
+      "tokens_seen": 1668349952
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000208,
+      "loss": 2.3655,
+      "theoretical_loss": 3.483576280789328,
+      "tokens_seen": 1668415488
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00021,
+      "loss": 2.3679,
+      "theoretical_loss": 3.483564461200885,
+      "tokens_seen": 1668481024
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000212,
+      "loss": 2.3781,
+      "theoretical_loss": 3.4835526422066785,
+      "tokens_seen": 1668546560
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000214,
+      "loss": 2.444,
+      "theoretical_loss": 3.4835408238066567,
+      "tokens_seen": 1668612096
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000216,
+      "loss": 2.3319,
+      "theoretical_loss": 3.483529006000766,
+      "tokens_seen": 1668677632
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000218,
+      "loss": 2.2344,
+      "theoretical_loss": 3.4835171887889533,
+      "tokens_seen": 1668743168
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00022,
+      "loss": 2.4793,
+      "theoretical_loss": 3.483505372171165,
+      "tokens_seen": 1668808704
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000222,
+      "loss": 2.4368,
+      "theoretical_loss": 3.483493556147349,
+      "tokens_seen": 1668874240
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000224,
+      "loss": 2.4653,
+      "theoretical_loss": 3.4834817407174503,
+      "tokens_seen": 1668939776
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00022600000000000002,
+      "loss": 2.2769,
+      "theoretical_loss": 3.4834699258814172,
+      "tokens_seen": 1669005312
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000228,
+      "loss": 2.5012,
+      "theoretical_loss": 3.4834581116391963,
+      "tokens_seen": 1669070848
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00023,
+      "loss": 2.7566,
+      "theoretical_loss": 3.483446297990734,
+      "tokens_seen": 1669136384
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00023200000000000003,
+      "loss": 2.427,
+      "theoretical_loss": 3.4834344849359775,
+      "tokens_seen": 1669201920
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00023400000000000002,
+      "loss": 2.5868,
+      "theoretical_loss": 3.4834226724748736,
+      "tokens_seen": 1669267456
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000236,
+      "loss": 2.5531,
+      "theoretical_loss": 3.483410860607369,
+      "tokens_seen": 1669332992
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00023799999999999998,
+      "loss": 2.5605,
+      "theoretical_loss": 3.4833990493334106,
+      "tokens_seen": 1669398528
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00024,
+      "loss": 2.2757,
+      "theoretical_loss": 3.4833872386529454,
+      "tokens_seen": 1669464064
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000242,
+      "loss": 2.5149,
+      "theoretical_loss": 3.48337542856592,
+      "tokens_seen": 1669529600
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000244,
+      "loss": 2.2792,
+      "theoretical_loss": 3.483363619072282,
+      "tokens_seen": 1669595136
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000246,
+      "loss": 2.4745,
+      "theoretical_loss": 3.4833518101719774,
+      "tokens_seen": 1669660672
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000248,
+      "loss": 2.37,
+      "theoretical_loss": 3.4833400018649536,
+      "tokens_seen": 1669726208
+    },
+    {
+      "epoch": 0.0,
+      "objective/train/docs_used": 921959,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.675856590270996,
+      "objective/train/theoretical_loss": 3.4833281941511576,
+      "objective/train/tokens_used": 28650976,
+      "theoretical_loss": 3.4833281941511576,
+      "tokens_seen": 1669791744
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00025,
+      "loss": 2.8538,
+      "theoretical_loss": 3.4833281941511576,
+      "tokens_seen": 1669791744
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000252,
+      "loss": 2.3873,
+      "theoretical_loss": 3.483316387030536,
+      "tokens_seen": 1669857280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000254,
+      "loss": 2.3762,
+      "theoretical_loss": 3.4833045805030363,
+      "tokens_seen": 1669922816
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000256,
+      "loss": 2.5253,
+      "theoretical_loss": 3.4832927745686044,
+      "tokens_seen": 1669988352
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00025800000000000004,
+      "loss": 2.489,
+      "theoretical_loss": 3.483280969227188,
+      "tokens_seen": 1670053888
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00026000000000000003,
+      "loss": 2.3014,
+      "theoretical_loss": 3.4832691644787337,
+      "tokens_seen": 1670119424
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000262,
+      "loss": 2.2745,
+      "theoretical_loss": 3.4832573603231887,
+      "tokens_seen": 1670184960
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000264,
+      "loss": 2.6538,
+      "theoretical_loss": 3.4832455567605,
+      "tokens_seen": 1670250496
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000266,
+      "loss": 2.3405,
+      "theoretical_loss": 3.4832337537906146,
+      "tokens_seen": 1670316032
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000268,
+      "loss": 2.4211,
+      "theoretical_loss": 3.483221951413479,
+      "tokens_seen": 1670381568
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00027,
+      "loss": 2.5642,
+      "theoretical_loss": 3.4832101496290404,
+      "tokens_seen": 1670447104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00027200000000000005,
+      "loss": 2.3653,
+      "theoretical_loss": 3.483198348437246,
+      "tokens_seen": 1670512640
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00027400000000000005,
+      "loss": 2.4734,
+      "theoretical_loss": 3.4831865478380433,
+      "tokens_seen": 1670578176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00027600000000000004,
+      "loss": 2.41,
+      "theoretical_loss": 3.483174747831378,
+      "tokens_seen": 1670643712
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00027800000000000004,
+      "loss": 2.2828,
+      "theoretical_loss": 3.4831629484171978,
+      "tokens_seen": 1670709248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028000000000000003,
+      "loss": 2.5743,
+      "theoretical_loss": 3.4831511495954497,
+      "tokens_seen": 1670774784
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028199999999999997,
+      "loss": 2.4478,
+      "theoretical_loss": 3.4831393513660807,
+      "tokens_seen": 1670840320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028399999999999996,
+      "loss": 2.5524,
+      "theoretical_loss": 3.483127553729038,
+      "tokens_seen": 1670905856
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028599999999999996,
+      "loss": 2.6412,
+      "theoretical_loss": 3.483115756684268,
+      "tokens_seen": 1670971392
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000288,
+      "loss": 2.3564,
+      "theoretical_loss": 3.483103960231719,
+      "tokens_seen": 1671036928
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029,
+      "loss": 2.5918,
+      "theoretical_loss": 3.4830921643713366,
+      "tokens_seen": 1671102464
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000292,
+      "loss": 2.2021,
+      "theoretical_loss": 3.4830803691030683,
+      "tokens_seen": 1671168000
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000294,
+      "loss": 2.3097,
+      "theoretical_loss": 3.483068574426862,
+      "tokens_seen": 1671233536
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000296,
+      "loss": 2.4126,
+      "theoretical_loss": 3.4830567803426637,
+      "tokens_seen": 1671299072
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000298,
+      "loss": 2.4376,
+      "theoretical_loss": 3.483044986850421,
+      "tokens_seen": 1671364608
+    },
+    {
+      "epoch": 0.01,
+      "objective/train/docs_used": 922598,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7050137519836426,
+      "objective/train/theoretical_loss": 3.4830331939500807,
+      "objective/train/tokens_used": 30289376,
+      "theoretical_loss": 3.4830331939500807,
+      "tokens_seen": 1671430144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003,
+      "loss": 2.5272,
+      "theoretical_loss": 3.4830331939500807,
+      "tokens_seen": 1671430144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000302,
+      "loss": 2.4159,
+      "theoretical_loss": 3.48302140164159,
+      "tokens_seen": 1671495680
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000304,
+      "loss": 2.3413,
+      "theoretical_loss": 3.483009609924897,
+      "tokens_seen": 1671561216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000306,
+      "loss": 2.7255,
+      "theoretical_loss": 3.482997818799947,
+      "tokens_seen": 1671626752
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000308,
+      "loss": 2.4114,
+      "theoretical_loss": 3.482986028266688,
+      "tokens_seen": 1671692288
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00031,
+      "loss": 2.3985,
+      "theoretical_loss": 3.4829742383250673,
+      "tokens_seen": 1671757824
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000312,
+      "loss": 2.4447,
+      "theoretical_loss": 3.4829624489750315,
+      "tokens_seen": 1671823360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000314,
+      "loss": 2.1292,
+      "theoretical_loss": 3.4829506602165283,
+      "tokens_seen": 1671888896
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000316,
+      "loss": 2.4829,
+      "theoretical_loss": 3.4829388720495045,
+      "tokens_seen": 1671954432
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00031800000000000003,
+      "loss": 2.3397,
+      "theoretical_loss": 3.482927084473907,
+      "tokens_seen": 1672019968
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00032,
+      "loss": 2.5797,
+      "theoretical_loss": 3.4829152974896838,
+      "tokens_seen": 1672085504
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000322,
+      "loss": 2.6122,
+      "theoretical_loss": 3.482903511096781,
+      "tokens_seen": 1672151040
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000324,
+      "loss": 2.3373,
+      "theoretical_loss": 3.482891725295146,
+      "tokens_seen": 1672216576
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000326,
+      "loss": 2.2181,
+      "theoretical_loss": 3.482879940084727,
+      "tokens_seen": 1672282112
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000328,
+      "loss": 2.5586,
+      "theoretical_loss": 3.48286815546547,
+      "tokens_seen": 1672347648
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00033,
+      "loss": 2.4485,
+      "theoretical_loss": 3.4828563714373226,
+      "tokens_seen": 1672413184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00033200000000000005,
+      "loss": 2.6228,
+      "theoretical_loss": 3.482844588000232,
+      "tokens_seen": 1672478720
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00033400000000000004,
+      "loss": 2.2578,
+      "theoretical_loss": 3.4828328051541453,
+      "tokens_seen": 1672544256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00033600000000000004,
+      "loss": 2.5647,
+      "theoretical_loss": 3.48282102289901,
+      "tokens_seen": 1672609792
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00033800000000000003,
+      "loss": 2.27,
+      "theoretical_loss": 3.482809241234773,
+      "tokens_seen": 1672675328
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00034,
+      "loss": 2.8122,
+      "theoretical_loss": 3.4827974601613816,
+      "tokens_seen": 1672740864
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000342,
+      "loss": 2.6542,
+      "theoretical_loss": 3.482785679678783,
+      "tokens_seen": 1672806400
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00034399999999999996,
+      "loss": 2.4875,
+      "theoretical_loss": 3.482773899786924,
+      "tokens_seen": 1672871936
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000346,
+      "loss": 2.5134,
+      "theoretical_loss": 3.482762120485753,
+      "tokens_seen": 1672937472
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000348,
+      "loss": 2.5356,
+      "theoretical_loss": 3.4827503417752164,
+      "tokens_seen": 1673003008
+    },
+    {
+      "epoch": 0.01,
+      "objective/train/docs_used": 923228,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5057578086853027,
+      "objective/train/theoretical_loss": 3.482738563655261,
+      "objective/train/tokens_used": 31927776,
+      "theoretical_loss": 3.482738563655261,
+      "tokens_seen": 1673068544
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00035,
+      "loss": 2.7017,
+      "theoretical_loss": 3.482738563655261,
+      "tokens_seen": 1673068544
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000352,
+      "loss": 2.3692,
+      "theoretical_loss": 3.482726786125835,
+      "tokens_seen": 1673134080
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000354,
+      "loss": 2.4674,
+      "theoretical_loss": 3.4827150091868853,
+      "tokens_seen": 1673199616
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000356,
+      "loss": 2.3015,
+      "theoretical_loss": 3.482703232838359,
+      "tokens_seen": 1673265152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000358,
+      "loss": 2.4719,
+      "theoretical_loss": 3.482691457080204,
+      "tokens_seen": 1673330688
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00035999999999999997,
+      "loss": 2.3731,
+      "theoretical_loss": 3.4826796819123667,
+      "tokens_seen": 1673396224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000362,
+      "loss": 2.3984,
+      "theoretical_loss": 3.482667907334795,
+      "tokens_seen": 1673461760
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000364,
+      "loss": 2.5499,
+      "theoretical_loss": 3.4826561333474357,
+      "tokens_seen": 1673527296
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000366,
+      "loss": 2.4005,
+      "theoretical_loss": 3.482644359950237,
+      "tokens_seen": 1673592832
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000368,
+      "loss": 2.5822,
+      "theoretical_loss": 3.482632587143145,
+      "tokens_seen": 1673658368
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00037,
+      "loss": 2.5947,
+      "theoretical_loss": 3.4826208149261078,
+      "tokens_seen": 1673723904
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000372,
+      "loss": 2.4242,
+      "theoretical_loss": 3.482609043299073,
+      "tokens_seen": 1673789440
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000374,
+      "loss": 2.5893,
+      "theoretical_loss": 3.482597272261987,
+      "tokens_seen": 1673854976
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00037600000000000003,
+      "loss": 2.7811,
+      "theoretical_loss": 3.482585501814798,
+      "tokens_seen": 1673920512
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000378,
+      "loss": 2.6416,
+      "theoretical_loss": 3.4825737319574523,
+      "tokens_seen": 1673986048
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00038,
+      "loss": 2.3689,
+      "theoretical_loss": 3.482561962689899,
+      "tokens_seen": 1674051584
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000382,
+      "loss": 2.4532,
+      "theoretical_loss": 3.4825501940120835,
+      "tokens_seen": 1674117120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000384,
+      "loss": 2.2893,
+      "theoretical_loss": 3.4825384259239547,
+      "tokens_seen": 1674182656
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000386,
+      "loss": 2.4148,
+      "theoretical_loss": 3.4825266584254586,
+      "tokens_seen": 1674248192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000388,
+      "loss": 2.4807,
+      "theoretical_loss": 3.482514891516544,
+      "tokens_seen": 1674313728
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00039000000000000005,
+      "loss": 2.5216,
+      "theoretical_loss": 3.482503125197157,
+      "tokens_seen": 1674379264
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00039200000000000004,
+      "loss": 2.2569,
+      "theoretical_loss": 3.4824913594672458,
+      "tokens_seen": 1674444800
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00039400000000000004,
+      "loss": 2.2831,
+      "theoretical_loss": 3.4824795943267577,
+      "tokens_seen": 1674510336
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00039600000000000003,
+      "loss": 2.5188,
+      "theoretical_loss": 3.48246782977564,
+      "tokens_seen": 1674575872
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000398,
+      "loss": 2.576,
+      "theoretical_loss": 3.4824560658138397,
+      "tokens_seen": 1674641408
+    },
+    {
+      "epoch": 0.01,
+      "objective/train/docs_used": 924425,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.702085256576538,
+      "objective/train/theoretical_loss": 3.482444302441305,
+      "objective/train/tokens_used": 33566176,
+      "theoretical_loss": 3.482444302441305,
+      "tokens_seen": 1674706944
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004,
+      "loss": 2.6435,
+      "theoretical_loss": 3.482444302441305,
+      "tokens_seen": 1674706944
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000402,
+      "loss": 2.4602,
+      "theoretical_loss": 3.482432539657983,
+      "tokens_seen": 1674772480
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000404,
+      "loss": 2.3951,
+      "theoretical_loss": 3.4824207774638207,
+      "tokens_seen": 1674838016
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00040600000000000006,
+      "loss": 2.4707,
+      "theoretical_loss": 3.4824090158587664,
+      "tokens_seen": 1674903552
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000408,
+      "loss": 2.4597,
+      "theoretical_loss": 3.482397254842767,
+      "tokens_seen": 1674969088
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00041,
+      "loss": 2.5413,
+      "theoretical_loss": 3.4823854944157695,
+      "tokens_seen": 1675034624
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000412,
+      "loss": 2.4902,
+      "theoretical_loss": 3.482373734577722,
+      "tokens_seen": 1675100160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000414,
+      "loss": 2.3291,
+      "theoretical_loss": 3.482361975328572,
+      "tokens_seen": 1675165696
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000416,
+      "loss": 2.4068,
+      "theoretical_loss": 3.4823502166682667,
+      "tokens_seen": 1675231232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00041799999999999997,
+      "loss": 2.4366,
+      "theoretical_loss": 3.482338458596754,
+      "tokens_seen": 1675296768
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00042,
+      "loss": 2.5455,
+      "theoretical_loss": 3.4823267011139807,
+      "tokens_seen": 1675362304
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000422,
+      "loss": 2.6574,
+      "theoretical_loss": 3.482314944219895,
+      "tokens_seen": 1675427840
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000424,
+      "loss": 2.3464,
+      "theoretical_loss": 3.482303187914444,
+      "tokens_seen": 1675493376
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000426,
+      "loss": 2.5067,
+      "theoretical_loss": 3.482291432197575,
+      "tokens_seen": 1675558912
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000428,
+      "loss": 2.4157,
+      "theoretical_loss": 3.4822796770692364,
+      "tokens_seen": 1675624448
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00043,
+      "loss": 2.5832,
+      "theoretical_loss": 3.482267922529374,
+      "tokens_seen": 1675689984
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000432,
+      "loss": 2.8137,
+      "theoretical_loss": 3.4822561685779374,
+      "tokens_seen": 1675755520
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00043400000000000003,
+      "loss": 2.3953,
+      "theoretical_loss": 3.482244415214873,
+      "tokens_seen": 1675821056
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000436,
+      "loss": 2.6011,
+      "theoretical_loss": 3.482232662440128,
+      "tokens_seen": 1675886592
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000438,
+      "loss": 2.2128,
+      "theoretical_loss": 3.482220910253651,
+      "tokens_seen": 1675952128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00044,
+      "loss": 2.4778,
+      "theoretical_loss": 3.4822091586553885,
+      "tokens_seen": 1676017664
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000442,
+      "loss": 2.2702,
+      "theoretical_loss": 3.482197407645289,
+      "tokens_seen": 1676083200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000444,
+      "loss": 2.5679,
+      "theoretical_loss": 3.482185657223299,
+      "tokens_seen": 1676148736
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000446,
+      "loss": 2.4653,
+      "theoretical_loss": 3.4821739073893676,
+      "tokens_seen": 1676214272
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000448,
+      "loss": 2.27,
+      "theoretical_loss": 3.482162158143441,
+      "tokens_seen": 1676279808
+    },
+    {
+      "epoch": 0.01,
+      "objective/train/docs_used": 924948,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.733900785446167,
+      "objective/train/theoretical_loss": 3.482150409485467,
+      "objective/train/tokens_used": 35204576,
+      "theoretical_loss": 3.482150409485467,
+      "tokens_seen": 1676345344
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00045000000000000004,
+      "loss": 2.5402,
+      "theoretical_loss": 3.482150409485467,
+      "tokens_seen": 1676345344
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00045200000000000004,
+      "loss": 2.5286,
+      "theoretical_loss": 3.4821386614153944,
+      "tokens_seen": 1676410880
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00045400000000000003,
+      "loss": 2.6377,
+      "theoretical_loss": 3.482126913933169,
+      "tokens_seen": 1676476416
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000456,
+      "loss": 2.5819,
+      "theoretical_loss": 3.4821151670387396,
+      "tokens_seen": 1676541952
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000458,
+      "loss": 2.5565,
+      "theoretical_loss": 3.482103420732053,
+      "tokens_seen": 1676607488
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00046,
+      "loss": 2.5833,
+      "theoretical_loss": 3.482091675013058,
+      "tokens_seen": 1676673024
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000462,
+      "loss": 2.6621,
+      "theoretical_loss": 3.482079929881701,
+      "tokens_seen": 1676738560
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00046400000000000006,
+      "loss": 2.3564,
+      "theoretical_loss": 3.4820681853379307,
+      "tokens_seen": 1676804096
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00046600000000000005,
+      "loss": 2.4111,
+      "theoretical_loss": 3.482056441381694,
+      "tokens_seen": 1676869632
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00046800000000000005,
+      "loss": 2.5978,
+      "theoretical_loss": 3.4820446980129383,
+      "tokens_seen": 1676935168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00047,
+      "loss": 2.5111,
+      "theoretical_loss": 3.4820329552316123,
+      "tokens_seen": 1677000704
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000472,
+      "loss": 2.5309,
+      "theoretical_loss": 3.4820212130376627,
+      "tokens_seen": 1677066240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000474,
+      "loss": 2.4254,
+      "theoretical_loss": 3.482009471431038,
+      "tokens_seen": 1677131776
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00047599999999999997,
+      "loss": 2.3422,
+      "theoretical_loss": 3.4819977304116847,
+      "tokens_seen": 1677197312
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00047799999999999996,
+      "loss": 2.5434,
+      "theoretical_loss": 3.4819859899795516,
+      "tokens_seen": 1677262848
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00048,
+      "loss": 2.5123,
+      "theoretical_loss": 3.481974250134586,
+      "tokens_seen": 1677328384
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000482,
+      "loss": 2.6815,
+      "theoretical_loss": 3.481962510876736,
+      "tokens_seen": 1677393920
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000484,
+      "loss": 2.3318,
+      "theoretical_loss": 3.481950772205948,
+      "tokens_seen": 1677459456
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000486,
+      "loss": 2.4658,
+      "theoretical_loss": 3.481939034122171,
+      "tokens_seen": 1677524992
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000488,
+      "loss": 2.3931,
+      "theoretical_loss": 3.4819272966253525,
+      "tokens_seen": 1677590528
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00049,
+      "loss": 2.5888,
+      "theoretical_loss": 3.4819155597154396,
+      "tokens_seen": 1677656064
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000492,
+      "loss": 2.5776,
+      "theoretical_loss": 3.481903823392381,
+      "tokens_seen": 1677721600
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000494,
+      "loss": 2.3611,
+      "theoretical_loss": 3.4818920876561235,
+      "tokens_seen": 1677787136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000496,
+      "loss": 2.501,
+      "theoretical_loss": 3.4818803525066153,
+      "tokens_seen": 1677852672
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000498,
+      "loss": 2.6304,
+      "theoretical_loss": 3.481868617943804,
+      "tokens_seen": 1677918208
+    },
+    {
+      "epoch": 0.01,
+      "objective/train/docs_used": 926151,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.59950590133667,
+      "objective/train/theoretical_loss": 3.4818568839676374,
+      "objective/train/tokens_used": 36842976,
+      "theoretical_loss": 3.4818568839676374,
+      "tokens_seen": 1677983744
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0005,
+      "loss": 2.582,
+      "theoretical_loss": 3.4818568839676374,
+      "tokens_seen": 1677983744
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000499979797979798,
+      "loss": 2.3636,
+      "theoretical_loss": 3.481845150578063,
+      "tokens_seen": 1678049280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004999595959595959,
+      "loss": 2.4696,
+      "theoretical_loss": 3.4818334177750296,
+      "tokens_seen": 1678114816
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004999393939393939,
+      "loss": 2.5003,
+      "theoretical_loss": 3.481821685558484,
+      "tokens_seen": 1678180352
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000499919191919192,
+      "loss": 2.6498,
+      "theoretical_loss": 3.481809953928374,
+      "tokens_seen": 1678245888
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00049989898989899,
+      "loss": 2.5553,
+      "theoretical_loss": 3.4817982228846476,
+      "tokens_seen": 1678311424
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004998787878787879,
+      "loss": 2.5759,
+      "theoretical_loss": 3.4817864924272524,
+      "tokens_seen": 1678376960
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004998585858585859,
+      "loss": 2.7333,
+      "theoretical_loss": 3.481774762556137,
+      "tokens_seen": 1678442496
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004998383838383839,
+      "loss": 2.6417,
+      "theoretical_loss": 3.4817630332712484,
+      "tokens_seen": 1678508032
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004998181818181818,
+      "loss": 2.8269,
+      "theoretical_loss": 3.4817513045725343,
+      "tokens_seen": 1678573568
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004997979797979798,
+      "loss": 2.4396,
+      "theoretical_loss": 3.481739576459943,
+      "tokens_seen": 1678639104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004997777777777778,
+      "loss": 2.6778,
+      "theoretical_loss": 3.4817278489334225,
+      "tokens_seen": 1678704640
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004997575757575757,
+      "loss": 2.4876,
+      "theoretical_loss": 3.48171612199292,
+      "tokens_seen": 1678770176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004997373737373737,
+      "loss": 2.5085,
+      "theoretical_loss": 3.481704395638383,
+      "tokens_seen": 1678835712
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004997171717171717,
+      "loss": 2.523,
+      "theoretical_loss": 3.4816926698697612,
+      "tokens_seen": 1678901248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004996969696969697,
+      "loss": 2.5983,
+      "theoretical_loss": 3.4816809446870005,
+      "tokens_seen": 1678966784
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004996767676767676,
+      "loss": 2.5985,
+      "theoretical_loss": 3.48166922009005,
+      "tokens_seen": 1679032320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004996565656565657,
+      "loss": 2.6887,
+      "theoretical_loss": 3.4816574960788564,
+      "tokens_seen": 1679097856
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004996363636363637,
+      "loss": 2.6201,
+      "theoretical_loss": 3.481645772653369,
+      "tokens_seen": 1679163392
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004996161616161617,
+      "loss": 2.6373,
+      "theoretical_loss": 3.4816340498135343,
+      "tokens_seen": 1679228928
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004995959595959596,
+      "loss": 2.5179,
+      "theoretical_loss": 3.4816223275593012,
+      "tokens_seen": 1679294464
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004995757575757576,
+      "loss": 2.514,
+      "theoretical_loss": 3.4816106058906175,
+      "tokens_seen": 1679360000
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004995555555555555,
+      "loss": 2.5432,
+      "theoretical_loss": 3.48159888480743,
+      "tokens_seen": 1679425536
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004995353535353536,
+      "loss": 2.5299,
+      "theoretical_loss": 3.481587164309688,
+      "tokens_seen": 1679491072
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004995151515151515,
+      "loss": 2.3886,
+      "theoretical_loss": 3.481575444397339,
+      "tokens_seen": 1679556608
+    },
+    {
+      "epoch": 0.01,
+      "objective/train/docs_used": 927219,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.202238082885742,
+      "objective/train/theoretical_loss": 3.4815637250703304,
+      "objective/train/tokens_used": 38481376,
+      "theoretical_loss": 3.4815637250703304,
+      "tokens_seen": 1679622144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004994949494949495,
+      "loss": 2.5746,
+      "theoretical_loss": 3.4815637250703304,
+      "tokens_seen": 1679622144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004994747474747474,
+      "loss": 2.6998,
+      "theoretical_loss": 3.4815520063286103,
+      "tokens_seen": 1679687680
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004994545454545454,
+      "loss": 2.4526,
+      "theoretical_loss": 3.4815402881721274,
+      "tokens_seen": 1679753216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004994343434343435,
+      "loss": 2.6013,
+      "theoretical_loss": 3.4815285706008288,
+      "tokens_seen": 1679818752
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004994141414141415,
+      "loss": 2.2639,
+      "theoretical_loss": 3.4815168536146626,
+      "tokens_seen": 1679884288
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004993939393939394,
+      "loss": 2.5337,
+      "theoretical_loss": 3.481505137213577,
+      "tokens_seen": 1679949824
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004993737373737374,
+      "loss": 2.6064,
+      "theoretical_loss": 3.48149342139752,
+      "tokens_seen": 1680015360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004993535353535354,
+      "loss": 2.4297,
+      "theoretical_loss": 3.4814817061664396,
+      "tokens_seen": 1680080896
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004993333333333334,
+      "loss": 2.548,
+      "theoretical_loss": 3.481469991520283,
+      "tokens_seen": 1680146432
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004993131313131313,
+      "loss": 2.5833,
+      "theoretical_loss": 3.481458277458999,
+      "tokens_seen": 1680211968
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004992929292929293,
+      "loss": 2.6257,
+      "theoretical_loss": 3.4814465639825354,
+      "tokens_seen": 1680277504
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004992727272727272,
+      "loss": 2.2464,
+      "theoretical_loss": 3.4814348510908397,
+      "tokens_seen": 1680343040
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004992525252525252,
+      "loss": 2.526,
+      "theoretical_loss": 3.481423138783861,
+      "tokens_seen": 1680408576
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004992323232323232,
+      "loss": 2.5034,
+      "theoretical_loss": 3.4814114270615466,
+      "tokens_seen": 1680474112
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004992121212121212,
+      "loss": 2.5144,
+      "theoretical_loss": 3.4813997159238443,
+      "tokens_seen": 1680539648
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004991919191919192,
+      "loss": 2.7348,
+      "theoretical_loss": 3.481388005370702,
+      "tokens_seen": 1680605184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004991717171717172,
+      "loss": 2.3679,
+      "theoretical_loss": 3.481376295402069,
+      "tokens_seen": 1680670720
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004991515151515152,
+      "loss": 2.5985,
+      "theoretical_loss": 3.481364586017892,
+      "tokens_seen": 1680736256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004991313131313132,
+      "loss": 2.6316,
+      "theoretical_loss": 3.4813528772181193,
+      "tokens_seen": 1680801792
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004991111111111111,
+      "loss": 2.601,
+      "theoretical_loss": 3.4813411690026994,
+      "tokens_seen": 1680867328
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004990909090909091,
+      "loss": 2.57,
+      "theoretical_loss": 3.4813294613715797,
+      "tokens_seen": 1680932864
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004990707070707071,
+      "loss": 2.4923,
+      "theoretical_loss": 3.4813177543247096,
+      "tokens_seen": 1680998400
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004990505050505051,
+      "loss": 2.5084,
+      "theoretical_loss": 3.4813060478620352,
+      "tokens_seen": 1681063936
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000499030303030303,
+      "loss": 2.6574,
+      "theoretical_loss": 3.4812943419835065,
+      "tokens_seen": 1681129472
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000499010101010101,
+      "loss": 2.7915,
+      "theoretical_loss": 3.48128263668907,
+      "tokens_seen": 1681195008
+    },
+    {
+      "epoch": 0.01,
+      "objective/train/docs_used": 927968,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5613365173339844,
+      "objective/train/theoretical_loss": 3.481270931978675,
+      "objective/train/tokens_used": 40119776,
+      "theoretical_loss": 3.481270931978675,
+      "tokens_seen": 1681260544
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000498989898989899,
+      "loss": 2.4737,
+      "theoretical_loss": 3.481270931978675,
+      "tokens_seen": 1681260544
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000498969696969697,
+      "loss": 2.6981,
+      "theoretical_loss": 3.4812592278522687,
+      "tokens_seen": 1681326080
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000498949494949495,
+      "loss": 2.6027,
+      "theoretical_loss": 3.4812475243098,
+      "tokens_seen": 1681391616
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000498929292929293,
+      "loss": 2.6348,
+      "theoretical_loss": 3.4812358213512162,
+      "tokens_seen": 1681457152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004989090909090909,
+      "loss": 2.6082,
+      "theoretical_loss": 3.4812241189764657,
+      "tokens_seen": 1681522688
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004988888888888889,
+      "loss": 2.6419,
+      "theoretical_loss": 3.4812124171854966,
+      "tokens_seen": 1681588224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004988686868686869,
+      "loss": 2.6168,
+      "theoretical_loss": 3.4812007159782574,
+      "tokens_seen": 1681653760
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004988484848484849,
+      "loss": 2.5625,
+      "theoretical_loss": 3.481189015354696,
+      "tokens_seen": 1681719296
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004988282828282828,
+      "loss": 2.6334,
+      "theoretical_loss": 3.4811773153147607,
+      "tokens_seen": 1681784832
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004988080808080808,
+      "loss": 2.6787,
+      "theoretical_loss": 3.481165615858399,
+      "tokens_seen": 1681850368
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004987878787878788,
+      "loss": 2.8212,
+      "theoretical_loss": 3.48115391698556,
+      "tokens_seen": 1681915904
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004987676767676767,
+      "loss": 2.5745,
+      "theoretical_loss": 3.481142218696191,
+      "tokens_seen": 1681981440
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004987474747474747,
+      "loss": 2.848,
+      "theoretical_loss": 3.4811305209902406,
+      "tokens_seen": 1682046976
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004987272727272728,
+      "loss": 2.4886,
+      "theoretical_loss": 3.481118823867657,
+      "tokens_seen": 1682112512
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004987070707070708,
+      "loss": 2.4485,
+      "theoretical_loss": 3.4811071273283885,
+      "tokens_seen": 1682178048
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004986868686868687,
+      "loss": 2.7483,
+      "theoretical_loss": 3.481095431372383,
+      "tokens_seen": 1682243584
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004986666666666667,
+      "loss": 2.472,
+      "theoretical_loss": 3.4810837359995883,
+      "tokens_seen": 1682309120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004986464646464647,
+      "loss": 2.6229,
+      "theoretical_loss": 3.481072041209954,
+      "tokens_seen": 1682374656
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004986262626262627,
+      "loss": 2.679,
+      "theoretical_loss": 3.4810603470034263,
+      "tokens_seen": 1682440192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004986060606060606,
+      "loss": 2.5162,
+      "theoretical_loss": 3.481048653379955,
+      "tokens_seen": 1682505728
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004985858585858586,
+      "loss": 2.6945,
+      "theoretical_loss": 3.4810369603394884,
+      "tokens_seen": 1682571264
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004985656565656565,
+      "loss": 2.6558,
+      "theoretical_loss": 3.4810252678819733,
+      "tokens_seen": 1682636800
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004985454545454545,
+      "loss": 2.6452,
+      "theoretical_loss": 3.481013576007359,
+      "tokens_seen": 1682702336
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004985252525252525,
+      "loss": 2.6641,
+      "theoretical_loss": 3.481001884715594,
+      "tokens_seen": 1682767872
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004985050505050506,
+      "loss": 2.5645,
+      "theoretical_loss": 3.4809901940066257,
+      "tokens_seen": 1682833408
+    },
+    {
+      "epoch": 0.01,
+      "objective/train/docs_used": 929249,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.0171878337860107,
+      "objective/train/theoretical_loss": 3.4809785038804026,
+      "objective/train/tokens_used": 41758176,
+      "theoretical_loss": 3.4809785038804026,
+      "tokens_seen": 1682898944
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004984848484848485,
+      "loss": 2.6663,
+      "theoretical_loss": 3.4809785038804026,
+      "tokens_seen": 1682898944
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004984646464646465,
+      "loss": 2.733,
+      "theoretical_loss": 3.480966814336873,
+      "tokens_seen": 1682964480
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004984444444444445,
+      "loss": 2.5084,
+      "theoretical_loss": 3.480955125375986,
+      "tokens_seen": 1683030016
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004984242424242425,
+      "loss": 2.7494,
+      "theoretical_loss": 3.4809434369976886,
+      "tokens_seen": 1683095552
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004984040404040404,
+      "loss": 2.3954,
+      "theoretical_loss": 3.48093174920193,
+      "tokens_seen": 1683161088
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004983838383838384,
+      "loss": 2.5031,
+      "theoretical_loss": 3.480920061988657,
+      "tokens_seen": 1683226624
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004983636363636363,
+      "loss": 2.5551,
+      "theoretical_loss": 3.48090837535782,
+      "tokens_seen": 1683292160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004983434343434344,
+      "loss": 2.5852,
+      "theoretical_loss": 3.480896689309366,
+      "tokens_seen": 1683357696
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004983232323232323,
+      "loss": 2.7372,
+      "theoretical_loss": 3.4808850038432437,
+      "tokens_seen": 1683423232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004983030303030303,
+      "loss": 2.6248,
+      "theoretical_loss": 3.480873318959401,
+      "tokens_seen": 1683488768
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004982828282828282,
+      "loss": 2.3415,
+      "theoretical_loss": 3.4808616346577868,
+      "tokens_seen": 1683554304
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004982626262626263,
+      "loss": 2.6582,
+      "theoretical_loss": 3.480849950938349,
+      "tokens_seen": 1683619840
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004982424242424243,
+      "loss": 2.5619,
+      "theoretical_loss": 3.4808382678010363,
+      "tokens_seen": 1683685376
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004982222222222223,
+      "loss": 2.7625,
+      "theoretical_loss": 3.480826585245797,
+      "tokens_seen": 1683750912
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004982020202020202,
+      "loss": 2.5896,
+      "theoretical_loss": 3.480814903272579,
+      "tokens_seen": 1683816448
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004981818181818182,
+      "loss": 2.7364,
+      "theoretical_loss": 3.4808032218813305,
+      "tokens_seen": 1683881984
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004981616161616162,
+      "loss": 2.7586,
+      "theoretical_loss": 3.4807915410720005,
+      "tokens_seen": 1683947520
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004981414141414142,
+      "loss": 2.5097,
+      "theoretical_loss": 3.480779860844538,
+      "tokens_seen": 1684013056
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004981212121212121,
+      "loss": 2.7616,
+      "theoretical_loss": 3.4807681811988895,
+      "tokens_seen": 1684078592
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004981010101010101,
+      "loss": 2.5229,
+      "theoretical_loss": 3.480756502135005,
+      "tokens_seen": 1684144128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000498080808080808,
+      "loss": 2.5937,
+      "theoretical_loss": 3.4807448236528318,
+      "tokens_seen": 1684209664
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000498060606060606,
+      "loss": 2.705,
+      "theoretical_loss": 3.480733145752319,
+      "tokens_seen": 1684275200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000498040404040404,
+      "loss": 2.6938,
+      "theoretical_loss": 3.4807214684334147,
+      "tokens_seen": 1684340736
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004980202020202021,
+      "loss": 2.6702,
+      "theoretical_loss": 3.480709791696068,
+      "tokens_seen": 1684406272
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000498,
+      "loss": 2.466,
+      "theoretical_loss": 3.4806981155402257,
+      "tokens_seen": 1684471808
+    },
+    {
+      "epoch": 0.01,
+      "objective/train/docs_used": 929812,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.769881010055542,
+      "objective/train/theoretical_loss": 3.4806864399658375,
+      "objective/train/tokens_used": 43396576,
+      "theoretical_loss": 3.4806864399658375,
+      "tokens_seen": 1684537344
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000497979797979798,
+      "loss": 2.5807,
+      "theoretical_loss": 3.4806864399658375,
+      "tokens_seen": 1684537344
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000497959595959596,
+      "loss": 2.532,
+      "theoretical_loss": 3.4806747649728518,
+      "tokens_seen": 1684602880
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000497939393939394,
+      "loss": 2.4939,
+      "theoretical_loss": 3.480663090561217,
+      "tokens_seen": 1684668416
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004979191919191919,
+      "loss": 2.6286,
+      "theoretical_loss": 3.480651416730881,
+      "tokens_seen": 1684733952
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004978989898989899,
+      "loss": 2.5486,
+      "theoretical_loss": 3.4806397434817926,
+      "tokens_seen": 1684799488
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004978787878787879,
+      "loss": 2.4135,
+      "theoretical_loss": 3.4806280708139,
+      "tokens_seen": 1684865024
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004978585858585858,
+      "loss": 2.487,
+      "theoretical_loss": 3.480616398727152,
+      "tokens_seen": 1684930560
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004978383838383838,
+      "loss": 2.5598,
+      "theoretical_loss": 3.480604727221497,
+      "tokens_seen": 1684996096
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004978181818181818,
+      "loss": 2.6736,
+      "theoretical_loss": 3.4805930562968834,
+      "tokens_seen": 1685061632
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004977979797979799,
+      "loss": 2.4844,
+      "theoretical_loss": 3.4805813859532595,
+      "tokens_seen": 1685127168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004977777777777778,
+      "loss": 2.7091,
+      "theoretical_loss": 3.4805697161905744,
+      "tokens_seen": 1685192704
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004977575757575758,
+      "loss": 2.5024,
+      "theoretical_loss": 3.480558047008776,
+      "tokens_seen": 1685258240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004977373737373738,
+      "loss": 2.5198,
+      "theoretical_loss": 3.480546378407812,
+      "tokens_seen": 1685323776
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004977171717171717,
+      "loss": 2.44,
+      "theoretical_loss": 3.4805347103876327,
+      "tokens_seen": 1685389312
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004976969696969697,
+      "loss": 2.524,
+      "theoretical_loss": 3.480523042948186,
+      "tokens_seen": 1685454848
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004976767676767677,
+      "loss": 2.6416,
+      "theoretical_loss": 3.48051137608942,
+      "tokens_seen": 1685520384
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004976565656565656,
+      "loss": 2.5126,
+      "theoretical_loss": 3.480499709811283,
+      "tokens_seen": 1685585920
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004976363636363636,
+      "loss": 2.6729,
+      "theoretical_loss": 3.4804880441137245,
+      "tokens_seen": 1685651456
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004976161616161616,
+      "loss": 2.4778,
+      "theoretical_loss": 3.480476378996692,
+      "tokens_seen": 1685716992
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004975959595959596,
+      "loss": 2.5328,
+      "theoretical_loss": 3.4804647144601346,
+      "tokens_seen": 1685782528
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004975757575757575,
+      "loss": 2.6505,
+      "theoretical_loss": 3.480453050504001,
+      "tokens_seen": 1685848064
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004975555555555556,
+      "loss": 2.5126,
+      "theoretical_loss": 3.4804413871282396,
+      "tokens_seen": 1685913600
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004975353535353536,
+      "loss": 2.6655,
+      "theoretical_loss": 3.480429724332798,
+      "tokens_seen": 1685979136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004975151515151516,
+      "loss": 2.5183,
+      "theoretical_loss": 3.480418062117627,
+      "tokens_seen": 1686044672
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004974949494949495,
+      "loss": 2.2442,
+      "theoretical_loss": 3.480406400482673,
+      "tokens_seen": 1686110208
+    },
+    {
+      "epoch": 0.01,
+      "objective/train/docs_used": 931049,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.4905221462249756,
+      "objective/train/theoretical_loss": 3.4803947394278856,
+      "objective/train/tokens_used": 45034976,
+      "theoretical_loss": 3.4803947394278856,
+      "tokens_seen": 1686175744
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004974747474747475,
+      "loss": 2.4943,
+      "theoretical_loss": 3.4803947394278856,
+      "tokens_seen": 1686175744
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004974545454545454,
+      "loss": 2.2533,
+      "theoretical_loss": 3.4803830789532126,
+      "tokens_seen": 1686241280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004974343434343435,
+      "loss": 2.7416,
+      "theoretical_loss": 3.4803714190586037,
+      "tokens_seen": 1686306816
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004974141414141414,
+      "loss": 2.4504,
+      "theoretical_loss": 3.4803597597440072,
+      "tokens_seen": 1686372352
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004973939393939394,
+      "loss": 2.7452,
+      "theoretical_loss": 3.4803481010093718,
+      "tokens_seen": 1686437888
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004973737373737373,
+      "loss": 2.6507,
+      "theoretical_loss": 3.4803364428546453,
+      "tokens_seen": 1686503424
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004973535353535353,
+      "loss": 2.2988,
+      "theoretical_loss": 3.4803247852797767,
+      "tokens_seen": 1686568960
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004973333333333334,
+      "loss": 2.5663,
+      "theoretical_loss": 3.480313128284715,
+      "tokens_seen": 1686634496
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004973131313131314,
+      "loss": 2.518,
+      "theoretical_loss": 3.4803014718694087,
+      "tokens_seen": 1686700032
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004972929292929293,
+      "loss": 2.4251,
+      "theoretical_loss": 3.4802898160338063,
+      "tokens_seen": 1686765568
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004972727272727273,
+      "loss": 2.5484,
+      "theoretical_loss": 3.480278160777856,
+      "tokens_seen": 1686831104
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004972525252525253,
+      "loss": 2.3676,
+      "theoretical_loss": 3.4802665061015077,
+      "tokens_seen": 1686896640
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004972323232323233,
+      "loss": 2.3593,
+      "theoretical_loss": 3.480254852004709,
+      "tokens_seen": 1686962176
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004972121212121212,
+      "loss": 2.5334,
+      "theoretical_loss": 3.4802431984874094,
+      "tokens_seen": 1687027712
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004971919191919192,
+      "loss": 2.5443,
+      "theoretical_loss": 3.4802315455495565,
+      "tokens_seen": 1687093248
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004971717171717171,
+      "loss": 2.473,
+      "theoretical_loss": 3.4802198931910993,
+      "tokens_seen": 1687158784
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004971515151515152,
+      "loss": 2.5458,
+      "theoretical_loss": 3.480208241411987,
+      "tokens_seen": 1687224320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004971313131313131,
+      "loss": 2.5911,
+      "theoretical_loss": 3.4801965902121683,
+      "tokens_seen": 1687289856
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004971111111111111,
+      "loss": 2.5036,
+      "theoretical_loss": 3.480184939591591,
+      "tokens_seen": 1687355392
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000497090909090909,
+      "loss": 2.452,
+      "theoretical_loss": 3.480173289550205,
+      "tokens_seen": 1687420928
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004970707070707071,
+      "loss": 2.3563,
+      "theoretical_loss": 3.4801616400879585,
+      "tokens_seen": 1687486464
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004970505050505051,
+      "loss": 2.6628,
+      "theoretical_loss": 3.4801499912048,
+      "tokens_seen": 1687552000
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004970303030303031,
+      "loss": 2.4553,
+      "theoretical_loss": 3.480138342900678,
+      "tokens_seen": 1687617536
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000497010101010101,
+      "loss": 2.6518,
+      "theoretical_loss": 3.480126695175542,
+      "tokens_seen": 1687683072
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000496989898989899,
+      "loss": 2.6693,
+      "theoretical_loss": 3.48011504802934,
+      "tokens_seen": 1687748608
+    },
+    {
+      "epoch": 0.02,
+      "objective/train/docs_used": 931336,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.531233072280884,
+      "objective/train/theoretical_loss": 3.4801034014620216,
+      "objective/train/tokens_used": 46673376,
+      "theoretical_loss": 3.4801034014620216,
+      "tokens_seen": 1687814144
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000496969696969697,
+      "loss": 2.5694,
+      "theoretical_loss": 3.4801034014620216,
+      "tokens_seen": 1687814144
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000496949494949495,
+      "loss": 2.4312,
+      "theoretical_loss": 3.4800917554735347,
+      "tokens_seen": 1687879680
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004969292929292929,
+      "loss": 2.6901,
+      "theoretical_loss": 3.4800801100638283,
+      "tokens_seen": 1687945216
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004969090909090909,
+      "loss": 2.7315,
+      "theoretical_loss": 3.480068465232852,
+      "tokens_seen": 1688010752
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004968888888888889,
+      "loss": 2.5066,
+      "theoretical_loss": 3.480056820980553,
+      "tokens_seen": 1688076288
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004968686868686868,
+      "loss": 2.4754,
+      "theoretical_loss": 3.4800451773068817,
+      "tokens_seen": 1688141824
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004968484848484849,
+      "loss": 2.8056,
+      "theoretical_loss": 3.4800335342117856,
+      "tokens_seen": 1688207360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004968282828282829,
+      "loss": 2.6707,
+      "theoretical_loss": 3.4800218916952144,
+      "tokens_seen": 1688272896
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004968080808080808,
+      "loss": 2.6028,
+      "theoretical_loss": 3.4800102497571164,
+      "tokens_seen": 1688338432
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004967878787878788,
+      "loss": 2.447,
+      "theoretical_loss": 3.47999860839744,
+      "tokens_seen": 1688403968
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004967676767676768,
+      "loss": 2.5797,
+      "theoretical_loss": 3.479986967616135,
+      "tokens_seen": 1688469504
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004967474747474748,
+      "loss": 2.6184,
+      "theoretical_loss": 3.4799753274131495,
+      "tokens_seen": 1688535040
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004967272727272727,
+      "loss": 2.651,
+      "theoretical_loss": 3.479963687788433,
+      "tokens_seen": 1688600576
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004967070707070707,
+      "loss": 2.6058,
+      "theoretical_loss": 3.4799520487419335,
+      "tokens_seen": 1688666112
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004966868686868687,
+      "loss": 2.5568,
+      "theoretical_loss": 3.4799404102736005,
+      "tokens_seen": 1688731648
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004966666666666666,
+      "loss": 2.6325,
+      "theoretical_loss": 3.4799287723833823,
+      "tokens_seen": 1688797184
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004966464646464646,
+      "loss": 2.4631,
+      "theoretical_loss": 3.4799171350712284,
+      "tokens_seen": 1688862720
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004966262626262627,
+      "loss": 2.8322,
+      "theoretical_loss": 3.4799054983370867,
+      "tokens_seen": 1688928256
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004966060606060607,
+      "loss": 2.5475,
+      "theoretical_loss": 3.4798938621809077,
+      "tokens_seen": 1688993792
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004965858585858586,
+      "loss": 2.3884,
+      "theoretical_loss": 3.4798822266026384,
+      "tokens_seen": 1689059328
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004965656565656566,
+      "loss": 2.6883,
+      "theoretical_loss": 3.4798705916022286,
+      "tokens_seen": 1689124864
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004965454545454546,
+      "loss": 2.4336,
+      "theoretical_loss": 3.4798589571796272,
+      "tokens_seen": 1689190400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004965252525252525,
+      "loss": 2.6315,
+      "theoretical_loss": 3.4798473233347833,
+      "tokens_seen": 1689255936
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004965050505050505,
+      "loss": 2.5791,
+      "theoretical_loss": 3.4798356900676444,
+      "tokens_seen": 1689321472
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004964848484848485,
+      "loss": 2.4727,
+      "theoretical_loss": 3.4798240573781616,
+      "tokens_seen": 1689387008
+    },
+    {
+      "epoch": 0.02,
+      "objective/train/docs_used": 932508,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.615687847137451,
+      "objective/train/theoretical_loss": 3.479812425266282,
+      "objective/train/tokens_used": 48311776,
+      "theoretical_loss": 3.479812425266282,
+      "tokens_seen": 1689452544
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004964646464646464,
+      "loss": 2.4249,
+      "theoretical_loss": 3.479812425266282,
+      "tokens_seen": 1689452544
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004964444444444444,
+      "loss": 2.4928,
+      "theoretical_loss": 3.4798007937319557,
+      "tokens_seen": 1689518080
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004964242424242424,
+      "loss": 2.7017,
+      "theoretical_loss": 3.479789162775131,
+      "tokens_seen": 1689583616
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004964040404040405,
+      "loss": 2.6564,
+      "theoretical_loss": 3.4797775323957567,
+      "tokens_seen": 1689649152
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004963838383838384,
+      "loss": 2.4665,
+      "theoretical_loss": 3.479765902593782,
+      "tokens_seen": 1689714688
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004963636363636364,
+      "loss": 2.7239,
+      "theoretical_loss": 3.479754273369156,
+      "tokens_seen": 1689780224
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004963434343434344,
+      "loss": 2.3918,
+      "theoretical_loss": 3.4797426447218274,
+      "tokens_seen": 1689845760
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004963232323232324,
+      "loss": 2.6125,
+      "theoretical_loss": 3.479731016651745,
+      "tokens_seen": 1689911296
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004963030303030303,
+      "loss": 2.9569,
+      "theoretical_loss": 3.4797193891588583,
+      "tokens_seen": 1689976832
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004962828282828283,
+      "loss": 2.43,
+      "theoretical_loss": 3.479707762243116,
+      "tokens_seen": 1690042368
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004962626262626262,
+      "loss": 2.442,
+      "theoretical_loss": 3.479696135904467,
+      "tokens_seen": 1690107904
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004962424242424243,
+      "loss": 2.9013,
+      "theoretical_loss": 3.47968451014286,
+      "tokens_seen": 1690173440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004962222222222222,
+      "loss": 2.6167,
+      "theoretical_loss": 3.479672884958245,
+      "tokens_seen": 1690238976
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004962020202020202,
+      "loss": 2.7755,
+      "theoretical_loss": 3.4796612603505697,
+      "tokens_seen": 1690304512
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004961818181818181,
+      "loss": 2.6197,
+      "theoretical_loss": 3.4796496363197837,
+      "tokens_seen": 1690370048
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004961616161616161,
+      "loss": 2.6195,
+      "theoretical_loss": 3.479638012865836,
+      "tokens_seen": 1690435584
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004961414141414142,
+      "loss": 2.691,
+      "theoretical_loss": 3.4796263899886757,
+      "tokens_seen": 1690501120
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004961212121212122,
+      "loss": 2.7274,
+      "theoretical_loss": 3.479614767688252,
+      "tokens_seen": 1690566656
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004961010101010101,
+      "loss": 2.7084,
+      "theoretical_loss": 3.4796031459645134,
+      "tokens_seen": 1690632192
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004960808080808081,
+      "loss": 2.6796,
+      "theoretical_loss": 3.479591524817409,
+      "tokens_seen": 1690697728
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004960606060606061,
+      "loss": 2.3627,
+      "theoretical_loss": 3.4795799042468882,
+      "tokens_seen": 1690763264
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004960404040404041,
+      "loss": 2.5495,
+      "theoretical_loss": 3.4795682842529,
+      "tokens_seen": 1690828800
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000496020202020202,
+      "loss": 2.6268,
+      "theoretical_loss": 3.4795566648353935,
+      "tokens_seen": 1690894336
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000496,
+      "loss": 2.614,
+      "theoretical_loss": 3.4795450459943167,
+      "tokens_seen": 1690959872
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004959797979797979,
+      "loss": 2.4627,
+      "theoretical_loss": 3.4795334277296206,
+      "tokens_seen": 1691025408
+    },
+    {
+      "epoch": 0.02,
+      "objective/train/docs_used": 933011,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.424670696258545,
+      "objective/train/theoretical_loss": 3.479521810041253,
+      "objective/train/tokens_used": 49950176,
+      "theoretical_loss": 3.479521810041253,
+      "tokens_seen": 1691090944
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004959595959595959,
+      "loss": 2.6558,
+      "theoretical_loss": 3.479521810041253,
+      "tokens_seen": 1691090944
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000495939393939394,
+      "loss": 2.5806,
+      "theoretical_loss": 3.4795101929291627,
+      "tokens_seen": 1691156480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000495919191919192,
+      "loss": 2.6571,
+      "theoretical_loss": 3.4794985763932997,
+      "tokens_seen": 1691222016
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004958989898989899,
+      "loss": 2.6959,
+      "theoretical_loss": 3.4794869604336123,
+      "tokens_seen": 1691287552
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004958787878787879,
+      "loss": 2.4706,
+      "theoretical_loss": 3.47947534505005,
+      "tokens_seen": 1691353088
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004958585858585859,
+      "loss": 2.572,
+      "theoretical_loss": 3.4794637302425624,
+      "tokens_seen": 1691418624
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004958383838383839,
+      "loss": 2.7139,
+      "theoretical_loss": 3.479452116011098,
+      "tokens_seen": 1691484160
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004958181818181818,
+      "loss": 2.6099,
+      "theoretical_loss": 3.4794405023556054,
+      "tokens_seen": 1691549696
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004957979797979798,
+      "loss": 2.4053,
+      "theoretical_loss": 3.4794288892760346,
+      "tokens_seen": 1691615232
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004957777777777778,
+      "loss": 2.8233,
+      "theoretical_loss": 3.4794172767723346,
+      "tokens_seen": 1691680768
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004957575757575757,
+      "loss": 2.7725,
+      "theoretical_loss": 3.479405664844454,
+      "tokens_seen": 1691746304
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004957373737373737,
+      "loss": 2.6491,
+      "theoretical_loss": 3.479394053492343,
+      "tokens_seen": 1691811840
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004957171717171717,
+      "loss": 2.516,
+      "theoretical_loss": 3.4793824427159494,
+      "tokens_seen": 1691877376
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004956969696969698,
+      "loss": 2.474,
+      "theoretical_loss": 3.4793708325152237,
+      "tokens_seen": 1691942912
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004956767676767677,
+      "loss": 2.3708,
+      "theoretical_loss": 3.479359222890114,
+      "tokens_seen": 1692008448
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004956565656565657,
+      "loss": 2.6737,
+      "theoretical_loss": 3.47934761384057,
+      "tokens_seen": 1692073984
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004956363636363637,
+      "loss": 2.5308,
+      "theoretical_loss": 3.479336005366541,
+      "tokens_seen": 1692139520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004956161616161616,
+      "loss": 2.4276,
+      "theoretical_loss": 3.4793243974679755,
+      "tokens_seen": 1692205056
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004955959595959596,
+      "loss": 2.6885,
+      "theoretical_loss": 3.479312790144823,
+      "tokens_seen": 1692270592
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004955757575757576,
+      "loss": 2.2714,
+      "theoretical_loss": 3.479301183397033,
+      "tokens_seen": 1692336128
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004955555555555556,
+      "loss": 2.4882,
+      "theoretical_loss": 3.479289577224555,
+      "tokens_seen": 1692401664
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004955353535353535,
+      "loss": 2.4553,
+      "theoretical_loss": 3.4792779716273365,
+      "tokens_seen": 1692467200
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004955151515151515,
+      "loss": 2.6182,
+      "theoretical_loss": 3.479266366605329,
+      "tokens_seen": 1692532736
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004954949494949495,
+      "loss": 2.5212,
+      "theoretical_loss": 3.4792547621584804,
+      "tokens_seen": 1692598272
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004954747474747474,
+      "loss": 2.4225,
+      "theoretical_loss": 3.47924315828674,
+      "tokens_seen": 1692663808
+    },
+    {
+      "epoch": 0.02,
+      "objective/train/docs_used": 934010,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.852196455001831,
+      "objective/train/theoretical_loss": 3.479231554990057,
+      "objective/train/tokens_used": 51588576,
+      "theoretical_loss": 3.479231554990057,
+      "tokens_seen": 1692729344
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004954545454545455,
+      "loss": 2.6898,
+      "theoretical_loss": 3.479231554990057,
+      "tokens_seen": 1692729344
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004954343434343435,
+      "loss": 2.6057,
+      "theoretical_loss": 3.479219952268381,
+      "tokens_seen": 1692794880
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004954141414141415,
+      "loss": 2.5534,
+      "theoretical_loss": 3.4792083501216613,
+      "tokens_seen": 1692860416
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004953939393939394,
+      "loss": 2.602,
+      "theoretical_loss": 3.479196748549847,
+      "tokens_seen": 1692925952
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004953737373737374,
+      "loss": 2.6322,
+      "theoretical_loss": 3.4791851475528874,
+      "tokens_seen": 1692991488
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004953535353535354,
+      "loss": 2.5945,
+      "theoretical_loss": 3.4791735471307312,
+      "tokens_seen": 1693057024
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004953333333333334,
+      "loss": 2.6971,
+      "theoretical_loss": 3.4791619472833286,
+      "tokens_seen": 1693122560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004953131313131313,
+      "loss": 2.6478,
+      "theoretical_loss": 3.479150348010628,
+      "tokens_seen": 1693188096
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004952929292929293,
+      "loss": 2.5666,
+      "theoretical_loss": 3.479138749312579,
+      "tokens_seen": 1693253632
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004952727272727272,
+      "loss": 2.5181,
+      "theoretical_loss": 3.4791271511891315,
+      "tokens_seen": 1693319168
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004952525252525252,
+      "loss": 2.566,
+      "theoretical_loss": 3.4791155536402343,
+      "tokens_seen": 1693384704
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004952323232323232,
+      "loss": 2.7651,
+      "theoretical_loss": 3.479103956665836,
+      "tokens_seen": 1693450240
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004952121212121213,
+      "loss": 2.5736,
+      "theoretical_loss": 3.479092360265887,
+      "tokens_seen": 1693515776
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004951919191919192,
+      "loss": 2.4578,
+      "theoretical_loss": 3.4790807644403365,
+      "tokens_seen": 1693581312
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004951717171717172,
+      "loss": 2.6557,
+      "theoretical_loss": 3.479069169189133,
+      "tokens_seen": 1693646848
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004951515151515152,
+      "loss": 2.6888,
+      "theoretical_loss": 3.479057574512227,
+      "tokens_seen": 1693712384
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004951313131313132,
+      "loss": 2.5443,
+      "theoretical_loss": 3.4790459804095666,
+      "tokens_seen": 1693777920
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004951111111111111,
+      "loss": 2.7016,
+      "theoretical_loss": 3.479034386881102,
+      "tokens_seen": 1693843456
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004950909090909091,
+      "loss": 2.4926,
+      "theoretical_loss": 3.4790227939267826,
+      "tokens_seen": 1693908992
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000495070707070707,
+      "loss": 2.7287,
+      "theoretical_loss": 3.479011201546557,
+      "tokens_seen": 1693974528
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004950505050505051,
+      "loss": 2.71,
+      "theoretical_loss": 3.478999609740375,
+      "tokens_seen": 1694040064
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000495030303030303,
+      "loss": 2.4208,
+      "theoretical_loss": 3.4789880185081863,
+      "tokens_seen": 1694105600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000495010101010101,
+      "loss": 2.4963,
+      "theoretical_loss": 3.4789764278499398,
+      "tokens_seen": 1694171136
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000494989898989899,
+      "loss": 2.7918,
+      "theoretical_loss": 3.478964837765585,
+      "tokens_seen": 1694236672
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000494969696969697,
+      "loss": 2.3404,
+      "theoretical_loss": 3.4789532482550714,
+      "tokens_seen": 1694302208
+    },
+    {
+      "epoch": 0.02,
+      "objective/train/docs_used": 934404,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7420923709869385,
+      "objective/train/theoretical_loss": 3.478941659318348,
+      "objective/train/tokens_used": 53226976,
+      "theoretical_loss": 3.478941659318348,
+      "tokens_seen": 1694367744
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000494949494949495,
+      "loss": 2.6966,
+      "theoretical_loss": 3.478941659318348,
+      "tokens_seen": 1694367744
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000494929292929293,
+      "loss": 2.652,
+      "theoretical_loss": 3.478930070955365,
+      "tokens_seen": 1694433280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004949090909090909,
+      "loss": 2.5731,
+      "theoretical_loss": 3.478918483166071,
+      "tokens_seen": 1694498816
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004948888888888889,
+      "loss": 2.5032,
+      "theoretical_loss": 3.478906895950416,
+      "tokens_seen": 1694564352
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004948686868686869,
+      "loss": 2.7563,
+      "theoretical_loss": 3.4788953093083492,
+      "tokens_seen": 1694629888
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004948484848484849,
+      "loss": 2.5857,
+      "theoretical_loss": 3.4788837232398198,
+      "tokens_seen": 1694695424
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004948282828282828,
+      "loss": 2.5759,
+      "theoretical_loss": 3.4788721377447773,
+      "tokens_seen": 1694760960
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004948080808080808,
+      "loss": 2.4183,
+      "theoretical_loss": 3.4788605528231713,
+      "tokens_seen": 1694826496
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004947878787878787,
+      "loss": 2.4682,
+      "theoretical_loss": 3.478848968474951,
+      "tokens_seen": 1694892032
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004947676767676767,
+      "loss": 2.7972,
+      "theoretical_loss": 3.4788373847000664,
+      "tokens_seen": 1694957568
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004947474747474748,
+      "loss": 2.5319,
+      "theoretical_loss": 3.4788258014984663,
+      "tokens_seen": 1695023104
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004947272727272728,
+      "loss": 2.4632,
+      "theoretical_loss": 3.478814218870101,
+      "tokens_seen": 1695088640
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004947070707070707,
+      "loss": 2.5513,
+      "theoretical_loss": 3.478802636814919,
+      "tokens_seen": 1695154176
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004946868686868687,
+      "loss": 2.5416,
+      "theoretical_loss": 3.4787910553328705,
+      "tokens_seen": 1695219712
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004946666666666667,
+      "loss": 2.6542,
+      "theoretical_loss": 3.4787794744239044,
+      "tokens_seen": 1695285248
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004946464646464647,
+      "loss": 2.5943,
+      "theoretical_loss": 3.4787678940879707,
+      "tokens_seen": 1695350784
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004946262626262626,
+      "loss": 2.5594,
+      "theoretical_loss": 3.4787563143250186,
+      "tokens_seen": 1695416320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004946060606060606,
+      "loss": 2.4973,
+      "theoretical_loss": 3.478744735134998,
+      "tokens_seen": 1695481856
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004945858585858586,
+      "loss": 2.5056,
+      "theoretical_loss": 3.4787331565178574,
+      "tokens_seen": 1695547392
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004945656565656565,
+      "loss": 2.5805,
+      "theoretical_loss": 3.4787215784735475,
+      "tokens_seen": 1695612928
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004945454545454545,
+      "loss": 2.5839,
+      "theoretical_loss": 3.4787100010020175,
+      "tokens_seen": 1695678464
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004945252525252526,
+      "loss": 2.551,
+      "theoretical_loss": 3.478698424103216,
+      "tokens_seen": 1695744000
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004945050505050506,
+      "loss": 2.7043,
+      "theoretical_loss": 3.4786868477770945,
+      "tokens_seen": 1695809536
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004944848484848485,
+      "loss": 2.6309,
+      "theoretical_loss": 3.4786752720236005,
+      "tokens_seen": 1695875072
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004944646464646465,
+      "loss": 2.7457,
+      "theoretical_loss": 3.4786636968426845,
+      "tokens_seen": 1695940608
+    },
+    {
+      "epoch": 0.02,
+      "objective/train/docs_used": 935560,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.167574882507324,
+      "objective/train/theoretical_loss": 3.478652122234296,
+      "objective/train/tokens_used": 54865376,
+      "theoretical_loss": 3.478652122234296,
+      "tokens_seen": 1696006144
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004944444444444445,
+      "loss": 2.4537,
+      "theoretical_loss": 3.478652122234296,
+      "tokens_seen": 1696006144
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004944242424242424,
+      "loss": 2.5082,
+      "theoretical_loss": 3.4786405481983844,
+      "tokens_seen": 1696071680
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004944040404040404,
+      "loss": 2.716,
+      "theoretical_loss": 3.4786289747348995,
+      "tokens_seen": 1696137216
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004943838383838384,
+      "loss": 2.5441,
+      "theoretical_loss": 3.478617401843791,
+      "tokens_seen": 1696202752
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004943636363636363,
+      "loss": 2.5488,
+      "theoretical_loss": 3.4786058295250077,
+      "tokens_seen": 1696268288
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004943434343434343,
+      "loss": 2.5682,
+      "theoretical_loss": 3.4785942577785,
+      "tokens_seen": 1696333824
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004943232323232323,
+      "loss": 2.4028,
+      "theoretical_loss": 3.4785826866042173,
+      "tokens_seen": 1696399360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004943030303030303,
+      "loss": 2.5064,
+      "theoretical_loss": 3.478571116002109,
+      "tokens_seen": 1696464896
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004942828282828283,
+      "loss": 2.4192,
+      "theoretical_loss": 3.4785595459721246,
+      "tokens_seen": 1696530432
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004942626262626263,
+      "loss": 2.6105,
+      "theoretical_loss": 3.478547976514214,
+      "tokens_seen": 1696595968
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004942424242424243,
+      "loss": 2.6893,
+      "theoretical_loss": 3.4785364076283267,
+      "tokens_seen": 1696661504
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004942222222222223,
+      "loss": 2.4376,
+      "theoretical_loss": 3.478524839314413,
+      "tokens_seen": 1696727040
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004942020202020202,
+      "loss": 2.7309,
+      "theoretical_loss": 3.478513271572421,
+      "tokens_seen": 1696792576
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004941818181818182,
+      "loss": 2.8496,
+      "theoretical_loss": 3.4785017044023014,
+      "tokens_seen": 1696858112
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004941616161616161,
+      "loss": 2.423,
+      "theoretical_loss": 3.4784901378040036,
+      "tokens_seen": 1696923648
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004941414141414142,
+      "loss": 2.494,
+      "theoretical_loss": 3.4784785717774778,
+      "tokens_seen": 1696989184
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004941212121212121,
+      "loss": 2.53,
+      "theoretical_loss": 3.4784670063226732,
+      "tokens_seen": 1697054720
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004941010101010101,
+      "loss": 2.749,
+      "theoretical_loss": 3.478455441439539,
+      "tokens_seen": 1697120256
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000494080808080808,
+      "loss": 2.3697,
+      "theoretical_loss": 3.4784438771280257,
+      "tokens_seen": 1697185792
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000494060606060606,
+      "loss": 2.6826,
+      "theoretical_loss": 3.4784323133880823,
+      "tokens_seen": 1697251328
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004940404040404041,
+      "loss": 2.5308,
+      "theoretical_loss": 3.4784207502196587,
+      "tokens_seen": 1697316864
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004940202020202021,
+      "loss": 2.4568,
+      "theoretical_loss": 3.4784091876227046,
+      "tokens_seen": 1697382400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000494,
+      "loss": 2.7751,
+      "theoretical_loss": 3.47839762559717,
+      "tokens_seen": 1697447936
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000493979797979798,
+      "loss": 2.5151,
+      "theoretical_loss": 3.478386064143004,
+      "tokens_seen": 1697513472
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000493959595959596,
+      "loss": 2.8479,
+      "theoretical_loss": 3.4783745032601567,
+      "tokens_seen": 1697579008
+    },
+    {
+      "epoch": 0.02,
+      "objective/train/docs_used": 936682,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.305478572845459,
+      "objective/train/theoretical_loss": 3.4783629429485785,
+      "objective/train/tokens_used": 56503776,
+      "theoretical_loss": 3.4783629429485785,
+      "tokens_seen": 1697644544
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000493939393939394,
+      "loss": 2.5086,
+      "theoretical_loss": 3.4783629429485785,
+      "tokens_seen": 1697644544
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004939191919191919,
+      "loss": 2.7043,
+      "theoretical_loss": 3.4783513832082176,
+      "tokens_seen": 1697710080
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004938989898989899,
+      "loss": 2.3848,
+      "theoretical_loss": 3.4783398240390246,
+      "tokens_seen": 1697775616
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004938787878787878,
+      "loss": 2.7298,
+      "theoretical_loss": 3.4783282654409495,
+      "tokens_seen": 1697841152
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004938585858585859,
+      "loss": 2.5754,
+      "theoretical_loss": 3.478316707413941,
+      "tokens_seen": 1697906688
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004938383838383838,
+      "loss": 2.7063,
+      "theoretical_loss": 3.4783051499579507,
+      "tokens_seen": 1697972224
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004938181818181819,
+      "loss": 2.6402,
+      "theoretical_loss": 3.478293593072926,
+      "tokens_seen": 1698037760
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004937979797979798,
+      "loss": 2.4452,
+      "theoretical_loss": 3.4782820367588183,
+      "tokens_seen": 1698103296
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004937777777777778,
+      "loss": 2.6044,
+      "theoretical_loss": 3.478270481015577,
+      "tokens_seen": 1698168832
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004937575757575758,
+      "loss": 2.6482,
+      "theoretical_loss": 3.4782589258431518,
+      "tokens_seen": 1698234368
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004937373737373738,
+      "loss": 2.7206,
+      "theoretical_loss": 3.4782473712414923,
+      "tokens_seen": 1698299904
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004937171717171717,
+      "loss": 2.4687,
+      "theoretical_loss": 3.4782358172105483,
+      "tokens_seen": 1698365440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004936969696969697,
+      "loss": 2.7044,
+      "theoretical_loss": 3.47822426375027,
+      "tokens_seen": 1698430976
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004936767676767677,
+      "loss": 2.5647,
+      "theoretical_loss": 3.478212710860607,
+      "tokens_seen": 1698496512
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004936565656565657,
+      "loss": 2.6277,
+      "theoretical_loss": 3.4782011585415087,
+      "tokens_seen": 1698562048
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004936363636363636,
+      "loss": 2.7133,
+      "theoretical_loss": 3.4781896067929257,
+      "tokens_seen": 1698627584
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004936161616161616,
+      "loss": 2.5221,
+      "theoretical_loss": 3.4781780556148068,
+      "tokens_seen": 1698693120
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004935959595959597,
+      "loss": 2.6434,
+      "theoretical_loss": 3.4781665050071027,
+      "tokens_seen": 1698758656
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004935757575757576,
+      "loss": 2.7506,
+      "theoretical_loss": 3.478154954969763,
+      "tokens_seen": 1698824192
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004935555555555556,
+      "loss": 2.578,
+      "theoretical_loss": 3.4781434055027365,
+      "tokens_seen": 1698889728
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004935353535353536,
+      "loss": 2.5255,
+      "theoretical_loss": 3.478131856605975,
+      "tokens_seen": 1698955264
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004935151515151515,
+      "loss": 2.4671,
+      "theoretical_loss": 3.4781203082794265,
+      "tokens_seen": 1699020800
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004934949494949495,
+      "loss": 2.5736,
+      "theoretical_loss": 3.4781087605230425,
+      "tokens_seen": 1699086336
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004934747474747475,
+      "loss": 2.5738,
+      "theoretical_loss": 3.4780972133367714,
+      "tokens_seen": 1699151872
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004934545454545455,
+      "loss": 2.6036,
+      "theoretical_loss": 3.478085666720564,
+      "tokens_seen": 1699217408
+    },
+    {
+      "epoch": 0.02,
+      "objective/train/docs_used": 937279,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.8765950202941895,
+      "objective/train/theoretical_loss": 3.47807412067437,
+      "objective/train/tokens_used": 58142176,
+      "theoretical_loss": 3.47807412067437,
+      "tokens_seen": 1699282944
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004934343434343434,
+      "loss": 2.6055,
+      "theoretical_loss": 3.47807412067437,
+      "tokens_seen": 1699282944
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004934141414141414,
+      "loss": 2.5349,
+      "theoretical_loss": 3.4780625751981384,
+      "tokens_seen": 1699348480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004933939393939394,
+      "loss": 2.7046,
+      "theoretical_loss": 3.478051030291821,
+      "tokens_seen": 1699414016
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004933737373737373,
+      "loss": 2.6394,
+      "theoretical_loss": 3.4780394859553656,
+      "tokens_seen": 1699479552
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004933535353535354,
+      "loss": 2.4959,
+      "theoretical_loss": 3.4780279421887235,
+      "tokens_seen": 1699545088
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004933333333333334,
+      "loss": 2.7816,
+      "theoretical_loss": 3.478016398991844,
+      "tokens_seen": 1699610624
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004933131313131314,
+      "loss": 2.622,
+      "theoretical_loss": 3.478004856364677,
+      "tokens_seen": 1699676160
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004932929292929293,
+      "loss": 2.6067,
+      "theoretical_loss": 3.477993314307173,
+      "tokens_seen": 1699741696
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004932727272727273,
+      "loss": 2.6708,
+      "theoretical_loss": 3.4779817728192803,
+      "tokens_seen": 1699807232
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004932525252525253,
+      "loss": 2.7044,
+      "theoretical_loss": 3.477970231900951,
+      "tokens_seen": 1699872768
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004932323232323232,
+      "loss": 2.5886,
+      "theoretical_loss": 3.477958691552134,
+      "tokens_seen": 1699938304
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004932121212121212,
+      "loss": 2.4632,
+      "theoretical_loss": 3.4779471517727787,
+      "tokens_seen": 1700003840
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004931919191919192,
+      "loss": 2.6896,
+      "theoretical_loss": 3.4779356125628365,
+      "tokens_seen": 1700069376
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004931717171717171,
+      "loss": 2.655,
+      "theoretical_loss": 3.4779240739222557,
+      "tokens_seen": 1700134912
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004931515151515151,
+      "loss": 2.6488,
+      "theoretical_loss": 3.4779125358509875,
+      "tokens_seen": 1700200448
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004931313131313131,
+      "loss": 2.6737,
+      "theoretical_loss": 3.4779009983489813,
+      "tokens_seen": 1700265984
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004931111111111112,
+      "loss": 2.5862,
+      "theoretical_loss": 3.4778894614161873,
+      "tokens_seen": 1700331520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004930909090909091,
+      "loss": 2.5534,
+      "theoretical_loss": 3.477877925052555,
+      "tokens_seen": 1700397056
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004930707070707071,
+      "loss": 2.6173,
+      "theoretical_loss": 3.4778663892580353,
+      "tokens_seen": 1700462592
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004930505050505051,
+      "loss": 2.7138,
+      "theoretical_loss": 3.477854854032577,
+      "tokens_seen": 1700528128
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004930303030303031,
+      "loss": 2.5778,
+      "theoretical_loss": 3.477843319376131,
+      "tokens_seen": 1700593664
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000493010101010101,
+      "loss": 2.5768,
+      "theoretical_loss": 3.4778317852886476,
+      "tokens_seen": 1700659200
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000492989898989899,
+      "loss": 2.8706,
+      "theoretical_loss": 3.4778202517700754,
+      "tokens_seen": 1700724736
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004929696969696969,
+      "loss": 2.4533,
+      "theoretical_loss": 3.4778087188203655,
+      "tokens_seen": 1700790272
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000492949494949495,
+      "loss": 2.6694,
+      "theoretical_loss": 3.477797186439468,
+      "tokens_seen": 1700855808
+    },
+    {
+      "epoch": 0.02,
+      "objective/train/docs_used": 938365,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.9837400913238525,
+      "objective/train/theoretical_loss": 3.4777856546273327,
+      "objective/train/tokens_used": 59780576,
+      "theoretical_loss": 3.4777856546273327,
+      "tokens_seen": 1700921344
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004929292929292929,
+      "loss": 2.7473,
+      "theoretical_loss": 3.4777856546273327,
+      "tokens_seen": 1700921344
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004929090909090909,
+      "loss": 2.4495,
+      "theoretical_loss": 3.477774123383909,
+      "tokens_seen": 1700986880
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004928888888888888,
+      "loss": 2.609,
+      "theoretical_loss": 3.477762592709148,
+      "tokens_seen": 1701052416
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004928686868686869,
+      "loss": 2.5167,
+      "theoretical_loss": 3.477751062602999,
+      "tokens_seen": 1701117952
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004928484848484849,
+      "loss": 2.4444,
+      "theoretical_loss": 3.477739533065412,
+      "tokens_seen": 1701183488
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004928282828282829,
+      "loss": 2.5933,
+      "theoretical_loss": 3.4777280040963383,
+      "tokens_seen": 1701249024
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004928080808080808,
+      "loss": 2.707,
+      "theoretical_loss": 3.4777164756957264,
+      "tokens_seen": 1701314560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004927878787878788,
+      "loss": 2.5917,
+      "theoretical_loss": 3.477704947863527,
+      "tokens_seen": 1701380096
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004927676767676768,
+      "loss": 2.5673,
+      "theoretical_loss": 3.4776934205996906,
+      "tokens_seen": 1701445632
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004927474747474748,
+      "loss": 2.8966,
+      "theoretical_loss": 3.4776818939041663,
+      "tokens_seen": 1701511168
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004927272727272727,
+      "loss": 2.6156,
+      "theoretical_loss": 3.477670367776905,
+      "tokens_seen": 1701576704
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004927070707070707,
+      "loss": 2.6147,
+      "theoretical_loss": 3.477658842217857,
+      "tokens_seen": 1701642240
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004926868686868686,
+      "loss": 2.613,
+      "theoretical_loss": 3.4776473172269715,
+      "tokens_seen": 1701707776
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004926666666666666,
+      "loss": 2.7808,
+      "theoretical_loss": 3.4776357928041994,
+      "tokens_seen": 1701773312
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004926464646464647,
+      "loss": 2.6945,
+      "theoretical_loss": 3.4776242689494903,
+      "tokens_seen": 1701838848
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004926262626262627,
+      "loss": 2.7847,
+      "theoretical_loss": 3.4776127456627948,
+      "tokens_seen": 1701904384
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004926060606060606,
+      "loss": 2.7205,
+      "theoretical_loss": 3.4776012229440623,
+      "tokens_seen": 1701969920
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004925858585858586,
+      "loss": 2.6992,
+      "theoretical_loss": 3.477589700793244,
+      "tokens_seen": 1702035456
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004925656565656566,
+      "loss": 2.6906,
+      "theoretical_loss": 3.4775781792102887,
+      "tokens_seen": 1702100992
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004925454545454546,
+      "loss": 2.659,
+      "theoretical_loss": 3.477566658195148,
+      "tokens_seen": 1702166528
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004925252525252525,
+      "loss": 2.727,
+      "theoretical_loss": 3.477555137747771,
+      "tokens_seen": 1702232064
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004925050505050505,
+      "loss": 2.7682,
+      "theoretical_loss": 3.4775436178681085,
+      "tokens_seen": 1702297600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004924848484848485,
+      "loss": 2.5353,
+      "theoretical_loss": 3.4775320985561105,
+      "tokens_seen": 1702363136
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004924646464646464,
+      "loss": 2.6144,
+      "theoretical_loss": 3.477520579811727,
+      "tokens_seen": 1702428672
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0004924444444444444,
+      "loss": 2.7138,
+      "theoretical_loss": 3.477509061634908,
+      "tokens_seen": 1702494208
+    },
+    {
+      "epoch": 0.02,
+      "objective/train/docs_used": 938945,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.2912395000457764,
+      "objective/train/theoretical_loss": 3.4774975440256037,
+      "objective/train/tokens_used": 61418976,
+      "theoretical_loss": 3.4774975440256037,
+      "tokens_seen": 1702559744
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004924242424242425,
+      "loss": 2.7635,
+      "theoretical_loss": 3.4774975440256037,
+      "tokens_seen": 1702559744
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004924040404040405,
+      "loss": 2.761,
+      "theoretical_loss": 3.4774860269837653,
+      "tokens_seen": 1702625280
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004923838383838384,
+      "loss": 2.6293,
+      "theoretical_loss": 3.4774745105093414,
+      "tokens_seen": 1702690816
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004923636363636364,
+      "loss": 2.6983,
+      "theoretical_loss": 3.4774629946022837,
+      "tokens_seen": 1702756352
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004923434343434344,
+      "loss": 2.6871,
+      "theoretical_loss": 3.477451479262541,
+      "tokens_seen": 1702821888
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004923232323232323,
+      "loss": 2.7305,
+      "theoretical_loss": 3.477439964490065,
+      "tokens_seen": 1702887424
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004923030303030303,
+      "loss": 2.6126,
+      "theoretical_loss": 3.477428450284805,
+      "tokens_seen": 1702952960
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004922828282828283,
+      "loss": 2.365,
+      "theoretical_loss": 3.4774169366467116,
+      "tokens_seen": 1703018496
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004922626262626262,
+      "loss": 2.6455,
+      "theoretical_loss": 3.4774054235757346,
+      "tokens_seen": 1703084032
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004922424242424242,
+      "loss": 2.5675,
+      "theoretical_loss": 3.4773939110718244,
+      "tokens_seen": 1703149568
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004922222222222222,
+      "loss": 2.5027,
+      "theoretical_loss": 3.4773823991349317,
+      "tokens_seen": 1703215104
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004922020202020202,
+      "loss": 2.7077,
+      "theoretical_loss": 3.4773708877650065,
+      "tokens_seen": 1703280640
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004921818181818182,
+      "loss": 2.5908,
+      "theoretical_loss": 3.477359376961999,
+      "tokens_seen": 1703346176
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004921616161616162,
+      "loss": 2.6174,
+      "theoretical_loss": 3.477347866725859,
+      "tokens_seen": 1703411712
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004921414141414142,
+      "loss": 2.6228,
+      "theoretical_loss": 3.4773363570565374,
+      "tokens_seen": 1703477248
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004921212121212122,
+      "loss": 2.8639,
+      "theoretical_loss": 3.4773248479539847,
+      "tokens_seen": 1703542784
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004921010101010101,
+      "loss": 2.75,
+      "theoretical_loss": 3.47731333941815,
+      "tokens_seen": 1703608320
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004920808080808081,
+      "loss": 2.4292,
+      "theoretical_loss": 3.477301831448985,
+      "tokens_seen": 1703673856
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004920606060606061,
+      "loss": 2.6518,
+      "theoretical_loss": 3.4772903240464395,
+      "tokens_seen": 1703739392
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004920404040404041,
+      "loss": 2.7381,
+      "theoretical_loss": 3.477278817210463,
+      "tokens_seen": 1703804928
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000492020202020202,
+      "loss": 2.7286,
+      "theoretical_loss": 3.4772673109410075,
+      "tokens_seen": 1703870464
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000492,
+      "loss": 2.7131,
+      "theoretical_loss": 3.4772558052380216,
+      "tokens_seen": 1703936000
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004919797979797979,
+      "loss": 2.8491,
+      "theoretical_loss": 3.4772443001014564,
+      "tokens_seen": 1704001536
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000491959595959596,
+      "loss": 2.6865,
+      "theoretical_loss": 3.4772327955312625,
+      "tokens_seen": 1704067072
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000491939393939394,
+      "loss": 2.4784,
+      "theoretical_loss": 3.47722129152739,
+      "tokens_seen": 1704132608
+    },
+    {
+      "epoch": 0.03,
+      "objective/train/docs_used": 940087,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.313035726547241,
+      "objective/train/theoretical_loss": 3.477209788089789,
+      "objective/train/tokens_used": 63057376,
+      "theoretical_loss": 3.477209788089789,
+      "tokens_seen": 1704198144
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000491919191919192,
+      "loss": 2.5144,
+      "theoretical_loss": 3.477209788089789,
+      "tokens_seen": 1704198144
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004918989898989899,
+      "loss": 2.4958,
+      "theoretical_loss": 3.4771982852184102,
+      "tokens_seen": 1704263680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004918787878787879,
+      "loss": 2.6901,
+      "theoretical_loss": 3.4771867829132033,
+      "tokens_seen": 1704329216
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004918585858585859,
+      "loss": 2.7283,
+      "theoretical_loss": 3.4771752811741194,
+      "tokens_seen": 1704394752
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004918383838383839,
+      "loss": 2.5867,
+      "theoretical_loss": 3.477163780001109,
+      "tokens_seen": 1704460288
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004918181818181818,
+      "loss": 2.6477,
+      "theoretical_loss": 3.477152279394122,
+      "tokens_seen": 1704525824
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004917979797979798,
+      "loss": 2.6709,
+      "theoretical_loss": 3.4771407793531086,
+      "tokens_seen": 1704591360
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004917777777777777,
+      "loss": 2.6038,
+      "theoretical_loss": 3.4771292798780196,
+      "tokens_seen": 1704656896
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004917575757575758,
+      "loss": 2.9226,
+      "theoretical_loss": 3.477117780968805,
+      "tokens_seen": 1704722432
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004917373737373737,
+      "loss": 2.5123,
+      "theoretical_loss": 3.4771062826254155,
+      "tokens_seen": 1704787968
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004917171717171718,
+      "loss": 2.5392,
+      "theoretical_loss": 3.477094784847802,
+      "tokens_seen": 1704853504
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004916969696969697,
+      "loss": 2.8546,
+      "theoretical_loss": 3.477083287635914,
+      "tokens_seen": 1704919040
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004916767676767677,
+      "loss": 2.5932,
+      "theoretical_loss": 3.4770717909897026,
+      "tokens_seen": 1704984576
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004916565656565657,
+      "loss": 2.6186,
+      "theoretical_loss": 3.4770602949091174,
+      "tokens_seen": 1705050112
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004916363636363637,
+      "loss": 2.6519,
+      "theoretical_loss": 3.47704879939411,
+      "tokens_seen": 1705115648
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004916161616161616,
+      "loss": 2.6116,
+      "theoretical_loss": 3.4770373044446297,
+      "tokens_seen": 1705181184
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004915959595959596,
+      "loss": 2.7875,
+      "theoretical_loss": 3.4770258100606277,
+      "tokens_seen": 1705246720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004915757575757576,
+      "loss": 2.621,
+      "theoretical_loss": 3.4770143162420544,
+      "tokens_seen": 1705312256
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004915555555555556,
+      "loss": 2.7483,
+      "theoretical_loss": 3.47700282298886,
+      "tokens_seen": 1705377792
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004915353535353535,
+      "loss": 2.6616,
+      "theoretical_loss": 3.4769913303009945,
+      "tokens_seen": 1705443328
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004915151515151515,
+      "loss": 2.8028,
+      "theoretical_loss": 3.4769798381784094,
+      "tokens_seen": 1705508864
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004914949494949494,
+      "loss": 2.3576,
+      "theoretical_loss": 3.4769683466210544,
+      "tokens_seen": 1705574400
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004914747474747475,
+      "loss": 2.6636,
+      "theoretical_loss": 3.47695685562888,
+      "tokens_seen": 1705639936
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004914545454545455,
+      "loss": 2.7262,
+      "theoretical_loss": 3.4769453652018374,
+      "tokens_seen": 1705705472
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004914343434343435,
+      "loss": 2.6114,
+      "theoretical_loss": 3.476933875339877,
+      "tokens_seen": 1705771008
+    },
+    {
+      "epoch": 0.03,
+      "objective/train/docs_used": 941279,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.1284117698669434,
+      "objective/train/theoretical_loss": 3.4769223860429483,
+      "objective/train/tokens_used": 64695776,
+      "theoretical_loss": 3.4769223860429483,
+      "tokens_seen": 1705836544
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004914141414141414,
+      "loss": 2.6038,
+      "theoretical_loss": 3.4769223860429483,
+      "tokens_seen": 1705836544
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004913939393939394,
+      "loss": 2.5564,
+      "theoretical_loss": 3.4769108973110026,
+      "tokens_seen": 1705902080
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004913737373737374,
+      "loss": 2.6878,
+      "theoretical_loss": 3.47689940914399,
+      "tokens_seen": 1705967616
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004913535353535354,
+      "loss": 2.7768,
+      "theoretical_loss": 3.476887921541862,
+      "tokens_seen": 1706033152
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004913333333333333,
+      "loss": 2.5874,
+      "theoretical_loss": 3.4768764345045677,
+      "tokens_seen": 1706098688
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004913131313131313,
+      "loss": 2.4851,
+      "theoretical_loss": 3.476864948032059,
+      "tokens_seen": 1706164224
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004912929292929293,
+      "loss": 2.3652,
+      "theoretical_loss": 3.4768534621242853,
+      "tokens_seen": 1706229760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004912727272727272,
+      "loss": 2.3602,
+      "theoretical_loss": 3.4768419767811976,
+      "tokens_seen": 1706295296
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004912525252525252,
+      "loss": 2.6307,
+      "theoretical_loss": 3.4768304920027466,
+      "tokens_seen": 1706360832
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004912323232323233,
+      "loss": 2.6657,
+      "theoretical_loss": 3.4768190077888823,
+      "tokens_seen": 1706426368
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004912121212121213,
+      "loss": 2.5756,
+      "theoretical_loss": 3.4768075241395566,
+      "tokens_seen": 1706491904
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004911919191919192,
+      "loss": 2.7722,
+      "theoretical_loss": 3.4767960410547185,
+      "tokens_seen": 1706557440
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004911717171717172,
+      "loss": 2.288,
+      "theoretical_loss": 3.4767845585343196,
+      "tokens_seen": 1706622976
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004911515151515152,
+      "loss": 2.4128,
+      "theoretical_loss": 3.4767730765783096,
+      "tokens_seen": 1706688512
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004911313131313131,
+      "loss": 2.4081,
+      "theoretical_loss": 3.47676159518664,
+      "tokens_seen": 1706754048
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004911111111111111,
+      "loss": 2.6869,
+      "theoretical_loss": 3.4767501143592607,
+      "tokens_seen": 1706819584
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004910909090909091,
+      "loss": 2.5965,
+      "theoretical_loss": 3.4767386340961233,
+      "tokens_seen": 1706885120
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000491070707070707,
+      "loss": 2.5827,
+      "theoretical_loss": 3.476727154397177,
+      "tokens_seen": 1706950656
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000491050505050505,
+      "loss": 2.7387,
+      "theoretical_loss": 3.4767156752623736,
+      "tokens_seen": 1707016192
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000491030303030303,
+      "loss": 2.7383,
+      "theoretical_loss": 3.4767041966916628,
+      "tokens_seen": 1707081728
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004910101010101011,
+      "loss": 2.7158,
+      "theoretical_loss": 3.476692718684996,
+      "tokens_seen": 1707147264
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000490989898989899,
+      "loss": 2.6993,
+      "theoretical_loss": 3.476681241242323,
+      "tokens_seen": 1707212800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000490969696969697,
+      "loss": 2.6435,
+      "theoretical_loss": 3.4766697643635958,
+      "tokens_seen": 1707278336
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000490949494949495,
+      "loss": 2.6064,
+      "theoretical_loss": 3.4766582880487635,
+      "tokens_seen": 1707343872
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000490929292929293,
+      "loss": 2.5374,
+      "theoretical_loss": 3.4766468122977776,
+      "tokens_seen": 1707409408
+    },
+    {
+      "epoch": 0.03,
+      "objective/train/docs_used": 941668,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.3641209602355957,
+      "objective/train/theoretical_loss": 3.4766353371105887,
+      "objective/train/tokens_used": 66334176,
+      "theoretical_loss": 3.4766353371105887,
+      "tokens_seen": 1707474944
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004909090909090909,
+      "loss": 2.6239,
+      "theoretical_loss": 3.4766353371105887,
+      "tokens_seen": 1707474944
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004908888888888889,
+      "loss": 2.5148,
+      "theoretical_loss": 3.4766238624871475,
+      "tokens_seen": 1707540480
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004908686868686868,
+      "loss": 2.5122,
+      "theoretical_loss": 3.4766123884274043,
+      "tokens_seen": 1707606016
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004908484848484849,
+      "loss": 2.9696,
+      "theoretical_loss": 3.4766009149313097,
+      "tokens_seen": 1707671552
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004908282828282828,
+      "loss": 2.4315,
+      "theoretical_loss": 3.476589441998815,
+      "tokens_seen": 1707737088
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004908080808080808,
+      "loss": 2.6819,
+      "theoretical_loss": 3.4765779696298704,
+      "tokens_seen": 1707802624
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004907878787878787,
+      "loss": 2.6707,
+      "theoretical_loss": 3.476566497824427,
+      "tokens_seen": 1707868160
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004907676767676768,
+      "loss": 2.6873,
+      "theoretical_loss": 3.476555026582435,
+      "tokens_seen": 1707933696
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004907474747474748,
+      "loss": 2.6117,
+      "theoretical_loss": 3.476543555903846,
+      "tokens_seen": 1707999232
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004907272727272728,
+      "loss": 2.7386,
+      "theoretical_loss": 3.4765320857886097,
+      "tokens_seen": 1708064768
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004907070707070707,
+      "loss": 2.7186,
+      "theoretical_loss": 3.476520616236677,
+      "tokens_seen": 1708130304
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004906868686868687,
+      "loss": 2.6979,
+      "theoretical_loss": 3.476509147247999,
+      "tokens_seen": 1708195840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004906666666666667,
+      "loss": 2.8907,
+      "theoretical_loss": 3.4764976788225264,
+      "tokens_seen": 1708261376
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004906464646464647,
+      "loss": 2.6496,
+      "theoretical_loss": 3.4764862109602097,
+      "tokens_seen": 1708326912
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004906262626262626,
+      "loss": 2.7096,
+      "theoretical_loss": 3.4764747436609995,
+      "tokens_seen": 1708392448
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004906060606060606,
+      "loss": 2.7149,
+      "theoretical_loss": 3.476463276924847,
+      "tokens_seen": 1708457984
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004905858585858585,
+      "loss": 2.8988,
+      "theoretical_loss": 3.476451810751703,
+      "tokens_seen": 1708523520
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004905656565656565,
+      "loss": 2.5287,
+      "theoretical_loss": 3.4764403451415173,
+      "tokens_seen": 1708589056
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004905454545454546,
+      "loss": 2.4729,
+      "theoretical_loss": 3.476428880094242,
+      "tokens_seen": 1708654592
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004905252525252526,
+      "loss": 2.7736,
+      "theoretical_loss": 3.476417415609827,
+      "tokens_seen": 1708720128
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004905050505050505,
+      "loss": 2.732,
+      "theoretical_loss": 3.476405951688223,
+      "tokens_seen": 1708785664
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004904848484848485,
+      "loss": 2.6687,
+      "theoretical_loss": 3.476394488329382,
+      "tokens_seen": 1708851200
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004904646464646465,
+      "loss": 2.7529,
+      "theoretical_loss": 3.476383025533253,
+      "tokens_seen": 1708916736
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004904444444444445,
+      "loss": 2.5937,
+      "theoretical_loss": 3.4763715632997885,
+      "tokens_seen": 1708982272
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004904242424242424,
+      "loss": 2.7642,
+      "theoretical_loss": 3.476360101628938,
+      "tokens_seen": 1709047808
+    },
+    {
+      "epoch": 0.03,
+      "objective/train/docs_used": 943105,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5795464515686035,
+      "objective/train/theoretical_loss": 3.476348640520653,
+      "objective/train/tokens_used": 67972576,
+      "theoretical_loss": 3.476348640520653,
+      "tokens_seen": 1709113344
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004904040404040404,
+      "loss": 2.6005,
+      "theoretical_loss": 3.476348640520653,
+      "tokens_seen": 1709113344
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004903838383838384,
+      "loss": 2.4409,
+      "theoretical_loss": 3.476337179974884,
+      "tokens_seen": 1709178880
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004903636363636364,
+      "loss": 2.6039,
+      "theoretical_loss": 3.476325719991582,
+      "tokens_seen": 1709244416
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004903434343434343,
+      "loss": 2.5914,
+      "theoretical_loss": 3.476314260570698,
+      "tokens_seen": 1709309952
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004903232323232323,
+      "loss": 2.4368,
+      "theoretical_loss": 3.4763028017121824,
+      "tokens_seen": 1709375488
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004903030303030304,
+      "loss": 2.7762,
+      "theoretical_loss": 3.4762913434159866,
+      "tokens_seen": 1709441024
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004902828282828283,
+      "loss": 2.6016,
+      "theoretical_loss": 3.476279885682061,
+      "tokens_seen": 1709506560
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004902626262626263,
+      "loss": 2.6904,
+      "theoretical_loss": 3.4762684285103562,
+      "tokens_seen": 1709572096
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004902424242424243,
+      "loss": 2.6499,
+      "theoretical_loss": 3.476256971900824,
+      "tokens_seen": 1709637632
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004902222222222222,
+      "loss": 2.5542,
+      "theoretical_loss": 3.4762455158534147,
+      "tokens_seen": 1709703168
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004902020202020202,
+      "loss": 2.8664,
+      "theoretical_loss": 3.476234060368079,
+      "tokens_seen": 1709768704
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004901818181818182,
+      "loss": 2.7259,
+      "theoretical_loss": 3.4762226054447676,
+      "tokens_seen": 1709834240
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004901616161616162,
+      "loss": 2.7595,
+      "theoretical_loss": 3.476211151083432,
+      "tokens_seen": 1709899776
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004901414141414141,
+      "loss": 2.7021,
+      "theoretical_loss": 3.4761996972840232,
+      "tokens_seen": 1709965312
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004901212121212121,
+      "loss": 2.5646,
+      "theoretical_loss": 3.476188244046491,
+      "tokens_seen": 1710030848
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004901010101010101,
+      "loss": 2.7032,
+      "theoretical_loss": 3.4761767913707877,
+      "tokens_seen": 1710096384
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000490080808080808,
+      "loss": 2.5891,
+      "theoretical_loss": 3.4761653392568634,
+      "tokens_seen": 1710161920
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004900606060606061,
+      "loss": 2.6411,
+      "theoretical_loss": 3.4761538877046694,
+      "tokens_seen": 1710227456
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004900404040404041,
+      "loss": 2.5804,
+      "theoretical_loss": 3.476142436714156,
+      "tokens_seen": 1710292992
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004900202020202021,
+      "loss": 2.6421,
+      "theoretical_loss": 3.4761309862852747,
+      "tokens_seen": 1710358528
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00049,
+      "loss": 2.891,
+      "theoretical_loss": 3.4761195364179764,
+      "tokens_seen": 1710424064
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000489979797979798,
+      "loss": 2.6098,
+      "theoretical_loss": 3.4761080871122116,
+      "tokens_seen": 1710489600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000489959595959596,
+      "loss": 2.8071,
+      "theoretical_loss": 3.476096638367932,
+      "tokens_seen": 1710555136
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004899393939393939,
+      "loss": 2.5849,
+      "theoretical_loss": 3.4760851901850875,
+      "tokens_seen": 1710620672
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004899191919191919,
+      "loss": 2.8232,
+      "theoretical_loss": 3.4760737425636297,
+      "tokens_seen": 1710686208
+    },
+    {
+      "epoch": 0.03,
+      "objective/train/docs_used": 943708,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.789628744125366,
+      "objective/train/theoretical_loss": 3.4760622955035103,
+      "objective/train/tokens_used": 69610976,
+      "theoretical_loss": 3.4760622955035103,
+      "tokens_seen": 1710751744
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004898989898989899,
+      "loss": 2.7948,
+      "theoretical_loss": 3.4760622955035103,
+      "tokens_seen": 1710751744
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004898787878787878,
+      "loss": 2.7579,
+      "theoretical_loss": 3.476050849004679,
+      "tokens_seen": 1710817280
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004898585858585858,
+      "loss": 2.8273,
+      "theoretical_loss": 3.4760394030670874,
+      "tokens_seen": 1710882816
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004898383838383839,
+      "loss": 2.5987,
+      "theoretical_loss": 3.476027957690686,
+      "tokens_seen": 1710948352
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004898181818181819,
+      "loss": 2.6575,
+      "theoretical_loss": 3.476016512875426,
+      "tokens_seen": 1711013888
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004897979797979798,
+      "loss": 2.792,
+      "theoretical_loss": 3.4760050686212596,
+      "tokens_seen": 1711079424
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004897777777777778,
+      "loss": 2.703,
+      "theoretical_loss": 3.475993624928136,
+      "tokens_seen": 1711144960
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004897575757575758,
+      "loss": 2.6028,
+      "theoretical_loss": 3.475982181796007,
+      "tokens_seen": 1711210496
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004897373737373738,
+      "loss": 2.8526,
+      "theoretical_loss": 3.475970739224824,
+      "tokens_seen": 1711276032
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004897171717171717,
+      "loss": 2.5888,
+      "theoretical_loss": 3.475959297214537,
+      "tokens_seen": 1711341568
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004896969696969697,
+      "loss": 2.7614,
+      "theoretical_loss": 3.475947855765098,
+      "tokens_seen": 1711407104
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004896767676767676,
+      "loss": 2.723,
+      "theoretical_loss": 3.4759364148764575,
+      "tokens_seen": 1711472640
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004896565656565657,
+      "loss": 2.5174,
+      "theoretical_loss": 3.4759249745485667,
+      "tokens_seen": 1711538176
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004896363636363636,
+      "loss": 2.6374,
+      "theoretical_loss": 3.4759135347813768,
+      "tokens_seen": 1711603712
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004896161616161617,
+      "loss": 2.7387,
+      "theoretical_loss": 3.4759020955748388,
+      "tokens_seen": 1711669248
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004895959595959596,
+      "loss": 2.6615,
+      "theoretical_loss": 3.4758906569289034,
+      "tokens_seen": 1711734784
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004895757575757576,
+      "loss": 2.6527,
+      "theoretical_loss": 3.475879218843522,
+      "tokens_seen": 1711800320
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004895555555555556,
+      "loss": 2.6785,
+      "theoretical_loss": 3.475867781318646,
+      "tokens_seen": 1711865856
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004895353535353536,
+      "loss": 2.4673,
+      "theoretical_loss": 3.4758563443542254,
+      "tokens_seen": 1711931392
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004895151515151515,
+      "loss": 2.6067,
+      "theoretical_loss": 3.4758449079502123,
+      "tokens_seen": 1711996928
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004894949494949495,
+      "loss": 2.6138,
+      "theoretical_loss": 3.475833472106557,
+      "tokens_seen": 1712062464
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004894747474747475,
+      "loss": 2.6808,
+      "theoretical_loss": 3.475822036823212,
+      "tokens_seen": 1712128000
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004894545454545455,
+      "loss": 2.6603,
+      "theoretical_loss": 3.4758106021001267,
+      "tokens_seen": 1712193536
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004894343434343434,
+      "loss": 2.4896,
+      "theoretical_loss": 3.4757991679372533,
+      "tokens_seen": 1712259072
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004894141414141414,
+      "loss": 2.5848,
+      "theoretical_loss": 3.4757877343345425,
+      "tokens_seen": 1712324608
+    },
+    {
+      "epoch": 0.03,
+      "objective/train/docs_used": 944906,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.562922239303589,
+      "objective/train/theoretical_loss": 3.4757763012919454,
+      "objective/train/tokens_used": 71249376,
+      "theoretical_loss": 3.4757763012919454,
+      "tokens_seen": 1712390144
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004893939393939393,
+      "loss": 2.6692,
+      "theoretical_loss": 3.4757763012919454,
+      "tokens_seen": 1712390144
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004893737373737374,
+      "loss": 2.848,
+      "theoretical_loss": 3.475764868809413,
+      "tokens_seen": 1712455680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004893535353535354,
+      "loss": 2.437,
+      "theoretical_loss": 3.4757534368868974,
+      "tokens_seen": 1712521216
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004893333333333334,
+      "loss": 2.7263,
+      "theoretical_loss": 3.4757420055243484,
+      "tokens_seen": 1712586752
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004893131313131313,
+      "loss": 2.6201,
+      "theoretical_loss": 3.4757305747217178,
+      "tokens_seen": 1712652288
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004892929292929293,
+      "loss": 2.774,
+      "theoretical_loss": 3.4757191444789566,
+      "tokens_seen": 1712717824
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004892727272727273,
+      "loss": 2.8183,
+      "theoretical_loss": 3.475707714796016,
+      "tokens_seen": 1712783360
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004892525252525253,
+      "loss": 2.6004,
+      "theoretical_loss": 3.4756962856728473,
+      "tokens_seen": 1712848896
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004892323232323232,
+      "loss": 2.8133,
+      "theoretical_loss": 3.4756848571094014,
+      "tokens_seen": 1712914432
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004892121212121212,
+      "loss": 2.8219,
+      "theoretical_loss": 3.4756734291056297,
+      "tokens_seen": 1712979968
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004891919191919192,
+      "loss": 2.826,
+      "theoretical_loss": 3.475662001661483,
+      "tokens_seen": 1713045504
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004891717171717171,
+      "loss": 2.6803,
+      "theoretical_loss": 3.475650574776914,
+      "tokens_seen": 1713111040
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004891515151515151,
+      "loss": 2.7393,
+      "theoretical_loss": 3.4756391484518714,
+      "tokens_seen": 1713176576
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004891313131313132,
+      "loss": 2.5757,
+      "theoretical_loss": 3.4756277226863084,
+      "tokens_seen": 1713242112
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004891111111111112,
+      "loss": 2.7433,
+      "theoretical_loss": 3.475616297480175,
+      "tokens_seen": 1713307648
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004890909090909091,
+      "loss": 2.5726,
+      "theoretical_loss": 3.475604872833424,
+      "tokens_seen": 1713373184
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004890707070707071,
+      "loss": 2.7851,
+      "theoretical_loss": 3.4755934487460047,
+      "tokens_seen": 1713438720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004890505050505051,
+      "loss": 2.6237,
+      "theoretical_loss": 3.475582025217869,
+      "tokens_seen": 1713504256
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000489030303030303,
+      "loss": 2.8517,
+      "theoretical_loss": 3.4755706022489687,
+      "tokens_seen": 1713569792
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000489010101010101,
+      "loss": 2.4496,
+      "theoretical_loss": 3.475559179839254,
+      "tokens_seen": 1713635328
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000488989898989899,
+      "loss": 2.4707,
+      "theoretical_loss": 3.4755477579886778,
+      "tokens_seen": 1713700864
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004889696969696969,
+      "loss": 2.7956,
+      "theoretical_loss": 3.4755363366971896,
+      "tokens_seen": 1713766400
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004889494949494949,
+      "loss": 2.6349,
+      "theoretical_loss": 3.4755249159647414,
+      "tokens_seen": 1713831936
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004889292929292929,
+      "loss": 2.6285,
+      "theoretical_loss": 3.4755134957912848,
+      "tokens_seen": 1713897472
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000488909090909091,
+      "loss": 2.7714,
+      "theoretical_loss": 3.4755020761767703,
+      "tokens_seen": 1713963008
+    },
+    {
+      "epoch": 0.03,
+      "objective/train/docs_used": 945396,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.1092090606689453,
+      "objective/train/theoretical_loss": 3.47549065712115,
+      "objective/train/tokens_used": 72887776,
+      "theoretical_loss": 3.47549065712115,
+      "tokens_seen": 1714028544
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004888888888888889,
+      "loss": 2.4456,
+      "theoretical_loss": 3.47549065712115,
+      "tokens_seen": 1714028544
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004888686868686869,
+      "loss": 2.5141,
+      "theoretical_loss": 3.4754792386243745,
+      "tokens_seen": 1714094080
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004888484848484849,
+      "loss": 2.7276,
+      "theoretical_loss": 3.4754678206863954,
+      "tokens_seen": 1714159616
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004888282828282829,
+      "loss": 2.5177,
+      "theoretical_loss": 3.475456403307164,
+      "tokens_seen": 1714225152
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004888080808080808,
+      "loss": 2.7962,
+      "theoretical_loss": 3.4754449864866315,
+      "tokens_seen": 1714290688
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004887878787878788,
+      "loss": 2.5457,
+      "theoretical_loss": 3.475433570224749,
+      "tokens_seen": 1714356224
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004887676767676768,
+      "loss": 2.6354,
+      "theoretical_loss": 3.475422154521468,
+      "tokens_seen": 1714421760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004887474747474748,
+      "loss": 2.6071,
+      "theoretical_loss": 3.47541073937674,
+      "tokens_seen": 1714487296
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004887272727272727,
+      "loss": 2.5188,
+      "theoretical_loss": 3.475399324790516,
+      "tokens_seen": 1714552832
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004887070707070707,
+      "loss": 2.874,
+      "theoretical_loss": 3.4753879107627483,
+      "tokens_seen": 1714618368
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004886868686868686,
+      "loss": 2.7343,
+      "theoretical_loss": 3.4753764972933867,
+      "tokens_seen": 1714683904
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004886666666666667,
+      "loss": 2.8105,
+      "theoretical_loss": 3.4753650843823833,
+      "tokens_seen": 1714749440
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004886464646464647,
+      "loss": 2.8456,
+      "theoretical_loss": 3.47535367202969,
+      "tokens_seen": 1714814976
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004886262626262627,
+      "loss": 2.5544,
+      "theoretical_loss": 3.475342260235257,
+      "tokens_seen": 1714880512
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004886060606060606,
+      "loss": 2.7038,
+      "theoretical_loss": 3.475330848999036,
+      "tokens_seen": 1714946048
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004885858585858586,
+      "loss": 2.4342,
+      "theoretical_loss": 3.4753194383209793,
+      "tokens_seen": 1715011584
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004885656565656566,
+      "loss": 2.8637,
+      "theoretical_loss": 3.475308028201037,
+      "tokens_seen": 1715077120
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004885454545454546,
+      "loss": 2.8579,
+      "theoretical_loss": 3.4752966186391614,
+      "tokens_seen": 1715142656
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004885252525252525,
+      "loss": 2.7374,
+      "theoretical_loss": 3.4752852096353033,
+      "tokens_seen": 1715208192
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004885050505050505,
+      "loss": 2.6466,
+      "theoretical_loss": 3.4752738011894144,
+      "tokens_seen": 1715273728
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004884848484848484,
+      "loss": 2.7546,
+      "theoretical_loss": 3.4752623933014464,
+      "tokens_seen": 1715339264
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004884646464646465,
+      "loss": 2.5621,
+      "theoretical_loss": 3.47525098597135,
+      "tokens_seen": 1715404800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004884444444444445,
+      "loss": 2.5583,
+      "theoretical_loss": 3.475239579199077,
+      "tokens_seen": 1715470336
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004884242424242425,
+      "loss": 2.6457,
+      "theoretical_loss": 3.4752281729845786,
+      "tokens_seen": 1715535872
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004884040404040404,
+      "loss": 2.7425,
+      "theoretical_loss": 3.4752167673278063,
+      "tokens_seen": 1715601408
+    },
+    {
+      "epoch": 0.03,
+      "objective/train/docs_used": 946450,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.578782796859741,
+      "objective/train/theoretical_loss": 3.475205362228712,
+      "objective/train/tokens_used": 74526176,
+      "theoretical_loss": 3.475205362228712,
+      "tokens_seen": 1715666944
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004883838383838384,
+      "loss": 2.544,
+      "theoretical_loss": 3.475205362228712,
+      "tokens_seen": 1715666944
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004883636363636364,
+      "loss": 2.5762,
+      "theoretical_loss": 3.475193957687246,
+      "tokens_seen": 1715732480
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004883434343434344,
+      "loss": 2.6404,
+      "theoretical_loss": 3.4751825537033607,
+      "tokens_seen": 1715798016
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004883232323232323,
+      "loss": 2.5441,
+      "theoretical_loss": 3.4751711502770077,
+      "tokens_seen": 1715863552
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004883030303030303,
+      "loss": 2.6666,
+      "theoretical_loss": 3.4751597474081377,
+      "tokens_seen": 1715929088
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004882828282828283,
+      "loss": 2.4691,
+      "theoretical_loss": 3.4751483450967027,
+      "tokens_seen": 1715994624
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004882626262626263,
+      "loss": 2.8491,
+      "theoretical_loss": 3.4751369433426538,
+      "tokens_seen": 1716060160
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004882424242424243,
+      "loss": 2.6214,
+      "theoretical_loss": 3.4751255421459426,
+      "tokens_seen": 1716125696
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00048822222222222225,
+      "loss": 2.7773,
+      "theoretical_loss": 3.4751141415065208,
+      "tokens_seen": 1716191232
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004882020202020202,
+      "loss": 2.4918,
+      "theoretical_loss": 3.4751027414243394,
+      "tokens_seen": 1716256768
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00048818181818181816,
+      "loss": 2.5795,
+      "theoretical_loss": 3.4750913418993505,
+      "tokens_seen": 1716322304
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00048816161616161617,
+      "loss": 2.5246,
+      "theoretical_loss": 3.4750799429315054,
+      "tokens_seen": 1716387840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004881414141414142,
+      "loss": 2.514,
+      "theoretical_loss": 3.475068544520755,
+      "tokens_seen": 1716453376
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00048812121212121214,
+      "loss": 2.639,
+      "theoretical_loss": 3.4750571466670515,
+      "tokens_seen": 1716518912
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004881010101010101,
+      "loss": 2.6768,
+      "theoretical_loss": 3.475045749370346,
+      "tokens_seen": 1716584448
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00048808080808080805,
+      "loss": 2.7541,
+      "theoretical_loss": 3.4750343526305905,
+      "tokens_seen": 1716649984
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004880606060606061,
+      "loss": 2.5554,
+      "theoretical_loss": 3.475022956447736,
+      "tokens_seen": 1716715520
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004880404040404041,
+      "loss": 2.7061,
+      "theoretical_loss": 3.4750115608217342,
+      "tokens_seen": 1716781056
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00048802020202020203,
+      "loss": 2.5479,
+      "theoretical_loss": 3.4750001657525367,
+      "tokens_seen": 1716846592
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000488,
+      "loss": 2.5462,
+      "theoretical_loss": 3.4749887712400955,
+      "tokens_seen": 1716912128
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00048797979797979795,
+      "loss": 2.4832,
+      "theoretical_loss": 3.474977377284361,
+      "tokens_seen": 1716977664
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000487959595959596,
+      "loss": 2.7024,
+      "theoretical_loss": 3.4749659838852858,
+      "tokens_seen": 1717043200
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00048793939393939397,
+      "loss": 2.841,
+      "theoretical_loss": 3.4749545910428212,
+      "tokens_seen": 1717108736
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004879191919191919,
+      "loss": 2.589,
+      "theoretical_loss": 3.474943198756918,
+      "tokens_seen": 1717174272
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004878989898989899,
+      "loss": 2.5626,
+      "theoretical_loss": 3.474931807027529,
+      "tokens_seen": 1717239808
+    },
+    {
+      "epoch": 0.03,
+      "objective/train/docs_used": 947071,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.492351531982422,
+      "objective/train/theoretical_loss": 3.474920415854605,
+      "objective/train/tokens_used": 76164576,
+      "theoretical_loss": 3.474920415854605,
+      "tokens_seen": 1717305344
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00048787878787878784,
+      "loss": 2.7781,
+      "theoretical_loss": 3.474920415854605,
+      "tokens_seen": 1717305344
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004878585858585859,
+      "loss": 2.6613,
+      "theoretical_loss": 3.474909025238098,
+      "tokens_seen": 1717370880
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00048783838383838386,
+      "loss": 2.6513,
+      "theoretical_loss": 3.474897635177959,
+      "tokens_seen": 1717436416
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004878181818181818,
+      "loss": 2.5682,
+      "theoretical_loss": 3.4748862456741403,
+      "tokens_seen": 1717501952
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004877979797979798,
+      "loss": 2.7969,
+      "theoretical_loss": 3.4748748567265926,
+      "tokens_seen": 1717567488
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004877777777777778,
+      "loss": 2.7713,
+      "theoretical_loss": 3.474863468335269,
+      "tokens_seen": 1717633024
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004877575757575758,
+      "loss": 2.7294,
+      "theoretical_loss": 3.4748520805001193,
+      "tokens_seen": 1717698560
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00048773737373737375,
+      "loss": 2.4451,
+      "theoretical_loss": 3.474840693221096,
+      "tokens_seen": 1717764096
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004877171717171717,
+      "loss": 2.5473,
+      "theoretical_loss": 3.4748293064981515,
+      "tokens_seen": 1717829632
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00048769696969696967,
+      "loss": 2.6743,
+      "theoretical_loss": 3.474817920331236,
+      "tokens_seen": 1717895168
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004876767676767677,
+      "loss": 2.4018,
+      "theoretical_loss": 3.4748065347203023,
+      "tokens_seen": 1717960704
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004876565656565657,
+      "loss": 2.5568,
+      "theoretical_loss": 3.474795149665301,
+      "tokens_seen": 1718026240
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00048763636363636365,
+      "loss": 2.4997,
+      "theoretical_loss": 3.474783765166185,
+      "tokens_seen": 1718091776
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004876161616161616,
+      "loss": 2.5956,
+      "theoretical_loss": 3.4747723812229045,
+      "tokens_seen": 1718157312
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004875959595959596,
+      "loss": 2.4705,
+      "theoretical_loss": 3.4747609978354124,
+      "tokens_seen": 1718222848
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00048757575757575757,
+      "loss": 2.8133,
+      "theoretical_loss": 3.4747496150036596,
+      "tokens_seen": 1718288384
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004875555555555556,
+      "loss": 2.6255,
+      "theoretical_loss": 3.4747382327275984,
+      "tokens_seen": 1718353920
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00048753535353535354,
+      "loss": 2.3969,
+      "theoretical_loss": 3.47472685100718,
+      "tokens_seen": 1718419456
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004875151515151515,
+      "loss": 2.8356,
+      "theoretical_loss": 3.474715469842356,
+      "tokens_seen": 1718484992
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004874949494949495,
+      "loss": 2.4337,
+      "theoretical_loss": 3.4747040892330787,
+      "tokens_seen": 1718550528
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00048747474747474746,
+      "loss": 2.5269,
+      "theoretical_loss": 3.474692709179299,
+      "tokens_seen": 1718616064
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004874545454545455,
+      "loss": 2.7374,
+      "theoretical_loss": 3.474681329680969,
+      "tokens_seen": 1718681600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00048743434343434343,
+      "loss": 2.7276,
+      "theoretical_loss": 3.4746699507380407,
+      "tokens_seen": 1718747136
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00048741414141414144,
+      "loss": 2.5347,
+      "theoretical_loss": 3.4746585723504655,
+      "tokens_seen": 1718812672
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0004873939393939394,
+      "loss": 2.6824,
+      "theoretical_loss": 3.4746471945181954,
+      "tokens_seen": 1718878208
+    },
+    {
+      "epoch": 0.03,
+      "objective/train/docs_used": 948336,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.850125789642334,
+      "objective/train/theoretical_loss": 3.4746358172411815,
+      "objective/train/tokens_used": 77802976,
+      "theoretical_loss": 3.4746358172411815,
+      "tokens_seen": 1718943744
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048737373737373736,
+      "loss": 2.6919,
+      "theoretical_loss": 3.4746358172411815,
+      "tokens_seen": 1718943744
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048735353535353537,
+      "loss": 2.537,
+      "theoretical_loss": 3.474624440519376,
+      "tokens_seen": 1719009280
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004873333333333333,
+      "loss": 2.7071,
+      "theoretical_loss": 3.4746130643527304,
+      "tokens_seen": 1719074816
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048731313131313134,
+      "loss": 2.3608,
+      "theoretical_loss": 3.474601688741197,
+      "tokens_seen": 1719140352
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004872929292929293,
+      "loss": 2.6354,
+      "theoretical_loss": 3.474590313684727,
+      "tokens_seen": 1719205888
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048727272727272725,
+      "loss": 2.7612,
+      "theoretical_loss": 3.474578939183272,
+      "tokens_seen": 1719271424
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048725252525252526,
+      "loss": 2.561,
+      "theoretical_loss": 3.474567565236785,
+      "tokens_seen": 1719336960
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048723232323232327,
+      "loss": 2.7526,
+      "theoretical_loss": 3.474556191845216,
+      "tokens_seen": 1719402496
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048721212121212123,
+      "loss": 2.8614,
+      "theoretical_loss": 3.4745448190085177,
+      "tokens_seen": 1719468032
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004871919191919192,
+      "loss": 2.3397,
+      "theoretical_loss": 3.4745334467266424,
+      "tokens_seen": 1719533568
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048717171717171714,
+      "loss": 2.4128,
+      "theoretical_loss": 3.4745220749995407,
+      "tokens_seen": 1719599104
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048715151515151515,
+      "loss": 2.5001,
+      "theoretical_loss": 3.474510703827165,
+      "tokens_seen": 1719664640
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048713131313131316,
+      "loss": 2.8698,
+      "theoretical_loss": 3.474499333209468,
+      "tokens_seen": 1719730176
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004871111111111111,
+      "loss": 2.8356,
+      "theoretical_loss": 3.4744879631464,
+      "tokens_seen": 1719795712
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004870909090909091,
+      "loss": 2.5725,
+      "theoretical_loss": 3.474476593637913,
+      "tokens_seen": 1719861248
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048707070707070704,
+      "loss": 2.4696,
+      "theoretical_loss": 3.47446522468396,
+      "tokens_seen": 1719926784
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004870505050505051,
+      "loss": 2.5063,
+      "theoretical_loss": 3.4744538562844913,
+      "tokens_seen": 1719992320
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048703030303030306,
+      "loss": 2.478,
+      "theoretical_loss": 3.4744424884394602,
+      "tokens_seen": 1720057856
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000487010101010101,
+      "loss": 2.609,
+      "theoretical_loss": 3.4744311211488172,
+      "tokens_seen": 1720123392
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048698989898989897,
+      "loss": 2.8236,
+      "theoretical_loss": 3.4744197544125153,
+      "tokens_seen": 1720188928
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000486969696969697,
+      "loss": 2.7664,
+      "theoretical_loss": 3.4744083882305055,
+      "tokens_seen": 1720254464
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000486949494949495,
+      "loss": 2.6229,
+      "theoretical_loss": 3.4743970226027403,
+      "tokens_seen": 1720320000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048692929292929295,
+      "loss": 2.8784,
+      "theoretical_loss": 3.474385657529171,
+      "tokens_seen": 1720385536
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004869090909090909,
+      "loss": 2.6972,
+      "theoretical_loss": 3.47437429300975,
+      "tokens_seen": 1720451072
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048688888888888886,
+      "loss": 2.9794,
+      "theoretical_loss": 3.4743629290444282,
+      "tokens_seen": 1720516608
+    },
+    {
+      "epoch": 0.04,
+      "objective/train/docs_used": 949353,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.0498993396759033,
+      "objective/train/theoretical_loss": 3.4743515656331594,
+      "objective/train/tokens_used": 79441376,
+      "theoretical_loss": 3.4743515656331594,
+      "tokens_seen": 1720582144
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048686868686868693,
+      "loss": 2.8742,
+      "theoretical_loss": 3.4743515656331594,
+      "tokens_seen": 1720582144
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004868484848484849,
+      "loss": 2.6368,
+      "theoretical_loss": 3.474340202775893,
+      "tokens_seen": 1720647680
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048682828282828284,
+      "loss": 2.6623,
+      "theoretical_loss": 3.474328840472583,
+      "tokens_seen": 1720713216
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004868080808080808,
+      "loss": 2.5302,
+      "theoretical_loss": 3.47431747872318,
+      "tokens_seen": 1720778752
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048678787878787876,
+      "loss": 2.827,
+      "theoretical_loss": 3.4743061175276364,
+      "tokens_seen": 1720844288
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004867676767676768,
+      "loss": 2.6887,
+      "theoretical_loss": 3.474294756885904,
+      "tokens_seen": 1720909824
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004867474747474748,
+      "loss": 2.6388,
+      "theoretical_loss": 3.4742833967979347,
+      "tokens_seen": 1720975360
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048672727272727274,
+      "loss": 2.5768,
+      "theoretical_loss": 3.474272037263681,
+      "tokens_seen": 1721040896
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004867070707070707,
+      "loss": 2.5206,
+      "theoretical_loss": 3.474260678283094,
+      "tokens_seen": 1721106432
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004866868686868687,
+      "loss": 2.7724,
+      "theoretical_loss": 3.474249319856126,
+      "tokens_seen": 1721171968
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004866666666666667,
+      "loss": 2.8046,
+      "theoretical_loss": 3.4742379619827286,
+      "tokens_seen": 1721237504
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048664646464646467,
+      "loss": 2.6154,
+      "theoretical_loss": 3.4742266046628543,
+      "tokens_seen": 1721303040
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048662626262626263,
+      "loss": 2.7199,
+      "theoretical_loss": 3.4742152478964545,
+      "tokens_seen": 1721368576
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004866060606060606,
+      "loss": 2.5633,
+      "theoretical_loss": 3.474203891683482,
+      "tokens_seen": 1721434112
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004865858585858586,
+      "loss": 2.7631,
+      "theoretical_loss": 3.474192536023888,
+      "tokens_seen": 1721499648
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004865656565656566,
+      "loss": 2.5908,
+      "theoretical_loss": 3.474181180917624,
+      "tokens_seen": 1721565184
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048654545454545456,
+      "loss": 2.601,
+      "theoretical_loss": 3.474169826364643,
+      "tokens_seen": 1721630720
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004865252525252525,
+      "loss": 2.8114,
+      "theoretical_loss": 3.4741584723648975,
+      "tokens_seen": 1721696256
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048650505050505053,
+      "loss": 2.7601,
+      "theoretical_loss": 3.4741471189183377,
+      "tokens_seen": 1721761792
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004864848484848485,
+      "loss": 2.546,
+      "theoretical_loss": 3.474135766024917,
+      "tokens_seen": 1721827328
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004864646464646465,
+      "loss": 2.4717,
+      "theoretical_loss": 3.4741244136845864,
+      "tokens_seen": 1721892864
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048644444444444446,
+      "loss": 2.68,
+      "theoretical_loss": 3.474113061897299,
+      "tokens_seen": 1721958400
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004864242424242424,
+      "loss": 2.5694,
+      "theoretical_loss": 3.474101710663006,
+      "tokens_seen": 1722023936
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004864040404040404,
+      "loss": 2.6604,
+      "theoretical_loss": 3.474090359981659,
+      "tokens_seen": 1722089472
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004863838383838384,
+      "loss": 2.6332,
+      "theoretical_loss": 3.4740790098532113,
+      "tokens_seen": 1722155008
+    },
+    {
+      "epoch": 0.04,
+      "objective/train/docs_used": 949989,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.493234634399414,
+      "objective/train/theoretical_loss": 3.4740676602776146,
+      "objective/train/tokens_used": 81079776,
+      "theoretical_loss": 3.4740676602776146,
+      "tokens_seen": 1722220544
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004863636363636364,
+      "loss": 2.5725,
+      "theoretical_loss": 3.4740676602776146,
+      "tokens_seen": 1722220544
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048634343434343435,
+      "loss": 2.5757,
+      "theoretical_loss": 3.47405631125482,
+      "tokens_seen": 1722286080
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048632323232323236,
+      "loss": 2.6726,
+      "theoretical_loss": 3.4740449627847805,
+      "tokens_seen": 1722351616
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004863030303030303,
+      "loss": 2.5449,
+      "theoretical_loss": 3.4740336148674476,
+      "tokens_seen": 1722417152
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004862828282828283,
+      "loss": 2.7329,
+      "theoretical_loss": 3.4740222675027734,
+      "tokens_seen": 1722482688
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004862626262626263,
+      "loss": 2.5414,
+      "theoretical_loss": 3.474010920690711,
+      "tokens_seen": 1722548224
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048624242424242424,
+      "loss": 2.5322,
+      "theoretical_loss": 3.4739995744312107,
+      "tokens_seen": 1722613760
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048622222222222225,
+      "loss": 2.6438,
+      "theoretical_loss": 3.4739882287242256,
+      "tokens_seen": 1722679296
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004862020202020202,
+      "loss": 2.7798,
+      "theoretical_loss": 3.473976883569708,
+      "tokens_seen": 1722744832
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048618181818181817,
+      "loss": 2.6148,
+      "theoretical_loss": 3.47396553896761,
+      "tokens_seen": 1722810368
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004861616161616162,
+      "loss": 2.7383,
+      "theoretical_loss": 3.4739541949178827,
+      "tokens_seen": 1722875904
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004861414141414142,
+      "loss": 2.5262,
+      "theoretical_loss": 3.4739428514204787,
+      "tokens_seen": 1722941440
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048612121212121215,
+      "loss": 2.6558,
+      "theoretical_loss": 3.47393150847535,
+      "tokens_seen": 1723006976
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004861010101010101,
+      "loss": 2.7145,
+      "theoretical_loss": 3.4739201660824497,
+      "tokens_seen": 1723072512
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048608080808080806,
+      "loss": 2.8424,
+      "theoretical_loss": 3.473908824241729,
+      "tokens_seen": 1723138048
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048606060606060607,
+      "loss": 2.4815,
+      "theoretical_loss": 3.4738974829531397,
+      "tokens_seen": 1723203584
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004860404040404041,
+      "loss": 2.5685,
+      "theoretical_loss": 3.4738861422166343,
+      "tokens_seen": 1723269120
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048602020202020204,
+      "loss": 2.5099,
+      "theoretical_loss": 3.4738748020321655,
+      "tokens_seen": 1723334656
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000486,
+      "loss": 2.5624,
+      "theoretical_loss": 3.4738634623996845,
+      "tokens_seen": 1723400192
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048597979797979795,
+      "loss": 2.8233,
+      "theoretical_loss": 3.473852123319144,
+      "tokens_seen": 1723465728
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048595959595959596,
+      "loss": 2.6669,
+      "theoretical_loss": 3.4738407847904966,
+      "tokens_seen": 1723531264
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000485939393939394,
+      "loss": 2.6042,
+      "theoretical_loss": 3.4738294468136934,
+      "tokens_seen": 1723596800
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048591919191919193,
+      "loss": 2.7076,
+      "theoretical_loss": 3.473818109388687,
+      "tokens_seen": 1723662336
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004858989898989899,
+      "loss": 2.6665,
+      "theoretical_loss": 3.47380677251543,
+      "tokens_seen": 1723727872
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048587878787878784,
+      "loss": 2.7535,
+      "theoretical_loss": 3.473795436193874,
+      "tokens_seen": 1723793408
+    },
+    {
+      "epoch": 0.04,
+      "objective/train/docs_used": 951217,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.653082847595215,
+      "objective/train/theoretical_loss": 3.473784100423971,
+      "objective/train/tokens_used": 82718176,
+      "theoretical_loss": 3.473784100423971,
+      "tokens_seen": 1723858944
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004858585858585859,
+      "loss": 2.8862,
+      "theoretical_loss": 3.473784100423971,
+      "tokens_seen": 1723858944
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048583838383838387,
+      "loss": 2.6551,
+      "theoretical_loss": 3.4737727652056742,
+      "tokens_seen": 1723924480
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004858181818181818,
+      "loss": 2.5733,
+      "theoretical_loss": 3.4737614305389344,
+      "tokens_seen": 1723990016
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004857979797979798,
+      "loss": 2.7681,
+      "theoretical_loss": 3.473750096423705,
+      "tokens_seen": 1724055552
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048577777777777774,
+      "loss": 2.5725,
+      "theoretical_loss": 3.473738762859938,
+      "tokens_seen": 1724121088
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004857575757575758,
+      "loss": 2.7263,
+      "theoretical_loss": 3.4737274298475844,
+      "tokens_seen": 1724186624
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048573737373737376,
+      "loss": 2.6403,
+      "theoretical_loss": 3.4737160973865984,
+      "tokens_seen": 1724252160
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004857171717171717,
+      "loss": 2.5917,
+      "theoretical_loss": 3.4737047654769304,
+      "tokens_seen": 1724317696
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004856969696969697,
+      "loss": 2.7386,
+      "theoretical_loss": 3.473693434118534,
+      "tokens_seen": 1724383232
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004856767676767677,
+      "loss": 2.7532,
+      "theoretical_loss": 3.4736821033113605,
+      "tokens_seen": 1724448768
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004856565656565657,
+      "loss": 2.6684,
+      "theoretical_loss": 3.4736707730553626,
+      "tokens_seen": 1724514304
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048563636363636365,
+      "loss": 2.4919,
+      "theoretical_loss": 3.473659443350493,
+      "tokens_seen": 1724579840
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004856161616161616,
+      "loss": 2.739,
+      "theoretical_loss": 3.4736481141967026,
+      "tokens_seen": 1724645376
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048559595959595957,
+      "loss": 2.8205,
+      "theoretical_loss": 3.4736367855939445,
+      "tokens_seen": 1724710912
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004855757575757576,
+      "loss": 2.8129,
+      "theoretical_loss": 3.4736254575421706,
+      "tokens_seen": 1724776448
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004855555555555556,
+      "loss": 2.5104,
+      "theoretical_loss": 3.4736141300413337,
+      "tokens_seen": 1724841984
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048553535353535355,
+      "loss": 2.5611,
+      "theoretical_loss": 3.4736028030913864,
+      "tokens_seen": 1724907520
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004855151515151515,
+      "loss": 2.564,
+      "theoretical_loss": 3.47359147669228,
+      "tokens_seen": 1724973056
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004854949494949495,
+      "loss": 2.6007,
+      "theoretical_loss": 3.473580150843967,
+      "tokens_seen": 1725038592
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048547474747474747,
+      "loss": 2.7037,
+      "theoretical_loss": 3.4735688255464003,
+      "tokens_seen": 1725104128
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004854545454545455,
+      "loss": 2.6524,
+      "theoretical_loss": 3.4735575007995316,
+      "tokens_seen": 1725169664
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048543434343434344,
+      "loss": 2.6171,
+      "theoretical_loss": 3.4735461766033136,
+      "tokens_seen": 1725235200
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004854141414141414,
+      "loss": 2.682,
+      "theoretical_loss": 3.473534852957698,
+      "tokens_seen": 1725300736
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004853939393939394,
+      "loss": 2.7294,
+      "theoretical_loss": 3.4735235298626375,
+      "tokens_seen": 1725366272
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048537373737373736,
+      "loss": 2.5625,
+      "theoretical_loss": 3.473512207318085,
+      "tokens_seen": 1725431808
+    },
+    {
+      "epoch": 0.04,
+      "objective/train/docs_used": 951846,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.9040403366088867,
+      "objective/train/theoretical_loss": 3.473500885323992,
+      "objective/train/tokens_used": 84356576,
+      "theoretical_loss": 3.473500885323992,
+      "tokens_seen": 1725497344
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004853535353535354,
+      "loss": 2.6504,
+      "theoretical_loss": 3.473500885323992,
+      "tokens_seen": 1725497344
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048533333333333333,
+      "loss": 2.8612,
+      "theoretical_loss": 3.473489563880311,
+      "tokens_seen": 1725562880
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048531313131313134,
+      "loss": 2.747,
+      "theoretical_loss": 3.473478242986994,
+      "tokens_seen": 1725628416
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004852929292929293,
+      "loss": 2.6916,
+      "theoretical_loss": 3.4734669226439943,
+      "tokens_seen": 1725693952
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004852727272727273,
+      "loss": 2.5991,
+      "theoretical_loss": 3.473455602851264,
+      "tokens_seen": 1725759488
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048525252525252527,
+      "loss": 2.5984,
+      "theoretical_loss": 3.473444283608755,
+      "tokens_seen": 1725825024
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004852323232323232,
+      "loss": 2.772,
+      "theoretical_loss": 3.4734329649164195,
+      "tokens_seen": 1725890560
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048521212121212123,
+      "loss": 2.6357,
+      "theoretical_loss": 3.4734216467742103,
+      "tokens_seen": 1725956096
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004851919191919192,
+      "loss": 2.6213,
+      "theoretical_loss": 3.47341032918208,
+      "tokens_seen": 1726021632
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004851717171717172,
+      "loss": 2.8269,
+      "theoretical_loss": 3.4733990121399807,
+      "tokens_seen": 1726087168
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048515151515151516,
+      "loss": 2.6231,
+      "theoretical_loss": 3.4733876956478644,
+      "tokens_seen": 1726152704
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048513131313131317,
+      "loss": 2.6762,
+      "theoretical_loss": 3.4733763797056842,
+      "tokens_seen": 1726218240
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004851111111111111,
+      "loss": 2.7135,
+      "theoretical_loss": 3.473365064313392,
+      "tokens_seen": 1726283776
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004850909090909091,
+      "loss": 2.7963,
+      "theoretical_loss": 3.4733537494709408,
+      "tokens_seen": 1726349312
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004850707070707071,
+      "loss": 2.7273,
+      "theoretical_loss": 3.4733424351782816,
+      "tokens_seen": 1726414848
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048505050505050505,
+      "loss": 2.9032,
+      "theoretical_loss": 3.4733311214353684,
+      "tokens_seen": 1726480384
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048503030303030306,
+      "loss": 2.7651,
+      "theoretical_loss": 3.4733198082421533,
+      "tokens_seen": 1726545920
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000485010101010101,
+      "loss": 2.5911,
+      "theoretical_loss": 3.473308495598588,
+      "tokens_seen": 1726611456
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000484989898989899,
+      "loss": 2.6404,
+      "theoretical_loss": 3.4732971835046254,
+      "tokens_seen": 1726676992
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000484969696969697,
+      "loss": 2.6003,
+      "theoretical_loss": 3.4732858719602175,
+      "tokens_seen": 1726742528
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000484949494949495,
+      "loss": 2.6377,
+      "theoretical_loss": 3.4732745609653177,
+      "tokens_seen": 1726808064
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048492929292929296,
+      "loss": 2.6006,
+      "theoretical_loss": 3.473263250519878,
+      "tokens_seen": 1726873600
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004849090909090909,
+      "loss": 2.6061,
+      "theoretical_loss": 3.4732519406238507,
+      "tokens_seen": 1726939136
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048488888888888887,
+      "loss": 2.6086,
+      "theoretical_loss": 3.4732406312771884,
+      "tokens_seen": 1727004672
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004848686868686869,
+      "loss": 2.8007,
+      "theoretical_loss": 3.473229322479843,
+      "tokens_seen": 1727070208
+    },
+    {
+      "epoch": 0.04,
+      "objective/train/docs_used": 953107,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.85353422164917,
+      "objective/train/theoretical_loss": 3.473218014231768,
+      "objective/train/tokens_used": 85994976,
+      "theoretical_loss": 3.473218014231768,
+      "tokens_seen": 1727135744
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004848484848484849,
+      "loss": 2.6263,
+      "theoretical_loss": 3.473218014231768,
+      "tokens_seen": 1727135744
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048482828282828285,
+      "loss": 2.5669,
+      "theoretical_loss": 3.473206706532915,
+      "tokens_seen": 1727201280
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004848080808080808,
+      "loss": 2.4935,
+      "theoretical_loss": 3.473195399383237,
+      "tokens_seen": 1727266816
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048478787878787876,
+      "loss": 2.5743,
+      "theoretical_loss": 3.4731840927826867,
+      "tokens_seen": 1727332352
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048476767676767683,
+      "loss": 2.5745,
+      "theoretical_loss": 3.473172786731216,
+      "tokens_seen": 1727397888
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004847474747474748,
+      "loss": 2.8035,
+      "theoretical_loss": 3.4731614812287774,
+      "tokens_seen": 1727463424
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048472727272727274,
+      "loss": 2.7708,
+      "theoretical_loss": 3.4731501762753236,
+      "tokens_seen": 1727528960
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004847070707070707,
+      "loss": 2.5373,
+      "theoretical_loss": 3.4731388718708076,
+      "tokens_seen": 1727594496
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048468686868686865,
+      "loss": 2.4428,
+      "theoretical_loss": 3.4731275680151814,
+      "tokens_seen": 1727660032
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004846666666666667,
+      "loss": 2.7221,
+      "theoretical_loss": 3.4731162647083975,
+      "tokens_seen": 1727725568
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004846464646464647,
+      "loss": 2.8533,
+      "theoretical_loss": 3.4731049619504084,
+      "tokens_seen": 1727791104
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048462626262626263,
+      "loss": 2.4632,
+      "theoretical_loss": 3.473093659741167,
+      "tokens_seen": 1727856640
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004846060606060606,
+      "loss": 2.6879,
+      "theoretical_loss": 3.4730823580806254,
+      "tokens_seen": 1727922176
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048458585858585855,
+      "loss": 2.7259,
+      "theoretical_loss": 3.473071056968737,
+      "tokens_seen": 1727987712
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004845656565656566,
+      "loss": 2.8179,
+      "theoretical_loss": 3.473059756405453,
+      "tokens_seen": 1728053248
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048454545454545457,
+      "loss": 2.7101,
+      "theoretical_loss": 3.4730484563907273,
+      "tokens_seen": 1728118784
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004845252525252525,
+      "loss": 2.7504,
+      "theoretical_loss": 3.4730371569245113,
+      "tokens_seen": 1728184320
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004845050505050505,
+      "loss": 2.3934,
+      "theoretical_loss": 3.473025858006759,
+      "tokens_seen": 1728249856
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004844848484848485,
+      "loss": 2.4794,
+      "theoretical_loss": 3.4730145596374213,
+      "tokens_seen": 1728315392
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004844646464646465,
+      "loss": 2.4167,
+      "theoretical_loss": 3.4730032618164524,
+      "tokens_seen": 1728380928
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048444444444444446,
+      "loss": 2.7904,
+      "theoretical_loss": 3.4729919645438034,
+      "tokens_seen": 1728446464
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004844242424242424,
+      "loss": 2.5223,
+      "theoretical_loss": 3.472980667819428,
+      "tokens_seen": 1728512000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004844040404040404,
+      "loss": 2.7771,
+      "theoretical_loss": 3.4729693716432783,
+      "tokens_seen": 1728577536
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004843838383838384,
+      "loss": 2.6938,
+      "theoretical_loss": 3.472958076015307,
+      "tokens_seen": 1728643072
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004843636363636364,
+      "loss": 2.6798,
+      "theoretical_loss": 3.472946780935467,
+      "tokens_seen": 1728708608
+    },
+    {
+      "epoch": 0.04,
+      "objective/train/docs_used": 954102,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.302818536758423,
+      "objective/train/theoretical_loss": 3.4729354864037107,
+      "objective/train/tokens_used": 87633376,
+      "theoretical_loss": 3.4729354864037107,
+      "tokens_seen": 1728774144
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048434343434343435,
+      "loss": 2.6855,
+      "theoretical_loss": 3.4729354864037107,
+      "tokens_seen": 1728774144
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004843232323232323,
+      "loss": 2.6738,
+      "theoretical_loss": 3.4729241924199905,
+      "tokens_seen": 1728839680
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004843030303030303,
+      "loss": 2.552,
+      "theoretical_loss": 3.4729128989842595,
+      "tokens_seen": 1728905216
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004842828282828283,
+      "loss": 2.682,
+      "theoretical_loss": 3.4729016060964697,
+      "tokens_seen": 1728970752
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004842626262626263,
+      "loss": 2.4425,
+      "theoretical_loss": 3.4728903137565745,
+      "tokens_seen": 1729036288
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048424242424242425,
+      "loss": 2.7817,
+      "theoretical_loss": 3.472879021964526,
+      "tokens_seen": 1729101824
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004842222222222222,
+      "loss": 2.825,
+      "theoretical_loss": 3.472867730720277,
+      "tokens_seen": 1729167360
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004842020202020202,
+      "loss": 2.5866,
+      "theoretical_loss": 3.4728564400237802,
+      "tokens_seen": 1729232896
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048418181818181817,
+      "loss": 2.9395,
+      "theoretical_loss": 3.4728451498749884,
+      "tokens_seen": 1729298432
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004841616161616162,
+      "loss": 2.7942,
+      "theoretical_loss": 3.4728338602738544,
+      "tokens_seen": 1729363968
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048414141414141414,
+      "loss": 2.5925,
+      "theoretical_loss": 3.4728225712203304,
+      "tokens_seen": 1729429504
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048412121212121215,
+      "loss": 2.5568,
+      "theoretical_loss": 3.4728112827143693,
+      "tokens_seen": 1729495040
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004841010101010101,
+      "loss": 2.7105,
+      "theoretical_loss": 3.472799994755924,
+      "tokens_seen": 1729560576
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048408080808080807,
+      "loss": 2.7117,
+      "theoretical_loss": 3.4727887073449466,
+      "tokens_seen": 1729626112
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004840606060606061,
+      "loss": 2.687,
+      "theoretical_loss": 3.4727774204813904,
+      "tokens_seen": 1729691648
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048404040404040403,
+      "loss": 2.7663,
+      "theoretical_loss": 3.472766134165208,
+      "tokens_seen": 1729757184
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048402020202020204,
+      "loss": 2.5287,
+      "theoretical_loss": 3.472754848396352,
+      "tokens_seen": 1729822720
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000484,
+      "loss": 2.6449,
+      "theoretical_loss": 3.4727435631747756,
+      "tokens_seen": 1729888256
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048397979797979796,
+      "loss": 2.7483,
+      "theoretical_loss": 3.4727322785004304,
+      "tokens_seen": 1729953792
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048395959595959597,
+      "loss": 2.5295,
+      "theoretical_loss": 3.4727209943732706,
+      "tokens_seen": 1730019328
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000483939393939394,
+      "loss": 2.597,
+      "theoretical_loss": 3.4727097107932474,
+      "tokens_seen": 1730084864
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048391919191919194,
+      "loss": 2.5403,
+      "theoretical_loss": 3.4726984277603146,
+      "tokens_seen": 1730150400
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004838989898989899,
+      "loss": 2.7501,
+      "theoretical_loss": 3.472687145274425,
+      "tokens_seen": 1730215936
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048387878787878785,
+      "loss": 2.6343,
+      "theoretical_loss": 3.472675863335531,
+      "tokens_seen": 1730281472
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048385858585858586,
+      "loss": 2.5461,
+      "theoretical_loss": 3.472664581943585,
+      "tokens_seen": 1730347008
+    },
+    {
+      "epoch": 0.04,
+      "objective/train/docs_used": 954813,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.2854347229003906,
+      "objective/train/theoretical_loss": 3.47265330109854,
+      "objective/train/tokens_used": 89271776,
+      "theoretical_loss": 3.47265330109854,
+      "tokens_seen": 1730412544
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048383838383838387,
+      "loss": 2.5293,
+      "theoretical_loss": 3.47265330109854,
+      "tokens_seen": 1730412544
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048381818181818183,
+      "loss": 2.7872,
+      "theoretical_loss": 3.4726420208003494,
+      "tokens_seen": 1730478080
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004837979797979798,
+      "loss": 2.7759,
+      "theoretical_loss": 3.472630741048966,
+      "tokens_seen": 1730543616
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048377777777777774,
+      "loss": 2.4633,
+      "theoretical_loss": 3.4726194618443413,
+      "tokens_seen": 1730609152
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004837575757575758,
+      "loss": 2.6828,
+      "theoretical_loss": 3.4726081831864293,
+      "tokens_seen": 1730674688
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048373737373737377,
+      "loss": 2.6382,
+      "theoretical_loss": 3.472596905075182,
+      "tokens_seen": 1730740224
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004837171717171717,
+      "loss": 2.6293,
+      "theoretical_loss": 3.4725856275105533,
+      "tokens_seen": 1730805760
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004836969696969697,
+      "loss": 2.421,
+      "theoretical_loss": 3.472574350492495,
+      "tokens_seen": 1730871296
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048367676767676764,
+      "loss": 2.5533,
+      "theoretical_loss": 3.47256307402096,
+      "tokens_seen": 1730936832
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004836565656565657,
+      "loss": 2.8264,
+      "theoretical_loss": 3.4725517980959015,
+      "tokens_seen": 1731002368
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048363636363636366,
+      "loss": 2.4918,
+      "theoretical_loss": 3.4725405227172725,
+      "tokens_seen": 1731067904
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004836161616161616,
+      "loss": 2.5235,
+      "theoretical_loss": 3.472529247885025,
+      "tokens_seen": 1731133440
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048359595959595957,
+      "loss": 2.5957,
+      "theoretical_loss": 3.472517973599113,
+      "tokens_seen": 1731198976
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048357575757575764,
+      "loss": 2.8658,
+      "theoretical_loss": 3.4725066998594882,
+      "tokens_seen": 1731264512
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004835555555555556,
+      "loss": 2.5194,
+      "theoretical_loss": 3.4724954266661046,
+      "tokens_seen": 1731330048
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048353535353535355,
+      "loss": 2.8079,
+      "theoretical_loss": 3.4724841540189137,
+      "tokens_seen": 1731395584
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004835151515151515,
+      "loss": 2.7591,
+      "theoretical_loss": 3.472472881917869,
+      "tokens_seen": 1731461120
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048349494949494946,
+      "loss": 2.6462,
+      "theoretical_loss": 3.4724616103629242,
+      "tokens_seen": 1731526656
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048347474747474753,
+      "loss": 2.4711,
+      "theoretical_loss": 3.4724503393540314,
+      "tokens_seen": 1731592192
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004834545454545455,
+      "loss": 2.5492,
+      "theoretical_loss": 3.472439068891143,
+      "tokens_seen": 1731657728
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048343434343434344,
+      "loss": 2.604,
+      "theoretical_loss": 3.4724277989742127,
+      "tokens_seen": 1731723264
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004834141414141414,
+      "loss": 2.6603,
+      "theoretical_loss": 3.4724165296031924,
+      "tokens_seen": 1731788800
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004833939393939394,
+      "loss": 2.5478,
+      "theoretical_loss": 3.4724052607780367,
+      "tokens_seen": 1731854336
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004833737373737374,
+      "loss": 2.4989,
+      "theoretical_loss": 3.472393992498697,
+      "tokens_seen": 1731919872
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004833535353535354,
+      "loss": 2.5639,
+      "theoretical_loss": 3.4723827247651267,
+      "tokens_seen": 1731985408
+    },
+    {
+      "epoch": 0.04,
+      "objective/train/docs_used": 956095,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.3975791931152344,
+      "objective/train/theoretical_loss": 3.472371457577279,
+      "objective/train/tokens_used": 90910176,
+      "theoretical_loss": 3.472371457577279,
+      "tokens_seen": 1732050944
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048333333333333334,
+      "loss": 2.5653,
+      "theoretical_loss": 3.472371457577279,
+      "tokens_seen": 1732050944
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004833131313131313,
+      "loss": 2.6249,
+      "theoretical_loss": 3.472360190935106,
+      "tokens_seen": 1732116480
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004832929292929293,
+      "loss": 2.7115,
+      "theoretical_loss": 3.472348924838562,
+      "tokens_seen": 1732182016
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004832727272727273,
+      "loss": 2.7388,
+      "theoretical_loss": 3.4723376592875983,
+      "tokens_seen": 1732247552
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048325252525252527,
+      "loss": 2.572,
+      "theoretical_loss": 3.4723263942821694,
+      "tokens_seen": 1732313088
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048323232323232323,
+      "loss": 2.7735,
+      "theoretical_loss": 3.472315129822227,
+      "tokens_seen": 1732378624
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048321212121212124,
+      "loss": 2.4005,
+      "theoretical_loss": 3.4723038659077248,
+      "tokens_seen": 1732444160
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004831919191919192,
+      "loss": 2.7573,
+      "theoretical_loss": 3.4722926025386154,
+      "tokens_seen": 1732509696
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004831717171717172,
+      "loss": 2.6348,
+      "theoretical_loss": 3.4722813397148515,
+      "tokens_seen": 1732575232
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048315151515151516,
+      "loss": 2.5441,
+      "theoretical_loss": 3.472270077436387,
+      "tokens_seen": 1732640768
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004831313131313131,
+      "loss": 2.6699,
+      "theoretical_loss": 3.4722588157031744,
+      "tokens_seen": 1732706304
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048311111111111113,
+      "loss": 2.6538,
+      "theoretical_loss": 3.472247554515166,
+      "tokens_seen": 1732771840
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004830909090909091,
+      "loss": 2.4593,
+      "theoretical_loss": 3.4722362938723155,
+      "tokens_seen": 1732837376
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004830707070707071,
+      "loss": 2.5461,
+      "theoretical_loss": 3.4722250337745764,
+      "tokens_seen": 1732902912
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048305050505050506,
+      "loss": 2.7509,
+      "theoretical_loss": 3.472213774221901,
+      "tokens_seen": 1732968448
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048303030303030307,
+      "loss": 2.5207,
+      "theoretical_loss": 3.4722025152142413,
+      "tokens_seen": 1733033984
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000483010101010101,
+      "loss": 2.7144,
+      "theoretical_loss": 3.472191256751552,
+      "tokens_seen": 1733099520
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000482989898989899,
+      "loss": 2.578,
+      "theoretical_loss": 3.472179998833786,
+      "tokens_seen": 1733165056
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000482969696969697,
+      "loss": 2.6586,
+      "theoretical_loss": 3.4721687414608953,
+      "tokens_seen": 1733230592
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048294949494949495,
+      "loss": 2.826,
+      "theoretical_loss": 3.4721574846328336,
+      "tokens_seen": 1733296128
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048292929292929296,
+      "loss": 2.6236,
+      "theoretical_loss": 3.472146228349554,
+      "tokens_seen": 1733361664
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004829090909090909,
+      "loss": 2.5089,
+      "theoretical_loss": 3.472134972611009,
+      "tokens_seen": 1733427200
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004828888888888889,
+      "loss": 2.6058,
+      "theoretical_loss": 3.472123717417152,
+      "tokens_seen": 1733492736
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004828686868686869,
+      "loss": 2.4047,
+      "theoretical_loss": 3.472112462767936,
+      "tokens_seen": 1733558272
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004828484848484849,
+      "loss": 2.6876,
+      "theoretical_loss": 3.4721012086633145,
+      "tokens_seen": 1733623808
+    },
+    {
+      "epoch": 0.04,
+      "objective/train/docs_used": 956643,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.931276798248291,
+      "objective/train/theoretical_loss": 3.4720899551032396,
+      "objective/train/tokens_used": 92548576,
+      "theoretical_loss": 3.4720899551032396,
+      "tokens_seen": 1733689344
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048282828282828285,
+      "loss": 2.5012,
+      "theoretical_loss": 3.4720899551032396,
+      "tokens_seen": 1733689344
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004828080808080808,
+      "loss": 2.5515,
+      "theoretical_loss": 3.472078702087665,
+      "tokens_seen": 1733754880
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048278787878787877,
+      "loss": 2.6655,
+      "theoretical_loss": 3.472067449616544,
+      "tokens_seen": 1733820416
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004827676767676768,
+      "loss": 2.7347,
+      "theoretical_loss": 3.4720561976898288,
+      "tokens_seen": 1733885952
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004827474747474748,
+      "loss": 2.7386,
+      "theoretical_loss": 3.4720449463074736,
+      "tokens_seen": 1733951488
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048272727272727275,
+      "loss": 2.4823,
+      "theoretical_loss": 3.47203369546943,
+      "tokens_seen": 1734017024
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004827070707070707,
+      "loss": 2.6493,
+      "theoretical_loss": 3.472022445175653,
+      "tokens_seen": 1734082560
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048268686868686866,
+      "loss": 2.746,
+      "theoretical_loss": 3.472011195426095,
+      "tokens_seen": 1734148096
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048266666666666667,
+      "loss": 2.4995,
+      "theoretical_loss": 3.4719999462207083,
+      "tokens_seen": 1734213632
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004826464646464647,
+      "loss": 2.6906,
+      "theoretical_loss": 3.4719886975594463,
+      "tokens_seen": 1734279168
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048262626262626264,
+      "loss": 2.3716,
+      "theoretical_loss": 3.471977449442263,
+      "tokens_seen": 1734344704
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004826060606060606,
+      "loss": 2.4103,
+      "theoretical_loss": 3.47196620186911,
+      "tokens_seen": 1734410240
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048258585858585855,
+      "loss": 2.7933,
+      "theoretical_loss": 3.471954954839942,
+      "tokens_seen": 1734475776
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004825656565656566,
+      "loss": 2.6078,
+      "theoretical_loss": 3.4719437083547113,
+      "tokens_seen": 1734541312
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004825454545454546,
+      "loss": 2.5246,
+      "theoretical_loss": 3.4719324624133714,
+      "tokens_seen": 1734606848
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048252525252525253,
+      "loss": 2.755,
+      "theoretical_loss": 3.4719212170158755,
+      "tokens_seen": 1734672384
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004825050505050505,
+      "loss": 2.2628,
+      "theoretical_loss": 3.471909972162176,
+      "tokens_seen": 1734737920
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048248484848484845,
+      "loss": 2.6222,
+      "theoretical_loss": 3.471898727852227,
+      "tokens_seen": 1734803456
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004824646464646465,
+      "loss": 2.6469,
+      "theoretical_loss": 3.471887484085981,
+      "tokens_seen": 1734868992
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048244444444444447,
+      "loss": 2.6871,
+      "theoretical_loss": 3.4718762408633914,
+      "tokens_seen": 1734934528
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004824242424242424,
+      "loss": 2.5698,
+      "theoretical_loss": 3.471864998184411,
+      "tokens_seen": 1735000064
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004824040404040404,
+      "loss": 2.8557,
+      "theoretical_loss": 3.471853756048994,
+      "tokens_seen": 1735065600
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004823838383838384,
+      "loss": 2.6444,
+      "theoretical_loss": 3.4718425144570926,
+      "tokens_seen": 1735131136
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004823636363636364,
+      "loss": 2.6187,
+      "theoretical_loss": 3.471831273408661,
+      "tokens_seen": 1735196672
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00048234343434343436,
+      "loss": 2.5576,
+      "theoretical_loss": 3.471820032903651,
+      "tokens_seen": 1735262208
+    },
+    {
+      "epoch": 0.04,
+      "objective/train/docs_used": 957795,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.0068306922912598,
+      "objective/train/theoretical_loss": 3.471808792942017,
+      "objective/train/tokens_used": 94186976,
+      "theoretical_loss": 3.471808792942017,
+      "tokens_seen": 1735327744
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0004823232323232323,
+      "loss": 2.839,
+      "theoretical_loss": 3.471808792942017,
+      "tokens_seen": 1735327744
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004823030303030303,
+      "loss": 2.612,
+      "theoretical_loss": 3.4717975535237118,
+      "tokens_seen": 1735393280
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004822828282828283,
+      "loss": 2.5873,
+      "theoretical_loss": 3.4717863146486887,
+      "tokens_seen": 1735458816
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004822626262626263,
+      "loss": 2.7101,
+      "theoretical_loss": 3.4717750763169004,
+      "tokens_seen": 1735524352
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048224242424242425,
+      "loss": 2.5618,
+      "theoretical_loss": 3.471763838528301,
+      "tokens_seen": 1735589888
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004822222222222222,
+      "loss": 2.7326,
+      "theoretical_loss": 3.471752601282843,
+      "tokens_seen": 1735655424
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004822020202020202,
+      "loss": 2.3858,
+      "theoretical_loss": 3.47174136458048,
+      "tokens_seen": 1735720960
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004821818181818182,
+      "loss": 2.6821,
+      "theoretical_loss": 3.4717301284211657,
+      "tokens_seen": 1735786496
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004821616161616162,
+      "loss": 2.6055,
+      "theoretical_loss": 3.4717188928048524,
+      "tokens_seen": 1735852032
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048214141414141415,
+      "loss": 2.8098,
+      "theoretical_loss": 3.4717076577314936,
+      "tokens_seen": 1735917568
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004821212121212121,
+      "loss": 2.555,
+      "theoretical_loss": 3.471696423201043,
+      "tokens_seen": 1735983104
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004821010101010101,
+      "loss": 2.7927,
+      "theoretical_loss": 3.4716851892134537,
+      "tokens_seen": 1736048640
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048208080808080807,
+      "loss": 2.7408,
+      "theoretical_loss": 3.4716739557686793,
+      "tokens_seen": 1736114176
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004820606060606061,
+      "loss": 2.9396,
+      "theoretical_loss": 3.471662722866672,
+      "tokens_seen": 1736179712
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048204040404040404,
+      "loss": 2.5873,
+      "theoretical_loss": 3.4716514905073863,
+      "tokens_seen": 1736245248
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048202020202020205,
+      "loss": 2.5494,
+      "theoretical_loss": 3.471640258690775,
+      "tokens_seen": 1736310784
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000482,
+      "loss": 2.6813,
+      "theoretical_loss": 3.4716290274167907,
+      "tokens_seen": 1736376320
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048197979797979796,
+      "loss": 2.6974,
+      "theoretical_loss": 3.471617796685388,
+      "tokens_seen": 1736441856
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000481959595959596,
+      "loss": 2.6864,
+      "theoretical_loss": 3.4716065664965194,
+      "tokens_seen": 1736507392
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048193939393939393,
+      "loss": 2.7289,
+      "theoretical_loss": 3.4715953368501387,
+      "tokens_seen": 1736572928
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048191919191919194,
+      "loss": 2.7123,
+      "theoretical_loss": 3.4715841077461986,
+      "tokens_seen": 1736638464
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004818989898989899,
+      "loss": 2.6387,
+      "theoretical_loss": 3.471572879184653,
+      "tokens_seen": 1736704000
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004818787878787879,
+      "loss": 2.763,
+      "theoretical_loss": 3.4715616511654552,
+      "tokens_seen": 1736769536
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048185858585858587,
+      "loss": 2.7991,
+      "theoretical_loss": 3.4715504236885577,
+      "tokens_seen": 1736835072
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004818383838383839,
+      "loss": 2.7678,
+      "theoretical_loss": 3.4715391967539153,
+      "tokens_seen": 1736900608
+    },
+    {
+      "epoch": 0.05,
+      "objective/train/docs_used": 958846,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5784518718719482,
+      "objective/train/theoretical_loss": 3.47152797036148,
+      "objective/train/tokens_used": 95825376,
+      "theoretical_loss": 3.47152797036148,
+      "tokens_seen": 1736966144
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048181818181818184,
+      "loss": 2.7097,
+      "theoretical_loss": 3.47152797036148,
+      "tokens_seen": 1736966144
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004817979797979798,
+      "loss": 2.537,
+      "theoretical_loss": 3.471516744511206,
+      "tokens_seen": 1737031680
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004817777777777778,
+      "loss": 2.6579,
+      "theoretical_loss": 3.471505519203046,
+      "tokens_seen": 1737097216
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048175757575757576,
+      "loss": 2.7483,
+      "theoretical_loss": 3.471494294436954,
+      "tokens_seen": 1737162752
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048173737373737377,
+      "loss": 2.5049,
+      "theoretical_loss": 3.471483070212883,
+      "tokens_seen": 1737228288
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048171717171717173,
+      "loss": 2.7261,
+      "theoretical_loss": 3.471471846530787,
+      "tokens_seen": 1737293824
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004816969696969697,
+      "loss": 2.6989,
+      "theoretical_loss": 3.471460623390618,
+      "tokens_seen": 1737359360
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004816767676767677,
+      "loss": 2.6896,
+      "theoretical_loss": 3.4714494007923307,
+      "tokens_seen": 1737424896
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004816565656565657,
+      "loss": 2.8564,
+      "theoretical_loss": 3.471438178735878,
+      "tokens_seen": 1737490432
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048163636363636366,
+      "loss": 2.6221,
+      "theoretical_loss": 3.4714269572212135,
+      "tokens_seen": 1737555968
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004816161616161616,
+      "loss": 2.733,
+      "theoretical_loss": 3.4714157362482907,
+      "tokens_seen": 1737621504
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004815959595959596,
+      "loss": 2.8006,
+      "theoretical_loss": 3.4714045158170626,
+      "tokens_seen": 1737687040
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004815757575757576,
+      "loss": 2.64,
+      "theoretical_loss": 3.4713932959274825,
+      "tokens_seen": 1737752576
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004815555555555556,
+      "loss": 2.7127,
+      "theoretical_loss": 3.4713820765795047,
+      "tokens_seen": 1737818112
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048153535353535356,
+      "loss": 2.929,
+      "theoretical_loss": 3.4713708577730817,
+      "tokens_seen": 1737883648
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004815151515151515,
+      "loss": 2.7397,
+      "theoretical_loss": 3.471359639508168,
+      "tokens_seen": 1737949184
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048149494949494947,
+      "loss": 2.7401,
+      "theoretical_loss": 3.4713484217847155,
+      "tokens_seen": 1738014720
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048147474747474754,
+      "loss": 2.6971,
+      "theoretical_loss": 3.471337204602679,
+      "tokens_seen": 1738080256
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004814545454545455,
+      "loss": 2.7005,
+      "theoretical_loss": 3.471325987962011,
+      "tokens_seen": 1738145792
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048143434343434345,
+      "loss": 2.5562,
+      "theoretical_loss": 3.471314771862666,
+      "tokens_seen": 1738211328
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004814141414141414,
+      "loss": 2.667,
+      "theoretical_loss": 3.4713035563045964,
+      "tokens_seen": 1738276864
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048139393939393936,
+      "loss": 2.7304,
+      "theoretical_loss": 3.4712923412877563,
+      "tokens_seen": 1738342400
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048137373737373743,
+      "loss": 2.521,
+      "theoretical_loss": 3.4712811268120993,
+      "tokens_seen": 1738407936
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004813535353535354,
+      "loss": 2.5026,
+      "theoretical_loss": 3.4712699128775784,
+      "tokens_seen": 1738473472
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048133333333333334,
+      "loss": 2.799,
+      "theoretical_loss": 3.4712586994841477,
+      "tokens_seen": 1738539008
+    },
+    {
+      "epoch": 0.05,
+      "objective/train/docs_used": 959288,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.807804584503174,
+      "objective/train/theoretical_loss": 3.4712474866317597,
+      "objective/train/tokens_used": 97463776,
+      "theoretical_loss": 3.4712474866317597,
+      "tokens_seen": 1738604544
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004813131313131313,
+      "loss": 2.7561,
+      "theoretical_loss": 3.4712474866317597,
+      "tokens_seen": 1738604544
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048129292929292926,
+      "loss": 2.7609,
+      "theoretical_loss": 3.4712362743203693,
+      "tokens_seen": 1738670080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004812727272727273,
+      "loss": 2.797,
+      "theoretical_loss": 3.4712250625499284,
+      "tokens_seen": 1738735616
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004812525252525253,
+      "loss": 3.041,
+      "theoretical_loss": 3.471213851320392,
+      "tokens_seen": 1738801152
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048123232323232323,
+      "loss": 2.8472,
+      "theoretical_loss": 3.471202640631713,
+      "tokens_seen": 1738866688
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004812121212121212,
+      "loss": 2.5309,
+      "theoretical_loss": 3.4711914304838443,
+      "tokens_seen": 1738932224
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004811919191919192,
+      "loss": 2.6006,
+      "theoretical_loss": 3.4711802208767404,
+      "tokens_seen": 1738997760
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004811717171717172,
+      "loss": 2.6453,
+      "theoretical_loss": 3.471169011810354,
+      "tokens_seen": 1739063296
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048115151515151517,
+      "loss": 2.7669,
+      "theoretical_loss": 3.4711578032846404,
+      "tokens_seen": 1739128832
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048113131313131313,
+      "loss": 2.3778,
+      "theoretical_loss": 3.4711465952995506,
+      "tokens_seen": 1739194368
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004811111111111111,
+      "loss": 2.7263,
+      "theoretical_loss": 3.47113538785504,
+      "tokens_seen": 1739259904
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004810909090909091,
+      "loss": 2.5269,
+      "theoretical_loss": 3.471124180951061,
+      "tokens_seen": 1739325440
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004810707070707071,
+      "loss": 2.8646,
+      "theoretical_loss": 3.471112974587568,
+      "tokens_seen": 1739390976
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048105050505050506,
+      "loss": 2.6309,
+      "theoretical_loss": 3.4711017687645147,
+      "tokens_seen": 1739456512
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000481030303030303,
+      "loss": 2.4525,
+      "theoretical_loss": 3.471090563481854,
+      "tokens_seen": 1739522048
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048101010101010103,
+      "loss": 2.6909,
+      "theoretical_loss": 3.4710793587395394,
+      "tokens_seen": 1739587584
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000480989898989899,
+      "loss": 2.8344,
+      "theoretical_loss": 3.4710681545375253,
+      "tokens_seen": 1739653120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000480969696969697,
+      "loss": 2.8379,
+      "theoretical_loss": 3.4710569508757647,
+      "tokens_seen": 1739718656
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048094949494949496,
+      "loss": 2.708,
+      "theoretical_loss": 3.471045747754211,
+      "tokens_seen": 1739784192
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004809292929292929,
+      "loss": 2.6413,
+      "theoretical_loss": 3.4710345451728184,
+      "tokens_seen": 1739849728
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004809090909090909,
+      "loss": 2.6477,
+      "theoretical_loss": 3.4710233431315407,
+      "tokens_seen": 1739915264
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004808888888888889,
+      "loss": 2.4457,
+      "theoretical_loss": 3.4710121416303306,
+      "tokens_seen": 1739980800
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004808686868686869,
+      "loss": 2.6397,
+      "theoretical_loss": 3.471000940669142,
+      "tokens_seen": 1740046336
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048084848484848485,
+      "loss": 2.6021,
+      "theoretical_loss": 3.470989740247929,
+      "tokens_seen": 1740111872
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048082828282828286,
+      "loss": 2.6974,
+      "theoretical_loss": 3.4709785403666453,
+      "tokens_seen": 1740177408
+    },
+    {
+      "epoch": 0.05,
+      "objective/train/docs_used": 960545,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.9208710193634033,
+      "objective/train/theoretical_loss": 3.4709673410252435,
+      "objective/train/tokens_used": 99102176,
+      "theoretical_loss": 3.4709673410252435,
+      "tokens_seen": 1740242944
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004808080808080808,
+      "loss": 2.7373,
+      "theoretical_loss": 3.4709673410252435,
+      "tokens_seen": 1740242944
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004807878787878788,
+      "loss": 2.8115,
+      "theoretical_loss": 3.4709561422236783,
+      "tokens_seen": 1740308480
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004807676767676768,
+      "loss": 2.7494,
+      "theoretical_loss": 3.470944943961903,
+      "tokens_seen": 1740374016
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048074747474747474,
+      "loss": 2.6765,
+      "theoretical_loss": 3.4709337462398713,
+      "tokens_seen": 1740439552
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048072727272727275,
+      "loss": 2.9857,
+      "theoretical_loss": 3.470922549057537,
+      "tokens_seen": 1740505088
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004807070707070707,
+      "loss": 2.443,
+      "theoretical_loss": 3.470911352414853,
+      "tokens_seen": 1740570624
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048068686868686867,
+      "loss": 2.6088,
+      "theoretical_loss": 3.4709001563117736,
+      "tokens_seen": 1740636160
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004806666666666667,
+      "loss": 2.8324,
+      "theoretical_loss": 3.470888960748253,
+      "tokens_seen": 1740701696
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004806464646464647,
+      "loss": 2.7392,
+      "theoretical_loss": 3.470877765724244,
+      "tokens_seen": 1740767232
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048062626262626264,
+      "loss": 2.5289,
+      "theoretical_loss": 3.4708665712397004,
+      "tokens_seen": 1740832768
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004806060606060606,
+      "loss": 2.7885,
+      "theoretical_loss": 3.4708553772945763,
+      "tokens_seen": 1740898304
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048058585858585856,
+      "loss": 2.5816,
+      "theoretical_loss": 3.4708441838888255,
+      "tokens_seen": 1740963840
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048056565656565657,
+      "loss": 2.7401,
+      "theoretical_loss": 3.4708329910224007,
+      "tokens_seen": 1741029376
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004805454545454546,
+      "loss": 2.6518,
+      "theoretical_loss": 3.4708217986952574,
+      "tokens_seen": 1741094912
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048052525252525254,
+      "loss": 2.6621,
+      "theoretical_loss": 3.470810606907347,
+      "tokens_seen": 1741160448
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004805050505050505,
+      "loss": 2.468,
+      "theoretical_loss": 3.4707994156586253,
+      "tokens_seen": 1741225984
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048048484848484845,
+      "loss": 2.56,
+      "theoretical_loss": 3.470788224949045,
+      "tokens_seen": 1741291520
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004804646464646465,
+      "loss": 2.6133,
+      "theoretical_loss": 3.47077703477856,
+      "tokens_seen": 1741357056
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004804444444444445,
+      "loss": 2.4266,
+      "theoretical_loss": 3.470765845147124,
+      "tokens_seen": 1741422592
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048042424242424243,
+      "loss": 2.5242,
+      "theoretical_loss": 3.470754656054691,
+      "tokens_seen": 1741488128
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004804040404040404,
+      "loss": 2.733,
+      "theoretical_loss": 3.4707434675012148,
+      "tokens_seen": 1741553664
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048038383838383834,
+      "loss": 2.8107,
+      "theoretical_loss": 3.470732279486649,
+      "tokens_seen": 1741619200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004803636363636364,
+      "loss": 2.6628,
+      "theoretical_loss": 3.4707210920109466,
+      "tokens_seen": 1741684736
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048034343434343437,
+      "loss": 2.7073,
+      "theoretical_loss": 3.470709905074062,
+      "tokens_seen": 1741750272
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004803232323232323,
+      "loss": 2.6392,
+      "theoretical_loss": 3.47069871867595,
+      "tokens_seen": 1741815808
+    },
+    {
+      "epoch": 0.05,
+      "objective/train/docs_used": 961825,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.8929381370544434,
+      "objective/train/theoretical_loss": 3.470687532816563,
+      "objective/train/tokens_used": 100740576,
+      "theoretical_loss": 3.470687532816563,
+      "tokens_seen": 1741881344
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004803030303030303,
+      "loss": 2.7658,
+      "theoretical_loss": 3.470687532816563,
+      "tokens_seen": 1741881344
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004802828282828283,
+      "loss": 2.7095,
+      "theoretical_loss": 3.4706763474958553,
+      "tokens_seen": 1741946880
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004802626262626263,
+      "loss": 2.6271,
+      "theoretical_loss": 3.4706651627137806,
+      "tokens_seen": 1742012416
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048024242424242426,
+      "loss": 2.6202,
+      "theoretical_loss": 3.4706539784702928,
+      "tokens_seen": 1742077952
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004802222222222222,
+      "loss": 2.6904,
+      "theoretical_loss": 3.470642794765345,
+      "tokens_seen": 1742143488
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048020202020202017,
+      "loss": 2.2801,
+      "theoretical_loss": 3.4706316115988924,
+      "tokens_seen": 1742209024
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004801818181818182,
+      "loss": 2.7515,
+      "theoretical_loss": 3.4706204289708875,
+      "tokens_seen": 1742274560
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004801616161616162,
+      "loss": 2.8955,
+      "theoretical_loss": 3.4706092468812852,
+      "tokens_seen": 1742340096
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048014141414141415,
+      "loss": 2.5899,
+      "theoretical_loss": 3.4705980653300386,
+      "tokens_seen": 1742405632
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004801212121212121,
+      "loss": 2.5155,
+      "theoretical_loss": 3.4705868843171017,
+      "tokens_seen": 1742471168
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004801010101010101,
+      "loss": 2.469,
+      "theoretical_loss": 3.4705757038424285,
+      "tokens_seen": 1742536704
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048008080808080813,
+      "loss": 2.6819,
+      "theoretical_loss": 3.4705645239059724,
+      "tokens_seen": 1742602240
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004800606060606061,
+      "loss": 2.9573,
+      "theoretical_loss": 3.470553344507688,
+      "tokens_seen": 1742667776
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048004040404040404,
+      "loss": 2.6073,
+      "theoretical_loss": 3.4705421656475286,
+      "tokens_seen": 1742733312
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000480020202020202,
+      "loss": 2.5603,
+      "theoretical_loss": 3.470530987325448,
+      "tokens_seen": 1742798848
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00048,
+      "loss": 2.5829,
+      "theoretical_loss": 3.4705198095414005,
+      "tokens_seen": 1742864384
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000479979797979798,
+      "loss": 2.6065,
+      "theoretical_loss": 3.4705086322953393,
+      "tokens_seen": 1742929920
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000479959595959596,
+      "loss": 2.7619,
+      "theoretical_loss": 3.4704974555872194,
+      "tokens_seen": 1742995456
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047993939393939394,
+      "loss": 2.5569,
+      "theoretical_loss": 3.4704862794169937,
+      "tokens_seen": 1743060992
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047991919191919195,
+      "loss": 2.7131,
+      "theoretical_loss": 3.470475103784616,
+      "tokens_seen": 1743126528
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004798989898989899,
+      "loss": 2.5933,
+      "theoretical_loss": 3.470463928690041,
+      "tokens_seen": 1743192064
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004798787878787879,
+      "loss": 2.505,
+      "theoretical_loss": 3.4704527541332224,
+      "tokens_seen": 1743257600
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047985858585858587,
+      "loss": 2.608,
+      "theoretical_loss": 3.4704415801141133,
+      "tokens_seen": 1743323136
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047983838383838383,
+      "loss": 2.523,
+      "theoretical_loss": 3.4704304066326683,
+      "tokens_seen": 1743388672
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047981818181818184,
+      "loss": 2.5742,
+      "theoretical_loss": 3.4704192336888413,
+      "tokens_seen": 1743454208
+    },
+    {
+      "epoch": 0.05,
+      "objective/train/docs_used": 962337,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5374813079833984,
+      "objective/train/theoretical_loss": 3.470408061282587,
+      "objective/train/tokens_used": 102378976,
+      "theoretical_loss": 3.470408061282587,
+      "tokens_seen": 1743519744
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004797979797979798,
+      "loss": 2.3709,
+      "theoretical_loss": 3.470408061282587,
+      "tokens_seen": 1743519744
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004797777777777778,
+      "loss": 2.8128,
+      "theoretical_loss": 3.4703968894138573,
+      "tokens_seen": 1743585280
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047975757575757577,
+      "loss": 2.4218,
+      "theoretical_loss": 3.470385718082608,
+      "tokens_seen": 1743650816
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004797373737373738,
+      "loss": 2.7803,
+      "theoretical_loss": 3.470374547288792,
+      "tokens_seen": 1743716352
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047971717171717173,
+      "loss": 2.8138,
+      "theoretical_loss": 3.470363377032364,
+      "tokens_seen": 1743781888
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004796969696969697,
+      "loss": 2.3758,
+      "theoretical_loss": 3.4703522073132773,
+      "tokens_seen": 1743847424
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004796767676767677,
+      "loss": 2.7284,
+      "theoretical_loss": 3.470341038131486,
+      "tokens_seen": 1743912960
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047965656565656566,
+      "loss": 2.6017,
+      "theoretical_loss": 3.470329869486944,
+      "tokens_seen": 1743978496
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047963636363636367,
+      "loss": 3.0104,
+      "theoretical_loss": 3.470318701379606,
+      "tokens_seen": 1744044032
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004796161616161616,
+      "loss": 2.6543,
+      "theoretical_loss": 3.4703075338094256,
+      "tokens_seen": 1744109568
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004795959595959596,
+      "loss": 2.7775,
+      "theoretical_loss": 3.4702963667763562,
+      "tokens_seen": 1744175104
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004795757575757576,
+      "loss": 2.5585,
+      "theoretical_loss": 3.470285200280352,
+      "tokens_seen": 1744240640
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004795555555555556,
+      "loss": 2.6517,
+      "theoretical_loss": 3.470274034321368,
+      "tokens_seen": 1744306176
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047953535353535356,
+      "loss": 2.394,
+      "theoretical_loss": 3.470262868899357,
+      "tokens_seen": 1744371712
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004795151515151515,
+      "loss": 2.7431,
+      "theoretical_loss": 3.470251704014273,
+      "tokens_seen": 1744437248
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004794949494949495,
+      "loss": 2.6337,
+      "theoretical_loss": 3.4702405396660705,
+      "tokens_seen": 1744502784
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004794747474747475,
+      "loss": 2.5813,
+      "theoretical_loss": 3.4702293758547036,
+      "tokens_seen": 1744568320
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004794545454545455,
+      "loss": 2.7258,
+      "theoretical_loss": 3.470218212580126,
+      "tokens_seen": 1744633856
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047943434343434345,
+      "loss": 2.6717,
+      "theoretical_loss": 3.4702070498422923,
+      "tokens_seen": 1744699392
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004794141414141414,
+      "loss": 2.3807,
+      "theoretical_loss": 3.4701958876411556,
+      "tokens_seen": 1744764928
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047939393939393937,
+      "loss": 2.7787,
+      "theoretical_loss": 3.4701847259766705,
+      "tokens_seen": 1744830464
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004793737373737374,
+      "loss": 2.4776,
+      "theoretical_loss": 3.4701735648487912,
+      "tokens_seen": 1744896000
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004793535353535354,
+      "loss": 2.5712,
+      "theoretical_loss": 3.4701624042574717,
+      "tokens_seen": 1744961536
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047933333333333335,
+      "loss": 2.6089,
+      "theoretical_loss": 3.4701512442026656,
+      "tokens_seen": 1745027072
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004793131313131313,
+      "loss": 2.6412,
+      "theoretical_loss": 3.470140084684327,
+      "tokens_seen": 1745092608
+    },
+    {
+      "epoch": 0.05,
+      "objective/train/docs_used": 963376,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.749994993209839,
+      "objective/train/theoretical_loss": 3.4701289257024106,
+      "objective/train/tokens_used": 104017376,
+      "theoretical_loss": 3.4701289257024106,
+      "tokens_seen": 1745158144
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047929292929292926,
+      "loss": 2.6664,
+      "theoretical_loss": 3.4701289257024106,
+      "tokens_seen": 1745158144
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004792727272727273,
+      "loss": 2.6718,
+      "theoretical_loss": 3.47011776725687,
+      "tokens_seen": 1745223680
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004792525252525253,
+      "loss": 2.7504,
+      "theoretical_loss": 3.47010660934766,
+      "tokens_seen": 1745289216
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047923232323232324,
+      "loss": 2.894,
+      "theoretical_loss": 3.470095451974733,
+      "tokens_seen": 1745354752
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004792121212121212,
+      "loss": 2.6854,
+      "theoretical_loss": 3.4700842951380446,
+      "tokens_seen": 1745420288
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047919191919191915,
+      "loss": 2.602,
+      "theoretical_loss": 3.4700731388375483,
+      "tokens_seen": 1745485824
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004791717171717172,
+      "loss": 2.5376,
+      "theoretical_loss": 3.4700619830731982,
+      "tokens_seen": 1745551360
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004791515151515152,
+      "loss": 2.6504,
+      "theoretical_loss": 3.470050827844949,
+      "tokens_seen": 1745616896
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047913131313131313,
+      "loss": 2.6948,
+      "theoretical_loss": 3.4700396731527543,
+      "tokens_seen": 1745682432
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004791111111111111,
+      "loss": 2.7571,
+      "theoretical_loss": 3.470028518996568,
+      "tokens_seen": 1745747968
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004790909090909091,
+      "loss": 2.4343,
+      "theoretical_loss": 3.4700173653763446,
+      "tokens_seen": 1745813504
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004790707070707071,
+      "loss": 2.7223,
+      "theoretical_loss": 3.470006212292038,
+      "tokens_seen": 1745879040
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047905050505050507,
+      "loss": 2.7366,
+      "theoretical_loss": 3.469995059743603,
+      "tokens_seen": 1745944576
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000479030303030303,
+      "loss": 2.7484,
+      "theoretical_loss": 3.4699839077309926,
+      "tokens_seen": 1746010112
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000479010101010101,
+      "loss": 2.5738,
+      "theoretical_loss": 3.469972756254162,
+      "tokens_seen": 1746075648
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000478989898989899,
+      "loss": 2.7647,
+      "theoretical_loss": 3.469961605313064,
+      "tokens_seen": 1746141184
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000478969696969697,
+      "loss": 2.8169,
+      "theoretical_loss": 3.4699504549076545,
+      "tokens_seen": 1746206720
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047894949494949496,
+      "loss": 2.7003,
+      "theoretical_loss": 3.4699393050378866,
+      "tokens_seen": 1746272256
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004789292929292929,
+      "loss": 2.7409,
+      "theoretical_loss": 3.4699281557037147,
+      "tokens_seen": 1746337792
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047890909090909093,
+      "loss": 2.7155,
+      "theoretical_loss": 3.469917006905093,
+      "tokens_seen": 1746403328
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004788888888888889,
+      "loss": 2.8076,
+      "theoretical_loss": 3.4699058586419755,
+      "tokens_seen": 1746468864
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004788686868686869,
+      "loss": 2.7611,
+      "theoretical_loss": 3.469894710914317,
+      "tokens_seen": 1746534400
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047884848484848485,
+      "loss": 2.6351,
+      "theoretical_loss": 3.46988356372207,
+      "tokens_seen": 1746599936
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004788282828282828,
+      "loss": 2.8063,
+      "theoretical_loss": 3.469872417065191,
+      "tokens_seen": 1746665472
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004788080808080808,
+      "loss": 2.7563,
+      "theoretical_loss": 3.4698612709436327,
+      "tokens_seen": 1746731008
+    },
+    {
+      "epoch": 0.05,
+      "objective/train/docs_used": 963934,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.45017409324646,
+      "objective/train/theoretical_loss": 3.46985012535735,
+      "objective/train/tokens_used": 105655776,
+      "theoretical_loss": 3.46985012535735,
+      "tokens_seen": 1746796544
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004787878787878788,
+      "loss": 2.6769,
+      "theoretical_loss": 3.46985012535735,
+      "tokens_seen": 1746796544
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004787676767676768,
+      "loss": 2.6498,
+      "theoretical_loss": 3.469838980306297,
+      "tokens_seen": 1746862080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047874747474747475,
+      "loss": 2.7676,
+      "theoretical_loss": 3.469827835790427,
+      "tokens_seen": 1746927616
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047872727272727276,
+      "loss": 2.7206,
+      "theoretical_loss": 3.4698166918096955,
+      "tokens_seen": 1746993152
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004787070707070707,
+      "loss": 2.6543,
+      "theoretical_loss": 3.469805548364056,
+      "tokens_seen": 1747058688
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047868686868686867,
+      "loss": 2.7194,
+      "theoretical_loss": 3.469794405453463,
+      "tokens_seen": 1747124224
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004786666666666667,
+      "loss": 2.7218,
+      "theoretical_loss": 3.4697832630778707,
+      "tokens_seen": 1747189760
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047864646464646464,
+      "loss": 2.6617,
+      "theoretical_loss": 3.4697721212372334,
+      "tokens_seen": 1747255296
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047862626262626265,
+      "loss": 2.4966,
+      "theoretical_loss": 3.4697609799315052,
+      "tokens_seen": 1747320832
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004786060606060606,
+      "loss": 2.5697,
+      "theoretical_loss": 3.4697498391606403,
+      "tokens_seen": 1747386368
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047858585858585856,
+      "loss": 2.792,
+      "theoretical_loss": 3.4697386989245933,
+      "tokens_seen": 1747451904
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004785656565656566,
+      "loss": 2.5246,
+      "theoretical_loss": 3.469727559223318,
+      "tokens_seen": 1747517440
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004785454545454546,
+      "loss": 2.7315,
+      "theoretical_loss": 3.469716420056769,
+      "tokens_seen": 1747582976
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047852525252525254,
+      "loss": 2.7625,
+      "theoretical_loss": 3.469705281424901,
+      "tokens_seen": 1747648512
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004785050505050505,
+      "loss": 2.4715,
+      "theoretical_loss": 3.4696941433276676,
+      "tokens_seen": 1747714048
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047848484848484846,
+      "loss": 2.7081,
+      "theoretical_loss": 3.469683005765023,
+      "tokens_seen": 1747779584
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047846464646464647,
+      "loss": 2.5652,
+      "theoretical_loss": 3.4696718687369223,
+      "tokens_seen": 1747845120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004784444444444445,
+      "loss": 2.6408,
+      "theoretical_loss": 3.469660732243319,
+      "tokens_seen": 1747910656
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047842424242424244,
+      "loss": 2.6097,
+      "theoretical_loss": 3.4696495962841674,
+      "tokens_seen": 1747976192
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004784040404040404,
+      "loss": 2.6966,
+      "theoretical_loss": 3.469638460859423,
+      "tokens_seen": 1748041728
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004783838383838384,
+      "loss": 2.6111,
+      "theoretical_loss": 3.4696273259690384,
+      "tokens_seen": 1748107264
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004783636363636364,
+      "loss": 2.7655,
+      "theoretical_loss": 3.4696161916129693,
+      "tokens_seen": 1748172800
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047834343434343437,
+      "loss": 2.5875,
+      "theoretical_loss": 3.4696050577911692,
+      "tokens_seen": 1748238336
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047832323232323233,
+      "loss": 2.698,
+      "theoretical_loss": 3.469593924503593,
+      "tokens_seen": 1748303872
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004783030303030303,
+      "loss": 2.7526,
+      "theoretical_loss": 3.4695827917501947,
+      "tokens_seen": 1748369408
+    },
+    {
+      "epoch": 0.05,
+      "objective/train/docs_used": 965147,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.617971897125244,
+      "objective/train/theoretical_loss": 3.4695716595309287,
+      "objective/train/tokens_used": 107294176,
+      "theoretical_loss": 3.4695716595309287,
+      "tokens_seen": 1748434944
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004782828282828283,
+      "loss": 2.5613,
+      "theoretical_loss": 3.4695716595309287,
+      "tokens_seen": 1748434944
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004782626262626263,
+      "loss": 2.9621,
+      "theoretical_loss": 3.4695605278457493,
+      "tokens_seen": 1748500480
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047824242424242426,
+      "loss": 2.642,
+      "theoretical_loss": 3.469549396694611,
+      "tokens_seen": 1748566016
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004782222222222222,
+      "loss": 2.7141,
+      "theoretical_loss": 3.4695382660774685,
+      "tokens_seen": 1748631552
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004782020202020202,
+      "loss": 2.5499,
+      "theoretical_loss": 3.469527135994275,
+      "tokens_seen": 1748697088
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047818181818181824,
+      "loss": 2.4784,
+      "theoretical_loss": 3.4695160064449864,
+      "tokens_seen": 1748762624
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004781616161616162,
+      "loss": 2.7111,
+      "theoretical_loss": 3.469504877429556,
+      "tokens_seen": 1748828160
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047814141414141416,
+      "loss": 2.5869,
+      "theoretical_loss": 3.4694937489479383,
+      "tokens_seen": 1748893696
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004781212121212121,
+      "loss": 2.4923,
+      "theoretical_loss": 3.4694826210000884,
+      "tokens_seen": 1748959232
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047810101010101007,
+      "loss": 2.6399,
+      "theoretical_loss": 3.46947149358596,
+      "tokens_seen": 1749024768
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047808080808080814,
+      "loss": 2.807,
+      "theoretical_loss": 3.4694603667055075,
+      "tokens_seen": 1749090304
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004780606060606061,
+      "loss": 2.3863,
+      "theoretical_loss": 3.469449240358686,
+      "tokens_seen": 1749155840
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047804040404040405,
+      "loss": 2.8256,
+      "theoretical_loss": 3.4694381145454494,
+      "tokens_seen": 1749221376
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000478020202020202,
+      "loss": 2.7606,
+      "theoretical_loss": 3.469426989265752,
+      "tokens_seen": 1749286912
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047799999999999996,
+      "loss": 2.6758,
+      "theoretical_loss": 3.469415864519548,
+      "tokens_seen": 1749352448
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047797979797979803,
+      "loss": 2.6428,
+      "theoretical_loss": 3.469404740306793,
+      "tokens_seen": 1749417984
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000477959595959596,
+      "loss": 2.7873,
+      "theoretical_loss": 3.469393616627441,
+      "tokens_seen": 1749483520
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047793939393939394,
+      "loss": 2.7165,
+      "theoretical_loss": 3.469382493481445,
+      "tokens_seen": 1749549056
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004779191919191919,
+      "loss": 2.7791,
+      "theoretical_loss": 3.4693713708687612,
+      "tokens_seen": 1749614592
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004778989898989899,
+      "loss": 2.8045,
+      "theoretical_loss": 3.469360248789343,
+      "tokens_seen": 1749680128
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004778787878787879,
+      "loss": 2.5269,
+      "theoretical_loss": 3.4693491272431456,
+      "tokens_seen": 1749745664
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004778585858585859,
+      "loss": 2.742,
+      "theoretical_loss": 3.469338006230123,
+      "tokens_seen": 1749811200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047783838383838384,
+      "loss": 2.7999,
+      "theoretical_loss": 3.46932688575023,
+      "tokens_seen": 1749876736
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004778181818181818,
+      "loss": 2.782,
+      "theoretical_loss": 3.4693157658034206,
+      "tokens_seen": 1749942272
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004777979797979798,
+      "loss": 2.8005,
+      "theoretical_loss": 3.46930464638965,
+      "tokens_seen": 1750007808
+    },
+    {
+      "epoch": 0.05,
+      "objective/train/docs_used": 966110,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6748225688934326,
+      "objective/train/theoretical_loss": 3.469293527508872,
+      "objective/train/tokens_used": 108932576,
+      "theoretical_loss": 3.469293527508872,
+      "tokens_seen": 1750073344
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004777777777777778,
+      "loss": 2.6659,
+      "theoretical_loss": 3.469293527508872,
+      "tokens_seen": 1750073344
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047775757575757577,
+      "loss": 2.7751,
+      "theoretical_loss": 3.4692824091610412,
+      "tokens_seen": 1750138880
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047773737373737373,
+      "loss": 2.6498,
+      "theoretical_loss": 3.4692712913461126,
+      "tokens_seen": 1750204416
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047771717171717174,
+      "loss": 2.6655,
+      "theoretical_loss": 3.4692601740640403,
+      "tokens_seen": 1750269952
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004776969696969697,
+      "loss": 2.6462,
+      "theoretical_loss": 3.4692490573147787,
+      "tokens_seen": 1750335488
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004776767676767677,
+      "loss": 2.8046,
+      "theoretical_loss": 3.4692379410982825,
+      "tokens_seen": 1750401024
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047765656565656566,
+      "loss": 2.5931,
+      "theoretical_loss": 3.4692268254145064,
+      "tokens_seen": 1750466560
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004776363636363636,
+      "loss": 2.7954,
+      "theoretical_loss": 3.4692157102634047,
+      "tokens_seen": 1750532096
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047761616161616163,
+      "loss": 2.6183,
+      "theoretical_loss": 3.469204595644932,
+      "tokens_seen": 1750597632
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004775959595959596,
+      "loss": 2.779,
+      "theoretical_loss": 3.4691934815590426,
+      "tokens_seen": 1750663168
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004775757575757576,
+      "loss": 2.8523,
+      "theoretical_loss": 3.4691823680056917,
+      "tokens_seen": 1750728704
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047755555555555556,
+      "loss": 2.5914,
+      "theoretical_loss": 3.4691712549848326,
+      "tokens_seen": 1750794240
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047753535353535357,
+      "loss": 2.6619,
+      "theoretical_loss": 3.4691601424964214,
+      "tokens_seen": 1750859776
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004775151515151515,
+      "loss": 2.845,
+      "theoretical_loss": 3.469149030540412,
+      "tokens_seen": 1750925312
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004774949494949495,
+      "loss": 2.6006,
+      "theoretical_loss": 3.469137919116758,
+      "tokens_seen": 1750990848
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004774747474747475,
+      "loss": 2.6609,
+      "theoretical_loss": 3.4691268082254156,
+      "tokens_seen": 1751056384
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047745454545454545,
+      "loss": 2.7653,
+      "theoretical_loss": 3.4691156978663384,
+      "tokens_seen": 1751121920
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047743434343434346,
+      "loss": 2.7009,
+      "theoretical_loss": 3.4691045880394817,
+      "tokens_seen": 1751187456
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004774141414141414,
+      "loss": 2.8441,
+      "theoretical_loss": 3.469093478744799,
+      "tokens_seen": 1751252992
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004773939393939394,
+      "loss": 2.6236,
+      "theoretical_loss": 3.4690823699822455,
+      "tokens_seen": 1751318528
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004773737373737374,
+      "loss": 2.6359,
+      "theoretical_loss": 3.4690712617517763,
+      "tokens_seen": 1751384064
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004773535353535354,
+      "loss": 2.6701,
+      "theoretical_loss": 3.469060154053345,
+      "tokens_seen": 1751449600
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047733333333333335,
+      "loss": 2.5259,
+      "theoretical_loss": 3.469049046886907,
+      "tokens_seen": 1751515136
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0004773131313131313,
+      "loss": 2.7385,
+      "theoretical_loss": 3.4690379402524165,
+      "tokens_seen": 1751580672
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00047729292929292927,
+      "loss": 2.4743,
+      "theoretical_loss": 3.4690268341498287,
+      "tokens_seen": 1751646208
+    },
+    {
+      "epoch": 0.05,
+      "objective/train/docs_used": 966555,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6258327960968018,
+      "objective/train/theoretical_loss": 3.4690157285790972,
+      "objective/train/tokens_used": 110570976,
+      "theoretical_loss": 3.4690157285790972,
+      "tokens_seen": 1751711744
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004772727272727273,
+      "loss": 2.6509,
+      "theoretical_loss": 3.4690157285790972,
+      "tokens_seen": 1751711744
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004772525252525253,
+      "loss": 2.809,
+      "theoretical_loss": 3.4690046235401777,
+      "tokens_seen": 1751777280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047723232323232325,
+      "loss": 2.6795,
+      "theoretical_loss": 3.468993519033024,
+      "tokens_seen": 1751842816
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004772121212121212,
+      "loss": 2.7073,
+      "theoretical_loss": 3.4689824150575914,
+      "tokens_seen": 1751908352
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047719191919191916,
+      "loss": 2.6694,
+      "theoretical_loss": 3.468971311613834,
+      "tokens_seen": 1751973888
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004771717171717172,
+      "loss": 2.6556,
+      "theoretical_loss": 3.4689602087017066,
+      "tokens_seen": 1752039424
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004771515151515152,
+      "loss": 2.7105,
+      "theoretical_loss": 3.468949106321164,
+      "tokens_seen": 1752104960
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047713131313131314,
+      "loss": 2.5831,
+      "theoretical_loss": 3.468938004472161,
+      "tokens_seen": 1752170496
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004771111111111111,
+      "loss": 2.644,
+      "theoretical_loss": 3.468926903154652,
+      "tokens_seen": 1752236032
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047709090909090905,
+      "loss": 2.5805,
+      "theoretical_loss": 3.468915802368592,
+      "tokens_seen": 1752301568
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004770707070707071,
+      "loss": 2.8613,
+      "theoretical_loss": 3.4689047021139356,
+      "tokens_seen": 1752367104
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004770505050505051,
+      "loss": 2.6725,
+      "theoretical_loss": 3.4688936023906374,
+      "tokens_seen": 1752432640
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047703030303030303,
+      "loss": 2.838,
+      "theoretical_loss": 3.4688825031986514,
+      "tokens_seen": 1752498176
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000477010101010101,
+      "loss": 2.4936,
+      "theoretical_loss": 3.468871404537934,
+      "tokens_seen": 1752563712
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000476989898989899,
+      "loss": 2.7542,
+      "theoretical_loss": 3.4688603064084376,
+      "tokens_seen": 1752629248
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000476969696969697,
+      "loss": 2.6908,
+      "theoretical_loss": 3.468849208810119,
+      "tokens_seen": 1752694784
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047694949494949497,
+      "loss": 2.6054,
+      "theoretical_loss": 3.4688381117429317,
+      "tokens_seen": 1752760320
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004769292929292929,
+      "loss": 2.5833,
+      "theoretical_loss": 3.468827015206831,
+      "tokens_seen": 1752825856
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004769090909090909,
+      "loss": 3.0322,
+      "theoretical_loss": 3.4688159192017713,
+      "tokens_seen": 1752891392
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004768888888888889,
+      "loss": 2.5106,
+      "theoretical_loss": 3.468804823727708,
+      "tokens_seen": 1752956928
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004768686868686869,
+      "loss": 2.5625,
+      "theoretical_loss": 3.4687937287845947,
+      "tokens_seen": 1753022464
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047684848484848486,
+      "loss": 2.8008,
+      "theoretical_loss": 3.468782634372387,
+      "tokens_seen": 1753088000
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004768282828282828,
+      "loss": 2.5919,
+      "theoretical_loss": 3.4687715404910398,
+      "tokens_seen": 1753153536
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047680808080808083,
+      "loss": 2.6055,
+      "theoretical_loss": 3.4687604471405065,
+      "tokens_seen": 1753219072
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004767878787878788,
+      "loss": 2.7941,
+      "theoretical_loss": 3.4687493543207433,
+      "tokens_seen": 1753284608
+    },
+    {
+      "epoch": 0.06,
+      "objective/train/docs_used": 967544,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.789055347442627,
+      "objective/train/theoretical_loss": 3.468738262031705,
+      "objective/train/tokens_used": 112209376,
+      "theoretical_loss": 3.468738262031705,
+      "tokens_seen": 1753350144
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004767676767676768,
+      "loss": 2.6272,
+      "theoretical_loss": 3.468738262031705,
+      "tokens_seen": 1753350144
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047674747474747475,
+      "loss": 2.7313,
+      "theoretical_loss": 3.468727170273345,
+      "tokens_seen": 1753415680
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004767272727272727,
+      "loss": 2.6265,
+      "theoretical_loss": 3.4687160790456195,
+      "tokens_seen": 1753481216
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004767070707070707,
+      "loss": 2.5517,
+      "theoretical_loss": 3.468704988348483,
+      "tokens_seen": 1753546752
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047668686868686873,
+      "loss": 2.6153,
+      "theoretical_loss": 3.4686938981818893,
+      "tokens_seen": 1753612288
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004766666666666667,
+      "loss": 2.698,
+      "theoretical_loss": 3.468682808545794,
+      "tokens_seen": 1753677824
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047664646464646465,
+      "loss": 2.5592,
+      "theoretical_loss": 3.4686717194401524,
+      "tokens_seen": 1753743360
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047662626262626266,
+      "loss": 2.6959,
+      "theoretical_loss": 3.4686606308649184,
+      "tokens_seen": 1753808896
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004766060606060606,
+      "loss": 2.7546,
+      "theoretical_loss": 3.468649542820047,
+      "tokens_seen": 1753874432
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004765858585858586,
+      "loss": 2.7334,
+      "theoretical_loss": 3.468638455305493,
+      "tokens_seen": 1753939968
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004765656565656566,
+      "loss": 2.7488,
+      "theoretical_loss": 3.4686273683212114,
+      "tokens_seen": 1754005504
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047654545454545454,
+      "loss": 2.5704,
+      "theoretical_loss": 3.468616281867157,
+      "tokens_seen": 1754071040
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047652525252525255,
+      "loss": 2.7951,
+      "theoretical_loss": 3.468605195943285,
+      "tokens_seen": 1754136576
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004765050505050505,
+      "loss": 2.7617,
+      "theoretical_loss": 3.46859411054955,
+      "tokens_seen": 1754202112
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004764848484848485,
+      "loss": 2.7249,
+      "theoretical_loss": 3.468583025685906,
+      "tokens_seen": 1754267648
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004764646464646465,
+      "loss": 2.8578,
+      "theoretical_loss": 3.468571941352309,
+      "tokens_seen": 1754333184
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004764444444444445,
+      "loss": 2.7968,
+      "theoretical_loss": 3.4685608575487135,
+      "tokens_seen": 1754398720
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047642424242424244,
+      "loss": 2.5969,
+      "theoretical_loss": 3.4685497742750737,
+      "tokens_seen": 1754464256
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004764040404040404,
+      "loss": 2.6532,
+      "theoretical_loss": 3.4685386915313456,
+      "tokens_seen": 1754529792
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004763838383838384,
+      "loss": 2.5712,
+      "theoretical_loss": 3.4685276093174835,
+      "tokens_seen": 1754595328
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047636363636363637,
+      "loss": 2.7184,
+      "theoretical_loss": 3.468516527633442,
+      "tokens_seen": 1754660864
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004763434343434344,
+      "loss": 2.666,
+      "theoretical_loss": 3.4685054464791767,
+      "tokens_seen": 1754726400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047632323232323233,
+      "loss": 2.704,
+      "theoretical_loss": 3.4684943658546414,
+      "tokens_seen": 1754791936
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004763030303030303,
+      "loss": 2.5184,
+      "theoretical_loss": 3.4684832857597923,
+      "tokens_seen": 1754857472
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004762828282828283,
+      "loss": 2.5821,
+      "theoretical_loss": 3.4684722061945834,
+      "tokens_seen": 1754923008
+    },
+    {
+      "epoch": 0.06,
+      "objective/train/docs_used": 968097,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.9240682125091553,
+      "objective/train/theoretical_loss": 3.46846112715897,
+      "objective/train/tokens_used": 113847776,
+      "theoretical_loss": 3.46846112715897,
+      "tokens_seen": 1754988544
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004762626262626263,
+      "loss": 2.6846,
+      "theoretical_loss": 3.46846112715897,
+      "tokens_seen": 1754988544
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047624242424242427,
+      "loss": 2.686,
+      "theoretical_loss": 3.4684500486529064,
+      "tokens_seen": 1755054080
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004762222222222222,
+      "loss": 2.6209,
+      "theoretical_loss": 3.468438970676348,
+      "tokens_seen": 1755119616
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004762020202020202,
+      "loss": 2.5909,
+      "theoretical_loss": 3.4684278932292503,
+      "tokens_seen": 1755185152
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004761818181818182,
+      "loss": 2.7275,
+      "theoretical_loss": 3.468416816311567,
+      "tokens_seen": 1755250688
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004761616161616162,
+      "loss": 2.7313,
+      "theoretical_loss": 3.4684057399232544,
+      "tokens_seen": 1755316224
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047614141414141416,
+      "loss": 2.5685,
+      "theoretical_loss": 3.468394664064266,
+      "tokens_seen": 1755381760
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004761212121212121,
+      "loss": 2.6424,
+      "theoretical_loss": 3.468383588734558,
+      "tokens_seen": 1755447296
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004761010101010101,
+      "loss": 2.3658,
+      "theoretical_loss": 3.468372513934084,
+      "tokens_seen": 1755512832
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004760808080808081,
+      "loss": 2.8822,
+      "theoretical_loss": 3.468361439662801,
+      "tokens_seen": 1755578368
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004760606060606061,
+      "loss": 2.768,
+      "theoretical_loss": 3.468350365920662,
+      "tokens_seen": 1755643904
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047604040404040406,
+      "loss": 2.4781,
+      "theoretical_loss": 3.468339292707623,
+      "tokens_seen": 1755709440
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000476020202020202,
+      "loss": 2.509,
+      "theoretical_loss": 3.468328220023638,
+      "tokens_seen": 1755774976
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047599999999999997,
+      "loss": 2.556,
+      "theoretical_loss": 3.468317147868663,
+      "tokens_seen": 1755840512
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047597979797979803,
+      "loss": 2.6309,
+      "theoretical_loss": 3.4683060762426523,
+      "tokens_seen": 1755906048
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000475959595959596,
+      "loss": 2.6173,
+      "theoretical_loss": 3.4682950051455617,
+      "tokens_seen": 1755971584
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047593939393939395,
+      "loss": 2.502,
+      "theoretical_loss": 3.4682839345773457,
+      "tokens_seen": 1756037120
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004759191919191919,
+      "loss": 2.7037,
+      "theoretical_loss": 3.468272864537959,
+      "tokens_seen": 1756102656
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047589898989898986,
+      "loss": 2.9021,
+      "theoretical_loss": 3.468261795027357,
+      "tokens_seen": 1756168192
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047587878787878793,
+      "loss": 2.6436,
+      "theoretical_loss": 3.4682507260454942,
+      "tokens_seen": 1756233728
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004758585858585859,
+      "loss": 2.5381,
+      "theoretical_loss": 3.4682396575923264,
+      "tokens_seen": 1756299264
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047583838383838384,
+      "loss": 2.6176,
+      "theoretical_loss": 3.4682285896678087,
+      "tokens_seen": 1756364800
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004758181818181818,
+      "loss": 2.5754,
+      "theoretical_loss": 3.468217522271895,
+      "tokens_seen": 1756430336
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004757979797979798,
+      "loss": 2.6936,
+      "theoretical_loss": 3.468206455404541,
+      "tokens_seen": 1756495872
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004757777777777778,
+      "loss": 2.4587,
+      "theoretical_loss": 3.468195389065702,
+      "tokens_seen": 1756561408
+    },
+    {
+      "epoch": 0.06,
+      "objective/train/docs_used": 969105,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5639123916625977,
+      "objective/train/theoretical_loss": 3.4681843232553327,
+      "objective/train/tokens_used": 115486176,
+      "theoretical_loss": 3.4681843232553327,
+      "tokens_seen": 1756626944
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004757575757575758,
+      "loss": 2.5915,
+      "theoretical_loss": 3.4681843232553327,
+      "tokens_seen": 1756626944
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047573737373737373,
+      "loss": 2.5335,
+      "theoretical_loss": 3.4681732579733886,
+      "tokens_seen": 1756692480
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004757171717171717,
+      "loss": 2.4869,
+      "theoretical_loss": 3.4681621932198237,
+      "tokens_seen": 1756758016
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004756969696969697,
+      "loss": 2.8309,
+      "theoretical_loss": 3.4681511289945943,
+      "tokens_seen": 1756823552
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004756767676767677,
+      "loss": 2.832,
+      "theoretical_loss": 3.4681400652976544,
+      "tokens_seen": 1756889088
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047565656565656567,
+      "loss": 2.6746,
+      "theoretical_loss": 3.46812900212896,
+      "tokens_seen": 1756954624
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004756363636363636,
+      "loss": 2.6248,
+      "theoretical_loss": 3.4681179394884656,
+      "tokens_seen": 1757020160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047561616161616164,
+      "loss": 2.5682,
+      "theoretical_loss": 3.468106877376126,
+      "tokens_seen": 1757085696
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004755959595959596,
+      "loss": 2.6822,
+      "theoretical_loss": 3.4680958157918975,
+      "tokens_seen": 1757151232
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004755757575757576,
+      "loss": 2.8153,
+      "theoretical_loss": 3.468084754735734,
+      "tokens_seen": 1757216768
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047555555555555556,
+      "loss": 2.7997,
+      "theoretical_loss": 3.4680736942075914,
+      "tokens_seen": 1757282304
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004755353535353535,
+      "loss": 2.6609,
+      "theoretical_loss": 3.468062634207424,
+      "tokens_seen": 1757347840
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047551515151515153,
+      "loss": 2.8563,
+      "theoretical_loss": 3.4680515747351874,
+      "tokens_seen": 1757413376
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004754949494949495,
+      "loss": 2.6059,
+      "theoretical_loss": 3.4680405157908365,
+      "tokens_seen": 1757478912
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004754747474747475,
+      "loss": 2.8125,
+      "theoretical_loss": 3.4680294573743264,
+      "tokens_seen": 1757544448
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047545454545454545,
+      "loss": 2.7202,
+      "theoretical_loss": 3.4680183994856133,
+      "tokens_seen": 1757609984
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047543434343434347,
+      "loss": 2.6924,
+      "theoretical_loss": 3.4680073421246504,
+      "tokens_seen": 1757675520
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004754141414141414,
+      "loss": 2.6198,
+      "theoretical_loss": 3.4679962852913944,
+      "tokens_seen": 1757741056
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004753939393939394,
+      "loss": 2.6276,
+      "theoretical_loss": 3.4679852289857998,
+      "tokens_seen": 1757806592
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004753737373737374,
+      "loss": 2.7586,
+      "theoretical_loss": 3.4679741732078213,
+      "tokens_seen": 1757872128
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047535353535353535,
+      "loss": 2.4646,
+      "theoretical_loss": 3.4679631179574155,
+      "tokens_seen": 1757937664
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047533333333333336,
+      "loss": 2.4745,
+      "theoretical_loss": 3.467952063234536,
+      "tokens_seen": 1758003200
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004753131313131313,
+      "loss": 2.7784,
+      "theoretical_loss": 3.467941009039139,
+      "tokens_seen": 1758068736
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047529292929292927,
+      "loss": 2.7118,
+      "theoretical_loss": 3.4679299553711793,
+      "tokens_seen": 1758134272
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004752727272727273,
+      "loss": 2.8808,
+      "theoretical_loss": 3.4679189022306116,
+      "tokens_seen": 1758199808
+    },
+    {
+      "epoch": 0.06,
+      "objective/train/docs_used": 970436,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.550333023071289,
+      "objective/train/theoretical_loss": 3.467907849617392,
+      "objective/train/tokens_used": 117124576,
+      "theoretical_loss": 3.467907849617392,
+      "tokens_seen": 1758265344
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004752525252525253,
+      "loss": 2.6393,
+      "theoretical_loss": 3.467907849617392,
+      "tokens_seen": 1758265344
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047523232323232325,
+      "loss": 2.488,
+      "theoretical_loss": 3.467896797531475,
+      "tokens_seen": 1758330880
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004752121212121212,
+      "loss": 2.7108,
+      "theoretical_loss": 3.467885745972816,
+      "tokens_seen": 1758396416
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047519191919191916,
+      "loss": 2.7059,
+      "theoretical_loss": 3.4678746949413704,
+      "tokens_seen": 1758461952
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004751717171717172,
+      "loss": 2.5608,
+      "theoretical_loss": 3.467863644437093,
+      "tokens_seen": 1758527488
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004751515151515152,
+      "loss": 2.7159,
+      "theoretical_loss": 3.467852594459939,
+      "tokens_seen": 1758593024
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047513131313131314,
+      "loss": 2.6068,
+      "theoretical_loss": 3.4678415450098643,
+      "tokens_seen": 1758658560
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004751111111111111,
+      "loss": 2.5219,
+      "theoretical_loss": 3.4678304960868234,
+      "tokens_seen": 1758724096
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047509090909090906,
+      "loss": 2.5771,
+      "theoretical_loss": 3.467819447690772,
+      "tokens_seen": 1758789632
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004750707070707071,
+      "loss": 2.6642,
+      "theoretical_loss": 3.4678083998216653,
+      "tokens_seen": 1758855168
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004750505050505051,
+      "loss": 2.6359,
+      "theoretical_loss": 3.467797352479458,
+      "tokens_seen": 1758920704
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047503030303030304,
+      "loss": 2.6904,
+      "theoretical_loss": 3.4677863056641054,
+      "tokens_seen": 1758986240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000475010101010101,
+      "loss": 2.5726,
+      "theoretical_loss": 3.4677752593755633,
+      "tokens_seen": 1759051776
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000474989898989899,
+      "loss": 2.689,
+      "theoretical_loss": 3.467764213613787,
+      "tokens_seen": 1759117312
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000474969696969697,
+      "loss": 2.5023,
+      "theoretical_loss": 3.4677531683787306,
+      "tokens_seen": 1759182848
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047494949494949497,
+      "loss": 2.5113,
+      "theoretical_loss": 3.4677421236703507,
+      "tokens_seen": 1759248384
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047492929292929293,
+      "loss": 2.7619,
+      "theoretical_loss": 3.4677310794886025,
+      "tokens_seen": 1759313920
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004749090909090909,
+      "loss": 2.5726,
+      "theoretical_loss": 3.46772003583344,
+      "tokens_seen": 1759379456
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047488888888888895,
+      "loss": 2.424,
+      "theoretical_loss": 3.46770899270482,
+      "tokens_seen": 1759444992
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004748686868686869,
+      "loss": 2.705,
+      "theoretical_loss": 3.467697950102697,
+      "tokens_seen": 1759510528
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047484848484848487,
+      "loss": 2.8012,
+      "theoretical_loss": 3.467686908027026,
+      "tokens_seen": 1759576064
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004748282828282828,
+      "loss": 2.6163,
+      "theoretical_loss": 3.467675866477763,
+      "tokens_seen": 1759641600
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004748080808080808,
+      "loss": 2.7308,
+      "theoretical_loss": 3.4676648254548628,
+      "tokens_seen": 1759707136
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047478787878787884,
+      "loss": 2.6315,
+      "theoretical_loss": 3.467653784958281,
+      "tokens_seen": 1759772672
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004747676767676768,
+      "loss": 2.8835,
+      "theoretical_loss": 3.467642744987973,
+      "tokens_seen": 1759838208
+    },
+    {
+      "epoch": 0.06,
+      "objective/train/docs_used": 971034,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.658268690109253,
+      "objective/train/theoretical_loss": 3.467631705543894,
+      "objective/train/tokens_used": 118762976,
+      "theoretical_loss": 3.467631705543894,
+      "tokens_seen": 1759903744
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047474747474747476,
+      "loss": 2.441,
+      "theoretical_loss": 3.467631705543894,
+      "tokens_seen": 1759903744
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004747272727272727,
+      "loss": 2.4756,
+      "theoretical_loss": 3.467620666625999,
+      "tokens_seen": 1759969280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047470707070707067,
+      "loss": 2.5877,
+      "theoretical_loss": 3.4676096282342432,
+      "tokens_seen": 1760034816
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047468686868686874,
+      "loss": 2.4157,
+      "theoretical_loss": 3.4675985903685826,
+      "tokens_seen": 1760100352
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004746666666666667,
+      "loss": 2.6185,
+      "theoretical_loss": 3.4675875530289724,
+      "tokens_seen": 1760165888
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047464646464646465,
+      "loss": 2.626,
+      "theoretical_loss": 3.4675765162153676,
+      "tokens_seen": 1760231424
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004746262626262626,
+      "loss": 2.6303,
+      "theoretical_loss": 3.467565479927724,
+      "tokens_seen": 1760296960
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004746060606060606,
+      "loss": 2.4665,
+      "theoretical_loss": 3.4675544441659962,
+      "tokens_seen": 1760362496
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047458585858585863,
+      "loss": 2.4859,
+      "theoretical_loss": 3.4675434089301405,
+      "tokens_seen": 1760428032
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004745656565656566,
+      "loss": 2.7329,
+      "theoretical_loss": 3.467532374220112,
+      "tokens_seen": 1760493568
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047454545454545454,
+      "loss": 2.7522,
+      "theoretical_loss": 3.4675213400358653,
+      "tokens_seen": 1760559104
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004745252525252525,
+      "loss": 2.6709,
+      "theoretical_loss": 3.4675103063773567,
+      "tokens_seen": 1760624640
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004745050505050505,
+      "loss": 2.3443,
+      "theoretical_loss": 3.4674992732445413,
+      "tokens_seen": 1760690176
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004744848484848485,
+      "loss": 2.5856,
+      "theoretical_loss": 3.4674882406373744,
+      "tokens_seen": 1760755712
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004744646464646465,
+      "loss": 2.8132,
+      "theoretical_loss": 3.4674772085558114,
+      "tokens_seen": 1760821248
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047444444444444444,
+      "loss": 2.691,
+      "theoretical_loss": 3.4674661769998076,
+      "tokens_seen": 1760886784
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047442424242424245,
+      "loss": 2.8995,
+      "theoretical_loss": 3.4674551459693186,
+      "tokens_seen": 1760952320
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004744040404040404,
+      "loss": 2.5649,
+      "theoretical_loss": 3.4674441154643,
+      "tokens_seen": 1761017856
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004743838383838384,
+      "loss": 2.7209,
+      "theoretical_loss": 3.467433085484706,
+      "tokens_seen": 1761083392
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047436363636363637,
+      "loss": 2.8425,
+      "theoretical_loss": 3.467422056030494,
+      "tokens_seen": 1761148928
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047434343434343433,
+      "loss": 2.7739,
+      "theoretical_loss": 3.467411027101618,
+      "tokens_seen": 1761214464
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047432323232323234,
+      "loss": 2.6837,
+      "theoretical_loss": 3.467399998698034,
+      "tokens_seen": 1761280000
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004743030303030303,
+      "loss": 2.7335,
+      "theoretical_loss": 3.467388970819697,
+      "tokens_seen": 1761345536
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004742828282828283,
+      "loss": 2.6222,
+      "theoretical_loss": 3.4673779434665635,
+      "tokens_seen": 1761411072
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047426262626262626,
+      "loss": 2.6726,
+      "theoretical_loss": 3.4673669166385874,
+      "tokens_seen": 1761476608
+    },
+    {
+      "epoch": 0.06,
+      "objective/train/docs_used": 972196,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6714515686035156,
+      "objective/train/theoretical_loss": 3.4673558903357247,
+      "objective/train/tokens_used": 120401376,
+      "theoretical_loss": 3.4673558903357247,
+      "tokens_seen": 1761542144
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004742424242424243,
+      "loss": 2.6306,
+      "theoretical_loss": 3.4673558903357247,
+      "tokens_seen": 1761542144
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047422222222222223,
+      "loss": 2.8951,
+      "theoretical_loss": 3.4673448645579317,
+      "tokens_seen": 1761607680
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004742020202020202,
+      "loss": 2.7546,
+      "theoretical_loss": 3.467333839305163,
+      "tokens_seen": 1761673216
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004741818181818182,
+      "loss": 2.5618,
+      "theoretical_loss": 3.4673228145773747,
+      "tokens_seen": 1761738752
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047416161616161616,
+      "loss": 2.5558,
+      "theoretical_loss": 3.467311790374521,
+      "tokens_seen": 1761804288
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047414141414141417,
+      "loss": 2.3849,
+      "theoretical_loss": 3.467300766696559,
+      "tokens_seen": 1761869824
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004741212121212121,
+      "loss": 2.6992,
+      "theoretical_loss": 3.467289743543443,
+      "tokens_seen": 1761935360
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004741010101010101,
+      "loss": 2.6239,
+      "theoretical_loss": 3.4672787209151292,
+      "tokens_seen": 1762000896
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004740808080808081,
+      "loss": 2.61,
+      "theoretical_loss": 3.4672676988115727,
+      "tokens_seen": 1762066432
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004740606060606061,
+      "loss": 2.7707,
+      "theoretical_loss": 3.4672566772327293,
+      "tokens_seen": 1762131968
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047404040404040406,
+      "loss": 2.5288,
+      "theoretical_loss": 3.4672456561785543,
+      "tokens_seen": 1762197504
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000474020202020202,
+      "loss": 2.5156,
+      "theoretical_loss": 3.467234635649003,
+      "tokens_seen": 1762263040
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000474,
+      "loss": 2.3634,
+      "theoretical_loss": 3.4672236156440315,
+      "tokens_seen": 1762328576
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000473979797979798,
+      "loss": 2.3909,
+      "theoretical_loss": 3.467212596163595,
+      "tokens_seen": 1762394112
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000473959595959596,
+      "loss": 2.6485,
+      "theoretical_loss": 3.467201577207649,
+      "tokens_seen": 1762459648
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047393939393939395,
+      "loss": 2.9205,
+      "theoretical_loss": 3.467190558776149,
+      "tokens_seen": 1762525184
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004739191919191919,
+      "loss": 2.4281,
+      "theoretical_loss": 3.4671795408690507,
+      "tokens_seen": 1762590720
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047389898989898987,
+      "loss": 2.6767,
+      "theoretical_loss": 3.4671685234863094,
+      "tokens_seen": 1762656256
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047387878787878793,
+      "loss": 2.6607,
+      "theoretical_loss": 3.467157506627881,
+      "tokens_seen": 1762721792
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004738585858585859,
+      "loss": 2.4989,
+      "theoretical_loss": 3.4671464902937204,
+      "tokens_seen": 1762787328
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047383838383838385,
+      "loss": 2.7566,
+      "theoretical_loss": 3.4671354744837837,
+      "tokens_seen": 1762852864
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004738181818181818,
+      "loss": 2.4784,
+      "theoretical_loss": 3.4671244591980264,
+      "tokens_seen": 1762918400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047379797979797976,
+      "loss": 2.6094,
+      "theoretical_loss": 3.4671134444364045,
+      "tokens_seen": 1762983936
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004737777777777778,
+      "loss": 2.3573,
+      "theoretical_loss": 3.467102430198873,
+      "tokens_seen": 1763049472
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004737575757575758,
+      "loss": 2.7262,
+      "theoretical_loss": 3.4670914164853874,
+      "tokens_seen": 1763115008
+    },
+    {
+      "epoch": 0.06,
+      "objective/train/docs_used": 972718,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.458674430847168,
+      "objective/train/theoretical_loss": 3.4670804032959035,
+      "objective/train/tokens_used": 122039776,
+      "theoretical_loss": 3.4670804032959035,
+      "tokens_seen": 1763180544
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047373737373737374,
+      "loss": 2.5665,
+      "theoretical_loss": 3.4670804032959035,
+      "tokens_seen": 1763180544
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004737171717171717,
+      "loss": 2.4811,
+      "theoretical_loss": 3.4670693906303773,
+      "tokens_seen": 1763246080
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004736969696969697,
+      "loss": 2.4546,
+      "theoretical_loss": 3.4670583784887636,
+      "tokens_seen": 1763311616
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004736767676767677,
+      "loss": 2.5793,
+      "theoretical_loss": 3.4670473668710184,
+      "tokens_seen": 1763377152
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004736565656565657,
+      "loss": 2.5163,
+      "theoretical_loss": 3.467036355777098,
+      "tokens_seen": 1763442688
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047363636363636363,
+      "loss": 2.5888,
+      "theoretical_loss": 3.467025345206956,
+      "tokens_seen": 1763508224
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004736161616161616,
+      "loss": 2.7923,
+      "theoretical_loss": 3.4670143351605502,
+      "tokens_seen": 1763573760
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004735959595959596,
+      "loss": 2.7194,
+      "theoretical_loss": 3.4670033256378354,
+      "tokens_seen": 1763639296
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004735757575757576,
+      "loss": 2.4734,
+      "theoretical_loss": 3.466992316638767,
+      "tokens_seen": 1763704832
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047355555555555557,
+      "loss": 2.7535,
+      "theoretical_loss": 3.4669813081633007,
+      "tokens_seen": 1763770368
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004735353535353535,
+      "loss": 2.7308,
+      "theoretical_loss": 3.466970300211393,
+      "tokens_seen": 1763835904
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047351515151515154,
+      "loss": 2.579,
+      "theoretical_loss": 3.466959292782998,
+      "tokens_seen": 1763901440
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004734949494949495,
+      "loss": 2.5358,
+      "theoretical_loss": 3.4669482858780727,
+      "tokens_seen": 1763966976
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004734747474747475,
+      "loss": 2.7295,
+      "theoretical_loss": 3.4669372794965723,
+      "tokens_seen": 1764032512
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047345454545454546,
+      "loss": 2.474,
+      "theoretical_loss": 3.4669262736384523,
+      "tokens_seen": 1764098048
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004734343434343434,
+      "loss": 2.7381,
+      "theoretical_loss": 3.4669152683036684,
+      "tokens_seen": 1764163584
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047341414141414143,
+      "loss": 2.7042,
+      "theoretical_loss": 3.466904263492176,
+      "tokens_seen": 1764229120
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004733939393939394,
+      "loss": 2.8351,
+      "theoretical_loss": 3.466893259203932,
+      "tokens_seen": 1764294656
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004733737373737374,
+      "loss": 2.7247,
+      "theoretical_loss": 3.4668822554388905,
+      "tokens_seen": 1764360192
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047335353535353535,
+      "loss": 2.5802,
+      "theoretical_loss": 3.4668712521970084,
+      "tokens_seen": 1764425728
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047333333333333336,
+      "loss": 2.6002,
+      "theoretical_loss": 3.4668602494782403,
+      "tokens_seen": 1764491264
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004733131313131313,
+      "loss": 2.6269,
+      "theoretical_loss": 3.466849247282543,
+      "tokens_seen": 1764556800
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004732929292929293,
+      "loss": 2.6215,
+      "theoretical_loss": 3.466838245609872,
+      "tokens_seen": 1764622336
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004732727272727273,
+      "loss": 2.6113,
+      "theoretical_loss": 3.466827244460182,
+      "tokens_seen": 1764687872
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047325252525252525,
+      "loss": 2.534,
+      "theoretical_loss": 3.46681624383343,
+      "tokens_seen": 1764753408
+    },
+    {
+      "epoch": 0.06,
+      "objective/train/docs_used": 973844,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.547422170639038,
+      "objective/train/theoretical_loss": 3.466805243729571,
+      "objective/train/tokens_used": 123678176,
+      "theoretical_loss": 3.466805243729571,
+      "tokens_seen": 1764818944
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047323232323232326,
+      "loss": 2.5461,
+      "theoretical_loss": 3.466805243729571,
+      "tokens_seen": 1764818944
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004732121212121212,
+      "loss": 2.4588,
+      "theoretical_loss": 3.466794244148561,
+      "tokens_seen": 1764884480
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004731919191919192,
+      "loss": 2.4346,
+      "theoretical_loss": 3.4667832450903555,
+      "tokens_seen": 1764950016
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004731717171717172,
+      "loss": 2.6804,
+      "theoretical_loss": 3.4667722465549105,
+      "tokens_seen": 1765015552
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004731515151515152,
+      "loss": 2.5802,
+      "theoretical_loss": 3.4667612485421815,
+      "tokens_seen": 1765081088
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047313131313131315,
+      "loss": 2.6148,
+      "theoretical_loss": 3.4667502510521246,
+      "tokens_seen": 1765146624
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004731111111111111,
+      "loss": 2.625,
+      "theoretical_loss": 3.466739254084695,
+      "tokens_seen": 1765212160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004730909090909091,
+      "loss": 2.6565,
+      "theoretical_loss": 3.466728257639849,
+      "tokens_seen": 1765277696
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004730707070707071,
+      "loss": 2.6265,
+      "theoretical_loss": 3.4667172617175424,
+      "tokens_seen": 1765343232
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004730505050505051,
+      "loss": 2.5729,
+      "theoretical_loss": 3.4667062663177304,
+      "tokens_seen": 1765408768
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047303030303030304,
+      "loss": 2.706,
+      "theoretical_loss": 3.4666952714403694,
+      "tokens_seen": 1765474304
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000473010101010101,
+      "loss": 2.5028,
+      "theoretical_loss": 3.4666842770854145,
+      "tokens_seen": 1765539840
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000472989898989899,
+      "loss": 2.7347,
+      "theoretical_loss": 3.466673283252822,
+      "tokens_seen": 1765605376
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000472969696969697,
+      "loss": 2.2926,
+      "theoretical_loss": 3.4666622899425477,
+      "tokens_seen": 1765670912
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000472949494949495,
+      "loss": 2.7329,
+      "theoretical_loss": 3.4666512971545473,
+      "tokens_seen": 1765736448
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047292929292929293,
+      "loss": 2.7155,
+      "theoretical_loss": 3.4666403048887764,
+      "tokens_seen": 1765801984
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004729090909090909,
+      "loss": 2.676,
+      "theoretical_loss": 3.4666293131451917,
+      "tokens_seen": 1765867520
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004728888888888889,
+      "loss": 2.5448,
+      "theoretical_loss": 3.4666183219237476,
+      "tokens_seen": 1765933056
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004728686868686869,
+      "loss": 2.6503,
+      "theoretical_loss": 3.4666073312244006,
+      "tokens_seen": 1765998592
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047284848484848487,
+      "loss": 2.7058,
+      "theoretical_loss": 3.466596341047107,
+      "tokens_seen": 1766064128
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047282828282828283,
+      "loss": 2.517,
+      "theoretical_loss": 3.4665853513918217,
+      "tokens_seen": 1766129664
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004728080808080808,
+      "loss": 2.626,
+      "theoretical_loss": 3.466574362258501,
+      "tokens_seen": 1766195200
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004727878787878788,
+      "loss": 2.6928,
+      "theoretical_loss": 3.4665633736471015,
+      "tokens_seen": 1766260736
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004727676767676768,
+      "loss": 2.6162,
+      "theoretical_loss": 3.466552385557578,
+      "tokens_seen": 1766326272
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047274747474747476,
+      "loss": 2.5867,
+      "theoretical_loss": 3.4665413979898863,
+      "tokens_seen": 1766391808
+    },
+    {
+      "epoch": 0.06,
+      "objective/train/docs_used": 974264,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.8634729385375977,
+      "objective/train/theoretical_loss": 3.466530410943983,
+      "objective/train/tokens_used": 125316576,
+      "theoretical_loss": 3.466530410943983,
+      "tokens_seen": 1766457344
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004727272727272727,
+      "loss": 2.755,
+      "theoretical_loss": 3.466530410943983,
+      "tokens_seen": 1766457344
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004727070707070707,
+      "loss": 2.6146,
+      "theoretical_loss": 3.466519424419823,
+      "tokens_seen": 1766522880
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047268686868686874,
+      "loss": 2.8209,
+      "theoretical_loss": 3.4665084384173634,
+      "tokens_seen": 1766588416
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004726666666666667,
+      "loss": 2.7291,
+      "theoretical_loss": 3.4664974529365593,
+      "tokens_seen": 1766653952
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047264646464646466,
+      "loss": 2.7675,
+      "theoretical_loss": 3.466486467977367,
+      "tokens_seen": 1766719488
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004726262626262626,
+      "loss": 2.4818,
+      "theoretical_loss": 3.4664754835397416,
+      "tokens_seen": 1766785024
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047260606060606057,
+      "loss": 2.5268,
+      "theoretical_loss": 3.4664644996236396,
+      "tokens_seen": 1766850560
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047258585858585864,
+      "loss": 2.5537,
+      "theoretical_loss": 3.4664535162290164,
+      "tokens_seen": 1766916096
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004725656565656566,
+      "loss": 2.643,
+      "theoretical_loss": 3.466442533355829,
+      "tokens_seen": 1766981632
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047254545454545455,
+      "loss": 2.6887,
+      "theoretical_loss": 3.466431551004032,
+      "tokens_seen": 1767047168
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004725252525252525,
+      "loss": 2.6244,
+      "theoretical_loss": 3.4664205691735823,
+      "tokens_seen": 1767112704
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004725050505050505,
+      "loss": 2.5747,
+      "theoretical_loss": 3.466409587864436,
+      "tokens_seen": 1767178240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047248484848484853,
+      "loss": 2.3783,
+      "theoretical_loss": 3.466398607076547,
+      "tokens_seen": 1767243776
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004724646464646465,
+      "loss": 2.6642,
+      "theoretical_loss": 3.4663876268098734,
+      "tokens_seen": 1767309312
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047244444444444444,
+      "loss": 2.4038,
+      "theoretical_loss": 3.466376647064371,
+      "tokens_seen": 1767374848
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004724242424242424,
+      "loss": 2.5637,
+      "theoretical_loss": 3.4663656678399946,
+      "tokens_seen": 1767440384
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004724040404040404,
+      "loss": 2.6683,
+      "theoretical_loss": 3.4663546891367005,
+      "tokens_seen": 1767505920
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004723838383838384,
+      "loss": 2.6663,
+      "theoretical_loss": 3.4663437109544453,
+      "tokens_seen": 1767571456
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004723636363636364,
+      "loss": 2.7499,
+      "theoretical_loss": 3.466332733293184,
+      "tokens_seen": 1767636992
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047234343434343433,
+      "loss": 2.6059,
+      "theoretical_loss": 3.4663217561528734,
+      "tokens_seen": 1767702528
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047232323232323235,
+      "loss": 2.8539,
+      "theoretical_loss": 3.466310779533469,
+      "tokens_seen": 1767768064
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004723030303030303,
+      "loss": 2.6061,
+      "theoretical_loss": 3.466299803434927,
+      "tokens_seen": 1767833600
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004722828282828283,
+      "loss": 2.7606,
+      "theoretical_loss": 3.4662888278572033,
+      "tokens_seen": 1767899136
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00047226262626262627,
+      "loss": 2.602,
+      "theoretical_loss": 3.4662778528002534,
+      "tokens_seen": 1767964672
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0004722424242424242,
+      "loss": 2.6908,
+      "theoretical_loss": 3.466266878264034,
+      "tokens_seen": 1768030208
+    },
+    {
+      "epoch": 0.06,
+      "objective/train/docs_used": 975251,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7586162090301514,
+      "objective/train/theoretical_loss": 3.4662559042485013,
+      "objective/train/tokens_used": 126954976,
+      "theoretical_loss": 3.4662559042485013,
+      "tokens_seen": 1768095744
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047222222222222224,
+      "loss": 2.6865,
+      "theoretical_loss": 3.4662559042485013,
+      "tokens_seen": 1768095744
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004722020202020202,
+      "loss": 2.5986,
+      "theoretical_loss": 3.4662449307536103,
+      "tokens_seen": 1768161280
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004721818181818182,
+      "loss": 2.5451,
+      "theoretical_loss": 3.466233957779318,
+      "tokens_seen": 1768226816
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047216161616161616,
+      "loss": 2.5286,
+      "theoretical_loss": 3.4662229853255795,
+      "tokens_seen": 1768292352
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004721414141414142,
+      "loss": 2.6367,
+      "theoretical_loss": 3.466212013392351,
+      "tokens_seen": 1768357888
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047212121212121213,
+      "loss": 2.6325,
+      "theoretical_loss": 3.4662010419795894,
+      "tokens_seen": 1768423424
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004721010101010101,
+      "loss": 2.5137,
+      "theoretical_loss": 3.46619007108725,
+      "tokens_seen": 1768488960
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004720808080808081,
+      "loss": 2.4045,
+      "theoretical_loss": 3.4661791007152885,
+      "tokens_seen": 1768554496
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047206060606060606,
+      "loss": 2.6393,
+      "theoretical_loss": 3.466168130863662,
+      "tokens_seen": 1768620032
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047204040404040407,
+      "loss": 2.4247,
+      "theoretical_loss": 3.4661571615323252,
+      "tokens_seen": 1768685568
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000472020202020202,
+      "loss": 2.4163,
+      "theoretical_loss": 3.466146192721235,
+      "tokens_seen": 1768751104
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000472,
+      "loss": 2.6915,
+      "theoretical_loss": 3.4661352244303476,
+      "tokens_seen": 1768816640
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000471979797979798,
+      "loss": 2.4705,
+      "theoretical_loss": 3.4661242566596187,
+      "tokens_seen": 1768882176
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000471959595959596,
+      "loss": 2.6915,
+      "theoretical_loss": 3.4661132894090043,
+      "tokens_seen": 1768947712
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047193939393939396,
+      "loss": 2.5054,
+      "theoretical_loss": 3.46610232267846,
+      "tokens_seen": 1769013248
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004719191919191919,
+      "loss": 2.6938,
+      "theoretical_loss": 3.4660913564679436,
+      "tokens_seen": 1769078784
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004718989898989899,
+      "loss": 2.5973,
+      "theoretical_loss": 3.4660803907774094,
+      "tokens_seen": 1769144320
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004718787878787879,
+      "loss": 2.6252,
+      "theoretical_loss": 3.4660694256068143,
+      "tokens_seen": 1769209856
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004718585858585859,
+      "loss": 2.7995,
+      "theoretical_loss": 3.4660584609561145,
+      "tokens_seen": 1769275392
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047183838383838385,
+      "loss": 2.5113,
+      "theoretical_loss": 3.466047496825265,
+      "tokens_seen": 1769340928
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004718181818181818,
+      "loss": 2.58,
+      "theoretical_loss": 3.4660365332142233,
+      "tokens_seen": 1769406464
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047179797979797977,
+      "loss": 2.6229,
+      "theoretical_loss": 3.4660255701229445,
+      "tokens_seen": 1769472000
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047177777777777783,
+      "loss": 2.625,
+      "theoretical_loss": 3.4660146075513856,
+      "tokens_seen": 1769537536
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004717575757575758,
+      "loss": 2.4516,
+      "theoretical_loss": 3.466003645499502,
+      "tokens_seen": 1769603072
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047173737373737374,
+      "loss": 2.6413,
+      "theoretical_loss": 3.46599268396725,
+      "tokens_seen": 1769668608
+    },
+    {
+      "epoch": 0.07,
+      "objective/train/docs_used": 976290,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.645479679107666,
+      "objective/train/theoretical_loss": 3.465981722954586,
+      "objective/train/tokens_used": 128593376,
+      "theoretical_loss": 3.465981722954586,
+      "tokens_seen": 1769734144
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004717171717171717,
+      "loss": 2.7062,
+      "theoretical_loss": 3.465981722954586,
+      "tokens_seen": 1769734144
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047169696969696966,
+      "loss": 2.7589,
+      "theoretical_loss": 3.465970762461466,
+      "tokens_seen": 1769799680
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004716767676767677,
+      "loss": 2.8116,
+      "theoretical_loss": 3.4659598024878457,
+      "tokens_seen": 1769865216
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004716565656565657,
+      "loss": 2.8078,
+      "theoretical_loss": 3.4659488430336816,
+      "tokens_seen": 1769930752
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047163636363636364,
+      "loss": 2.2874,
+      "theoretical_loss": 3.46593788409893,
+      "tokens_seen": 1769996288
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004716161616161616,
+      "loss": 2.2027,
+      "theoretical_loss": 3.4659269256835468,
+      "tokens_seen": 1770061824
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004715959595959596,
+      "loss": 2.7392,
+      "theoretical_loss": 3.4659159677874882,
+      "tokens_seen": 1770127360
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004715757575757576,
+      "loss": 2.8195,
+      "theoretical_loss": 3.4659050104107103,
+      "tokens_seen": 1770192896
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004715555555555556,
+      "loss": 2.6863,
+      "theoretical_loss": 3.4658940535531695,
+      "tokens_seen": 1770258432
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047153535353535353,
+      "loss": 2.4416,
+      "theoretical_loss": 3.465883097214822,
+      "tokens_seen": 1770323968
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004715151515151515,
+      "loss": 2.5736,
+      "theoretical_loss": 3.4658721413956237,
+      "tokens_seen": 1770389504
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047149494949494955,
+      "loss": 2.5487,
+      "theoretical_loss": 3.465861186095531,
+      "tokens_seen": 1770455040
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004714747474747475,
+      "loss": 2.9009,
+      "theoretical_loss": 3.4658502313144997,
+      "tokens_seen": 1770520576
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047145454545454547,
+      "loss": 2.8561,
+      "theoretical_loss": 3.4658392770524866,
+      "tokens_seen": 1770586112
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004714343434343434,
+      "loss": 2.646,
+      "theoretical_loss": 3.4658283233094473,
+      "tokens_seen": 1770651648
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004714141414141414,
+      "loss": 2.6522,
+      "theoretical_loss": 3.4658173700853387,
+      "tokens_seen": 1770717184
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047139393939393944,
+      "loss": 2.7383,
+      "theoretical_loss": 3.4658064173801164,
+      "tokens_seen": 1770782720
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004713737373737374,
+      "loss": 2.4971,
+      "theoretical_loss": 3.465795465193737,
+      "tokens_seen": 1770848256
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047135353535353536,
+      "loss": 2.8238,
+      "theoretical_loss": 3.465784513526156,
+      "tokens_seen": 1770913792
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004713333333333333,
+      "loss": 2.5457,
+      "theoretical_loss": 3.465773562377331,
+      "tokens_seen": 1770979328
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004713131313131313,
+      "loss": 2.7584,
+      "theoretical_loss": 3.465762611747217,
+      "tokens_seen": 1771044864
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047129292929292934,
+      "loss": 2.5373,
+      "theoretical_loss": 3.4657516616357706,
+      "tokens_seen": 1771110400
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004712727272727273,
+      "loss": 2.7518,
+      "theoretical_loss": 3.4657407120429484,
+      "tokens_seen": 1771175936
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047125252525252525,
+      "loss": 2.6246,
+      "theoretical_loss": 3.4657297629687056,
+      "tokens_seen": 1771241472
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004712323232323232,
+      "loss": 2.6083,
+      "theoretical_loss": 3.465718814413,
+      "tokens_seen": 1771307008
+    },
+    {
+      "epoch": 0.07,
+      "objective/train/docs_used": 976909,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.4419713020324707,
+      "objective/train/theoretical_loss": 3.4657078663757868,
+      "objective/train/tokens_used": 130231776,
+      "theoretical_loss": 3.4657078663757868,
+      "tokens_seen": 1771372544
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004712121212121212,
+      "loss": 2.683,
+      "theoretical_loss": 3.4657078663757868,
+      "tokens_seen": 1771372544
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047119191919191923,
+      "loss": 2.7196,
+      "theoretical_loss": 3.4656969188570224,
+      "tokens_seen": 1771438080
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004711717171717172,
+      "loss": 2.5106,
+      "theoretical_loss": 3.4656859718566633,
+      "tokens_seen": 1771503616
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047115151515151514,
+      "loss": 2.4445,
+      "theoretical_loss": 3.465675025374666,
+      "tokens_seen": 1771569152
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047113131313131316,
+      "loss": 2.4538,
+      "theoretical_loss": 3.4656640794109856,
+      "tokens_seen": 1771634688
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004711111111111111,
+      "loss": 2.6856,
+      "theoretical_loss": 3.4656531339655796,
+      "tokens_seen": 1771700224
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004710909090909091,
+      "loss": 2.771,
+      "theoretical_loss": 3.4656421890384044,
+      "tokens_seen": 1771765760
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004710707070707071,
+      "loss": 2.5379,
+      "theoretical_loss": 3.4656312446294155,
+      "tokens_seen": 1771831296
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047105050505050504,
+      "loss": 2.587,
+      "theoretical_loss": 3.465620300738569,
+      "tokens_seen": 1771896832
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047103030303030305,
+      "loss": 2.5692,
+      "theoretical_loss": 3.4656093573658224,
+      "tokens_seen": 1771962368
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000471010101010101,
+      "loss": 2.7043,
+      "theoretical_loss": 3.465598414511131,
+      "tokens_seen": 1772027904
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000470989898989899,
+      "loss": 2.6211,
+      "theoretical_loss": 3.4655874721744517,
+      "tokens_seen": 1772093440
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047096969696969697,
+      "loss": 2.6399,
+      "theoretical_loss": 3.46557653035574,
+      "tokens_seen": 1772158976
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000470949494949495,
+      "loss": 2.5832,
+      "theoretical_loss": 3.465565589054954,
+      "tokens_seen": 1772224512
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047092929292929294,
+      "loss": 2.6915,
+      "theoretical_loss": 3.4655546482720476,
+      "tokens_seen": 1772290048
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004709090909090909,
+      "loss": 2.5416,
+      "theoretical_loss": 3.4655437080069786,
+      "tokens_seen": 1772355584
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004708888888888889,
+      "loss": 2.6132,
+      "theoretical_loss": 3.4655327682597035,
+      "tokens_seen": 1772421120
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047086868686868687,
+      "loss": 2.4983,
+      "theoretical_loss": 3.4655218290301786,
+      "tokens_seen": 1772486656
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004708484848484849,
+      "loss": 2.6079,
+      "theoretical_loss": 3.465510890318359,
+      "tokens_seen": 1772552192
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047082828282828283,
+      "loss": 2.7404,
+      "theoretical_loss": 3.465499952124203,
+      "tokens_seen": 1772617728
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004708080808080808,
+      "loss": 2.7041,
+      "theoretical_loss": 3.465489014447665,
+      "tokens_seen": 1772683264
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004707878787878788,
+      "loss": 2.6753,
+      "theoretical_loss": 3.4654780772887026,
+      "tokens_seen": 1772748800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004707676767676768,
+      "loss": 2.8614,
+      "theoretical_loss": 3.465467140647272,
+      "tokens_seen": 1772814336
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047074747474747477,
+      "loss": 2.5925,
+      "theoretical_loss": 3.4654562045233295,
+      "tokens_seen": 1772879872
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004707272727272727,
+      "loss": 2.7646,
+      "theoretical_loss": 3.465445268916832,
+      "tokens_seen": 1772945408
+    },
+    {
+      "epoch": 0.07,
+      "objective/train/docs_used": 977811,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.4676389694213867,
+      "objective/train/theoretical_loss": 3.4654343338277345,
+      "objective/train/tokens_used": 131870176,
+      "theoretical_loss": 3.4654343338277345,
+      "tokens_seen": 1773010944
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004707070707070707,
+      "loss": 2.5732,
+      "theoretical_loss": 3.4654343338277345,
+      "tokens_seen": 1773010944
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004706868686868687,
+      "loss": 2.8103,
+      "theoretical_loss": 3.4654233992559944,
+      "tokens_seen": 1773076480
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004706666666666667,
+      "loss": 2.4505,
+      "theoretical_loss": 3.465412465201568,
+      "tokens_seen": 1773142016
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047064646464646466,
+      "loss": 2.3781,
+      "theoretical_loss": 3.465401531664412,
+      "tokens_seen": 1773207552
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004706262626262626,
+      "loss": 2.8003,
+      "theoretical_loss": 3.4653905986444817,
+      "tokens_seen": 1773273088
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004706060606060606,
+      "loss": 2.6426,
+      "theoretical_loss": 3.465379666141735,
+      "tokens_seen": 1773338624
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047058585858585864,
+      "loss": 2.5942,
+      "theoretical_loss": 3.4653687341561277,
+      "tokens_seen": 1773404160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004705656565656566,
+      "loss": 2.628,
+      "theoretical_loss": 3.4653578026876155,
+      "tokens_seen": 1773469696
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047054545454545455,
+      "loss": 2.6492,
+      "theoretical_loss": 3.4653468717361555,
+      "tokens_seen": 1773535232
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004705252525252525,
+      "loss": 2.6411,
+      "theoretical_loss": 3.4653359413017046,
+      "tokens_seen": 1773600768
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047050505050505047,
+      "loss": 2.7541,
+      "theoretical_loss": 3.465325011384219,
+      "tokens_seen": 1773666304
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047048484848484853,
+      "loss": 2.4253,
+      "theoretical_loss": 3.465314081983654,
+      "tokens_seen": 1773731840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004704646464646465,
+      "loss": 2.7165,
+      "theoretical_loss": 3.4653031530999674,
+      "tokens_seen": 1773797376
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047044444444444445,
+      "loss": 2.7943,
+      "theoretical_loss": 3.4652922247331155,
+      "tokens_seen": 1773862912
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004704242424242424,
+      "loss": 2.5949,
+      "theoretical_loss": 3.465281296883054,
+      "tokens_seen": 1773928448
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004704040404040404,
+      "loss": 2.5194,
+      "theoretical_loss": 3.4652703695497404,
+      "tokens_seen": 1773993984
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004703838383838384,
+      "loss": 2.8284,
+      "theoretical_loss": 3.46525944273313,
+      "tokens_seen": 1774059520
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004703636363636364,
+      "loss": 2.5426,
+      "theoretical_loss": 3.4652485164331805,
+      "tokens_seen": 1774125056
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047034343434343434,
+      "loss": 2.6549,
+      "theoretical_loss": 3.4652375906498474,
+      "tokens_seen": 1774190592
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004703232323232323,
+      "loss": 2.5567,
+      "theoretical_loss": 3.4652266653830877,
+      "tokens_seen": 1774256128
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004703030303030303,
+      "loss": 2.6891,
+      "theoretical_loss": 3.465215740632858,
+      "tokens_seen": 1774321664
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004702828282828283,
+      "loss": 2.6164,
+      "theoretical_loss": 3.465204816399114,
+      "tokens_seen": 1774387200
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004702626262626263,
+      "loss": 2.6243,
+      "theoretical_loss": 3.465193892681813,
+      "tokens_seen": 1774452736
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047024242424242423,
+      "loss": 2.6201,
+      "theoretical_loss": 3.4651829694809115,
+      "tokens_seen": 1774518272
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047022222222222224,
+      "loss": 2.6683,
+      "theoretical_loss": 3.465172046796366,
+      "tokens_seen": 1774583808
+    },
+    {
+      "epoch": 0.07,
+      "objective/train/docs_used": 978517,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.493600368499756,
+      "objective/train/theoretical_loss": 3.4651611246281324,
+      "objective/train/tokens_used": 133508576,
+      "theoretical_loss": 3.4651611246281324,
+      "tokens_seen": 1774649344
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004702020202020202,
+      "loss": 2.5756,
+      "theoretical_loss": 3.4651611246281324,
+      "tokens_seen": 1774649344
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004701818181818182,
+      "loss": 2.7666,
+      "theoretical_loss": 3.465150202976168,
+      "tokens_seen": 1774714880
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047016161616161617,
+      "loss": 2.6612,
+      "theoretical_loss": 3.465139281840429,
+      "tokens_seen": 1774780416
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004701414141414141,
+      "loss": 2.5653,
+      "theoretical_loss": 3.465128361220872,
+      "tokens_seen": 1774845952
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047012121212121214,
+      "loss": 2.6465,
+      "theoretical_loss": 3.465117441117453,
+      "tokens_seen": 1774911488
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004701010101010101,
+      "loss": 2.5144,
+      "theoretical_loss": 3.4651065215301298,
+      "tokens_seen": 1774977024
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004700808080808081,
+      "loss": 2.7176,
+      "theoretical_loss": 3.4650956024588577,
+      "tokens_seen": 1775042560
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047006060606060606,
+      "loss": 2.7295,
+      "theoretical_loss": 3.4650846839035943,
+      "tokens_seen": 1775108096
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047004040404040407,
+      "loss": 2.5483,
+      "theoretical_loss": 3.465073765864295,
+      "tokens_seen": 1775173632
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047002020202020203,
+      "loss": 2.5907,
+      "theoretical_loss": 3.465062848340917,
+      "tokens_seen": 1775239168
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00047,
+      "loss": 2.6084,
+      "theoretical_loss": 3.4650519313334174,
+      "tokens_seen": 1775304704
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000469979797979798,
+      "loss": 2.4807,
+      "theoretical_loss": 3.465041014841751,
+      "tokens_seen": 1775370240
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046995959595959595,
+      "loss": 2.6201,
+      "theoretical_loss": 3.465030098865877,
+      "tokens_seen": 1775435776
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046993939393939396,
+      "loss": 2.4878,
+      "theoretical_loss": 3.4650191834057504,
+      "tokens_seen": 1775501312
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004699191919191919,
+      "loss": 2.6224,
+      "theoretical_loss": 3.465008268461328,
+      "tokens_seen": 1775566848
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004698989898989899,
+      "loss": 2.653,
+      "theoretical_loss": 3.4649973540325663,
+      "tokens_seen": 1775632384
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004698787878787879,
+      "loss": 2.5147,
+      "theoretical_loss": 3.4649864401194215,
+      "tokens_seen": 1775697920
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004698585858585859,
+      "loss": 2.4981,
+      "theoretical_loss": 3.4649755267218514,
+      "tokens_seen": 1775763456
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046983838383838386,
+      "loss": 2.6706,
+      "theoretical_loss": 3.464964613839812,
+      "tokens_seen": 1775828992
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004698181818181818,
+      "loss": 2.6469,
+      "theoretical_loss": 3.4649537014732594,
+      "tokens_seen": 1775894528
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004697979797979798,
+      "loss": 2.6877,
+      "theoretical_loss": 3.464942789622151,
+      "tokens_seen": 1775960064
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004697777777777778,
+      "loss": 2.6401,
+      "theoretical_loss": 3.464931878286443,
+      "tokens_seen": 1776025600
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004697575757575758,
+      "loss": 2.7416,
+      "theoretical_loss": 3.4649209674660923,
+      "tokens_seen": 1776091136
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046973737373737375,
+      "loss": 2.716,
+      "theoretical_loss": 3.464910057161056,
+      "tokens_seen": 1776156672
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004697171717171717,
+      "loss": 2.775,
+      "theoretical_loss": 3.4648991473712893,
+      "tokens_seen": 1776222208
+    },
+    {
+      "epoch": 0.07,
+      "objective/train/docs_used": 979554,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.655245542526245,
+      "objective/train/theoretical_loss": 3.4648882380967505,
+      "objective/train/tokens_used": 135146976,
+      "theoretical_loss": 3.4648882380967505,
+      "tokens_seen": 1776287744
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004696969696969697,
+      "loss": 2.735,
+      "theoretical_loss": 3.4648882380967505,
+      "tokens_seen": 1776287744
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046967676767676773,
+      "loss": 2.7324,
+      "theoretical_loss": 3.464877329337395,
+      "tokens_seen": 1776353280
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004696565656565657,
+      "loss": 2.6467,
+      "theoretical_loss": 3.4648664210931797,
+      "tokens_seen": 1776418816
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046963636363636364,
+      "loss": 2.6666,
+      "theoretical_loss": 3.4648555133640624,
+      "tokens_seen": 1776484352
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004696161616161616,
+      "loss": 2.694,
+      "theoretical_loss": 3.464844606149998,
+      "tokens_seen": 1776549888
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004695959595959596,
+      "loss": 2.5421,
+      "theoretical_loss": 3.4648336994509448,
+      "tokens_seen": 1776615424
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004695757575757576,
+      "loss": 2.4765,
+      "theoretical_loss": 3.4648227932668583,
+      "tokens_seen": 1776680960
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004695555555555556,
+      "loss": 2.7016,
+      "theoretical_loss": 3.464811887597696,
+      "tokens_seen": 1776746496
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046953535353535354,
+      "loss": 2.3606,
+      "theoretical_loss": 3.464800982443414,
+      "tokens_seen": 1776812032
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004695151515151515,
+      "loss": 2.7383,
+      "theoretical_loss": 3.4647900778039693,
+      "tokens_seen": 1776877568
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004694949494949495,
+      "loss": 2.5567,
+      "theoretical_loss": 3.464779173679319,
+      "tokens_seen": 1776943104
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004694747474747475,
+      "loss": 2.6411,
+      "theoretical_loss": 3.4647682700694187,
+      "tokens_seen": 1777008640
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046945454545454547,
+      "loss": 2.4005,
+      "theoretical_loss": 3.464757366974226,
+      "tokens_seen": 1777074176
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046943434343434343,
+      "loss": 2.7334,
+      "theoretical_loss": 3.4647464643936976,
+      "tokens_seen": 1777139712
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004694141414141414,
+      "loss": 2.5828,
+      "theoretical_loss": 3.4647355623277902,
+      "tokens_seen": 1777205248
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046939393939393945,
+      "loss": 2.3715,
+      "theoretical_loss": 3.4647246607764606,
+      "tokens_seen": 1777270784
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004693737373737374,
+      "loss": 2.6547,
+      "theoretical_loss": 3.4647137597396647,
+      "tokens_seen": 1777336320
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046935353535353536,
+      "loss": 2.4004,
+      "theoretical_loss": 3.4647028592173603,
+      "tokens_seen": 1777401856
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004693333333333333,
+      "loss": 2.7303,
+      "theoretical_loss": 3.464691959209503,
+      "tokens_seen": 1777467392
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004693131313131313,
+      "loss": 2.5245,
+      "theoretical_loss": 3.4646810597160513,
+      "tokens_seen": 1777532928
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046929292929292934,
+      "loss": 2.6391,
+      "theoretical_loss": 3.46467016073696,
+      "tokens_seen": 1777598464
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004692727272727273,
+      "loss": 2.6976,
+      "theoretical_loss": 3.4646592622721872,
+      "tokens_seen": 1777664000
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046925252525252526,
+      "loss": 2.5603,
+      "theoretical_loss": 3.4646483643216897,
+      "tokens_seen": 1777729536
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004692323232323232,
+      "loss": 2.4618,
+      "theoretical_loss": 3.4646374668854234,
+      "tokens_seen": 1777795072
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004692121212121212,
+      "loss": 2.8756,
+      "theoretical_loss": 3.4646265699633454,
+      "tokens_seen": 1777860608
+    },
+    {
+      "epoch": 0.07,
+      "objective/train/docs_used": 979848,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.122527599334717,
+      "objective/train/theoretical_loss": 3.464615673555413,
+      "objective/train/tokens_used": 136785376,
+      "theoretical_loss": 3.464615673555413,
+      "tokens_seen": 1777926144
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046919191919191924,
+      "loss": 2.7605,
+      "theoretical_loss": 3.464615673555413,
+      "tokens_seen": 1777926144
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004691717171717172,
+      "loss": 2.6706,
+      "theoretical_loss": 3.4646047776615823,
+      "tokens_seen": 1777991680
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046915151515151515,
+      "loss": 2.5914,
+      "theoretical_loss": 3.4645938822818105,
+      "tokens_seen": 1778057216
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004691313131313131,
+      "loss": 2.8862,
+      "theoretical_loss": 3.4645829874160543,
+      "tokens_seen": 1778122752
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004691111111111111,
+      "loss": 2.6125,
+      "theoretical_loss": 3.4645720930642705,
+      "tokens_seen": 1778188288
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046909090909090913,
+      "loss": 2.6488,
+      "theoretical_loss": 3.4645611992264156,
+      "tokens_seen": 1778253824
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004690707070707071,
+      "loss": 2.5569,
+      "theoretical_loss": 3.4645503059024474,
+      "tokens_seen": 1778319360
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046905050505050504,
+      "loss": 2.4418,
+      "theoretical_loss": 3.4645394130923215,
+      "tokens_seen": 1778384896
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046903030303030305,
+      "loss": 2.6375,
+      "theoretical_loss": 3.4645285207959953,
+      "tokens_seen": 1778450432
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000469010101010101,
+      "loss": 2.5087,
+      "theoretical_loss": 3.464517629013426,
+      "tokens_seen": 1778515968
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000468989898989899,
+      "loss": 2.4809,
+      "theoretical_loss": 3.4645067377445695,
+      "tokens_seen": 1778581504
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000468969696969697,
+      "loss": 2.696,
+      "theoretical_loss": 3.464495846989384,
+      "tokens_seen": 1778647040
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046894949494949494,
+      "loss": 2.7445,
+      "theoretical_loss": 3.4644849567478246,
+      "tokens_seen": 1778712576
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046892929292929295,
+      "loss": 2.7525,
+      "theoretical_loss": 3.4644740670198497,
+      "tokens_seen": 1778778112
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004689090909090909,
+      "loss": 2.6765,
+      "theoretical_loss": 3.464463177805415,
+      "tokens_seen": 1778843648
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004688888888888889,
+      "loss": 2.6821,
+      "theoretical_loss": 3.4644522891044787,
+      "tokens_seen": 1778909184
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046886868686868687,
+      "loss": 2.8012,
+      "theoretical_loss": 3.4644414009169964,
+      "tokens_seen": 1778974720
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004688484848484849,
+      "loss": 2.573,
+      "theoretical_loss": 3.4644305132429256,
+      "tokens_seen": 1779040256
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046882828282828284,
+      "loss": 2.8178,
+      "theoretical_loss": 3.464419626082223,
+      "tokens_seen": 1779105792
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004688080808080808,
+      "loss": 2.7468,
+      "theoretical_loss": 3.4644087394348455,
+      "tokens_seen": 1779171328
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004687878787878788,
+      "loss": 2.5975,
+      "theoretical_loss": 3.46439785330075,
+      "tokens_seen": 1779236864
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046876767676767676,
+      "loss": 2.7253,
+      "theoretical_loss": 3.4643869676798937,
+      "tokens_seen": 1779302400
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004687474747474748,
+      "loss": 2.6003,
+      "theoretical_loss": 3.464376082572233,
+      "tokens_seen": 1779367936
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046872727272727273,
+      "loss": 2.6038,
+      "theoretical_loss": 3.464365197977725,
+      "tokens_seen": 1779433472
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004687070707070707,
+      "loss": 2.7135,
+      "theoretical_loss": 3.464354313896327,
+      "tokens_seen": 1779499008
+    },
+    {
+      "epoch": 0.07,
+      "objective/train/docs_used": 981168,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.4408981800079346,
+      "objective/train/theoretical_loss": 3.464343430327995,
+      "objective/train/tokens_used": 138423776,
+      "theoretical_loss": 3.464343430327995,
+      "tokens_seen": 1779564544
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004686868686868687,
+      "loss": 2.6659,
+      "theoretical_loss": 3.464343430327995,
+      "tokens_seen": 1779564544
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004686666666666667,
+      "loss": 2.545,
+      "theoretical_loss": 3.4643325472726865,
+      "tokens_seen": 1779630080
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046864646464646467,
+      "loss": 2.5786,
+      "theoretical_loss": 3.4643216647303583,
+      "tokens_seen": 1779695616
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004686262626262626,
+      "loss": 2.6326,
+      "theoretical_loss": 3.4643107827009674,
+      "tokens_seen": 1779761152
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004686060606060606,
+      "loss": 2.476,
+      "theoretical_loss": 3.4642999011844715,
+      "tokens_seen": 1779826688
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004685858585858586,
+      "loss": 2.5703,
+      "theoretical_loss": 3.464289020180826,
+      "tokens_seen": 1779892224
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004685656565656566,
+      "loss": 2.5252,
+      "theoretical_loss": 3.464278139689989,
+      "tokens_seen": 1779957760
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046854545454545456,
+      "loss": 2.653,
+      "theoretical_loss": 3.4642672597119173,
+      "tokens_seen": 1780023296
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004685252525252525,
+      "loss": 2.5044,
+      "theoretical_loss": 3.4642563802465673,
+      "tokens_seen": 1780088832
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004685050505050505,
+      "loss": 2.5998,
+      "theoretical_loss": 3.464245501293896,
+      "tokens_seen": 1780154368
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046848484848484854,
+      "loss": 2.5963,
+      "theoretical_loss": 3.464234622853861,
+      "tokens_seen": 1780219904
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004684646464646465,
+      "loss": 2.6996,
+      "theoretical_loss": 3.4642237449264193,
+      "tokens_seen": 1780285440
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046844444444444445,
+      "loss": 2.702,
+      "theoretical_loss": 3.4642128675115273,
+      "tokens_seen": 1780350976
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004684242424242424,
+      "loss": 2.7233,
+      "theoretical_loss": 3.4642019906091424,
+      "tokens_seen": 1780416512
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046840404040404037,
+      "loss": 2.7893,
+      "theoretical_loss": 3.464191114219221,
+      "tokens_seen": 1780482048
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046838383838383843,
+      "loss": 2.4928,
+      "theoretical_loss": 3.464180238341721,
+      "tokens_seen": 1780547584
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004683636363636364,
+      "loss": 2.7807,
+      "theoretical_loss": 3.4641693629765986,
+      "tokens_seen": 1780613120
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046834343434343435,
+      "loss": 2.4966,
+      "theoretical_loss": 3.464158488123811,
+      "tokens_seen": 1780678656
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004683232323232323,
+      "loss": 2.6617,
+      "theoretical_loss": 3.4641476137833154,
+      "tokens_seen": 1780744192
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046830303030303026,
+      "loss": 2.8356,
+      "theoretical_loss": 3.464136739955069,
+      "tokens_seen": 1780809728
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004682828282828283,
+      "loss": 2.577,
+      "theoretical_loss": 3.464125866639028,
+      "tokens_seen": 1780875264
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004682626262626263,
+      "loss": 2.6046,
+      "theoretical_loss": 3.4641149938351505,
+      "tokens_seen": 1780940800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046824242424242424,
+      "loss": 2.7569,
+      "theoretical_loss": 3.464104121543393,
+      "tokens_seen": 1781006336
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004682222222222222,
+      "loss": 2.6343,
+      "theoretical_loss": 3.4640932497637125,
+      "tokens_seen": 1781071872
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004682020202020202,
+      "loss": 2.9327,
+      "theoretical_loss": 3.464082378496066,
+      "tokens_seen": 1781137408
+    },
+    {
+      "epoch": 0.07,
+      "objective/train/docs_used": 982366,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5046448707580566,
+      "objective/train/theoretical_loss": 3.4640715077404103,
+      "objective/train/tokens_used": 140062176,
+      "theoretical_loss": 3.4640715077404103,
+      "tokens_seen": 1781202944
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004681818181818182,
+      "loss": 2.4701,
+      "theoretical_loss": 3.4640715077404103,
+      "tokens_seen": 1781202944
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004681616161616162,
+      "loss": 2.5085,
+      "theoretical_loss": 3.4640606374967033,
+      "tokens_seen": 1781268480
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046814141414141413,
+      "loss": 2.7434,
+      "theoretical_loss": 3.464049767764901,
+      "tokens_seen": 1781334016
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004681212121212121,
+      "loss": 2.681,
+      "theoretical_loss": 3.4640388985449615,
+      "tokens_seen": 1781399552
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046810101010101015,
+      "loss": 2.8738,
+      "theoretical_loss": 3.4640280298368413,
+      "tokens_seen": 1781465088
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004680808080808081,
+      "loss": 2.5953,
+      "theoretical_loss": 3.4640171616404976,
+      "tokens_seen": 1781530624
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046806060606060607,
+      "loss": 2.8082,
+      "theoretical_loss": 3.4640062939558867,
+      "tokens_seen": 1781596160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000468040404040404,
+      "loss": 2.6266,
+      "theoretical_loss": 3.4639954267829673,
+      "tokens_seen": 1781661696
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046802020202020203,
+      "loss": 2.7488,
+      "theoretical_loss": 3.4639845601216948,
+      "tokens_seen": 1781727232
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046800000000000005,
+      "loss": 2.6267,
+      "theoretical_loss": 3.4639736939720276,
+      "tokens_seen": 1781792768
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000467979797979798,
+      "loss": 2.8528,
+      "theoretical_loss": 3.4639628283339223,
+      "tokens_seen": 1781858304
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046795959595959596,
+      "loss": 2.4593,
+      "theoretical_loss": 3.463951963207336,
+      "tokens_seen": 1781923840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004679393939393939,
+      "loss": 2.7433,
+      "theoretical_loss": 3.463941098592225,
+      "tokens_seen": 1781989376
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046791919191919193,
+      "loss": 2.7575,
+      "theoretical_loss": 3.463930234488548,
+      "tokens_seen": 1782054912
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046789898989898994,
+      "loss": 2.8066,
+      "theoretical_loss": 3.4639193708962615,
+      "tokens_seen": 1782120448
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004678787878787879,
+      "loss": 2.6911,
+      "theoretical_loss": 3.4639085078153222,
+      "tokens_seen": 1782185984
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046785858585858585,
+      "loss": 2.5327,
+      "theoretical_loss": 3.4638976452456873,
+      "tokens_seen": 1782251520
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046783838383838386,
+      "loss": 2.713,
+      "theoretical_loss": 3.4638867831873137,
+      "tokens_seen": 1782317056
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004678181818181818,
+      "loss": 2.7197,
+      "theoretical_loss": 3.4638759216401596,
+      "tokens_seen": 1782382592
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046779797979797983,
+      "loss": 2.7073,
+      "theoretical_loss": 3.463865060604181,
+      "tokens_seen": 1782448128
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004677777777777778,
+      "loss": 2.5768,
+      "theoretical_loss": 3.4638542000793358,
+      "tokens_seen": 1782513664
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046775757575757574,
+      "loss": 2.7761,
+      "theoretical_loss": 3.463843340065581,
+      "tokens_seen": 1782579200
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046773737373737376,
+      "loss": 2.67,
+      "theoretical_loss": 3.463832480562873,
+      "tokens_seen": 1782644736
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004677171717171717,
+      "loss": 2.6575,
+      "theoretical_loss": 3.463821621571171,
+      "tokens_seen": 1782710272
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004676969696969697,
+      "loss": 2.5326,
+      "theoretical_loss": 3.4638107630904296,
+      "tokens_seen": 1782775808
+    },
+    {
+      "epoch": 0.07,
+      "objective/train/docs_used": 982914,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.4014761447906494,
+      "objective/train/theoretical_loss": 3.4637999051206076,
+      "objective/train/tokens_used": 141700576,
+      "theoretical_loss": 3.4637999051206076,
+      "tokens_seen": 1782841344
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004676767676767677,
+      "loss": 2.5757,
+      "theoretical_loss": 3.4637999051206076,
+      "tokens_seen": 1782841344
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004676565656565657,
+      "loss": 2.7185,
+      "theoretical_loss": 3.4637890476616615,
+      "tokens_seen": 1782906880
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046763636363636365,
+      "loss": 2.4657,
+      "theoretical_loss": 3.4637781907135485,
+      "tokens_seen": 1782972416
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004676161616161616,
+      "loss": 2.753,
+      "theoretical_loss": 3.4637673342762265,
+      "tokens_seen": 1783037952
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004675959595959596,
+      "loss": 2.6718,
+      "theoretical_loss": 3.463756478349652,
+      "tokens_seen": 1783103488
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004675757575757576,
+      "loss": 2.8216,
+      "theoretical_loss": 3.463745622933782,
+      "tokens_seen": 1783169024
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004675555555555556,
+      "loss": 2.5559,
+      "theoretical_loss": 3.4637347680285746,
+      "tokens_seen": 1783234560
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046753535353535354,
+      "loss": 2.7446,
+      "theoretical_loss": 3.463723913633987,
+      "tokens_seen": 1783300096
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004675151515151515,
+      "loss": 2.7602,
+      "theoretical_loss": 3.463713059749975,
+      "tokens_seen": 1783365632
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004674949494949495,
+      "loss": 2.5877,
+      "theoretical_loss": 3.4637022063764977,
+      "tokens_seen": 1783431168
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004674747474747475,
+      "loss": 2.7978,
+      "theoretical_loss": 3.4636913535135108,
+      "tokens_seen": 1783496704
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004674545454545455,
+      "loss": 2.547,
+      "theoretical_loss": 3.4636805011609724,
+      "tokens_seen": 1783562240
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046743434343434343,
+      "loss": 2.8632,
+      "theoretical_loss": 3.463669649318839,
+      "tokens_seen": 1783627776
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004674141414141414,
+      "loss": 2.57,
+      "theoretical_loss": 3.4636587979870685,
+      "tokens_seen": 1783693312
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004673939393939394,
+      "loss": 2.5764,
+      "theoretical_loss": 3.463647947165618,
+      "tokens_seen": 1783758848
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004673737373737374,
+      "loss": 2.9032,
+      "theoretical_loss": 3.4636370968544448,
+      "tokens_seen": 1783824384
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046735353535353537,
+      "loss": 2.7317,
+      "theoretical_loss": 3.463626247053506,
+      "tokens_seen": 1783889920
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004673333333333333,
+      "loss": 2.6338,
+      "theoretical_loss": 3.4636153977627586,
+      "tokens_seen": 1783955456
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004673131313131313,
+      "loss": 2.7822,
+      "theoretical_loss": 3.4636045489821607,
+      "tokens_seen": 1784020992
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046729292929292935,
+      "loss": 2.497,
+      "theoretical_loss": 3.4635937007116686,
+      "tokens_seen": 1784086528
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004672727272727273,
+      "loss": 2.509,
+      "theoretical_loss": 3.4635828529512405,
+      "tokens_seen": 1784152064
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046725252525252526,
+      "loss": 2.6947,
+      "theoretical_loss": 3.4635720057008332,
+      "tokens_seen": 1784217600
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004672323232323232,
+      "loss": 2.5064,
+      "theoretical_loss": 3.463561158960404,
+      "tokens_seen": 1784283136
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004672121212121212,
+      "loss": 2.3541,
+      "theoretical_loss": 3.4635503127299097,
+      "tokens_seen": 1784348672
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00046719191919191924,
+      "loss": 2.5577,
+      "theoretical_loss": 3.4635394670093085,
+      "tokens_seen": 1784414208
+    },
+    {
+      "epoch": 0.07,
+      "objective/train/docs_used": 984019,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.804720401763916,
+      "objective/train/theoretical_loss": 3.4635286217985572,
+      "objective/train/tokens_used": 143338976,
+      "theoretical_loss": 3.4635286217985572,
+      "tokens_seen": 1784479744
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0004671717171717172,
+      "loss": 2.8734,
+      "theoretical_loss": 3.4635286217985572,
+      "tokens_seen": 1784479744
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046715151515151516,
+      "loss": 2.841,
+      "theoretical_loss": 3.4635177770976133,
+      "tokens_seen": 1784545280
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004671313131313131,
+      "loss": 2.4586,
+      "theoretical_loss": 3.463506932906434,
+      "tokens_seen": 1784610816
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004671111111111111,
+      "loss": 2.3573,
+      "theoretical_loss": 3.463496089224977,
+      "tokens_seen": 1784676352
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046709090909090913,
+      "loss": 2.7561,
+      "theoretical_loss": 3.4634852460531986,
+      "tokens_seen": 1784741888
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004670707070707071,
+      "loss": 2.8304,
+      "theoretical_loss": 3.463474403391057,
+      "tokens_seen": 1784807424
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046705050505050505,
+      "loss": 2.566,
+      "theoretical_loss": 3.46346356123851,
+      "tokens_seen": 1784872960
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000467030303030303,
+      "loss": 2.7447,
+      "theoretical_loss": 3.4634527195955136,
+      "tokens_seen": 1784938496
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000467010101010101,
+      "loss": 2.6607,
+      "theoretical_loss": 3.463441878462026,
+      "tokens_seen": 1785004032
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000466989898989899,
+      "loss": 2.4519,
+      "theoretical_loss": 3.463431037838004,
+      "tokens_seen": 1785069568
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000466969696969697,
+      "loss": 2.7172,
+      "theoretical_loss": 3.463420197723406,
+      "tokens_seen": 1785135104
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046694949494949494,
+      "loss": 2.4123,
+      "theoretical_loss": 3.463409358118188,
+      "tokens_seen": 1785200640
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046692929292929295,
+      "loss": 2.7201,
+      "theoretical_loss": 3.4633985190223084,
+      "tokens_seen": 1785266176
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004669090909090909,
+      "loss": 2.7307,
+      "theoretical_loss": 3.4633876804357246,
+      "tokens_seen": 1785331712
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004668888888888889,
+      "loss": 2.5424,
+      "theoretical_loss": 3.463376842358393,
+      "tokens_seen": 1785397248
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004668686868686869,
+      "loss": 2.8041,
+      "theoretical_loss": 3.463366004790272,
+      "tokens_seen": 1785462784
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046684848484848483,
+      "loss": 2.7058,
+      "theoretical_loss": 3.4633551677313186,
+      "tokens_seen": 1785528320
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046682828282828284,
+      "loss": 2.6283,
+      "theoretical_loss": 3.46334433118149,
+      "tokens_seen": 1785593856
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004668080808080808,
+      "loss": 2.4156,
+      "theoretical_loss": 3.463333495140744,
+      "tokens_seen": 1785659392
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004667878787878788,
+      "loss": 2.4195,
+      "theoretical_loss": 3.4633226596090374,
+      "tokens_seen": 1785724928
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046676767676767677,
+      "loss": 2.4871,
+      "theoretical_loss": 3.463311824586328,
+      "tokens_seen": 1785790464
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004667474747474748,
+      "loss": 2.4098,
+      "theoretical_loss": 3.4633009900725735,
+      "tokens_seen": 1785856000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046672727272727274,
+      "loss": 2.8379,
+      "theoretical_loss": 3.463290156067731,
+      "tokens_seen": 1785921536
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004667070707070707,
+      "loss": 2.5884,
+      "theoretical_loss": 3.4632793225717573,
+      "tokens_seen": 1785987072
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004666868686868687,
+      "loss": 2.7167,
+      "theoretical_loss": 3.463268489584611,
+      "tokens_seen": 1786052608
+    },
+    {
+      "epoch": 0.08,
+      "objective/train/docs_used": 984652,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.4918596744537354,
+      "objective/train/theoretical_loss": 3.463257657106249,
+      "objective/train/tokens_used": 144977376,
+      "theoretical_loss": 3.463257657106249,
+      "tokens_seen": 1786118144
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046666666666666666,
+      "loss": 2.6791,
+      "theoretical_loss": 3.463257657106249,
+      "tokens_seen": 1786118144
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004666464646464647,
+      "loss": 2.5787,
+      "theoretical_loss": 3.463246825136628,
+      "tokens_seen": 1786183680
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046662626262626263,
+      "loss": 2.4831,
+      "theoretical_loss": 3.463235993675707,
+      "tokens_seen": 1786249216
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004666060606060606,
+      "loss": 2.6653,
+      "theoretical_loss": 3.4632251627234423,
+      "tokens_seen": 1786314752
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004665858585858586,
+      "loss": 2.604,
+      "theoretical_loss": 3.4632143322797915,
+      "tokens_seen": 1786380288
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004665656565656566,
+      "loss": 2.458,
+      "theoretical_loss": 3.4632035023447125,
+      "tokens_seen": 1786445824
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046654545454545457,
+      "loss": 2.6402,
+      "theoretical_loss": 3.4631926729181624,
+      "tokens_seen": 1786511360
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004665252525252525,
+      "loss": 2.5807,
+      "theoretical_loss": 3.4631818440000988,
+      "tokens_seen": 1786576896
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004665050505050505,
+      "loss": 2.7028,
+      "theoretical_loss": 3.4631710155904787,
+      "tokens_seen": 1786642432
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004664848484848485,
+      "loss": 2.6881,
+      "theoretical_loss": 3.4631601876892604,
+      "tokens_seen": 1786707968
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004664646464646465,
+      "loss": 2.6738,
+      "theoretical_loss": 3.4631493602964007,
+      "tokens_seen": 1786773504
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046644444444444446,
+      "loss": 2.5745,
+      "theoretical_loss": 3.4631385334118576,
+      "tokens_seen": 1786839040
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004664242424242424,
+      "loss": 2.6133,
+      "theoretical_loss": 3.4631277070355884,
+      "tokens_seen": 1786904576
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046640404040404037,
+      "loss": 2.6516,
+      "theoretical_loss": 3.463116881167551,
+      "tokens_seen": 1786970112
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004663838383838384,
+      "loss": 2.6627,
+      "theoretical_loss": 3.4631060558077014,
+      "tokens_seen": 1787035648
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004663636363636364,
+      "loss": 2.6789,
+      "theoretical_loss": 3.463095230955999,
+      "tokens_seen": 1787101184
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046634343434343435,
+      "loss": 2.7396,
+      "theoretical_loss": 3.4630844066124,
+      "tokens_seen": 1787166720
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004663232323232323,
+      "loss": 2.5693,
+      "theoretical_loss": 3.463073582776863,
+      "tokens_seen": 1787232256
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004663030303030303,
+      "loss": 2.8965,
+      "theoretical_loss": 3.463062759449344,
+      "tokens_seen": 1787297792
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046628282828282833,
+      "loss": 2.7113,
+      "theoretical_loss": 3.4630519366298023,
+      "tokens_seen": 1787363328
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004662626262626263,
+      "loss": 2.7535,
+      "theoretical_loss": 3.4630411143181945,
+      "tokens_seen": 1787428864
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046624242424242424,
+      "loss": 2.6914,
+      "theoretical_loss": 3.4630302925144782,
+      "tokens_seen": 1787494400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004662222222222222,
+      "loss": 2.4684,
+      "theoretical_loss": 3.463019471218611,
+      "tokens_seen": 1787559936
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004662020202020202,
+      "loss": 2.4305,
+      "theoretical_loss": 3.4630086504305497,
+      "tokens_seen": 1787625472
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004661818181818182,
+      "loss": 2.5705,
+      "theoretical_loss": 3.462997830150253,
+      "tokens_seen": 1787691008
+    },
+    {
+      "epoch": 0.08,
+      "objective/train/docs_used": 985931,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.233638286590576,
+      "objective/train/theoretical_loss": 3.4629870103776783,
+      "objective/train/tokens_used": 146615776,
+      "theoretical_loss": 3.4629870103776783,
+      "tokens_seen": 1787756544
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004661616161616162,
+      "loss": 2.5084,
+      "theoretical_loss": 3.4629870103776783,
+      "tokens_seen": 1787756544
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046614141414141414,
+      "loss": 2.4923,
+      "theoretical_loss": 3.462976191112783,
+      "tokens_seen": 1787822080
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004661212121212121,
+      "loss": 2.4434,
+      "theoretical_loss": 3.4629653723555247,
+      "tokens_seen": 1787887616
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046610101010101016,
+      "loss": 2.669,
+      "theoretical_loss": 3.4629545541058606,
+      "tokens_seen": 1787953152
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004660808080808081,
+      "loss": 2.6602,
+      "theoretical_loss": 3.4629437363637483,
+      "tokens_seen": 1788018688
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046606060606060607,
+      "loss": 2.6326,
+      "theoretical_loss": 3.462932919129146,
+      "tokens_seen": 1788084224
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046604040404040403,
+      "loss": 2.5248,
+      "theoretical_loss": 3.4629221024020107,
+      "tokens_seen": 1788149760
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000466020202020202,
+      "loss": 2.4801,
+      "theoretical_loss": 3.4629112861823,
+      "tokens_seen": 1788215296
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046600000000000005,
+      "loss": 2.6761,
+      "theoretical_loss": 3.4629004704699726,
+      "tokens_seen": 1788280832
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000465979797979798,
+      "loss": 2.6667,
+      "theoretical_loss": 3.462889655264984,
+      "tokens_seen": 1788346368
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046595959595959597,
+      "loss": 2.8018,
+      "theoretical_loss": 3.462878840567294,
+      "tokens_seen": 1788411904
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004659393939393939,
+      "loss": 2.7818,
+      "theoretical_loss": 3.4628680263768588,
+      "tokens_seen": 1788477440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046591919191919193,
+      "loss": 2.6761,
+      "theoretical_loss": 3.4628572126936366,
+      "tokens_seen": 1788542976
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046589898989898994,
+      "loss": 2.5054,
+      "theoretical_loss": 3.4628463995175847,
+      "tokens_seen": 1788608512
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004658787878787879,
+      "loss": 2.6396,
+      "theoretical_loss": 3.4628355868486613,
+      "tokens_seen": 1788674048
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046585858585858586,
+      "loss": 2.595,
+      "theoretical_loss": 3.4628247746868235,
+      "tokens_seen": 1788739584
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004658383838383838,
+      "loss": 2.6262,
+      "theoretical_loss": 3.462813963032029,
+      "tokens_seen": 1788805120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004658181818181818,
+      "loss": 2.6693,
+      "theoretical_loss": 3.4628031518842355,
+      "tokens_seen": 1788870656
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046579797979797984,
+      "loss": 2.5535,
+      "theoretical_loss": 3.462792341243401,
+      "tokens_seen": 1788936192
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004657777777777778,
+      "loss": 2.4625,
+      "theoretical_loss": 3.4627815311094823,
+      "tokens_seen": 1789001728
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046575757575757575,
+      "loss": 2.4874,
+      "theoretical_loss": 3.462770721482438,
+      "tokens_seen": 1789067264
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046573737373737376,
+      "loss": 2.3284,
+      "theoretical_loss": 3.4627599123622255,
+      "tokens_seen": 1789132800
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004657171717171717,
+      "loss": 2.6288,
+      "theoretical_loss": 3.462749103748802,
+      "tokens_seen": 1789198336
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046569696969696973,
+      "loss": 2.4654,
+      "theoretical_loss": 3.4627382956421258,
+      "tokens_seen": 1789263872
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004656767676767677,
+      "loss": 2.5071,
+      "theoretical_loss": 3.4627274880421544,
+      "tokens_seen": 1789329408
+    },
+    {
+      "epoch": 0.08,
+      "objective/train/docs_used": 986490,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.360241413116455,
+      "objective/train/theoretical_loss": 3.462716680948845,
+      "objective/train/tokens_used": 148254176,
+      "theoretical_loss": 3.462716680948845,
+      "tokens_seen": 1789394944
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046565656565656564,
+      "loss": 2.8045,
+      "theoretical_loss": 3.462716680948845,
+      "tokens_seen": 1789394944
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046563636363636365,
+      "loss": 2.7073,
+      "theoretical_loss": 3.462705874362156,
+      "tokens_seen": 1789460480
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004656161616161616,
+      "loss": 2.4757,
+      "theoretical_loss": 3.462695068282045,
+      "tokens_seen": 1789526016
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004655959595959596,
+      "loss": 2.5969,
+      "theoretical_loss": 3.4626842627084695,
+      "tokens_seen": 1789591552
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004655757575757576,
+      "loss": 2.7488,
+      "theoretical_loss": 3.4626734576413867,
+      "tokens_seen": 1789657088
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004655555555555556,
+      "loss": 2.6564,
+      "theoretical_loss": 3.4626626530807547,
+      "tokens_seen": 1789722624
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046553535353535355,
+      "loss": 2.5831,
+      "theoretical_loss": 3.462651849026532,
+      "tokens_seen": 1789788160
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004655151515151515,
+      "loss": 2.6404,
+      "theoretical_loss": 3.4626410454786756,
+      "tokens_seen": 1789853696
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004654949494949495,
+      "loss": 2.4886,
+      "theoretical_loss": 3.462630242437143,
+      "tokens_seen": 1789919232
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046547474747474747,
+      "loss": 2.7851,
+      "theoretical_loss": 3.462619439901892,
+      "tokens_seen": 1789984768
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004654545454545455,
+      "loss": 2.4352,
+      "theoretical_loss": 3.462608637872881,
+      "tokens_seen": 1790050304
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046543434343434344,
+      "loss": 2.5899,
+      "theoretical_loss": 3.462597836350067,
+      "tokens_seen": 1790115840
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004654141414141414,
+      "loss": 2.4442,
+      "theoretical_loss": 3.4625870353334083,
+      "tokens_seen": 1790181376
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004653939393939394,
+      "loss": 2.4023,
+      "theoretical_loss": 3.462576234822863,
+      "tokens_seen": 1790246912
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004653737373737374,
+      "loss": 2.5182,
+      "theoretical_loss": 3.462565434818387,
+      "tokens_seen": 1790312448
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004653535353535354,
+      "loss": 2.6444,
+      "theoretical_loss": 3.46255463531994,
+      "tokens_seen": 1790377984
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046533333333333333,
+      "loss": 2.5226,
+      "theoretical_loss": 3.4625438363274794,
+      "tokens_seen": 1790443520
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004653131313131313,
+      "loss": 2.7895,
+      "theoretical_loss": 3.462533037840962,
+      "tokens_seen": 1790509056
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004652929292929293,
+      "loss": 2.5776,
+      "theoretical_loss": 3.4625222398603466,
+      "tokens_seen": 1790574592
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004652727272727273,
+      "loss": 2.5917,
+      "theoretical_loss": 3.4625114423855905,
+      "tokens_seen": 1790640128
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046525252525252527,
+      "loss": 2.7743,
+      "theoretical_loss": 3.4625006454166516,
+      "tokens_seen": 1790705664
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004652323232323232,
+      "loss": 2.7714,
+      "theoretical_loss": 3.4624898489534877,
+      "tokens_seen": 1790771200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004652121212121212,
+      "loss": 2.5768,
+      "theoretical_loss": 3.4624790529960574,
+      "tokens_seen": 1790836736
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046519191919191925,
+      "loss": 2.751,
+      "theoretical_loss": 3.462468257544317,
+      "tokens_seen": 1790902272
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004651717171717172,
+      "loss": 2.5792,
+      "theoretical_loss": 3.4624574625982247,
+      "tokens_seen": 1790967808
+    },
+    {
+      "epoch": 0.08,
+      "objective/train/docs_used": 987679,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5805740356445312,
+      "objective/train/theoretical_loss": 3.462446668157739,
+      "objective/train/tokens_used": 149892576,
+      "theoretical_loss": 3.462446668157739,
+      "tokens_seen": 1791033344
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046515151515151516,
+      "loss": 2.4389,
+      "theoretical_loss": 3.462446668157739,
+      "tokens_seen": 1791033344
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004651313131313131,
+      "loss": 2.5047,
+      "theoretical_loss": 3.4624358742228174,
+      "tokens_seen": 1791098880
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004651111111111111,
+      "loss": 2.6406,
+      "theoretical_loss": 3.462425080793418,
+      "tokens_seen": 1791164416
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046509090909090914,
+      "loss": 2.5138,
+      "theoretical_loss": 3.462414287869498,
+      "tokens_seen": 1791229952
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004650707070707071,
+      "loss": 2.7199,
+      "theoretical_loss": 3.4624034954510154,
+      "tokens_seen": 1791295488
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046505050505050505,
+      "loss": 2.5726,
+      "theoretical_loss": 3.4623927035379283,
+      "tokens_seen": 1791361024
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000465030303030303,
+      "loss": 2.5712,
+      "theoretical_loss": 3.462381912130194,
+      "tokens_seen": 1791426560
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046501010101010097,
+      "loss": 2.4468,
+      "theoretical_loss": 3.462371121227772,
+      "tokens_seen": 1791492096
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046498989898989903,
+      "loss": 2.628,
+      "theoretical_loss": 3.462360330830618,
+      "tokens_seen": 1791557632
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000464969696969697,
+      "loss": 2.8937,
+      "theoretical_loss": 3.462349540938691,
+      "tokens_seen": 1791623168
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046494949494949495,
+      "loss": 2.5733,
+      "theoretical_loss": 3.4623387515519486,
+      "tokens_seen": 1791688704
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004649292929292929,
+      "loss": 2.7154,
+      "theoretical_loss": 3.462327962670349,
+      "tokens_seen": 1791754240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004649090909090909,
+      "loss": 2.3878,
+      "theoretical_loss": 3.462317174293849,
+      "tokens_seen": 1791819776
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004648888888888889,
+      "loss": 2.773,
+      "theoretical_loss": 3.462306386422408,
+      "tokens_seen": 1791885312
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004648686868686869,
+      "loss": 2.6715,
+      "theoretical_loss": 3.462295599055983,
+      "tokens_seen": 1791950848
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046484848484848484,
+      "loss": 2.6902,
+      "theoretical_loss": 3.4622848121945324,
+      "tokens_seen": 1792016384
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004648282828282828,
+      "loss": 2.551,
+      "theoretical_loss": 3.4622740258380134,
+      "tokens_seen": 1792081920
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004648080808080808,
+      "loss": 2.4135,
+      "theoretical_loss": 3.4622632399863846,
+      "tokens_seen": 1792147456
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004647878787878788,
+      "loss": 2.5175,
+      "theoretical_loss": 3.4622524546396027,
+      "tokens_seen": 1792212992
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004647676767676768,
+      "loss": 2.7417,
+      "theoretical_loss": 3.4622416697976277,
+      "tokens_seen": 1792278528
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046474747474747473,
+      "loss": 2.4755,
+      "theoretical_loss": 3.4622308854604156,
+      "tokens_seen": 1792344064
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046472727272727274,
+      "loss": 2.5535,
+      "theoretical_loss": 3.462220101627925,
+      "tokens_seen": 1792409600
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004647070707070707,
+      "loss": 2.6122,
+      "theoretical_loss": 3.462209318300114,
+      "tokens_seen": 1792475136
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004646868686868687,
+      "loss": 2.7892,
+      "theoretical_loss": 3.46219853547694,
+      "tokens_seen": 1792540672
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046466666666666667,
+      "loss": 2.6861,
+      "theoretical_loss": 3.4621877531583616,
+      "tokens_seen": 1792606208
+    },
+    {
+      "epoch": 0.08,
+      "objective/train/docs_used": 988330,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5130808353424072,
+      "objective/train/theoretical_loss": 3.462176971344337,
+      "objective/train/tokens_used": 151530976,
+      "theoretical_loss": 3.462176971344337,
+      "tokens_seen": 1792671744
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004646464646464646,
+      "loss": 2.5728,
+      "theoretical_loss": 3.462176971344337,
+      "tokens_seen": 1792671744
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046462626262626264,
+      "loss": 2.4717,
+      "theoretical_loss": 3.4621661900348224,
+      "tokens_seen": 1792737280
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046460606060606065,
+      "loss": 2.6032,
+      "theoretical_loss": 3.4621554092297777,
+      "tokens_seen": 1792802816
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004645858585858586,
+      "loss": 2.6182,
+      "theoretical_loss": 3.46214462892916,
+      "tokens_seen": 1792868352
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046456565656565656,
+      "loss": 2.7285,
+      "theoretical_loss": 3.462133849132927,
+      "tokens_seen": 1792933888
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046454545454545457,
+      "loss": 2.4959,
+      "theoretical_loss": 3.462123069841038,
+      "tokens_seen": 1792999424
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046452525252525253,
+      "loss": 2.5876,
+      "theoretical_loss": 3.462112291053449,
+      "tokens_seen": 1793064960
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046450505050505054,
+      "loss": 2.52,
+      "theoretical_loss": 3.4621015127701193,
+      "tokens_seen": 1793130496
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004644848484848485,
+      "loss": 2.5387,
+      "theoretical_loss": 3.4620907349910066,
+      "tokens_seen": 1793196032
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046446464646464645,
+      "loss": 2.4564,
+      "theoretical_loss": 3.462079957716069,
+      "tokens_seen": 1793261568
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046444444444444446,
+      "loss": 2.7585,
+      "theoretical_loss": 3.4620691809452637,
+      "tokens_seen": 1793327104
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004644242424242424,
+      "loss": 2.5675,
+      "theoretical_loss": 3.46205840467855,
+      "tokens_seen": 1793392640
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046440404040404043,
+      "loss": 2.6589,
+      "theoretical_loss": 3.462047628915885,
+      "tokens_seen": 1793458176
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004643838383838384,
+      "loss": 2.5341,
+      "theoretical_loss": 3.462036853657227,
+      "tokens_seen": 1793523712
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004643636363636364,
+      "loss": 2.81,
+      "theoretical_loss": 3.4620260789025337,
+      "tokens_seen": 1793589248
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046434343434343436,
+      "loss": 2.6829,
+      "theoretical_loss": 3.4620153046517634,
+      "tokens_seen": 1793654784
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004643232323232323,
+      "loss": 2.7504,
+      "theoretical_loss": 3.462004530904874,
+      "tokens_seen": 1793720320
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004643030303030303,
+      "loss": 2.6757,
+      "theoretical_loss": 3.461993757661824,
+      "tokens_seen": 1793785856
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004642828282828283,
+      "loss": 2.5854,
+      "theoretical_loss": 3.461982984922571,
+      "tokens_seen": 1793851392
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004642626262626263,
+      "loss": 2.7475,
+      "theoretical_loss": 3.4619722126870727,
+      "tokens_seen": 1793916928
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046424242424242425,
+      "loss": 2.6829,
+      "theoretical_loss": 3.461961440955288,
+      "tokens_seen": 1793982464
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004642222222222222,
+      "loss": 2.5497,
+      "theoretical_loss": 3.461950669727174,
+      "tokens_seen": 1794048000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004642020202020202,
+      "loss": 2.5674,
+      "theoretical_loss": 3.4619398990026893,
+      "tokens_seen": 1794113536
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046418181818181823,
+      "loss": 2.6553,
+      "theoretical_loss": 3.461929128781792,
+      "tokens_seen": 1794179072
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004641616161616162,
+      "loss": 2.7199,
+      "theoretical_loss": 3.4619183590644402,
+      "tokens_seen": 1794244608
+    },
+    {
+      "epoch": 0.08,
+      "objective/train/docs_used": 989671,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.687277317047119,
+      "objective/train/theoretical_loss": 3.4619075898505915,
+      "objective/train/tokens_used": 153169376,
+      "theoretical_loss": 3.4619075898505915,
+      "tokens_seen": 1794310144
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046414141414141414,
+      "loss": 2.7491,
+      "theoretical_loss": 3.4619075898505915,
+      "tokens_seen": 1794310144
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004641212121212121,
+      "loss": 2.6619,
+      "theoretical_loss": 3.461896821140204,
+      "tokens_seen": 1794375680
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004641010101010101,
+      "loss": 2.6628,
+      "theoretical_loss": 3.4618860529332363,
+      "tokens_seen": 1794441216
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004640808080808081,
+      "loss": 2.4893,
+      "theoretical_loss": 3.461875285229646,
+      "tokens_seen": 1794506752
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004640606060606061,
+      "loss": 2.6323,
+      "theoretical_loss": 3.461864518029392,
+      "tokens_seen": 1794572288
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046404040404040403,
+      "loss": 2.7328,
+      "theoretical_loss": 3.4618537513324315,
+      "tokens_seen": 1794637824
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000464020202020202,
+      "loss": 2.6513,
+      "theoretical_loss": 3.4618429851387233,
+      "tokens_seen": 1794703360
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046400000000000006,
+      "loss": 2.5925,
+      "theoretical_loss": 3.4618322194482243,
+      "tokens_seen": 1794768896
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000463979797979798,
+      "loss": 2.4511,
+      "theoretical_loss": 3.4618214542608943,
+      "tokens_seen": 1794834432
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046395959595959597,
+      "loss": 2.7825,
+      "theoretical_loss": 3.46181068957669,
+      "tokens_seen": 1794899968
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046393939393939393,
+      "loss": 2.6791,
+      "theoretical_loss": 3.46179992539557,
+      "tokens_seen": 1794965504
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004639191919191919,
+      "loss": 2.6882,
+      "theoretical_loss": 3.4617891617174923,
+      "tokens_seen": 1795031040
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046389898989898995,
+      "loss": 2.7291,
+      "theoretical_loss": 3.4617783985424158,
+      "tokens_seen": 1795096576
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004638787878787879,
+      "loss": 2.7468,
+      "theoretical_loss": 3.4617676358702973,
+      "tokens_seen": 1795162112
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046385858585858586,
+      "loss": 2.6849,
+      "theoretical_loss": 3.461756873701096,
+      "tokens_seen": 1795227648
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004638383838383838,
+      "loss": 2.5844,
+      "theoretical_loss": 3.4617461120347697,
+      "tokens_seen": 1795293184
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046381818181818183,
+      "loss": 2.5013,
+      "theoretical_loss": 3.461735350871277,
+      "tokens_seen": 1795358720
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046379797979797984,
+      "loss": 2.564,
+      "theoretical_loss": 3.461724590210575,
+      "tokens_seen": 1795424256
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004637777777777778,
+      "loss": 2.6563,
+      "theoretical_loss": 3.4617138300526222,
+      "tokens_seen": 1795489792
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046375757575757576,
+      "loss": 2.4432,
+      "theoretical_loss": 3.4617030703973777,
+      "tokens_seen": 1795555328
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004637373737373737,
+      "loss": 2.7619,
+      "theoretical_loss": 3.461692311244798,
+      "tokens_seen": 1795620864
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004637171717171717,
+      "loss": 2.8329,
+      "theoretical_loss": 3.4616815525948432,
+      "tokens_seen": 1795686400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046369696969696974,
+      "loss": 2.434,
+      "theoretical_loss": 3.4616707944474703,
+      "tokens_seen": 1795751936
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004636767676767677,
+      "loss": 2.5695,
+      "theoretical_loss": 3.4616600368026376,
+      "tokens_seen": 1795817472
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046365656565656565,
+      "loss": 2.3648,
+      "theoretical_loss": 3.4616492796603033,
+      "tokens_seen": 1795883008
+    },
+    {
+      "epoch": 0.08,
+      "objective/train/docs_used": 990185,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.496757984161377,
+      "objective/train/theoretical_loss": 3.461638523020426,
+      "objective/train/tokens_used": 154807776,
+      "theoretical_loss": 3.461638523020426,
+      "tokens_seen": 1795948544
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046363636363636366,
+      "loss": 2.5464,
+      "theoretical_loss": 3.461638523020426,
+      "tokens_seen": 1795948544
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004636161616161616,
+      "loss": 2.6964,
+      "theoretical_loss": 3.4616277668829634,
+      "tokens_seen": 1796014080
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046359595959595963,
+      "loss": 2.5699,
+      "theoretical_loss": 3.461617011247874,
+      "tokens_seen": 1796079616
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004635757575757576,
+      "loss": 2.6995,
+      "theoretical_loss": 3.461606256115116,
+      "tokens_seen": 1796145152
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046355555555555554,
+      "loss": 2.5355,
+      "theoretical_loss": 3.461595501484647,
+      "tokens_seen": 1796210688
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046353535353535355,
+      "loss": 2.4066,
+      "theoretical_loss": 3.4615847473564263,
+      "tokens_seen": 1796276224
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004635151515151515,
+      "loss": 2.573,
+      "theoretical_loss": 3.4615739937304113,
+      "tokens_seen": 1796341760
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004634949494949495,
+      "loss": 2.5721,
+      "theoretical_loss": 3.4615632406065604,
+      "tokens_seen": 1796407296
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004634747474747475,
+      "loss": 2.782,
+      "theoretical_loss": 3.461552487984832,
+      "tokens_seen": 1796472832
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004634545454545455,
+      "loss": 2.3226,
+      "theoretical_loss": 3.4615417358651843,
+      "tokens_seen": 1796538368
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046343434343434345,
+      "loss": 2.6606,
+      "theoretical_loss": 3.461530984247575,
+      "tokens_seen": 1796603904
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004634141414141414,
+      "loss": 2.6738,
+      "theoretical_loss": 3.4615202331319637,
+      "tokens_seen": 1796669440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004633939393939394,
+      "loss": 2.6571,
+      "theoretical_loss": 3.4615094825183066,
+      "tokens_seen": 1796734976
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046337373737373737,
+      "loss": 2.5698,
+      "theoretical_loss": 3.4614987324065645,
+      "tokens_seen": 1796800512
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004633535353535354,
+      "loss": 2.6218,
+      "theoretical_loss": 3.4614879827966933,
+      "tokens_seen": 1796866048
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046333333333333334,
+      "loss": 2.6216,
+      "theoretical_loss": 3.461477233688653,
+      "tokens_seen": 1796931584
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004633131313131313,
+      "loss": 2.6791,
+      "theoretical_loss": 3.4614664850824,
+      "tokens_seen": 1796997120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004632929292929293,
+      "loss": 2.6627,
+      "theoretical_loss": 3.4614557369778947,
+      "tokens_seen": 1797062656
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004632727272727273,
+      "loss": 2.7609,
+      "theoretical_loss": 3.461444989375094,
+      "tokens_seen": 1797128192
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004632525252525253,
+      "loss": 2.689,
+      "theoretical_loss": 3.461434242273957,
+      "tokens_seen": 1797193728
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046323232323232323,
+      "loss": 2.352,
+      "theoretical_loss": 3.461423495674441,
+      "tokens_seen": 1797259264
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004632121212121212,
+      "loss": 2.7198,
+      "theoretical_loss": 3.461412749576505,
+      "tokens_seen": 1797324800
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004631919191919192,
+      "loss": 2.5821,
+      "theoretical_loss": 3.4614020039801074,
+      "tokens_seen": 1797390336
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004631717171717172,
+      "loss": 2.5262,
+      "theoretical_loss": 3.4613912588852065,
+      "tokens_seen": 1797455872
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046315151515151517,
+      "loss": 2.5769,
+      "theoretical_loss": 3.46138051429176,
+      "tokens_seen": 1797521408
+    },
+    {
+      "epoch": 0.08,
+      "objective/train/docs_used": 991489,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6060218811035156,
+      "objective/train/theoretical_loss": 3.4613697701997266,
+      "objective/train/tokens_used": 156446176,
+      "theoretical_loss": 3.4613697701997266,
+      "tokens_seen": 1797586944
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004631313131313131,
+      "loss": 2.5123,
+      "theoretical_loss": 3.4613697701997266,
+      "tokens_seen": 1797586944
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004631111111111111,
+      "loss": 2.7132,
+      "theoretical_loss": 3.461359026609065,
+      "tokens_seen": 1797652480
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004630909090909091,
+      "loss": 2.5873,
+      "theoretical_loss": 3.4613482835197322,
+      "tokens_seen": 1797718016
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004630707070707071,
+      "loss": 2.737,
+      "theoretical_loss": 3.461337540931688,
+      "tokens_seen": 1797783552
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046305050505050506,
+      "loss": 2.5273,
+      "theoretical_loss": 3.4613267988448913,
+      "tokens_seen": 1797849088
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000463030303030303,
+      "loss": 2.561,
+      "theoretical_loss": 3.4613160572592983,
+      "tokens_seen": 1797914624
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046301010101010097,
+      "loss": 2.7165,
+      "theoretical_loss": 3.4613053161748684,
+      "tokens_seen": 1797980160
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046298989898989904,
+      "loss": 2.7665,
+      "theoretical_loss": 3.46129457559156,
+      "tokens_seen": 1798045696
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000462969696969697,
+      "loss": 2.846,
+      "theoretical_loss": 3.461283835509332,
+      "tokens_seen": 1798111232
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046294949494949495,
+      "loss": 2.7031,
+      "theoretical_loss": 3.4612730959281413,
+      "tokens_seen": 1798176768
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004629292929292929,
+      "loss": 2.7652,
+      "theoretical_loss": 3.4612623568479477,
+      "tokens_seen": 1798242304
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004629090909090909,
+      "loss": 2.5421,
+      "theoretical_loss": 3.461251618268709,
+      "tokens_seen": 1798307840
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046288888888888893,
+      "loss": 2.7416,
+      "theoretical_loss": 3.4612408801903833,
+      "tokens_seen": 1798373376
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004628686868686869,
+      "loss": 2.6124,
+      "theoretical_loss": 3.4612301426129295,
+      "tokens_seen": 1798438912
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046284848484848484,
+      "loss": 2.8337,
+      "theoretical_loss": 3.4612194055363057,
+      "tokens_seen": 1798504448
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004628282828282828,
+      "loss": 2.5391,
+      "theoretical_loss": 3.4612086689604706,
+      "tokens_seen": 1798569984
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046280808080808087,
+      "loss": 2.6909,
+      "theoretical_loss": 3.461197932885382,
+      "tokens_seen": 1798635520
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004627878787878788,
+      "loss": 2.4735,
+      "theoretical_loss": 3.461187197310999,
+      "tokens_seen": 1798701056
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004627676767676768,
+      "loss": 2.4016,
+      "theoretical_loss": 3.4611764622372796,
+      "tokens_seen": 1798766592
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046274747474747474,
+      "loss": 2.7274,
+      "theoretical_loss": 3.4611657276641816,
+      "tokens_seen": 1798832128
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004627272727272727,
+      "loss": 2.4681,
+      "theoretical_loss": 3.4611549935916646,
+      "tokens_seen": 1798897664
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046270707070707076,
+      "loss": 2.6583,
+      "theoretical_loss": 3.4611442600196867,
+      "tokens_seen": 1798963200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004626868686868687,
+      "loss": 2.5414,
+      "theoretical_loss": 3.461133526948206,
+      "tokens_seen": 1799028736
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004626666666666667,
+      "loss": 2.7138,
+      "theoretical_loss": 3.4611227943771814,
+      "tokens_seen": 1799094272
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046264646464646463,
+      "loss": 2.6,
+      "theoretical_loss": 3.4611120623065705,
+      "tokens_seen": 1799159808
+    },
+    {
+      "epoch": 0.08,
+      "objective/train/docs_used": 992118,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7015321254730225,
+      "objective/train/theoretical_loss": 3.4611013307363323,
+      "objective/train/tokens_used": 158084576,
+      "theoretical_loss": 3.4611013307363323,
+      "tokens_seen": 1799225344
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046262626262626264,
+      "loss": 2.5245,
+      "theoretical_loss": 3.4611013307363323,
+      "tokens_seen": 1799225344
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046260606060606065,
+      "loss": 2.5428,
+      "theoretical_loss": 3.4610905996664254,
+      "tokens_seen": 1799290880
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004625858585858586,
+      "loss": 2.9118,
+      "theoretical_loss": 3.4610798690968076,
+      "tokens_seen": 1799356416
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046256565656565657,
+      "loss": 2.7124,
+      "theoretical_loss": 3.461069139027438,
+      "tokens_seen": 1799421952
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004625454545454545,
+      "loss": 2.6946,
+      "theoretical_loss": 3.461058409458275,
+      "tokens_seen": 1799487488
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046252525252525253,
+      "loss": 2.7679,
+      "theoretical_loss": 3.461047680389277,
+      "tokens_seen": 1799553024
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046250505050505054,
+      "loss": 2.8121,
+      "theoretical_loss": 3.461036951820402,
+      "tokens_seen": 1799618560
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004624848484848485,
+      "loss": 2.5779,
+      "theoretical_loss": 3.4610262237516096,
+      "tokens_seen": 1799684096
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046246464646464646,
+      "loss": 2.6062,
+      "theoretical_loss": 3.4610154961828568,
+      "tokens_seen": 1799749632
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046244444444444447,
+      "loss": 2.6283,
+      "theoretical_loss": 3.461004769114103,
+      "tokens_seen": 1799815168
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004624242424242424,
+      "loss": 2.6631,
+      "theoretical_loss": 3.4609940425453067,
+      "tokens_seen": 1799880704
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046240404040404044,
+      "loss": 2.6806,
+      "theoretical_loss": 3.460983316476426,
+      "tokens_seen": 1799946240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004623838383838384,
+      "loss": 2.7087,
+      "theoretical_loss": 3.46097259090742,
+      "tokens_seen": 1800011776
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046236363636363635,
+      "loss": 2.4683,
+      "theoretical_loss": 3.4609618658382466,
+      "tokens_seen": 1800077312
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046234343434343436,
+      "loss": 2.8184,
+      "theoretical_loss": 3.4609511412688647,
+      "tokens_seen": 1800142848
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004623232323232323,
+      "loss": 2.5615,
+      "theoretical_loss": 3.4609404171992324,
+      "tokens_seen": 1800208384
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046230303030303033,
+      "loss": 2.5703,
+      "theoretical_loss": 3.460929693629309,
+      "tokens_seen": 1800273920
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004622828282828283,
+      "loss": 2.5853,
+      "theoretical_loss": 3.4609189705590513,
+      "tokens_seen": 1800339456
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004622626262626263,
+      "loss": 2.8325,
+      "theoretical_loss": 3.46090824798842,
+      "tokens_seen": 1800404992
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046224242424242425,
+      "loss": 2.4077,
+      "theoretical_loss": 3.460897525917373,
+      "tokens_seen": 1800470528
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004622222222222222,
+      "loss": 2.45,
+      "theoretical_loss": 3.4608868043458676,
+      "tokens_seen": 1800536064
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004622020202020202,
+      "loss": 2.7897,
+      "theoretical_loss": 3.4608760832738636,
+      "tokens_seen": 1800601600
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004621818181818182,
+      "loss": 2.7197,
+      "theoretical_loss": 3.4608653627013193,
+      "tokens_seen": 1800667136
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0004621616161616162,
+      "loss": 2.5448,
+      "theoretical_loss": 3.4608546426281928,
+      "tokens_seen": 1800732672
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00046214141414141415,
+      "loss": 2.6301,
+      "theoretical_loss": 3.460843923054443,
+      "tokens_seen": 1800798208
+    },
+    {
+      "epoch": 0.08,
+      "objective/train/docs_used": 993246,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.732495069503784,
+      "objective/train/theoretical_loss": 3.4608332039800294,
+      "objective/train/tokens_used": 159722976,
+      "theoretical_loss": 3.4608332039800294,
+      "tokens_seen": 1800863744
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004621212121212121,
+      "loss": 2.7341,
+      "theoretical_loss": 3.4608332039800294,
+      "tokens_seen": 1800863744
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004621010101010101,
+      "loss": 2.6709,
+      "theoretical_loss": 3.460822485404909,
+      "tokens_seen": 1800929280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004620808080808081,
+      "loss": 2.6878,
+      "theoretical_loss": 3.460811767329041,
+      "tokens_seen": 1800994816
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004620606060606061,
+      "loss": 2.8197,
+      "theoretical_loss": 3.460801049752384,
+      "tokens_seen": 1801060352
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046204040404040404,
+      "loss": 2.4737,
+      "theoretical_loss": 3.4607903326748968,
+      "tokens_seen": 1801125888
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000462020202020202,
+      "loss": 2.6015,
+      "theoretical_loss": 3.4607796160965374,
+      "tokens_seen": 1801191424
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000462,
+      "loss": 2.7105,
+      "theoretical_loss": 3.460768900017265,
+      "tokens_seen": 1801256960
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000461979797979798,
+      "loss": 2.7153,
+      "theoretical_loss": 3.4607581844370383,
+      "tokens_seen": 1801322496
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000461959595959596,
+      "loss": 2.7215,
+      "theoretical_loss": 3.460747469355815,
+      "tokens_seen": 1801388032
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046193939393939393,
+      "loss": 2.7523,
+      "theoretical_loss": 3.4607367547735546,
+      "tokens_seen": 1801453568
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004619191919191919,
+      "loss": 2.6356,
+      "theoretical_loss": 3.460726040690215,
+      "tokens_seen": 1801519104
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046189898989898996,
+      "loss": 2.8216,
+      "theoretical_loss": 3.4607153271057554,
+      "tokens_seen": 1801584640
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004618787878787879,
+      "loss": 2.6657,
+      "theoretical_loss": 3.460704614020135,
+      "tokens_seen": 1801650176
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046185858585858587,
+      "loss": 2.5136,
+      "theoretical_loss": 3.460693901433311,
+      "tokens_seen": 1801715712
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004618383838383838,
+      "loss": 2.7718,
+      "theoretical_loss": 3.4606831893452425,
+      "tokens_seen": 1801781248
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004618181818181818,
+      "loss": 2.645,
+      "theoretical_loss": 3.460672477755889,
+      "tokens_seen": 1801846784
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046179797979797985,
+      "loss": 2.637,
+      "theoretical_loss": 3.4606617666652078,
+      "tokens_seen": 1801912320
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004617777777777778,
+      "loss": 2.6257,
+      "theoretical_loss": 3.4606510560731585,
+      "tokens_seen": 1801977856
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046175757575757576,
+      "loss": 2.5434,
+      "theoretical_loss": 3.4606403459796997,
+      "tokens_seen": 1802043392
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004617373737373737,
+      "loss": 2.5597,
+      "theoretical_loss": 3.4606296363847897,
+      "tokens_seen": 1802108928
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004617171717171717,
+      "loss": 2.4782,
+      "theoretical_loss": 3.4606189272883876,
+      "tokens_seen": 1802174464
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046169696969696974,
+      "loss": 2.5543,
+      "theoretical_loss": 3.460608218690451,
+      "tokens_seen": 1802240000
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004616767676767677,
+      "loss": 2.6303,
+      "theoretical_loss": 3.46059751059094,
+      "tokens_seen": 1802305536
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046165656565656565,
+      "loss": 2.3844,
+      "theoretical_loss": 3.4605868029898126,
+      "tokens_seen": 1802371072
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004616363636363636,
+      "loss": 2.6324,
+      "theoretical_loss": 3.4605760958870277,
+      "tokens_seen": 1802436608
+    },
+    {
+      "epoch": 0.09,
+      "objective/train/docs_used": 994266,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.712080717086792,
+      "objective/train/theoretical_loss": 3.460565389282543,
+      "objective/train/tokens_used": 161361376,
+      "theoretical_loss": 3.460565389282543,
+      "tokens_seen": 1802502144
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004616161616161616,
+      "loss": 2.6,
+      "theoretical_loss": 3.460565389282543,
+      "tokens_seen": 1802502144
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046159595959595963,
+      "loss": 2.8756,
+      "theoretical_loss": 3.4605546831763188,
+      "tokens_seen": 1802567680
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004615757575757576,
+      "loss": 2.7354,
+      "theoretical_loss": 3.460543977568313,
+      "tokens_seen": 1802633216
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046155555555555555,
+      "loss": 2.6013,
+      "theoretical_loss": 3.460533272458484,
+      "tokens_seen": 1802698752
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004615353535353535,
+      "loss": 2.8307,
+      "theoretical_loss": 3.4605225678467906,
+      "tokens_seen": 1802764288
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004615151515151515,
+      "loss": 2.7824,
+      "theoretical_loss": 3.460511863733192,
+      "tokens_seen": 1802829824
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004614949494949495,
+      "loss": 2.4709,
+      "theoretical_loss": 3.460501160117647,
+      "tokens_seen": 1802895360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004614747474747475,
+      "loss": 2.6541,
+      "theoretical_loss": 3.4604904570001134,
+      "tokens_seen": 1802960896
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046145454545454544,
+      "loss": 2.684,
+      "theoretical_loss": 3.4604797543805503,
+      "tokens_seen": 1803026432
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046143434343434345,
+      "loss": 2.6394,
+      "theoretical_loss": 3.460469052258917,
+      "tokens_seen": 1803091968
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004614141414141414,
+      "loss": 2.5222,
+      "theoretical_loss": 3.460458350635172,
+      "tokens_seen": 1803157504
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004613939393939394,
+      "loss": 2.4386,
+      "theoretical_loss": 3.460447649509274,
+      "tokens_seen": 1803223040
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004613737373737374,
+      "loss": 2.5735,
+      "theoretical_loss": 3.4604369488811813,
+      "tokens_seen": 1803288576
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046135353535353533,
+      "loss": 2.5963,
+      "theoretical_loss": 3.4604262487508533,
+      "tokens_seen": 1803354112
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046133333333333334,
+      "loss": 2.6435,
+      "theoretical_loss": 3.460415549118248,
+      "tokens_seen": 1803419648
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004613131313131313,
+      "loss": 2.5554,
+      "theoretical_loss": 3.460404849983325,
+      "tokens_seen": 1803485184
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004612929292929293,
+      "loss": 2.5415,
+      "theoretical_loss": 3.4603941513460423,
+      "tokens_seen": 1803550720
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046127272727272727,
+      "loss": 2.6456,
+      "theoretical_loss": 3.4603834532063598,
+      "tokens_seen": 1803616256
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004612525252525253,
+      "loss": 2.6177,
+      "theoretical_loss": 3.460372755564235,
+      "tokens_seen": 1803681792
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046123232323232324,
+      "loss": 2.3797,
+      "theoretical_loss": 3.4603620584196273,
+      "tokens_seen": 1803747328
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004612121212121212,
+      "loss": 2.5797,
+      "theoretical_loss": 3.4603513617724957,
+      "tokens_seen": 1803812864
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004611919191919192,
+      "loss": 2.6529,
+      "theoretical_loss": 3.460340665622798,
+      "tokens_seen": 1803878400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046117171717171716,
+      "loss": 2.5033,
+      "theoretical_loss": 3.460329969970495,
+      "tokens_seen": 1803943936
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046115151515151517,
+      "loss": 2.7475,
+      "theoretical_loss": 3.460319274815543,
+      "tokens_seen": 1804009472
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046113131313131313,
+      "loss": 2.4978,
+      "theoretical_loss": 3.460308580157902,
+      "tokens_seen": 1804075008
+    },
+    {
+      "epoch": 0.09,
+      "objective/train/docs_used": 994874,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.3973050117492676,
+      "objective/train/theoretical_loss": 3.460297885997531,
+      "objective/train/tokens_used": 162999776,
+      "theoretical_loss": 3.460297885997531,
+      "tokens_seen": 1804140544
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046111111111111114,
+      "loss": 2.5142,
+      "theoretical_loss": 3.460297885997531,
+      "tokens_seen": 1804140544
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004610909090909091,
+      "loss": 2.5792,
+      "theoretical_loss": 3.4602871923343894,
+      "tokens_seen": 1804206080
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004610707070707071,
+      "loss": 2.6572,
+      "theoretical_loss": 3.4602764991684345,
+      "tokens_seen": 1804271616
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046105050505050506,
+      "loss": 2.7006,
+      "theoretical_loss": 3.460265806499626,
+      "tokens_seen": 1804337152
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000461030303030303,
+      "loss": 2.5644,
+      "theoretical_loss": 3.4602551143279228,
+      "tokens_seen": 1804402688
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046101010101010103,
+      "loss": 2.5553,
+      "theoretical_loss": 3.4602444226532834,
+      "tokens_seen": 1804468224
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000460989898989899,
+      "loss": 2.4872,
+      "theoretical_loss": 3.4602337314756664,
+      "tokens_seen": 1804533760
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000460969696969697,
+      "loss": 2.5919,
+      "theoretical_loss": 3.460223040795032,
+      "tokens_seen": 1804599296
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046094949494949496,
+      "loss": 2.5868,
+      "theoretical_loss": 3.4602123506113376,
+      "tokens_seen": 1804664832
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004609292929292929,
+      "loss": 2.557,
+      "theoretical_loss": 3.4602016609245423,
+      "tokens_seen": 1804730368
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004609090909090909,
+      "loss": 2.4903,
+      "theoretical_loss": 3.460190971734605,
+      "tokens_seen": 1804795904
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046088888888888894,
+      "loss": 2.6423,
+      "theoretical_loss": 3.4601802830414856,
+      "tokens_seen": 1804861440
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004608686868686869,
+      "loss": 2.607,
+      "theoretical_loss": 3.4601695948451416,
+      "tokens_seen": 1804926976
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046084848484848485,
+      "loss": 2.4623,
+      "theoretical_loss": 3.4601589071455323,
+      "tokens_seen": 1804992512
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004608282828282828,
+      "loss": 2.5671,
+      "theoretical_loss": 3.460148219942617,
+      "tokens_seen": 1805058048
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004608080808080808,
+      "loss": 2.7849,
+      "theoretical_loss": 3.4601375332363538,
+      "tokens_seen": 1805123584
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046078787878787883,
+      "loss": 2.4045,
+      "theoretical_loss": 3.4601268470267024,
+      "tokens_seen": 1805189120
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004607676767676768,
+      "loss": 2.6255,
+      "theoretical_loss": 3.4601161613136213,
+      "tokens_seen": 1805254656
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046074747474747474,
+      "loss": 2.3319,
+      "theoretical_loss": 3.4601054760970698,
+      "tokens_seen": 1805320192
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004607272727272727,
+      "loss": 2.6209,
+      "theoretical_loss": 3.460094791377006,
+      "tokens_seen": 1805385728
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046070707070707077,
+      "loss": 2.5799,
+      "theoretical_loss": 3.4600841071533894,
+      "tokens_seen": 1805451264
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004606868686868687,
+      "loss": 2.6091,
+      "theoretical_loss": 3.460073423426179,
+      "tokens_seen": 1805516800
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004606666666666667,
+      "loss": 2.8417,
+      "theoretical_loss": 3.460062740195333,
+      "tokens_seen": 1805582336
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046064646464646464,
+      "loss": 2.5454,
+      "theoretical_loss": 3.4600520574608113,
+      "tokens_seen": 1805647872
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004606262626262626,
+      "loss": 2.7244,
+      "theoretical_loss": 3.4600413752225725,
+      "tokens_seen": 1805713408
+    },
+    {
+      "epoch": 0.09,
+      "objective/train/docs_used": 995886,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.640719413757324,
+      "objective/train/theoretical_loss": 3.460030693480575,
+      "objective/train/tokens_used": 164638176,
+      "theoretical_loss": 3.460030693480575,
+      "tokens_seen": 1805778944
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046060606060606066,
+      "loss": 2.7838,
+      "theoretical_loss": 3.460030693480575,
+      "tokens_seen": 1805778944
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004605858585858586,
+      "loss": 2.5454,
+      "theoretical_loss": 3.460020012234778,
+      "tokens_seen": 1805844480
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046056565656565657,
+      "loss": 2.737,
+      "theoretical_loss": 3.460009331485141,
+      "tokens_seen": 1805910016
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046054545454545453,
+      "loss": 2.6839,
+      "theoretical_loss": 3.459998651231622,
+      "tokens_seen": 1805975552
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046052525252525254,
+      "loss": 2.5493,
+      "theoretical_loss": 3.459987971474181,
+      "tokens_seen": 1806041088
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046050505050505055,
+      "loss": 2.4706,
+      "theoretical_loss": 3.4599772922127765,
+      "tokens_seen": 1806106624
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004604848484848485,
+      "loss": 2.7013,
+      "theoretical_loss": 3.4599666134473672,
+      "tokens_seen": 1806172160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046046464646464646,
+      "loss": 2.7035,
+      "theoretical_loss": 3.459955935177912,
+      "tokens_seen": 1806237696
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004604444444444444,
+      "loss": 2.6302,
+      "theoretical_loss": 3.45994525740437,
+      "tokens_seen": 1806303232
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046042424242424243,
+      "loss": 2.5569,
+      "theoretical_loss": 3.459934580126701,
+      "tokens_seen": 1806368768
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046040404040404044,
+      "loss": 2.6291,
+      "theoretical_loss": 3.459923903344863,
+      "tokens_seen": 1806434304
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004603838383838384,
+      "loss": 2.6722,
+      "theoretical_loss": 3.4599132270588155,
+      "tokens_seen": 1806499840
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046036363636363636,
+      "loss": 2.8284,
+      "theoretical_loss": 3.4599025512685166,
+      "tokens_seen": 1806565376
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046034343434343437,
+      "loss": 2.836,
+      "theoretical_loss": 3.459891875973927,
+      "tokens_seen": 1806630912
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004603232323232323,
+      "loss": 2.5876,
+      "theoretical_loss": 3.459881201175004,
+      "tokens_seen": 1806696448
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046030303030303034,
+      "loss": 2.5903,
+      "theoretical_loss": 3.4598705268717076,
+      "tokens_seen": 1806761984
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004602828282828283,
+      "loss": 2.5872,
+      "theoretical_loss": 3.459859853063996,
+      "tokens_seen": 1806827520
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046026262626262625,
+      "loss": 2.7085,
+      "theoretical_loss": 3.4598491797518287,
+      "tokens_seen": 1806893056
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046024242424242426,
+      "loss": 2.6058,
+      "theoretical_loss": 3.4598385069351654,
+      "tokens_seen": 1806958592
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004602222222222222,
+      "loss": 2.431,
+      "theoretical_loss": 3.459827834613964,
+      "tokens_seen": 1807024128
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046020202020202023,
+      "loss": 2.6627,
+      "theoretical_loss": 3.459817162788184,
+      "tokens_seen": 1807089664
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004601818181818182,
+      "loss": 2.6105,
+      "theoretical_loss": 3.4598064914577846,
+      "tokens_seen": 1807155200
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004601616161616162,
+      "loss": 2.6375,
+      "theoretical_loss": 3.4597958206227246,
+      "tokens_seen": 1807220736
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046014141414141415,
+      "loss": 2.6245,
+      "theoretical_loss": 3.459785150282963,
+      "tokens_seen": 1807286272
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004601212121212121,
+      "loss": 2.7088,
+      "theoretical_loss": 3.459774480438459,
+      "tokens_seen": 1807351808
+    },
+    {
+      "epoch": 0.09,
+      "objective/train/docs_used": 996582,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.828779935836792,
+      "objective/train/theoretical_loss": 3.4597638110891715,
+      "objective/train/tokens_used": 166276576,
+      "theoretical_loss": 3.4597638110891715,
+      "tokens_seen": 1807417344
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004601010101010101,
+      "loss": 2.6118,
+      "theoretical_loss": 3.4597638110891715,
+      "tokens_seen": 1807417344
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004600808080808081,
+      "loss": 2.55,
+      "theoretical_loss": 3.4597531422350603,
+      "tokens_seen": 1807482880
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004600606060606061,
+      "loss": 2.9119,
+      "theoretical_loss": 3.459742473876083,
+      "tokens_seen": 1807548416
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046004040404040405,
+      "loss": 2.4969,
+      "theoretical_loss": 3.4597318060122,
+      "tokens_seen": 1807613952
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000460020202020202,
+      "loss": 2.3569,
+      "theoretical_loss": 3.4597211386433697,
+      "tokens_seen": 1807679488
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00046,
+      "loss": 2.8525,
+      "theoretical_loss": 3.4597104717695517,
+      "tokens_seen": 1807745024
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000459979797979798,
+      "loss": 2.6927,
+      "theoretical_loss": 3.459699805390705,
+      "tokens_seen": 1807810560
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000459959595959596,
+      "loss": 2.6665,
+      "theoretical_loss": 3.4596891395067875,
+      "tokens_seen": 1807876096
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045993939393939394,
+      "loss": 2.6101,
+      "theoretical_loss": 3.45967847411776,
+      "tokens_seen": 1807941632
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004599191919191919,
+      "loss": 2.6743,
+      "theoretical_loss": 3.4596678092235806,
+      "tokens_seen": 1808007168
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004598989898989899,
+      "loss": 2.7123,
+      "theoretical_loss": 3.4596571448242086,
+      "tokens_seen": 1808072704
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004598787878787879,
+      "loss": 2.589,
+      "theoretical_loss": 3.459646480919603,
+      "tokens_seen": 1808138240
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004598585858585859,
+      "loss": 2.4773,
+      "theoretical_loss": 3.4596358175097235,
+      "tokens_seen": 1808203776
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045983838383838383,
+      "loss": 2.4748,
+      "theoretical_loss": 3.4596251545945282,
+      "tokens_seen": 1808269312
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004598181818181818,
+      "loss": 2.5673,
+      "theoretical_loss": 3.4596144921739773,
+      "tokens_seen": 1808334848
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004597979797979798,
+      "loss": 2.4766,
+      "theoretical_loss": 3.45960383024803,
+      "tokens_seen": 1808400384
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004597777777777778,
+      "loss": 2.7642,
+      "theoretical_loss": 3.459593168816644,
+      "tokens_seen": 1808465920
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045975757575757577,
+      "loss": 2.7496,
+      "theoretical_loss": 3.4595825078797793,
+      "tokens_seen": 1808531456
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004597373737373737,
+      "loss": 2.8011,
+      "theoretical_loss": 3.4595718474373953,
+      "tokens_seen": 1808596992
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004597171717171717,
+      "loss": 2.8709,
+      "theoretical_loss": 3.459561187489451,
+      "tokens_seen": 1808662528
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045969696969696975,
+      "loss": 2.6683,
+      "theoretical_loss": 3.459550528035905,
+      "tokens_seen": 1808728064
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004596767676767677,
+      "loss": 2.6207,
+      "theoretical_loss": 3.4595398690767176,
+      "tokens_seen": 1808793600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045965656565656566,
+      "loss": 2.4637,
+      "theoretical_loss": 3.459529210611847,
+      "tokens_seen": 1808859136
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004596363636363636,
+      "loss": 2.7937,
+      "theoretical_loss": 3.4595185526412524,
+      "tokens_seen": 1808924672
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004596161616161616,
+      "loss": 2.6215,
+      "theoretical_loss": 3.4595078951648937,
+      "tokens_seen": 1808990208
+    },
+    {
+      "epoch": 0.09,
+      "objective/train/docs_used": 997711,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.4244730472564697,
+      "objective/train/theoretical_loss": 3.4594972381827294,
+      "objective/train/tokens_used": 167914976,
+      "theoretical_loss": 3.4594972381827294,
+      "tokens_seen": 1809055744
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045959595959595964,
+      "loss": 2.64,
+      "theoretical_loss": 3.4594972381827294,
+      "tokens_seen": 1809055744
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004595757575757576,
+      "loss": 2.6338,
+      "theoretical_loss": 3.4594865816947182,
+      "tokens_seen": 1809121280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045955555555555555,
+      "loss": 2.5629,
+      "theoretical_loss": 3.4594759257008207,
+      "tokens_seen": 1809186816
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004595353535353535,
+      "loss": 2.5733,
+      "theoretical_loss": 3.4594652702009956,
+      "tokens_seen": 1809252352
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004595151515151515,
+      "loss": 2.6077,
+      "theoretical_loss": 3.4594546151952015,
+      "tokens_seen": 1809317888
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045949494949494953,
+      "loss": 2.4091,
+      "theoretical_loss": 3.4594439606833975,
+      "tokens_seen": 1809383424
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004594747474747475,
+      "loss": 2.6934,
+      "theoretical_loss": 3.459433306665544,
+      "tokens_seen": 1809448960
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045945454545454545,
+      "loss": 2.6156,
+      "theoretical_loss": 3.4594226531415986,
+      "tokens_seen": 1809514496
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004594343434343434,
+      "loss": 2.8258,
+      "theoretical_loss": 3.4594120001115223,
+      "tokens_seen": 1809580032
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045941414141414147,
+      "loss": 2.7045,
+      "theoretical_loss": 3.459401347575273,
+      "tokens_seen": 1809645568
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004593939393939394,
+      "loss": 2.6075,
+      "theoretical_loss": 3.4593906955328104,
+      "tokens_seen": 1809711104
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004593737373737374,
+      "loss": 2.6631,
+      "theoretical_loss": 3.4593800439840936,
+      "tokens_seen": 1809776640
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045935353535353534,
+      "loss": 2.8016,
+      "theoretical_loss": 3.459369392929082,
+      "tokens_seen": 1809842176
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045933333333333335,
+      "loss": 2.7707,
+      "theoretical_loss": 3.4593587423677348,
+      "tokens_seen": 1809907712
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045931313131313136,
+      "loss": 2.5944,
+      "theoretical_loss": 3.459348092300011,
+      "tokens_seen": 1809973248
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004592929292929293,
+      "loss": 2.536,
+      "theoretical_loss": 3.45933744272587,
+      "tokens_seen": 1810038784
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004592727272727273,
+      "loss": 2.6021,
+      "theoretical_loss": 3.459326793645271,
+      "tokens_seen": 1810104320
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045925252525252523,
+      "loss": 2.6998,
+      "theoretical_loss": 3.4593161450581738,
+      "tokens_seen": 1810169856
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045923232323232324,
+      "loss": 2.8461,
+      "theoretical_loss": 3.4593054969645367,
+      "tokens_seen": 1810235392
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045921212121212125,
+      "loss": 2.7473,
+      "theoretical_loss": 3.45929484936432,
+      "tokens_seen": 1810300928
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004591919191919192,
+      "loss": 2.6926,
+      "theoretical_loss": 3.459284202257482,
+      "tokens_seen": 1810366464
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045917171717171717,
+      "loss": 2.6265,
+      "theoretical_loss": 3.459273555643983,
+      "tokens_seen": 1810432000
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004591515151515152,
+      "loss": 2.5741,
+      "theoretical_loss": 3.4592629095237815,
+      "tokens_seen": 1810497536
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045913131313131313,
+      "loss": 2.6005,
+      "theoretical_loss": 3.459252263896837,
+      "tokens_seen": 1810563072
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045911111111111115,
+      "loss": 2.6592,
+      "theoretical_loss": 3.4592416187631088,
+      "tokens_seen": 1810628608
+    },
+    {
+      "epoch": 0.09,
+      "objective/train/docs_used": 998243,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6315853595733643,
+      "objective/train/theoretical_loss": 3.4592309741225558,
+      "objective/train/tokens_used": 169553376,
+      "theoretical_loss": 3.4592309741225558,
+      "tokens_seen": 1810694144
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004590909090909091,
+      "loss": 2.5426,
+      "theoretical_loss": 3.4592309741225558,
+      "tokens_seen": 1810694144
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045907070707070706,
+      "loss": 2.5075,
+      "theoretical_loss": 3.459220329975138,
+      "tokens_seen": 1810759680
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045905050505050507,
+      "loss": 2.7955,
+      "theoretical_loss": 3.4592096863208144,
+      "tokens_seen": 1810825216
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045903030303030303,
+      "loss": 2.5196,
+      "theoretical_loss": 3.4591990431595443,
+      "tokens_seen": 1810890752
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045901010101010104,
+      "loss": 2.6317,
+      "theoretical_loss": 3.4591884004912874,
+      "tokens_seen": 1810956288
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000458989898989899,
+      "loss": 2.3922,
+      "theoretical_loss": 3.4591777583160024,
+      "tokens_seen": 1811021824
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000458969696969697,
+      "loss": 2.9019,
+      "theoretical_loss": 3.4591671166336484,
+      "tokens_seen": 1811087360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045894949494949496,
+      "loss": 2.7095,
+      "theoretical_loss": 3.459156475444186,
+      "tokens_seen": 1811152896
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004589292929292929,
+      "loss": 2.7239,
+      "theoretical_loss": 3.4591458347475736,
+      "tokens_seen": 1811218432
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045890909090909093,
+      "loss": 2.7473,
+      "theoretical_loss": 3.4591351945437703,
+      "tokens_seen": 1811283968
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004588888888888889,
+      "loss": 2.6645,
+      "theoretical_loss": 3.4591245548327363,
+      "tokens_seen": 1811349504
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004588686868686869,
+      "loss": 2.8029,
+      "theoretical_loss": 3.45911391561443,
+      "tokens_seen": 1811415040
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045884848484848486,
+      "loss": 2.5806,
+      "theoretical_loss": 3.4591032768888117,
+      "tokens_seen": 1811480576
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004588282828282828,
+      "loss": 2.6284,
+      "theoretical_loss": 3.4590926386558403,
+      "tokens_seen": 1811546112
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004588080808080808,
+      "loss": 2.767,
+      "theoretical_loss": 3.459082000915475,
+      "tokens_seen": 1811611648
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045878787878787883,
+      "loss": 2.792,
+      "theoretical_loss": 3.4590713636676753,
+      "tokens_seen": 1811677184
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004587676767676768,
+      "loss": 2.7077,
+      "theoretical_loss": 3.4590607269124005,
+      "tokens_seen": 1811742720
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045874747474747475,
+      "loss": 2.6499,
+      "theoretical_loss": 3.459050090649611,
+      "tokens_seen": 1811808256
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004587272727272727,
+      "loss": 2.8369,
+      "theoretical_loss": 3.4590394548792642,
+      "tokens_seen": 1811873792
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004587070707070707,
+      "loss": 2.3474,
+      "theoretical_loss": 3.459028819601321,
+      "tokens_seen": 1811939328
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045868686868686873,
+      "loss": 2.7419,
+      "theoretical_loss": 3.459018184815741,
+      "tokens_seen": 1812004864
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004586666666666667,
+      "loss": 2.7241,
+      "theoretical_loss": 3.459007550522482,
+      "tokens_seen": 1812070400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045864646464646464,
+      "loss": 2.5917,
+      "theoretical_loss": 3.458996916721505,
+      "tokens_seen": 1812135936
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004586262626262626,
+      "loss": 2.5461,
+      "theoretical_loss": 3.4589862834127683,
+      "tokens_seen": 1812201472
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045860606060606066,
+      "loss": 2.429,
+      "theoretical_loss": 3.458975650596232,
+      "tokens_seen": 1812267008
+    },
+    {
+      "epoch": 0.09,
+      "objective/train/docs_used": 999229,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6778602600097656,
+      "objective/train/theoretical_loss": 3.458965018271855,
+      "objective/train/tokens_used": 171191776,
+      "theoretical_loss": 3.458965018271855,
+      "tokens_seen": 1812332544
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004585858585858586,
+      "loss": 2.6921,
+      "theoretical_loss": 3.458965018271855,
+      "tokens_seen": 1812332544
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004585656565656566,
+      "loss": 2.6781,
+      "theoretical_loss": 3.458954386439598,
+      "tokens_seen": 1812398080
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045854545454545453,
+      "loss": 2.8004,
+      "theoretical_loss": 3.458943755099419,
+      "tokens_seen": 1812463616
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004585252525252525,
+      "loss": 2.5061,
+      "theoretical_loss": 3.4589331242512777,
+      "tokens_seen": 1812529152
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045850505050505056,
+      "loss": 2.6991,
+      "theoretical_loss": 3.458922493895134,
+      "tokens_seen": 1812594688
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004584848484848485,
+      "loss": 2.59,
+      "theoretical_loss": 3.458911864030947,
+      "tokens_seen": 1812660224
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045846464646464647,
+      "loss": 2.7665,
+      "theoretical_loss": 3.4589012346586765,
+      "tokens_seen": 1812725760
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004584444444444444,
+      "loss": 2.5644,
+      "theoretical_loss": 3.458890605778281,
+      "tokens_seen": 1812791296
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004584242424242424,
+      "loss": 2.6733,
+      "theoretical_loss": 3.4588799773897216,
+      "tokens_seen": 1812856832
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045840404040404045,
+      "loss": 2.495,
+      "theoretical_loss": 3.4588693494929563,
+      "tokens_seen": 1812922368
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004583838383838384,
+      "loss": 2.4922,
+      "theoretical_loss": 3.458858722087945,
+      "tokens_seen": 1812987904
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045836363636363636,
+      "loss": 2.4791,
+      "theoretical_loss": 3.4588480951746474,
+      "tokens_seen": 1813053440
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004583434343434343,
+      "loss": 2.6036,
+      "theoretical_loss": 3.4588374687530226,
+      "tokens_seen": 1813118976
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045832323232323233,
+      "loss": 2.4228,
+      "theoretical_loss": 3.4588268428230307,
+      "tokens_seen": 1813184512
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045830303030303034,
+      "loss": 2.747,
+      "theoretical_loss": 3.4588162173846304,
+      "tokens_seen": 1813250048
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004582828282828283,
+      "loss": 2.4541,
+      "theoretical_loss": 3.458805592437782,
+      "tokens_seen": 1813315584
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045826262626262626,
+      "loss": 2.4754,
+      "theoretical_loss": 3.458794967982444,
+      "tokens_seen": 1813381120
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004582424242424242,
+      "loss": 2.3365,
+      "theoretical_loss": 3.458784344018577,
+      "tokens_seen": 1813446656
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004582222222222222,
+      "loss": 2.8325,
+      "theoretical_loss": 3.45877372054614,
+      "tokens_seen": 1813512192
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045820202020202023,
+      "loss": 2.7678,
+      "theoretical_loss": 3.4587630975650923,
+      "tokens_seen": 1813577728
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004581818181818182,
+      "loss": 2.5188,
+      "theoretical_loss": 3.4587524750753937,
+      "tokens_seen": 1813643264
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045816161616161615,
+      "loss": 2.6459,
+      "theoretical_loss": 3.4587418530770035,
+      "tokens_seen": 1813708800
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045814141414141416,
+      "loss": 2.477,
+      "theoretical_loss": 3.4587312315698813,
+      "tokens_seen": 1813774336
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004581212121212121,
+      "loss": 2.5432,
+      "theoretical_loss": 3.458720610553987,
+      "tokens_seen": 1813839872
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004581010101010101,
+      "loss": 2.7704,
+      "theoretical_loss": 3.4587099900292797,
+      "tokens_seen": 1813905408
+    },
+    {
+      "epoch": 0.09,
+      "objective/train/docs_used": 999785,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5763378143310547,
+      "objective/train/theoretical_loss": 3.4586993699957187,
+      "objective/train/tokens_used": 172830176,
+      "theoretical_loss": 3.4586993699957187,
+      "tokens_seen": 1813970944
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004580808080808081,
+      "loss": 2.5818,
+      "theoretical_loss": 3.4586993699957187,
+      "tokens_seen": 1813970944
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045806060606060604,
+      "loss": 2.3931,
+      "theoretical_loss": 3.4586887504532644,
+      "tokens_seen": 1814036480
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045804040404040405,
+      "loss": 2.455,
+      "theoretical_loss": 3.458678131401876,
+      "tokens_seen": 1814102016
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000458020202020202,
+      "loss": 2.5927,
+      "theoretical_loss": 3.4586675128415125,
+      "tokens_seen": 1814167552
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000458,
+      "loss": 2.9255,
+      "theoretical_loss": 3.4586568947721337,
+      "tokens_seen": 1814233088
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000457979797979798,
+      "loss": 2.5883,
+      "theoretical_loss": 3.4586462771936994,
+      "tokens_seen": 1814298624
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000457959595959596,
+      "loss": 2.618,
+      "theoretical_loss": 3.4586356601061694,
+      "tokens_seen": 1814364160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045793939393939394,
+      "loss": 2.4486,
+      "theoretical_loss": 3.4586250435095027,
+      "tokens_seen": 1814429696
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004579191919191919,
+      "loss": 2.7142,
+      "theoretical_loss": 3.4586144274036594,
+      "tokens_seen": 1814495232
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004578989898989899,
+      "loss": 2.4711,
+      "theoretical_loss": 3.4586038117885987,
+      "tokens_seen": 1814560768
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045787878787878787,
+      "loss": 2.5087,
+      "theoretical_loss": 3.4585931966642804,
+      "tokens_seen": 1814626304
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004578585858585859,
+      "loss": 2.622,
+      "theoretical_loss": 3.458582582030664,
+      "tokens_seen": 1814691840
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045783838383838384,
+      "loss": 2.472,
+      "theoretical_loss": 3.4585719678877087,
+      "tokens_seen": 1814757376
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004578181818181818,
+      "loss": 2.8238,
+      "theoretical_loss": 3.458561354235375,
+      "tokens_seen": 1814822912
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004577979797979798,
+      "loss": 2.4024,
+      "theoretical_loss": 3.458550741073622,
+      "tokens_seen": 1814888448
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004577777777777778,
+      "loss": 2.6186,
+      "theoretical_loss": 3.458540128402409,
+      "tokens_seen": 1814953984
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045775757575757577,
+      "loss": 2.609,
+      "theoretical_loss": 3.458529516221696,
+      "tokens_seen": 1815019520
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045773737373737373,
+      "loss": 2.565,
+      "theoretical_loss": 3.4585189045314424,
+      "tokens_seen": 1815085056
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045771717171717174,
+      "loss": 2.7391,
+      "theoretical_loss": 3.458508293331609,
+      "tokens_seen": 1815150592
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004576969696969697,
+      "loss": 2.5499,
+      "theoretical_loss": 3.458497682622153,
+      "tokens_seen": 1815216128
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004576767676767677,
+      "loss": 2.7469,
+      "theoretical_loss": 3.4584870724030363,
+      "tokens_seen": 1815281664
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045765656565656567,
+      "loss": 2.6686,
+      "theoretical_loss": 3.4584764626742173,
+      "tokens_seen": 1815347200
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004576363636363636,
+      "loss": 2.6245,
+      "theoretical_loss": 3.4584658534356567,
+      "tokens_seen": 1815412736
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045761616161616163,
+      "loss": 2.6962,
+      "theoretical_loss": 3.458455244687313,
+      "tokens_seen": 1815478272
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045759595959595964,
+      "loss": 2.6647,
+      "theoretical_loss": 3.458444636429146,
+      "tokens_seen": 1815543808
+    },
+    {
+      "epoch": 0.09,
+      "objective/train/docs_used": 1000504,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.835871458053589,
+      "objective/train/theoretical_loss": 3.458434028661116,
+      "objective/train/tokens_used": 174468576,
+      "theoretical_loss": 3.458434028661116,
+      "tokens_seen": 1815609344
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004575757575757576,
+      "loss": 2.7497,
+      "theoretical_loss": 3.458434028661116,
+      "tokens_seen": 1815609344
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045755555555555556,
+      "loss": 2.6998,
+      "theoretical_loss": 3.4584234213831824,
+      "tokens_seen": 1815674880
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004575353535353535,
+      "loss": 2.5959,
+      "theoretical_loss": 3.458412814595305,
+      "tokens_seen": 1815740416
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004575151515151515,
+      "loss": 2.6133,
+      "theoretical_loss": 3.458402208297443,
+      "tokens_seen": 1815805952
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045749494949494954,
+      "loss": 2.6405,
+      "theoretical_loss": 3.4583916024895567,
+      "tokens_seen": 1815871488
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004574747474747475,
+      "loss": 2.7544,
+      "theoretical_loss": 3.458380997171605,
+      "tokens_seen": 1815937024
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045745454545454545,
+      "loss": 2.6652,
+      "theoretical_loss": 3.4583703923435483,
+      "tokens_seen": 1816002560
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004574343434343434,
+      "loss": 2.7187,
+      "theoretical_loss": 3.458359788005346,
+      "tokens_seen": 1816068096
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004574141414141415,
+      "loss": 2.6787,
+      "theoretical_loss": 3.458349184156958,
+      "tokens_seen": 1816133632
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045739393939393943,
+      "loss": 2.6822,
+      "theoretical_loss": 3.458338580798343,
+      "tokens_seen": 1816199168
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004573737373737374,
+      "loss": 2.6664,
+      "theoretical_loss": 3.4583279779294624,
+      "tokens_seen": 1816264704
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045735353535353534,
+      "loss": 2.6565,
+      "theoretical_loss": 3.458317375550275,
+      "tokens_seen": 1816330240
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004573333333333333,
+      "loss": 2.6583,
+      "theoretical_loss": 3.4583067736607402,
+      "tokens_seen": 1816395776
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045731313131313137,
+      "loss": 2.9378,
+      "theoretical_loss": 3.4582961722608188,
+      "tokens_seen": 1816461312
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004572929292929293,
+      "loss": 2.6101,
+      "theoretical_loss": 3.458285571350469,
+      "tokens_seen": 1816526848
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004572727272727273,
+      "loss": 2.8678,
+      "theoretical_loss": 3.4582749709296516,
+      "tokens_seen": 1816592384
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045725252525252524,
+      "loss": 2.5882,
+      "theoretical_loss": 3.458264370998326,
+      "tokens_seen": 1816657920
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045723232323232325,
+      "loss": 2.5968,
+      "theoretical_loss": 3.4582537715564525,
+      "tokens_seen": 1816723456
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045721212121212126,
+      "loss": 2.7255,
+      "theoretical_loss": 3.45824317260399,
+      "tokens_seen": 1816788992
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004571919191919192,
+      "loss": 2.7617,
+      "theoretical_loss": 3.458232574140899,
+      "tokens_seen": 1816854528
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045717171717171717,
+      "loss": 2.4575,
+      "theoretical_loss": 3.4582219761671382,
+      "tokens_seen": 1816920064
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045715151515151513,
+      "loss": 2.66,
+      "theoretical_loss": 3.4582113786826683,
+      "tokens_seen": 1816985600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045713131313131314,
+      "loss": 2.527,
+      "theoretical_loss": 3.4582007816874487,
+      "tokens_seen": 1817051136
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00045711111111111115,
+      "loss": 2.4134,
+      "theoretical_loss": 3.4581901851814396,
+      "tokens_seen": 1817116672
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0004570909090909091,
+      "loss": 2.8572,
+      "theoretical_loss": 3.4581795891646,
+      "tokens_seen": 1817182208
+    },
+    {
+      "epoch": 0.09,
+      "objective/train/docs_used": 1001884,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.327115058898926,
+      "objective/train/theoretical_loss": 3.4581689936368907,
+      "objective/train/tokens_used": 176106976,
+      "theoretical_loss": 3.4581689936368907,
+      "tokens_seen": 1817247744
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045707070707070706,
+      "loss": 2.5736,
+      "theoretical_loss": 3.4581689936368907,
+      "tokens_seen": 1817247744
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004570505050505051,
+      "loss": 2.8025,
+      "theoretical_loss": 3.45815839859827,
+      "tokens_seen": 1817313280
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045703030303030303,
+      "loss": 2.6386,
+      "theoretical_loss": 3.458147804048699,
+      "tokens_seen": 1817378816
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045701010101010104,
+      "loss": 2.719,
+      "theoretical_loss": 3.458137209988138,
+      "tokens_seen": 1817444352
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000456989898989899,
+      "loss": 2.5823,
+      "theoretical_loss": 3.4581266164165445,
+      "tokens_seen": 1817509888
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045696969696969696,
+      "loss": 2.7439,
+      "theoretical_loss": 3.45811602333388,
+      "tokens_seen": 1817575424
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045694949494949497,
+      "loss": 2.4384,
+      "theoretical_loss": 3.4581054307401047,
+      "tokens_seen": 1817640960
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004569292929292929,
+      "loss": 2.8483,
+      "theoretical_loss": 3.458094838635177,
+      "tokens_seen": 1817706496
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045690909090909094,
+      "loss": 2.6419,
+      "theoretical_loss": 3.458084247019057,
+      "tokens_seen": 1817772032
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004568888888888889,
+      "loss": 2.7798,
+      "theoretical_loss": 3.458073655891706,
+      "tokens_seen": 1817837568
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004568686868686869,
+      "loss": 2.5721,
+      "theoretical_loss": 3.458063065253082,
+      "tokens_seen": 1817903104
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045684848484848486,
+      "loss": 2.8557,
+      "theoretical_loss": 3.4580524751031456,
+      "tokens_seen": 1817968640
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004568282828282828,
+      "loss": 2.6035,
+      "theoretical_loss": 3.4580418854418573,
+      "tokens_seen": 1818034176
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045680808080808083,
+      "loss": 2.5388,
+      "theoretical_loss": 3.4580312962691755,
+      "tokens_seen": 1818099712
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004567878787878788,
+      "loss": 2.5895,
+      "theoretical_loss": 3.458020707585061,
+      "tokens_seen": 1818165248
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004567676767676768,
+      "loss": 2.4687,
+      "theoretical_loss": 3.4580101193894732,
+      "tokens_seen": 1818230784
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045674747474747475,
+      "loss": 2.461,
+      "theoretical_loss": 3.457999531682373,
+      "tokens_seen": 1818296320
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004567272727272727,
+      "loss": 2.5961,
+      "theoretical_loss": 3.4579889444637186,
+      "tokens_seen": 1818361856
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004567070707070707,
+      "loss": 2.5384,
+      "theoretical_loss": 3.457978357733471,
+      "tokens_seen": 1818427392
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045668686868686873,
+      "loss": 2.729,
+      "theoretical_loss": 3.4579677714915897,
+      "tokens_seen": 1818492928
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004566666666666667,
+      "loss": 2.8327,
+      "theoretical_loss": 3.457957185738035,
+      "tokens_seen": 1818558464
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045664646464646465,
+      "loss": 2.4631,
+      "theoretical_loss": 3.4579466004727664,
+      "tokens_seen": 1818624000
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004566262626262626,
+      "loss": 2.7424,
+      "theoretical_loss": 3.4579360156957435,
+      "tokens_seen": 1818689536
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004566060606060606,
+      "loss": 2.7302,
+      "theoretical_loss": 3.4579254314069265,
+      "tokens_seen": 1818755072
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004565858585858586,
+      "loss": 2.8242,
+      "theoretical_loss": 3.4579148476062755,
+      "tokens_seen": 1818820608
+    },
+    {
+      "epoch": 0.1,
+      "objective/train/docs_used": 1002311,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.1334288120269775,
+      "objective/train/theoretical_loss": 3.4579042642937505,
+      "objective/train/tokens_used": 177745376,
+      "theoretical_loss": 3.4579042642937505,
+      "tokens_seen": 1818886144
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004565656565656566,
+      "loss": 2.9767,
+      "theoretical_loss": 3.4579042642937505,
+      "tokens_seen": 1818886144
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045654545454545454,
+      "loss": 2.3271,
+      "theoretical_loss": 3.4578936814693106,
+      "tokens_seen": 1818951680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004565252525252525,
+      "loss": 2.7679,
+      "theoretical_loss": 3.4578830991329164,
+      "tokens_seen": 1819017216
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004565050505050505,
+      "loss": 2.4984,
+      "theoretical_loss": 3.4578725172845273,
+      "tokens_seen": 1819082752
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004564848484848485,
+      "loss": 2.6605,
+      "theoretical_loss": 3.457861935924104,
+      "tokens_seen": 1819148288
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004564646464646465,
+      "loss": 2.582,
+      "theoretical_loss": 3.4578513550516057,
+      "tokens_seen": 1819213824
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045644444444444443,
+      "loss": 2.6528,
+      "theoretical_loss": 3.457840774666993,
+      "tokens_seen": 1819279360
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004564242424242424,
+      "loss": 2.6102,
+      "theoretical_loss": 3.457830194770225,
+      "tokens_seen": 1819344896
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045640404040404045,
+      "loss": 2.647,
+      "theoretical_loss": 3.4578196153612617,
+      "tokens_seen": 1819410432
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004563838383838384,
+      "loss": 2.8564,
+      "theoretical_loss": 3.457809036440064,
+      "tokens_seen": 1819475968
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045636363636363637,
+      "loss": 2.581,
+      "theoretical_loss": 3.4577984580065912,
+      "tokens_seen": 1819541504
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004563434343434343,
+      "loss": 2.8138,
+      "theoretical_loss": 3.4577878800608026,
+      "tokens_seen": 1819607040
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004563232323232323,
+      "loss": 2.4587,
+      "theoretical_loss": 3.4577773026026595,
+      "tokens_seen": 1819672576
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045630303030303035,
+      "loss": 2.6125,
+      "theoretical_loss": 3.457766725632121,
+      "tokens_seen": 1819738112
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004562828282828283,
+      "loss": 2.7351,
+      "theoretical_loss": 3.457756149149147,
+      "tokens_seen": 1819803648
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045626262626262626,
+      "loss": 2.8924,
+      "theoretical_loss": 3.457745573153698,
+      "tokens_seen": 1819869184
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004562424242424242,
+      "loss": 2.6131,
+      "theoretical_loss": 3.4577349976457334,
+      "tokens_seen": 1819934720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045622222222222223,
+      "loss": 2.7737,
+      "theoretical_loss": 3.4577244226252137,
+      "tokens_seen": 1820000256
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045620202020202024,
+      "loss": 2.6232,
+      "theoretical_loss": 3.457713848092099,
+      "tokens_seen": 1820065792
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004561818181818182,
+      "loss": 2.7592,
+      "theoretical_loss": 3.4577032740463483,
+      "tokens_seen": 1820131328
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045616161616161615,
+      "loss": 2.7347,
+      "theoretical_loss": 3.4576927004879225,
+      "tokens_seen": 1820196864
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004561414141414141,
+      "loss": 2.6675,
+      "theoretical_loss": 3.4576821274167813,
+      "tokens_seen": 1820262400
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004561212121212121,
+      "loss": 2.815,
+      "theoretical_loss": 3.4576715548328845,
+      "tokens_seen": 1820327936
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045610101010101013,
+      "loss": 2.8066,
+      "theoretical_loss": 3.4576609827361926,
+      "tokens_seen": 1820393472
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004560808080808081,
+      "loss": 2.6566,
+      "theoretical_loss": 3.4576504111266653,
+      "tokens_seen": 1820459008
+    },
+    {
+      "epoch": 0.1,
+      "objective/train/docs_used": 1003206,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.652801990509033,
+      "objective/train/theoretical_loss": 3.457639840004262,
+      "objective/train/tokens_used": 179383776,
+      "theoretical_loss": 3.457639840004262,
+      "tokens_seen": 1820524544
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045606060606060605,
+      "loss": 2.7231,
+      "theoretical_loss": 3.457639840004262,
+      "tokens_seen": 1820524544
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045604040404040406,
+      "loss": 2.5342,
+      "theoretical_loss": 3.4576292693689443,
+      "tokens_seen": 1820590080
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045602020202020207,
+      "loss": 2.6447,
+      "theoretical_loss": 3.4576186992206708,
+      "tokens_seen": 1820655616
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000456,
+      "loss": 2.5253,
+      "theoretical_loss": 3.4576081295594023,
+      "tokens_seen": 1820721152
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000455979797979798,
+      "loss": 2.5932,
+      "theoretical_loss": 3.457597560385098,
+      "tokens_seen": 1820786688
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045595959595959594,
+      "loss": 2.4873,
+      "theoretical_loss": 3.4575869916977195,
+      "tokens_seen": 1820852224
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045593939393939395,
+      "loss": 2.7011,
+      "theoretical_loss": 3.4575764234972253,
+      "tokens_seen": 1820917760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045591919191919196,
+      "loss": 2.6617,
+      "theoretical_loss": 3.4575658557835753,
+      "tokens_seen": 1820983296
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004558989898989899,
+      "loss": 2.7763,
+      "theoretical_loss": 3.457555288556731,
+      "tokens_seen": 1821048832
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004558787878787879,
+      "loss": 2.813,
+      "theoretical_loss": 3.4575447218166517,
+      "tokens_seen": 1821114368
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004558585858585859,
+      "loss": 2.7815,
+      "theoretical_loss": 3.4575341555632972,
+      "tokens_seen": 1821179904
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045583838383838384,
+      "loss": 2.755,
+      "theoretical_loss": 3.4575235897966277,
+      "tokens_seen": 1821245440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045581818181818185,
+      "loss": 2.4572,
+      "theoretical_loss": 3.457513024516604,
+      "tokens_seen": 1821310976
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004557979797979798,
+      "loss": 2.9435,
+      "theoretical_loss": 3.457502459723185,
+      "tokens_seen": 1821376512
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045577777777777777,
+      "loss": 2.6106,
+      "theoretical_loss": 3.4574918954163314,
+      "tokens_seen": 1821442048
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004557575757575758,
+      "loss": 2.4757,
+      "theoretical_loss": 3.4574813315960036,
+      "tokens_seen": 1821507584
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045573737373737374,
+      "loss": 2.5581,
+      "theoretical_loss": 3.4574707682621613,
+      "tokens_seen": 1821573120
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045571717171717175,
+      "loss": 2.8554,
+      "theoretical_loss": 3.4574602054147645,
+      "tokens_seen": 1821638656
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004556969696969697,
+      "loss": 2.6493,
+      "theoretical_loss": 3.4574496430537733,
+      "tokens_seen": 1821704192
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004556767676767677,
+      "loss": 2.9531,
+      "theoretical_loss": 3.4574390811791478,
+      "tokens_seen": 1821769728
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045565656565656567,
+      "loss": 2.811,
+      "theoretical_loss": 3.4574285197908483,
+      "tokens_seen": 1821835264
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045563636363636363,
+      "loss": 2.8732,
+      "theoretical_loss": 3.4574179588888345,
+      "tokens_seen": 1821900800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045561616161616164,
+      "loss": 2.7586,
+      "theoretical_loss": 3.4574073984730678,
+      "tokens_seen": 1821966336
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004555959595959596,
+      "loss": 2.6515,
+      "theoretical_loss": 3.4573968385435068,
+      "tokens_seen": 1822031872
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004555757575757576,
+      "loss": 3.0081,
+      "theoretical_loss": 3.4573862791001124,
+      "tokens_seen": 1822097408
+    },
+    {
+      "epoch": 0.1,
+      "objective/train/docs_used": 1004309,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.1638598442077637,
+      "objective/train/theoretical_loss": 3.4573757201428443,
+      "objective/train/tokens_used": 181022176,
+      "theoretical_loss": 3.4573757201428443,
+      "tokens_seen": 1822162944
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045555555555555556,
+      "loss": 2.8326,
+      "theoretical_loss": 3.4573757201428443,
+      "tokens_seen": 1822162944
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004555353535353535,
+      "loss": 2.7456,
+      "theoretical_loss": 3.457365161671663,
+      "tokens_seen": 1822228480
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045551515151515153,
+      "loss": 2.6156,
+      "theoretical_loss": 3.457354603686529,
+      "tokens_seen": 1822294016
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045549494949494954,
+      "loss": 2.7395,
+      "theoretical_loss": 3.457344046187401,
+      "tokens_seen": 1822359552
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004554747474747475,
+      "loss": 2.901,
+      "theoretical_loss": 3.457333489174241,
+      "tokens_seen": 1822425088
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045545454545454546,
+      "loss": 2.5662,
+      "theoretical_loss": 3.4573229326470076,
+      "tokens_seen": 1822490624
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004554343434343434,
+      "loss": 2.5802,
+      "theoretical_loss": 3.457312376605662,
+      "tokens_seen": 1822556160
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004554141414141414,
+      "loss": 2.6943,
+      "theoretical_loss": 3.457301821050164,
+      "tokens_seen": 1822621696
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045539393939393944,
+      "loss": 2.5732,
+      "theoretical_loss": 3.4572912659804738,
+      "tokens_seen": 1822687232
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004553737373737374,
+      "loss": 2.783,
+      "theoretical_loss": 3.4572807113965514,
+      "tokens_seen": 1822752768
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045535353535353535,
+      "loss": 2.7854,
+      "theoretical_loss": 3.4572701572983573,
+      "tokens_seen": 1822818304
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004553333333333333,
+      "loss": 2.3922,
+      "theoretical_loss": 3.457259603685851,
+      "tokens_seen": 1822883840
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045531313131313137,
+      "loss": 2.6926,
+      "theoretical_loss": 3.4572490505589943,
+      "tokens_seen": 1822949376
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045529292929292933,
+      "loss": 2.8008,
+      "theoretical_loss": 3.4572384979177455,
+      "tokens_seen": 1823014912
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004552727272727273,
+      "loss": 2.3679,
+      "theoretical_loss": 3.4572279457620656,
+      "tokens_seen": 1823080448
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045525252525252524,
+      "loss": 2.6327,
+      "theoretical_loss": 3.457217394091915,
+      "tokens_seen": 1823145984
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004552323232323232,
+      "loss": 2.7428,
+      "theoretical_loss": 3.4572068429072536,
+      "tokens_seen": 1823211520
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045521212121212126,
+      "loss": 2.7718,
+      "theoretical_loss": 3.457196292208042,
+      "tokens_seen": 1823277056
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004551919191919192,
+      "loss": 2.8718,
+      "theoretical_loss": 3.4571857419942402,
+      "tokens_seen": 1823342592
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004551717171717172,
+      "loss": 2.5766,
+      "theoretical_loss": 3.457175192265808,
+      "tokens_seen": 1823408128
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045515151515151513,
+      "loss": 2.8171,
+      "theoretical_loss": 3.457164643022706,
+      "tokens_seen": 1823473664
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004551313131313131,
+      "loss": 2.6179,
+      "theoretical_loss": 3.4571540942648946,
+      "tokens_seen": 1823539200
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045511111111111116,
+      "loss": 2.6814,
+      "theoretical_loss": 3.4571435459923334,
+      "tokens_seen": 1823604736
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004550909090909091,
+      "loss": 2.4929,
+      "theoretical_loss": 3.457132998204984,
+      "tokens_seen": 1823670272
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045507070707070707,
+      "loss": 2.7795,
+      "theoretical_loss": 3.457122450902805,
+      "tokens_seen": 1823735808
+    },
+    {
+      "epoch": 0.1,
+      "objective/train/docs_used": 1004868,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.2328946590423584,
+      "objective/train/theoretical_loss": 3.4571119040857576,
+      "objective/train/tokens_used": 182660576,
+      "theoretical_loss": 3.4571119040857576,
+      "tokens_seen": 1823801344
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045505050505050503,
+      "loss": 2.337,
+      "theoretical_loss": 3.4571119040857576,
+      "tokens_seen": 1823801344
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045503030303030304,
+      "loss": 2.9013,
+      "theoretical_loss": 3.4571013577538015,
+      "tokens_seen": 1823866880
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045501010101010105,
+      "loss": 2.7025,
+      "theoretical_loss": 3.457090811906898,
+      "tokens_seen": 1823932416
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000454989898989899,
+      "loss": 2.9696,
+      "theoretical_loss": 3.457080266545006,
+      "tokens_seen": 1823997952
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045496969696969696,
+      "loss": 2.6379,
+      "theoretical_loss": 3.457069721668087,
+      "tokens_seen": 1824063488
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004549494949494949,
+      "loss": 2.7066,
+      "theoretical_loss": 3.4570591772761006,
+      "tokens_seen": 1824129024
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045492929292929293,
+      "loss": 2.756,
+      "theoretical_loss": 3.4570486333690065,
+      "tokens_seen": 1824194560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045490909090909094,
+      "loss": 2.7721,
+      "theoretical_loss": 3.4570380899467663,
+      "tokens_seen": 1824260096
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004548888888888889,
+      "loss": 2.7186,
+      "theoretical_loss": 3.4570275470093397,
+      "tokens_seen": 1824325632
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045486868686868686,
+      "loss": 2.7226,
+      "theoretical_loss": 3.457017004556687,
+      "tokens_seen": 1824391168
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045484848484848487,
+      "loss": 2.6073,
+      "theoretical_loss": 3.457006462588768,
+      "tokens_seen": 1824456704
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004548282828282828,
+      "loss": 2.5764,
+      "theoretical_loss": 3.456995921105544,
+      "tokens_seen": 1824522240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045480808080808083,
+      "loss": 2.7873,
+      "theoretical_loss": 3.4569853801069743,
+      "tokens_seen": 1824587776
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004547878787878788,
+      "loss": 2.6969,
+      "theoretical_loss": 3.4569748395930198,
+      "tokens_seen": 1824653312
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045476767676767675,
+      "loss": 2.7108,
+      "theoretical_loss": 3.4569642995636407,
+      "tokens_seen": 1824718848
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045474747474747476,
+      "loss": 2.8133,
+      "theoretical_loss": 3.4569537600187976,
+      "tokens_seen": 1824784384
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004547272727272727,
+      "loss": 2.5824,
+      "theoretical_loss": 3.4569432209584496,
+      "tokens_seen": 1824849920
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045470707070707073,
+      "loss": 2.7374,
+      "theoretical_loss": 3.4569326823825586,
+      "tokens_seen": 1824915456
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004546868686868687,
+      "loss": 2.4365,
+      "theoretical_loss": 3.4569221442910845,
+      "tokens_seen": 1824980992
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004546666666666667,
+      "loss": 2.9807,
+      "theoretical_loss": 3.4569116066839873,
+      "tokens_seen": 1825046528
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045464646464646465,
+      "loss": 2.7103,
+      "theoretical_loss": 3.4569010695612272,
+      "tokens_seen": 1825112064
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004546262626262626,
+      "loss": 2.7703,
+      "theoretical_loss": 3.456890532922765,
+      "tokens_seen": 1825177600
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004546060606060606,
+      "loss": 2.5735,
+      "theoretical_loss": 3.456879996768561,
+      "tokens_seen": 1825243136
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004545858585858586,
+      "loss": 2.5735,
+      "theoretical_loss": 3.4568694610985755,
+      "tokens_seen": 1825308672
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004545656565656566,
+      "loss": 2.5422,
+      "theoretical_loss": 3.456858925912768,
+      "tokens_seen": 1825374208
+    },
+    {
+      "debugging/Self-BLEU-5": 0.5101952161435601,
+      "debugging/distinct-1-grams": 0.7564583571701602,
+      "debugging/distinct-2-grams": 0.9365913780414189,
+      "debugging/entropy-1-grams": 5.753082263870947,
+      "debugging/entropy-2-grams": 6.573169152318852,
+      "debugging/length": 571.4,
+      "debugging/num_segments": 10,
+      "epoch": 0.1,
+      "objective/train/docs_used": 1006237,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5774552822113037,
+      "objective/train/theoretical_loss": 3.4568483912111008,
+      "objective/train/tokens_used": 184298976,
+      "theoretical_loss": 3.4568483912111008,
+      "tokens_seen": 1825439744
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045454545454545455,
+      "loss": 2.8626,
+      "theoretical_loss": 3.4568483912111008,
+      "tokens_seen": 1825439744
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004545252525252525,
+      "loss": 2.7059,
+      "theoretical_loss": 3.4568378569935323,
+      "tokens_seen": 1825505280
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004545050505050505,
+      "loss": 2.4361,
+      "theoretical_loss": 3.4568273232600237,
+      "tokens_seen": 1825570816
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004544848484848485,
+      "loss": 2.6962,
+      "theoretical_loss": 3.456816790010536,
+      "tokens_seen": 1825636352
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004544646464646465,
+      "loss": 2.8448,
+      "theoretical_loss": 3.4568062572450287,
+      "tokens_seen": 1825701888
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045444444444444444,
+      "loss": 2.7327,
+      "theoretical_loss": 3.456795724963462,
+      "tokens_seen": 1825767424
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004544242424242424,
+      "loss": 2.7395,
+      "theoretical_loss": 3.456785193165797,
+      "tokens_seen": 1825832960
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004544040404040404,
+      "loss": 2.7404,
+      "theoretical_loss": 3.456774661851995,
+      "tokens_seen": 1825898496
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004543838383838384,
+      "loss": 2.6062,
+      "theoretical_loss": 3.4567641310220134,
+      "tokens_seen": 1825964032
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004543636363636364,
+      "loss": 2.7898,
+      "theoretical_loss": 3.4567536006758157,
+      "tokens_seen": 1826029568
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045434343434343433,
+      "loss": 2.6872,
+      "theoretical_loss": 3.456743070813361,
+      "tokens_seen": 1826095104
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004543232323232323,
+      "loss": 2.5771,
+      "theoretical_loss": 3.4567325414346093,
+      "tokens_seen": 1826160640
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045430303030303035,
+      "loss": 2.8704,
+      "theoretical_loss": 3.4567220125395215,
+      "tokens_seen": 1826226176
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004542828282828283,
+      "loss": 2.7221,
+      "theoretical_loss": 3.456711484128059,
+      "tokens_seen": 1826291712
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045426262626262627,
+      "loss": 2.7092,
+      "theoretical_loss": 3.45670095620018,
+      "tokens_seen": 1826357248
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004542424242424242,
+      "loss": 2.6278,
+      "theoretical_loss": 3.456690428755847,
+      "tokens_seen": 1826422784
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045422222222222223,
+      "loss": 2.6251,
+      "theoretical_loss": 3.4566799017950194,
+      "tokens_seen": 1826488320
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045420202020202025,
+      "loss": 2.5685,
+      "theoretical_loss": 3.4566693753176576,
+      "tokens_seen": 1826553856
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004541818181818182,
+      "loss": 2.5096,
+      "theoretical_loss": 3.4566588493237234,
+      "tokens_seen": 1826619392
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045416161616161616,
+      "loss": 2.6342,
+      "theoretical_loss": 3.4566483238131753,
+      "tokens_seen": 1826684928
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004541414141414141,
+      "loss": 2.8359,
+      "theoretical_loss": 3.456637798785975,
+      "tokens_seen": 1826750464
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004541212121212122,
+      "loss": 2.5948,
+      "theoretical_loss": 3.4566272742420825,
+      "tokens_seen": 1826816000
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045410101010101014,
+      "loss": 2.586,
+      "theoretical_loss": 3.4566167501814586,
+      "tokens_seen": 1826881536
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004540808080808081,
+      "loss": 2.5007,
+      "theoretical_loss": 3.4566062266040634,
+      "tokens_seen": 1826947072
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045406060606060605,
+      "loss": 2.7515,
+      "theoretical_loss": 3.4565957035098576,
+      "tokens_seen": 1827012608
+    },
+    {
+      "epoch": 0.1,
+      "objective/train/docs_used": 1006887,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7742507457733154,
+      "objective/train/theoretical_loss": 3.4565851808988013,
+      "objective/train/tokens_used": 185937376,
+      "theoretical_loss": 3.4565851808988013,
+      "tokens_seen": 1827078144
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000454040404040404,
+      "loss": 2.6239,
+      "theoretical_loss": 3.4565851808988013,
+      "tokens_seen": 1827078144
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004540202020202021,
+      "loss": 2.68,
+      "theoretical_loss": 3.4565746587708563,
+      "tokens_seen": 1827143680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045400000000000003,
+      "loss": 2.5783,
+      "theoretical_loss": 3.456564137125981,
+      "tokens_seen": 1827209216
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000453979797979798,
+      "loss": 2.852,
+      "theoretical_loss": 3.456553615964138,
+      "tokens_seen": 1827274752
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045395959595959594,
+      "loss": 2.5331,
+      "theoretical_loss": 3.4565430952852862,
+      "tokens_seen": 1827340288
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045393939393939396,
+      "loss": 2.4804,
+      "theoretical_loss": 3.456532575089387,
+      "tokens_seen": 1827405824
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045391919191919197,
+      "loss": 2.7907,
+      "theoretical_loss": 3.4565220553764004,
+      "tokens_seen": 1827471360
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004538989898989899,
+      "loss": 2.5951,
+      "theoretical_loss": 3.4565115361462873,
+      "tokens_seen": 1827536896
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004538787878787879,
+      "loss": 2.5612,
+      "theoretical_loss": 3.456501017399008,
+      "tokens_seen": 1827602432
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045385858585858584,
+      "loss": 2.5128,
+      "theoretical_loss": 3.4564904991345236,
+      "tokens_seen": 1827667968
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045383838383838385,
+      "loss": 2.7096,
+      "theoretical_loss": 3.4564799813527936,
+      "tokens_seen": 1827733504
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045381818181818186,
+      "loss": 2.6215,
+      "theoretical_loss": 3.4564694640537788,
+      "tokens_seen": 1827799040
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004537979797979798,
+      "loss": 2.6144,
+      "theoretical_loss": 3.4564589472374405,
+      "tokens_seen": 1827864576
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004537777777777778,
+      "loss": 2.4467,
+      "theoretical_loss": 3.4564484309037384,
+      "tokens_seen": 1827930112
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004537575757575758,
+      "loss": 2.8435,
+      "theoretical_loss": 3.456437915052634,
+      "tokens_seen": 1827995648
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045373737373737374,
+      "loss": 2.7307,
+      "theoretical_loss": 3.456427399684087,
+      "tokens_seen": 1828061184
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045371717171717175,
+      "loss": 2.4665,
+      "theoretical_loss": 3.456416884798058,
+      "tokens_seen": 1828126720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004536969696969697,
+      "loss": 2.6755,
+      "theoretical_loss": 3.456406370394508,
+      "tokens_seen": 1828192256
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045367676767676767,
+      "loss": 2.6087,
+      "theoretical_loss": 3.456395856473397,
+      "tokens_seen": 1828257792
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004536565656565657,
+      "loss": 2.4562,
+      "theoretical_loss": 3.456385343034686,
+      "tokens_seen": 1828323328
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045363636363636363,
+      "loss": 2.5869,
+      "theoretical_loss": 3.456374830078336,
+      "tokens_seen": 1828388864
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045361616161616164,
+      "loss": 2.6877,
+      "theoretical_loss": 3.456364317604306,
+      "tokens_seen": 1828454400
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004535959595959596,
+      "loss": 2.7657,
+      "theoretical_loss": 3.456353805612559,
+      "tokens_seen": 1828519936
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004535757575757576,
+      "loss": 2.8022,
+      "theoretical_loss": 3.4563432941030534,
+      "tokens_seen": 1828585472
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045355555555555557,
+      "loss": 2.6839,
+      "theoretical_loss": 3.4563327830757506,
+      "tokens_seen": 1828651008
+    },
+    {
+      "epoch": 0.1,
+      "objective/train/docs_used": 1008261,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.3774466514587402,
+      "objective/train/theoretical_loss": 3.4563222725306115,
+      "objective/train/tokens_used": 187575776,
+      "theoretical_loss": 3.4563222725306115,
+      "tokens_seen": 1828716544
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004535353535353535,
+      "loss": 2.6512,
+      "theoretical_loss": 3.4563222725306115,
+      "tokens_seen": 1828716544
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045351515151515154,
+      "loss": 2.8056,
+      "theoretical_loss": 3.4563117624675965,
+      "tokens_seen": 1828782080
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004534949494949495,
+      "loss": 2.4355,
+      "theoretical_loss": 3.456301252886666,
+      "tokens_seen": 1828847616
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004534747474747475,
+      "loss": 2.7577,
+      "theoretical_loss": 3.456290743787781,
+      "tokens_seen": 1828913152
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045345454545454546,
+      "loss": 2.6925,
+      "theoretical_loss": 3.456280235170902,
+      "tokens_seen": 1828978688
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004534343434343434,
+      "loss": 2.8243,
+      "theoretical_loss": 3.4562697270359886,
+      "tokens_seen": 1829044224
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045341414141414143,
+      "loss": 2.7169,
+      "theoretical_loss": 3.456259219383003,
+      "tokens_seen": 1829109760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045339393939393944,
+      "loss": 2.5994,
+      "theoretical_loss": 3.456248712211905,
+      "tokens_seen": 1829175296
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004533737373737374,
+      "loss": 2.7856,
+      "theoretical_loss": 3.4562382055226557,
+      "tokens_seen": 1829240832
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045335353535353535,
+      "loss": 2.584,
+      "theoretical_loss": 3.456227699315215,
+      "tokens_seen": 1829306368
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004533333333333333,
+      "loss": 2.7333,
+      "theoretical_loss": 3.456217193589544,
+      "tokens_seen": 1829371904
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004533131313131313,
+      "loss": 2.5758,
+      "theoretical_loss": 3.4562066883456035,
+      "tokens_seen": 1829437440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045329292929292933,
+      "loss": 2.6453,
+      "theoretical_loss": 3.456196183583354,
+      "tokens_seen": 1829502976
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004532727272727273,
+      "loss": 2.7702,
+      "theoretical_loss": 3.4561856793027568,
+      "tokens_seen": 1829568512
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045325252525252525,
+      "loss": 2.65,
+      "theoretical_loss": 3.4561751755037706,
+      "tokens_seen": 1829634048
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004532323232323232,
+      "loss": 2.7699,
+      "theoretical_loss": 3.4561646721863584,
+      "tokens_seen": 1829699584
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004532121212121212,
+      "loss": 2.6576,
+      "theoretical_loss": 3.4561541693504796,
+      "tokens_seen": 1829765120
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004531919191919192,
+      "loss": 2.5966,
+      "theoretical_loss": 3.4561436669960948,
+      "tokens_seen": 1829830656
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004531717171717172,
+      "loss": 2.5938,
+      "theoretical_loss": 3.4561331651231653,
+      "tokens_seen": 1829896192
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045315151515151514,
+      "loss": 2.7886,
+      "theoretical_loss": 3.4561226637316516,
+      "tokens_seen": 1829961728
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004531313131313131,
+      "loss": 2.7663,
+      "theoretical_loss": 3.456112162821514,
+      "tokens_seen": 1830027264
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045311111111111116,
+      "loss": 2.7715,
+      "theoretical_loss": 3.456101662392714,
+      "tokens_seen": 1830092800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004530909090909091,
+      "loss": 2.5694,
+      "theoretical_loss": 3.4560911624452113,
+      "tokens_seen": 1830158336
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004530707070707071,
+      "loss": 2.5395,
+      "theoretical_loss": 3.456080662978967,
+      "tokens_seen": 1830223872
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045305050505050503,
+      "loss": 2.4467,
+      "theoretical_loss": 3.4560701639939424,
+      "tokens_seen": 1830289408
+    },
+    {
+      "epoch": 0.1,
+      "objective/train/docs_used": 1009482,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.900181770324707,
+      "objective/train/theoretical_loss": 3.4560596654900975,
+      "objective/train/tokens_used": 189214176,
+      "theoretical_loss": 3.4560596654900975,
+      "tokens_seen": 1830354944
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000453030303030303,
+      "loss": 2.6643,
+      "theoretical_loss": 3.4560596654900975,
+      "tokens_seen": 1830354944
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045301010101010106,
+      "loss": 2.6989,
+      "theoretical_loss": 3.456049167467393,
+      "tokens_seen": 1830420480
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000452989898989899,
+      "loss": 2.6867,
+      "theoretical_loss": 3.45603866992579,
+      "tokens_seen": 1830486016
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045296969696969697,
+      "loss": 2.452,
+      "theoretical_loss": 3.456028172865249,
+      "tokens_seen": 1830551552
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004529494949494949,
+      "loss": 2.9193,
+      "theoretical_loss": 3.4560176762857315,
+      "tokens_seen": 1830617088
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045292929292929294,
+      "loss": 2.5696,
+      "theoretical_loss": 3.456007180187197,
+      "tokens_seen": 1830682624
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045290909090909095,
+      "loss": 2.7878,
+      "theoretical_loss": 3.455996684569607,
+      "tokens_seen": 1830748160
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004528888888888889,
+      "loss": 2.4411,
+      "theoretical_loss": 3.4559861894329216,
+      "tokens_seen": 1830813696
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045286868686868686,
+      "loss": 2.7437,
+      "theoretical_loss": 3.4559756947771025,
+      "tokens_seen": 1830879232
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004528484848484848,
+      "loss": 2.7311,
+      "theoretical_loss": 3.45596520060211,
+      "tokens_seen": 1830944768
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045282828282828283,
+      "loss": 2.6225,
+      "theoretical_loss": 3.4559547069079044,
+      "tokens_seen": 1831010304
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045280808080808084,
+      "loss": 2.6697,
+      "theoretical_loss": 3.455944213694447,
+      "tokens_seen": 1831075840
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004527878787878788,
+      "loss": 2.5926,
+      "theoretical_loss": 3.4559337209616987,
+      "tokens_seen": 1831141376
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045276767676767675,
+      "loss": 2.7966,
+      "theoretical_loss": 3.45592322870962,
+      "tokens_seen": 1831206912
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045274747474747477,
+      "loss": 2.5301,
+      "theoretical_loss": 3.4559127369381715,
+      "tokens_seen": 1831272448
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004527272727272727,
+      "loss": 2.6866,
+      "theoretical_loss": 3.4559022456473145,
+      "tokens_seen": 1831337984
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045270707070707073,
+      "loss": 2.681,
+      "theoretical_loss": 3.455891754837009,
+      "tokens_seen": 1831403520
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004526868686868687,
+      "loss": 2.3233,
+      "theoretical_loss": 3.4558812645072168,
+      "tokens_seen": 1831469056
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045266666666666665,
+      "loss": 2.6848,
+      "theoretical_loss": 3.455870774657898,
+      "tokens_seen": 1831534592
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045264646464646466,
+      "loss": 2.6255,
+      "theoretical_loss": 3.4558602852890132,
+      "tokens_seen": 1831600128
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004526262626262626,
+      "loss": 2.7053,
+      "theoretical_loss": 3.4558497964005244,
+      "tokens_seen": 1831665664
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004526060606060606,
+      "loss": 2.6355,
+      "theoretical_loss": 3.455839307992391,
+      "tokens_seen": 1831731200
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004525858585858586,
+      "loss": 2.7105,
+      "theoretical_loss": 3.455828820064575,
+      "tokens_seen": 1831796736
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004525656565656566,
+      "loss": 2.8606,
+      "theoretical_loss": 3.455818332617036,
+      "tokens_seen": 1831862272
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045254545454545455,
+      "loss": 2.9341,
+      "theoretical_loss": 3.455807845649735,
+      "tokens_seen": 1831927808
+    },
+    {
+      "epoch": 0.1,
+      "objective/train/docs_used": 1009995,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7280309200286865,
+      "objective/train/theoretical_loss": 3.4557973591626343,
+      "objective/train/tokens_used": 190852576,
+      "theoretical_loss": 3.4557973591626343,
+      "tokens_seen": 1831993344
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045252525252525256,
+      "loss": 2.636,
+      "theoretical_loss": 3.4557973591626343,
+      "tokens_seen": 1831993344
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004525050505050505,
+      "loss": 2.6817,
+      "theoretical_loss": 3.4557868731556933,
+      "tokens_seen": 1832058880
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004524848484848485,
+      "loss": 2.5711,
+      "theoretical_loss": 3.455776387628873,
+      "tokens_seen": 1832124416
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004524646464646465,
+      "loss": 2.5362,
+      "theoretical_loss": 3.4557659025821352,
+      "tokens_seen": 1832189952
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045244444444444444,
+      "loss": 2.7667,
+      "theoretical_loss": 3.45575541801544,
+      "tokens_seen": 1832255488
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045242424242424245,
+      "loss": 2.78,
+      "theoretical_loss": 3.455744933928748,
+      "tokens_seen": 1832321024
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004524040404040404,
+      "loss": 2.6392,
+      "theoretical_loss": 3.45573445032202,
+      "tokens_seen": 1832386560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004523838383838384,
+      "loss": 2.5771,
+      "theoretical_loss": 3.455723967195218,
+      "tokens_seen": 1832452096
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004523636363636364,
+      "loss": 2.3533,
+      "theoretical_loss": 3.4557134845483013,
+      "tokens_seen": 1832517632
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045234343434343434,
+      "loss": 2.5183,
+      "theoretical_loss": 3.4557030023812323,
+      "tokens_seen": 1832583168
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045232323232323235,
+      "loss": 2.7558,
+      "theoretical_loss": 3.4556925206939706,
+      "tokens_seen": 1832648704
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004523030303030303,
+      "loss": 2.6891,
+      "theoretical_loss": 3.455682039486478,
+      "tokens_seen": 1832714240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004522828282828283,
+      "loss": 2.5169,
+      "theoretical_loss": 3.455671558758715,
+      "tokens_seen": 1832779776
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045226262626262627,
+      "loss": 2.6283,
+      "theoretical_loss": 3.4556610785106425,
+      "tokens_seen": 1832845312
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045224242424242423,
+      "loss": 2.7893,
+      "theoretical_loss": 3.455650598742221,
+      "tokens_seen": 1832910848
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045222222222222224,
+      "loss": 2.8487,
+      "theoretical_loss": 3.4556401194534123,
+      "tokens_seen": 1832976384
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045220202020202025,
+      "loss": 2.5464,
+      "theoretical_loss": 3.4556296406441764,
+      "tokens_seen": 1833041920
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004521818181818182,
+      "loss": 2.7207,
+      "theoretical_loss": 3.4556191623144747,
+      "tokens_seen": 1833107456
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045216161616161616,
+      "loss": 2.5345,
+      "theoretical_loss": 3.455608684464268,
+      "tokens_seen": 1833172992
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004521414141414141,
+      "loss": 2.5418,
+      "theoretical_loss": 3.4555982070935176,
+      "tokens_seen": 1833238528
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045212121212121213,
+      "loss": 2.877,
+      "theoretical_loss": 3.4555877302021836,
+      "tokens_seen": 1833304064
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045210101010101014,
+      "loss": 2.577,
+      "theoretical_loss": 3.455577253790228,
+      "tokens_seen": 1833369600
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004520808080808081,
+      "loss": 2.6027,
+      "theoretical_loss": 3.4555667778576105,
+      "tokens_seen": 1833435136
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00045206060606060606,
+      "loss": 2.8754,
+      "theoretical_loss": 3.455556302404293,
+      "tokens_seen": 1833500672
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000452040404040404,
+      "loss": 2.8005,
+      "theoretical_loss": 3.4555458274302353,
+      "tokens_seen": 1833566208
+    },
+    {
+      "epoch": 0.1,
+      "objective/train/docs_used": 1010832,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6558239459991455,
+      "objective/train/theoretical_loss": 3.4555353529354003,
+      "objective/train/tokens_used": 192490976,
+      "theoretical_loss": 3.4555353529354003,
+      "tokens_seen": 1833631744
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0004520202020202021,
+      "loss": 2.4985,
+      "theoretical_loss": 3.4555353529354003,
+      "tokens_seen": 1833631744
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045200000000000004,
+      "loss": 2.7386,
+      "theoretical_loss": 3.4555248789197472,
+      "tokens_seen": 1833697280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000451979797979798,
+      "loss": 2.6204,
+      "theoretical_loss": 3.455514405383237,
+      "tokens_seen": 1833762816
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045195959595959595,
+      "loss": 2.7539,
+      "theoretical_loss": 3.455503932325832,
+      "tokens_seen": 1833828352
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004519393939393939,
+      "loss": 2.7639,
+      "theoretical_loss": 3.4554934597474922,
+      "tokens_seen": 1833893888
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045191919191919197,
+      "loss": 2.6223,
+      "theoretical_loss": 3.4554829876481787,
+      "tokens_seen": 1833959424
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045189898989898993,
+      "loss": 2.7164,
+      "theoretical_loss": 3.4554725160278528,
+      "tokens_seen": 1834024960
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004518787878787879,
+      "loss": 2.3732,
+      "theoretical_loss": 3.4554620448864743,
+      "tokens_seen": 1834090496
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045185858585858584,
+      "loss": 2.6354,
+      "theoretical_loss": 3.4554515742240057,
+      "tokens_seen": 1834156032
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004518383838383838,
+      "loss": 2.6698,
+      "theoretical_loss": 3.4554411040404074,
+      "tokens_seen": 1834221568
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045181818181818186,
+      "loss": 2.6226,
+      "theoretical_loss": 3.45543063433564,
+      "tokens_seen": 1834287104
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004517979797979798,
+      "loss": 2.6863,
+      "theoretical_loss": 3.4554201651096648,
+      "tokens_seen": 1834352640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004517777777777778,
+      "loss": 2.6369,
+      "theoretical_loss": 3.455409696362443,
+      "tokens_seen": 1834418176
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045175757575757574,
+      "loss": 2.5681,
+      "theoretical_loss": 3.4553992280939356,
+      "tokens_seen": 1834483712
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045173737373737375,
+      "loss": 2.6862,
+      "theoretical_loss": 3.4553887603041034,
+      "tokens_seen": 1834549248
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045171717171717176,
+      "loss": 2.5591,
+      "theoretical_loss": 3.4553782929929073,
+      "tokens_seen": 1834614784
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004516969696969697,
+      "loss": 2.6327,
+      "theoretical_loss": 3.4553678261603085,
+      "tokens_seen": 1834680320
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045167676767676767,
+      "loss": 2.6442,
+      "theoretical_loss": 3.455357359806268,
+      "tokens_seen": 1834745856
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045165656565656563,
+      "loss": 2.6768,
+      "theoretical_loss": 3.455346893930747,
+      "tokens_seen": 1834811392
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045163636363636364,
+      "loss": 2.4703,
+      "theoretical_loss": 3.4553364285337063,
+      "tokens_seen": 1834876928
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045161616161616165,
+      "loss": 2.7648,
+      "theoretical_loss": 3.4553259636151066,
+      "tokens_seen": 1834942464
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004515959595959596,
+      "loss": 2.6158,
+      "theoretical_loss": 3.45531549917491,
+      "tokens_seen": 1835008000
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045157575757575756,
+      "loss": 2.8335,
+      "theoretical_loss": 3.455305035213076,
+      "tokens_seen": 1835073536
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004515555555555556,
+      "loss": 2.7982,
+      "theoretical_loss": 3.455294571729567,
+      "tokens_seen": 1835139072
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045153535353535353,
+      "loss": 2.8067,
+      "theoretical_loss": 3.455284108724344,
+      "tokens_seen": 1835204608
+    },
+    {
+      "epoch": 0.11,
+      "objective/train/docs_used": 1011110,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.309302806854248,
+      "objective/train/theoretical_loss": 3.4552736461973668,
+      "objective/train/tokens_used": 194129376,
+      "theoretical_loss": 3.4552736461973668,
+      "tokens_seen": 1835270144
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045151515151515154,
+      "loss": 2.5007,
+      "theoretical_loss": 3.4552736461973668,
+      "tokens_seen": 1835270144
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004514949494949495,
+      "loss": 2.5452,
+      "theoretical_loss": 3.455263184148598,
+      "tokens_seen": 1835335680
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045147474747474746,
+      "loss": 2.6689,
+      "theoretical_loss": 3.455252722577997,
+      "tokens_seen": 1835401216
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045145454545454547,
+      "loss": 2.6253,
+      "theoretical_loss": 3.455242261485527,
+      "tokens_seen": 1835466752
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004514343434343434,
+      "loss": 2.4733,
+      "theoretical_loss": 3.4552318008711476,
+      "tokens_seen": 1835532288
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045141414141414144,
+      "loss": 2.4469,
+      "theoretical_loss": 3.45522134073482,
+      "tokens_seen": 1835597824
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004513939393939394,
+      "loss": 2.7587,
+      "theoretical_loss": 3.455210881076505,
+      "tokens_seen": 1835663360
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004513737373737374,
+      "loss": 2.5127,
+      "theoretical_loss": 3.455200421896165,
+      "tokens_seen": 1835728896
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045135353535353536,
+      "loss": 2.8253,
+      "theoretical_loss": 3.45518996319376,
+      "tokens_seen": 1835794432
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004513333333333333,
+      "loss": 2.5292,
+      "theoretical_loss": 3.4551795049692515,
+      "tokens_seen": 1835859968
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045131313131313133,
+      "loss": 2.7594,
+      "theoretical_loss": 3.4551690472226007,
+      "tokens_seen": 1835925504
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004512929292929293,
+      "loss": 2.6391,
+      "theoretical_loss": 3.455158589953768,
+      "tokens_seen": 1835991040
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004512727272727273,
+      "loss": 2.6793,
+      "theoretical_loss": 3.455148133162715,
+      "tokens_seen": 1836056576
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045125252525252525,
+      "loss": 2.7743,
+      "theoretical_loss": 3.455137676849403,
+      "tokens_seen": 1836122112
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004512323232323232,
+      "loss": 2.7776,
+      "theoretical_loss": 3.455127221013793,
+      "tokens_seen": 1836187648
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004512121212121212,
+      "loss": 2.7312,
+      "theoretical_loss": 3.4551167656558457,
+      "tokens_seen": 1836253184
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045119191919191923,
+      "loss": 2.8942,
+      "theoretical_loss": 3.4551063107755233,
+      "tokens_seen": 1836318720
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004511717171717172,
+      "loss": 2.6964,
+      "theoretical_loss": 3.4550958563727856,
+      "tokens_seen": 1836384256
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045115151515151515,
+      "loss": 2.4926,
+      "theoretical_loss": 3.4550854024475948,
+      "tokens_seen": 1836449792
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004511313131313131,
+      "loss": 2.8051,
+      "theoretical_loss": 3.4550749489999113,
+      "tokens_seen": 1836515328
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004511111111111111,
+      "loss": 2.7998,
+      "theoretical_loss": 3.4550644960296966,
+      "tokens_seen": 1836580864
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004510909090909091,
+      "loss": 2.4606,
+      "theoretical_loss": 3.455054043536912,
+      "tokens_seen": 1836646400
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004510707070707071,
+      "loss": 2.5874,
+      "theoretical_loss": 3.455043591521518,
+      "tokens_seen": 1836711936
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045105050505050504,
+      "loss": 2.5794,
+      "theoretical_loss": 3.4550331399834766,
+      "tokens_seen": 1836777472
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000451030303030303,
+      "loss": 2.7082,
+      "theoretical_loss": 3.4550226889227487,
+      "tokens_seen": 1836843008
+    },
+    {
+      "epoch": 0.11,
+      "objective/train/docs_used": 1012365,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6850903034210205,
+      "objective/train/theoretical_loss": 3.455012238339296,
+      "objective/train/tokens_used": 195767776,
+      "theoretical_loss": 3.455012238339296,
+      "tokens_seen": 1836908544
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045101010101010106,
+      "loss": 2.617,
+      "theoretical_loss": 3.455012238339296,
+      "tokens_seen": 1836908544
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000450989898989899,
+      "loss": 2.8187,
+      "theoretical_loss": 3.455001788233078,
+      "tokens_seen": 1836974080
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000450969696969697,
+      "loss": 2.7156,
+      "theoretical_loss": 3.4549913386040574,
+      "tokens_seen": 1837039616
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045094949494949493,
+      "loss": 2.5967,
+      "theoretical_loss": 3.454980889452195,
+      "tokens_seen": 1837105152
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004509292929292929,
+      "loss": 2.4446,
+      "theoretical_loss": 3.4549704407774517,
+      "tokens_seen": 1837170688
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045090909090909095,
+      "loss": 2.6417,
+      "theoretical_loss": 3.454959992579789,
+      "tokens_seen": 1837236224
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004508888888888889,
+      "loss": 2.592,
+      "theoretical_loss": 3.4549495448591685,
+      "tokens_seen": 1837301760
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045086868686868687,
+      "loss": 2.8509,
+      "theoretical_loss": 3.4549390976155503,
+      "tokens_seen": 1837367296
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004508484848484848,
+      "loss": 2.6198,
+      "theoretical_loss": 3.4549286508488963,
+      "tokens_seen": 1837432832
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004508282828282829,
+      "loss": 2.4963,
+      "theoretical_loss": 3.454918204559168,
+      "tokens_seen": 1837498368
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045080808080808085,
+      "loss": 2.5869,
+      "theoretical_loss": 3.4549077587463257,
+      "tokens_seen": 1837563904
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004507878787878788,
+      "loss": 2.6174,
+      "theoretical_loss": 3.4548973134103322,
+      "tokens_seen": 1837629440
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045076767676767676,
+      "loss": 2.6871,
+      "theoretical_loss": 3.4548868685511467,
+      "tokens_seen": 1837694976
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004507474747474747,
+      "loss": 2.8478,
+      "theoretical_loss": 3.4548764241687318,
+      "tokens_seen": 1837760512
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004507272727272728,
+      "loss": 2.5321,
+      "theoretical_loss": 3.4548659802630484,
+      "tokens_seen": 1837826048
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045070707070707074,
+      "loss": 2.7471,
+      "theoretical_loss": 3.454855536834058,
+      "tokens_seen": 1837891584
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004506868686868687,
+      "loss": 2.8312,
+      "theoretical_loss": 3.4548450938817212,
+      "tokens_seen": 1837957120
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045066666666666665,
+      "loss": 2.5958,
+      "theoretical_loss": 3.454834651406,
+      "tokens_seen": 1838022656
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045064646464646466,
+      "loss": 2.6896,
+      "theoretical_loss": 3.454824209406855,
+      "tokens_seen": 1838088192
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004506262626262627,
+      "loss": 2.7956,
+      "theoretical_loss": 3.4548137678842474,
+      "tokens_seen": 1838153728
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045060606060606063,
+      "loss": 2.626,
+      "theoretical_loss": 3.4548033268381397,
+      "tokens_seen": 1838219264
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004505858585858586,
+      "loss": 2.7092,
+      "theoretical_loss": 3.454792886268492,
+      "tokens_seen": 1838284800
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045056565656565655,
+      "loss": 2.8204,
+      "theoretical_loss": 3.4547824461752654,
+      "tokens_seen": 1838350336
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045054545454545456,
+      "loss": 2.9183,
+      "theoretical_loss": 3.4547720065584215,
+      "tokens_seen": 1838415872
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045052525252525257,
+      "loss": 2.7341,
+      "theoretical_loss": 3.454761567417922,
+      "tokens_seen": 1838481408
+    },
+    {
+      "epoch": 0.11,
+      "objective/train/docs_used": 1013519,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.807917356491089,
+      "objective/train/theoretical_loss": 3.454751128753728,
+      "objective/train/tokens_used": 197406176,
+      "theoretical_loss": 3.454751128753728,
+      "tokens_seen": 1838546944
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004505050505050505,
+      "loss": 2.8162,
+      "theoretical_loss": 3.454751128753728,
+      "tokens_seen": 1838546944
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004504848484848485,
+      "loss": 2.6302,
+      "theoretical_loss": 3.4547406905658002,
+      "tokens_seen": 1838612480
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004504646464646465,
+      "loss": 2.6506,
+      "theoretical_loss": 3.454730252854101,
+      "tokens_seen": 1838678016
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045044444444444445,
+      "loss": 2.7298,
+      "theoretical_loss": 3.454719815618591,
+      "tokens_seen": 1838743552
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045042424242424246,
+      "loss": 2.7674,
+      "theoretical_loss": 3.4547093788592314,
+      "tokens_seen": 1838809088
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004504040404040404,
+      "loss": 2.5375,
+      "theoretical_loss": 3.4546989425759835,
+      "tokens_seen": 1838874624
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004503838383838384,
+      "loss": 2.5675,
+      "theoretical_loss": 3.454688506768809,
+      "tokens_seen": 1838940160
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004503636363636364,
+      "loss": 2.8516,
+      "theoretical_loss": 3.4546780714376695,
+      "tokens_seen": 1839005696
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045034343434343434,
+      "loss": 2.5366,
+      "theoretical_loss": 3.454667636582525,
+      "tokens_seen": 1839071232
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045032323232323235,
+      "loss": 2.511,
+      "theoretical_loss": 3.4546572022033377,
+      "tokens_seen": 1839136768
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004503030303030303,
+      "loss": 2.8897,
+      "theoretical_loss": 3.4546467683000692,
+      "tokens_seen": 1839202304
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004502828282828283,
+      "loss": 2.7794,
+      "theoretical_loss": 3.4546363348726805,
+      "tokens_seen": 1839267840
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004502626262626263,
+      "loss": 2.832,
+      "theoretical_loss": 3.4546259019211334,
+      "tokens_seen": 1839333376
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045024242424242423,
+      "loss": 2.6955,
+      "theoretical_loss": 3.4546154694453883,
+      "tokens_seen": 1839398912
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045022222222222225,
+      "loss": 2.6539,
+      "theoretical_loss": 3.454605037445407,
+      "tokens_seen": 1839464448
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004502020202020202,
+      "loss": 2.4324,
+      "theoretical_loss": 3.454594605921151,
+      "tokens_seen": 1839529984
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004501818181818182,
+      "loss": 2.6089,
+      "theoretical_loss": 3.454584174872582,
+      "tokens_seen": 1839595520
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045016161616161617,
+      "loss": 2.7772,
+      "theoretical_loss": 3.454573744299661,
+      "tokens_seen": 1839661056
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004501414141414141,
+      "loss": 2.7491,
+      "theoretical_loss": 3.454563314202349,
+      "tokens_seen": 1839726592
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045012121212121214,
+      "loss": 2.7735,
+      "theoretical_loss": 3.454552884580607,
+      "tokens_seen": 1839792128
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045010101010101015,
+      "loss": 2.666,
+      "theoretical_loss": 3.4545424554343978,
+      "tokens_seen": 1839857664
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004500808080808081,
+      "loss": 2.7511,
+      "theoretical_loss": 3.454532026763682,
+      "tokens_seen": 1839923200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045006060606060606,
+      "loss": 2.4584,
+      "theoretical_loss": 3.4545215985684212,
+      "tokens_seen": 1839988736
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000450040404040404,
+      "loss": 2.7286,
+      "theoretical_loss": 3.4545111708485763,
+      "tokens_seen": 1840054272
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045002020202020203,
+      "loss": 2.8811,
+      "theoretical_loss": 3.454500743604109,
+      "tokens_seen": 1840119808
+    },
+    {
+      "epoch": 0.11,
+      "objective/train/docs_used": 1014020,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.2581570148468018,
+      "objective/train/theoretical_loss": 3.4544903168349808,
+      "objective/train/tokens_used": 199044576,
+      "theoretical_loss": 3.4544903168349808,
+      "tokens_seen": 1840185344
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00045000000000000004,
+      "loss": 2.6365,
+      "theoretical_loss": 3.4544903168349808,
+      "tokens_seen": 1840185344
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000449979797979798,
+      "loss": 2.6191,
+      "theoretical_loss": 3.454479890541153,
+      "tokens_seen": 1840250880
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044995959595959596,
+      "loss": 2.6941,
+      "theoretical_loss": 3.454469464722587,
+      "tokens_seen": 1840316416
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004499393939393939,
+      "loss": 2.8638,
+      "theoretical_loss": 3.4544590393792443,
+      "tokens_seen": 1840381952
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004499191919191919,
+      "loss": 2.5423,
+      "theoretical_loss": 3.4544486145110858,
+      "tokens_seen": 1840447488
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044989898989898993,
+      "loss": 2.8338,
+      "theoretical_loss": 3.454438190118074,
+      "tokens_seen": 1840513024
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004498787878787879,
+      "loss": 2.854,
+      "theoretical_loss": 3.4544277662001694,
+      "tokens_seen": 1840578560
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044985858585858585,
+      "loss": 2.5746,
+      "theoretical_loss": 3.454417342757334,
+      "tokens_seen": 1840644096
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004498383838383838,
+      "loss": 2.5932,
+      "theoretical_loss": 3.4544069197895286,
+      "tokens_seen": 1840709632
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044981818181818187,
+      "loss": 2.7428,
+      "theoretical_loss": 3.454396497296715,
+      "tokens_seen": 1840775168
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044979797979797983,
+      "loss": 2.5032,
+      "theoretical_loss": 3.454386075278855,
+      "tokens_seen": 1840840704
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004497777777777778,
+      "loss": 2.6914,
+      "theoretical_loss": 3.4543756537359096,
+      "tokens_seen": 1840906240
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044975757575757574,
+      "loss": 2.4781,
+      "theoretical_loss": 3.4543652326678402,
+      "tokens_seen": 1840971776
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004497373737373737,
+      "loss": 2.8693,
+      "theoretical_loss": 3.454354812074609,
+      "tokens_seen": 1841037312
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044971717171717176,
+      "loss": 2.5425,
+      "theoretical_loss": 3.454344391956176,
+      "tokens_seen": 1841102848
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004496969696969697,
+      "loss": 2.6198,
+      "theoretical_loss": 3.4543339723125035,
+      "tokens_seen": 1841168384
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004496767676767677,
+      "loss": 2.6949,
+      "theoretical_loss": 3.4543235531435537,
+      "tokens_seen": 1841233920
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044965656565656563,
+      "loss": 2.6307,
+      "theoretical_loss": 3.4543131344492872,
+      "tokens_seen": 1841299456
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044963636363636364,
+      "loss": 2.632,
+      "theoretical_loss": 3.454302716229665,
+      "tokens_seen": 1841364992
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044961616161616166,
+      "loss": 2.7287,
+      "theoretical_loss": 3.45429229848465,
+      "tokens_seen": 1841430528
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004495959595959596,
+      "loss": 2.521,
+      "theoretical_loss": 3.454281881214203,
+      "tokens_seen": 1841496064
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044957575757575757,
+      "loss": 2.6339,
+      "theoretical_loss": 3.454271464418285,
+      "tokens_seen": 1841561600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004495555555555555,
+      "loss": 2.7453,
+      "theoretical_loss": 3.454261048096858,
+      "tokens_seen": 1841627136
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044953535353535354,
+      "loss": 2.5324,
+      "theoretical_loss": 3.4542506322498836,
+      "tokens_seen": 1841692672
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044951515151515155,
+      "loss": 2.5193,
+      "theoretical_loss": 3.454240216877323,
+      "tokens_seen": 1841758208
+    },
+    {
+      "epoch": 0.11,
+      "objective/train/docs_used": 1015263,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.0066089630126953,
+      "objective/train/theoretical_loss": 3.454229801979138,
+      "objective/train/tokens_used": 200682976,
+      "theoretical_loss": 3.454229801979138,
+      "tokens_seen": 1841823744
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004494949494949495,
+      "loss": 2.759,
+      "theoretical_loss": 3.454229801979138,
+      "tokens_seen": 1841823744
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044947474747474746,
+      "loss": 2.5906,
+      "theoretical_loss": 3.45421938755529,
+      "tokens_seen": 1841889280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004494545454545455,
+      "loss": 2.5523,
+      "theoretical_loss": 3.4542089736057404,
+      "tokens_seen": 1841954816
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044943434343434343,
+      "loss": 2.7433,
+      "theoretical_loss": 3.4541985601304503,
+      "tokens_seen": 1842020352
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044941414141414144,
+      "loss": 2.6943,
+      "theoretical_loss": 3.4541881471293827,
+      "tokens_seen": 1842085888
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004493939393939394,
+      "loss": 2.7566,
+      "theoretical_loss": 3.4541777346024976,
+      "tokens_seen": 1842151424
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044937373737373735,
+      "loss": 2.6265,
+      "theoretical_loss": 3.454167322549757,
+      "tokens_seen": 1842216960
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044935353535353537,
+      "loss": 2.7595,
+      "theoretical_loss": 3.4541569109711228,
+      "tokens_seen": 1842282496
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004493333333333333,
+      "loss": 2.3543,
+      "theoretical_loss": 3.454146499866556,
+      "tokens_seen": 1842348032
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044931313131313133,
+      "loss": 2.6629,
+      "theoretical_loss": 3.4541360892360187,
+      "tokens_seen": 1842413568
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004492929292929293,
+      "loss": 2.6116,
+      "theoretical_loss": 3.4541256790794725,
+      "tokens_seen": 1842479104
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004492727272727273,
+      "loss": 2.6179,
+      "theoretical_loss": 3.4541152693968784,
+      "tokens_seen": 1842544640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044925252525252526,
+      "loss": 2.8316,
+      "theoretical_loss": 3.454104860188198,
+      "tokens_seen": 1842610176
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004492323232323232,
+      "loss": 2.542,
+      "theoretical_loss": 3.4540944514533933,
+      "tokens_seen": 1842675712
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004492121212121212,
+      "loss": 2.6138,
+      "theoretical_loss": 3.454084043192426,
+      "tokens_seen": 1842741248
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004491919191919192,
+      "loss": 2.7504,
+      "theoretical_loss": 3.4540736354052575,
+      "tokens_seen": 1842806784
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004491717171717172,
+      "loss": 2.663,
+      "theoretical_loss": 3.454063228091848,
+      "tokens_seen": 1842872320
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044915151515151515,
+      "loss": 2.6304,
+      "theoretical_loss": 3.4540528212521617,
+      "tokens_seen": 1842937856
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044913131313131316,
+      "loss": 2.5554,
+      "theoretical_loss": 3.4540424148861586,
+      "tokens_seen": 1843003392
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004491111111111111,
+      "loss": 2.6124,
+      "theoretical_loss": 3.4540320089938,
+      "tokens_seen": 1843068928
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044909090909090913,
+      "loss": 2.8087,
+      "theoretical_loss": 3.454021603575048,
+      "tokens_seen": 1843134464
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004490707070707071,
+      "loss": 2.9358,
+      "theoretical_loss": 3.454011198629865,
+      "tokens_seen": 1843200000
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044905050505050504,
+      "loss": 2.5626,
+      "theoretical_loss": 3.454000794158211,
+      "tokens_seen": 1843265536
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044903030303030306,
+      "loss": 2.4122,
+      "theoretical_loss": 3.453990390160049,
+      "tokens_seen": 1843331072
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000449010101010101,
+      "loss": 2.8471,
+      "theoretical_loss": 3.4539799866353396,
+      "tokens_seen": 1843396608
+    },
+    {
+      "epoch": 0.11,
+      "objective/train/docs_used": 1015947,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.9678730964660645,
+      "objective/train/theoretical_loss": 3.453969583584045,
+      "objective/train/tokens_used": 202321376,
+      "theoretical_loss": 3.453969583584045,
+      "tokens_seen": 1843462144
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000448989898989899,
+      "loss": 2.7793,
+      "theoretical_loss": 3.453969583584045,
+      "tokens_seen": 1843462144
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000448969696969697,
+      "loss": 2.5612,
+      "theoretical_loss": 3.4539591810061268,
+      "tokens_seen": 1843527680
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044894949494949494,
+      "loss": 2.6931,
+      "theoretical_loss": 3.453948778901547,
+      "tokens_seen": 1843593216
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044892929292929295,
+      "loss": 2.5578,
+      "theoretical_loss": 3.4539383772702665,
+      "tokens_seen": 1843658752
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044890909090909096,
+      "loss": 2.6837,
+      "theoretical_loss": 3.4539279761122472,
+      "tokens_seen": 1843724288
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004488888888888889,
+      "loss": 2.615,
+      "theoretical_loss": 3.4539175754274507,
+      "tokens_seen": 1843789824
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044886868686868687,
+      "loss": 2.6294,
+      "theoretical_loss": 3.453907175215839,
+      "tokens_seen": 1843855360
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044884848484848483,
+      "loss": 2.5685,
+      "theoretical_loss": 3.4538967754773733,
+      "tokens_seen": 1843920896
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044882828282828284,
+      "loss": 2.6948,
+      "theoretical_loss": 3.453886376212015,
+      "tokens_seen": 1843986432
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044880808080808085,
+      "loss": 2.6051,
+      "theoretical_loss": 3.4538759774197265,
+      "tokens_seen": 1844051968
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004487878787878788,
+      "loss": 2.7497,
+      "theoretical_loss": 3.4538655791004693,
+      "tokens_seen": 1844117504
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044876767676767677,
+      "loss": 2.9335,
+      "theoretical_loss": 3.4538551812542053,
+      "tokens_seen": 1844183040
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004487474747474747,
+      "loss": 2.4906,
+      "theoretical_loss": 3.4538447838808954,
+      "tokens_seen": 1844248576
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004487272727272728,
+      "loss": 2.6569,
+      "theoretical_loss": 3.453834386980502,
+      "tokens_seen": 1844314112
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044870707070707074,
+      "loss": 2.7522,
+      "theoretical_loss": 3.453823990552986,
+      "tokens_seen": 1844379648
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004486868686868687,
+      "loss": 2.2554,
+      "theoretical_loss": 3.4538135945983104,
+      "tokens_seen": 1844445184
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044866666666666666,
+      "loss": 2.5178,
+      "theoretical_loss": 3.453803199116435,
+      "tokens_seen": 1844510720
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004486464646464646,
+      "loss": 2.5563,
+      "theoretical_loss": 3.453792804107324,
+      "tokens_seen": 1844576256
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004486262626262627,
+      "loss": 2.7275,
+      "theoretical_loss": 3.4537824095709366,
+      "tokens_seen": 1844641792
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044860606060606064,
+      "loss": 2.2752,
+      "theoretical_loss": 3.4537720155072362,
+      "tokens_seen": 1844707328
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004485858585858586,
+      "loss": 2.7008,
+      "theoretical_loss": 3.4537616219161835,
+      "tokens_seen": 1844772864
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044856565656565655,
+      "loss": 2.7969,
+      "theoretical_loss": 3.4537512287977403,
+      "tokens_seen": 1844838400
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004485454545454545,
+      "loss": 2.7409,
+      "theoretical_loss": 3.4537408361518693,
+      "tokens_seen": 1844903936
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004485252525252526,
+      "loss": 2.4061,
+      "theoretical_loss": 3.4537304439785315,
+      "tokens_seen": 1844969472
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044850505050505053,
+      "loss": 2.663,
+      "theoretical_loss": 3.4537200522776885,
+      "tokens_seen": 1845035008
+    },
+    {
+      "epoch": 0.11,
+      "objective/train/docs_used": 1017203,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.1169590950012207,
+      "objective/train/theoretical_loss": 3.4537096610493023,
+      "objective/train/tokens_used": 203959776,
+      "theoretical_loss": 3.4537096610493023,
+      "tokens_seen": 1845100544
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004484848484848485,
+      "loss": 2.6011,
+      "theoretical_loss": 3.4537096610493023,
+      "tokens_seen": 1845100544
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044846464646464644,
+      "loss": 2.5262,
+      "theoretical_loss": 3.4536992702933347,
+      "tokens_seen": 1845166080
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044844444444444445,
+      "loss": 2.5678,
+      "theoretical_loss": 3.453688880009747,
+      "tokens_seen": 1845231616
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044842424242424247,
+      "loss": 2.5909,
+      "theoretical_loss": 3.4536784901985014,
+      "tokens_seen": 1845297152
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004484040404040404,
+      "loss": 2.6497,
+      "theoretical_loss": 3.4536681008595593,
+      "tokens_seen": 1845362688
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004483838383838384,
+      "loss": 2.7736,
+      "theoretical_loss": 3.4536577119928835,
+      "tokens_seen": 1845428224
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044836363636363634,
+      "loss": 2.6749,
+      "theoretical_loss": 3.4536473235984344,
+      "tokens_seen": 1845493760
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044834343434343435,
+      "loss": 2.6772,
+      "theoretical_loss": 3.4536369356761742,
+      "tokens_seen": 1845559296
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044832323232323236,
+      "loss": 2.6142,
+      "theoretical_loss": 3.4536265482260653,
+      "tokens_seen": 1845624832
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004483030303030303,
+      "loss": 2.6519,
+      "theoretical_loss": 3.4536161612480685,
+      "tokens_seen": 1845690368
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044828282828282827,
+      "loss": 2.735,
+      "theoretical_loss": 3.453605774742146,
+      "tokens_seen": 1845755904
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004482626262626263,
+      "loss": 2.825,
+      "theoretical_loss": 3.4535953887082598,
+      "tokens_seen": 1845821440
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044824242424242424,
+      "loss": 2.8881,
+      "theoretical_loss": 3.4535850031463715,
+      "tokens_seen": 1845886976
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044822222222222225,
+      "loss": 2.6401,
+      "theoretical_loss": 3.453574618056443,
+      "tokens_seen": 1845952512
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004482020202020202,
+      "loss": 2.6842,
+      "theoretical_loss": 3.4535642334384358,
+      "tokens_seen": 1846018048
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044818181818181816,
+      "loss": 2.7116,
+      "theoretical_loss": 3.4535538492923123,
+      "tokens_seen": 1846083584
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004481616161616162,
+      "loss": 2.5188,
+      "theoretical_loss": 3.4535434656180337,
+      "tokens_seen": 1846149120
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044814141414141413,
+      "loss": 2.5895,
+      "theoretical_loss": 3.4535330824155617,
+      "tokens_seen": 1846214656
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044812121212121214,
+      "loss": 2.5826,
+      "theoretical_loss": 3.453522699684859,
+      "tokens_seen": 1846280192
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004481010101010101,
+      "loss": 2.6438,
+      "theoretical_loss": 3.4535123174258864,
+      "tokens_seen": 1846345728
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004480808080808081,
+      "loss": 2.6586,
+      "theoretical_loss": 3.4535019356386067,
+      "tokens_seen": 1846411264
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044806060606060607,
+      "loss": 2.6629,
+      "theoretical_loss": 3.4534915543229805,
+      "tokens_seen": 1846476800
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000448040404040404,
+      "loss": 2.7671,
+      "theoretical_loss": 3.4534811734789708,
+      "tokens_seen": 1846542336
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044802020202020204,
+      "loss": 2.7187,
+      "theoretical_loss": 3.453470793106539,
+      "tokens_seen": 1846607872
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000448,
+      "loss": 2.6632,
+      "theoretical_loss": 3.453460413205647,
+      "tokens_seen": 1846673408
+    },
+    {
+      "epoch": 0.11,
+      "objective/train/docs_used": 1017618,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.3263816833496094,
+      "objective/train/theoretical_loss": 3.453450033776256,
+      "objective/train/tokens_used": 205598176,
+      "theoretical_loss": 3.453450033776256,
+      "tokens_seen": 1846738944
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000447979797979798,
+      "loss": 2.5091,
+      "theoretical_loss": 3.453450033776256,
+      "tokens_seen": 1846738944
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044795959595959596,
+      "loss": 2.5978,
+      "theoretical_loss": 3.453439654818329,
+      "tokens_seen": 1846804480
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004479393939393939,
+      "loss": 2.7581,
+      "theoretical_loss": 3.4534292763318275,
+      "tokens_seen": 1846870016
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044791919191919193,
+      "loss": 2.7108,
+      "theoretical_loss": 3.4534188983167127,
+      "tokens_seen": 1846935552
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044789898989898994,
+      "loss": 2.6206,
+      "theoretical_loss": 3.453408520772947,
+      "tokens_seen": 1847001088
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004478787878787879,
+      "loss": 2.4462,
+      "theoretical_loss": 3.453398143700492,
+      "tokens_seen": 1847066624
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044785858585858585,
+      "loss": 2.6832,
+      "theoretical_loss": 3.4533877670993096,
+      "tokens_seen": 1847132160
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004478383838383838,
+      "loss": 2.5257,
+      "theoretical_loss": 3.4533773909693624,
+      "tokens_seen": 1847197696
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004478181818181818,
+      "loss": 2.721,
+      "theoretical_loss": 3.4533670153106115,
+      "tokens_seen": 1847263232
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044779797979797983,
+      "loss": 2.7503,
+      "theoretical_loss": 3.453356640123019,
+      "tokens_seen": 1847328768
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004477777777777778,
+      "loss": 2.8506,
+      "theoretical_loss": 3.4533462654065463,
+      "tokens_seen": 1847394304
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044775757575757575,
+      "loss": 2.7111,
+      "theoretical_loss": 3.4533358911611565,
+      "tokens_seen": 1847459840
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004477373737373737,
+      "loss": 2.7537,
+      "theoretical_loss": 3.4533255173868103,
+      "tokens_seen": 1847525376
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044771717171717177,
+      "loss": 2.7598,
+      "theoretical_loss": 3.4533151440834704,
+      "tokens_seen": 1847590912
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004476969696969697,
+      "loss": 2.5996,
+      "theoretical_loss": 3.4533047712510982,
+      "tokens_seen": 1847656448
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004476767676767677,
+      "loss": 2.5975,
+      "theoretical_loss": 3.453294398889656,
+      "tokens_seen": 1847721984
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044765656565656564,
+      "loss": 2.8062,
+      "theoretical_loss": 3.453284026999105,
+      "tokens_seen": 1847787520
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004476363636363636,
+      "loss": 2.7529,
+      "theoretical_loss": 3.4532736555794084,
+      "tokens_seen": 1847853056
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044761616161616166,
+      "loss": 2.457,
+      "theoretical_loss": 3.453263284630527,
+      "tokens_seen": 1847918592
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004475959595959596,
+      "loss": 2.3476,
+      "theoretical_loss": 3.453252914152423,
+      "tokens_seen": 1847984128
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004475757575757576,
+      "loss": 2.5465,
+      "theoretical_loss": 3.4532425441450587,
+      "tokens_seen": 1848049664
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044755555555555553,
+      "loss": 2.7043,
+      "theoretical_loss": 3.4532321746083956,
+      "tokens_seen": 1848115200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044753535353535354,
+      "loss": 2.527,
+      "theoretical_loss": 3.4532218055423956,
+      "tokens_seen": 1848180736
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044751515151515155,
+      "loss": 2.6198,
+      "theoretical_loss": 3.4532114369470213,
+      "tokens_seen": 1848246272
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004474949494949495,
+      "loss": 2.7901,
+      "theoretical_loss": 3.453201068822234,
+      "tokens_seen": 1848311808
+    },
+    {
+      "epoch": 0.11,
+      "objective/train/docs_used": 1018803,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.287393093109131,
+      "objective/train/theoretical_loss": 3.453190701167996,
+      "objective/train/tokens_used": 207236576,
+      "theoretical_loss": 3.453190701167996,
+      "tokens_seen": 1848377344
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044747474747474747,
+      "loss": 2.4596,
+      "theoretical_loss": 3.453190701167996,
+      "tokens_seen": 1848377344
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004474545454545454,
+      "loss": 2.6234,
+      "theoretical_loss": 3.4531803339842693,
+      "tokens_seen": 1848442880
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044743434343434344,
+      "loss": 2.9293,
+      "theoretical_loss": 3.453169967271015,
+      "tokens_seen": 1848508416
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044741414141414145,
+      "loss": 2.6174,
+      "theoretical_loss": 3.4531596010281964,
+      "tokens_seen": 1848573952
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004473939393939394,
+      "loss": 2.5937,
+      "theoretical_loss": 3.4531492352557747,
+      "tokens_seen": 1848639488
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044737373737373736,
+      "loss": 2.6619,
+      "theoretical_loss": 3.453138869953712,
+      "tokens_seen": 1848705024
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044735353535353537,
+      "loss": 2.7539,
+      "theoretical_loss": 3.4531285051219704,
+      "tokens_seen": 1848770560
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004473333333333334,
+      "loss": 2.8116,
+      "theoretical_loss": 3.4531181407605116,
+      "tokens_seen": 1848836096
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044731313131313134,
+      "loss": 2.7512,
+      "theoretical_loss": 3.453107776869298,
+      "tokens_seen": 1848901632
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004472929292929293,
+      "loss": 2.8565,
+      "theoretical_loss": 3.4530974134482917,
+      "tokens_seen": 1848967168
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044727272727272725,
+      "loss": 2.6797,
+      "theoretical_loss": 3.4530870504974542,
+      "tokens_seen": 1849032704
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044725252525252526,
+      "loss": 2.5755,
+      "theoretical_loss": 3.4530766880167474,
+      "tokens_seen": 1849098240
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004472323232323233,
+      "loss": 2.5315,
+      "theoretical_loss": 3.453066326006134,
+      "tokens_seen": 1849163776
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044721212121212123,
+      "loss": 2.7439,
+      "theoretical_loss": 3.4530559644655754,
+      "tokens_seen": 1849229312
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004471919191919192,
+      "loss": 2.7021,
+      "theoretical_loss": 3.453045603395034,
+      "tokens_seen": 1849294848
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004471717171717172,
+      "loss": 2.4656,
+      "theoretical_loss": 3.4530352427944715,
+      "tokens_seen": 1849360384
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044715151515151516,
+      "loss": 2.5479,
+      "theoretical_loss": 3.45302488266385,
+      "tokens_seen": 1849425920
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044713131313131317,
+      "loss": 2.778,
+      "theoretical_loss": 3.453014523003132,
+      "tokens_seen": 1849491456
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004471111111111111,
+      "loss": 2.9284,
+      "theoretical_loss": 3.453004163812279,
+      "tokens_seen": 1849556992
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004470909090909091,
+      "loss": 2.6908,
+      "theoretical_loss": 3.452993805091253,
+      "tokens_seen": 1849622528
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0004470707070707071,
+      "loss": 2.711,
+      "theoretical_loss": 3.4529834468400162,
+      "tokens_seen": 1849688064
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044705050505050505,
+      "loss": 2.7749,
+      "theoretical_loss": 3.4529730890585313,
+      "tokens_seen": 1849753600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044703030303030306,
+      "loss": 2.4908,
+      "theoretical_loss": 3.4529627317467595,
+      "tokens_seen": 1849819136
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000447010101010101,
+      "loss": 2.5087,
+      "theoretical_loss": 3.4529523749046627,
+      "tokens_seen": 1849884672
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00044698989898989903,
+      "loss": 2.6053,
+      "theoretical_loss": 3.4529420185322035,
+      "tokens_seen": 1849950208
+    },
+    {
+      "epoch": 0.11,
+      "objective/train/docs_used": 1019318,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6232709884643555,
+      "objective/train/theoretical_loss": 3.4529316626293443,
+      "objective/train/tokens_used": 208874976,
+      "theoretical_loss": 3.4529316626293443,
+      "tokens_seen": 1850015744
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000446969696969697,
+      "loss": 2.5128,
+      "theoretical_loss": 3.4529316626293443,
+      "tokens_seen": 1850015744
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044694949494949494,
+      "loss": 2.6452,
+      "theoretical_loss": 3.4529213071960463,
+      "tokens_seen": 1850081280
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044692929292929295,
+      "loss": 2.74,
+      "theoretical_loss": 3.4529109522322727,
+      "tokens_seen": 1850146816
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004469090909090909,
+      "loss": 2.6177,
+      "theoretical_loss": 3.452900597737984,
+      "tokens_seen": 1850212352
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004468888888888889,
+      "loss": 2.609,
+      "theoretical_loss": 3.452890243713144,
+      "tokens_seen": 1850277888
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004468686868686869,
+      "loss": 2.3597,
+      "theoretical_loss": 3.452879890157713,
+      "tokens_seen": 1850343424
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044684848484848484,
+      "loss": 2.6272,
+      "theoretical_loss": 3.452869537071655,
+      "tokens_seen": 1850408960
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044682828282828285,
+      "loss": 2.591,
+      "theoretical_loss": 3.4528591844549306,
+      "tokens_seen": 1850474496
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004468080808080808,
+      "loss": 2.7254,
+      "theoretical_loss": 3.4528488323075024,
+      "tokens_seen": 1850540032
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004467878787878788,
+      "loss": 2.5987,
+      "theoretical_loss": 3.4528384806293326,
+      "tokens_seen": 1850605568
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044676767676767677,
+      "loss": 2.7431,
+      "theoretical_loss": 3.452828129420384,
+      "tokens_seen": 1850671104
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044674747474747473,
+      "loss": 2.4767,
+      "theoretical_loss": 3.452817778680617,
+      "tokens_seen": 1850736640
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044672727272727274,
+      "loss": 2.5841,
+      "theoretical_loss": 3.4528074284099954,
+      "tokens_seen": 1850802176
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044670707070707075,
+      "loss": 2.5436,
+      "theoretical_loss": 3.45279707860848,
+      "tokens_seen": 1850867712
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004466868686868687,
+      "loss": 2.7338,
+      "theoretical_loss": 3.452786729276034,
+      "tokens_seen": 1850933248
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044666666666666666,
+      "loss": 2.629,
+      "theoretical_loss": 3.452776380412619,
+      "tokens_seen": 1850998784
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004466464646464646,
+      "loss": 2.5704,
+      "theoretical_loss": 3.4527660320181974,
+      "tokens_seen": 1851064320
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044662626262626263,
+      "loss": 2.6984,
+      "theoretical_loss": 3.4527556840927307,
+      "tokens_seen": 1851129856
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044660606060606064,
+      "loss": 2.5546,
+      "theoretical_loss": 3.452745336636182,
+      "tokens_seen": 1851195392
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004465858585858586,
+      "loss": 2.7072,
+      "theoretical_loss": 3.452734989648513,
+      "tokens_seen": 1851260928
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044656565656565656,
+      "loss": 2.5535,
+      "theoretical_loss": 3.4527246431296854,
+      "tokens_seen": 1851326464
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004465454545454545,
+      "loss": 2.678,
+      "theoretical_loss": 3.4527142970796625,
+      "tokens_seen": 1851392000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004465252525252526,
+      "loss": 2.683,
+      "theoretical_loss": 3.452703951498405,
+      "tokens_seen": 1851457536
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044650505050505054,
+      "loss": 2.7782,
+      "theoretical_loss": 3.452693606385876,
+      "tokens_seen": 1851523072
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004464848484848485,
+      "loss": 2.79,
+      "theoretical_loss": 3.452683261742038,
+      "tokens_seen": 1851588608
+    },
+    {
+      "epoch": 0.12,
+      "objective/train/docs_used": 1019593,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.985586643218994,
+      "objective/train/theoretical_loss": 3.4526729175668525,
+      "objective/train/tokens_used": 210513376,
+      "theoretical_loss": 3.4526729175668525,
+      "tokens_seen": 1851654144
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044646464646464645,
+      "loss": 2.7467,
+      "theoretical_loss": 3.4526729175668525,
+      "tokens_seen": 1851654144
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004464444444444444,
+      "loss": 2.6811,
+      "theoretical_loss": 3.452662573860281,
+      "tokens_seen": 1851719680
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044642424242424247,
+      "loss": 2.4987,
+      "theoretical_loss": 3.452652230622288,
+      "tokens_seen": 1851785216
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044640404040404043,
+      "loss": 2.6987,
+      "theoretical_loss": 3.4526418878528333,
+      "tokens_seen": 1851850752
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004463838383838384,
+      "loss": 2.4518,
+      "theoretical_loss": 3.4526315455518803,
+      "tokens_seen": 1851916288
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044636363636363634,
+      "loss": 2.7949,
+      "theoretical_loss": 3.452621203719391,
+      "tokens_seen": 1851981824
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044634343434343435,
+      "loss": 2.6532,
+      "theoretical_loss": 3.452610862355327,
+      "tokens_seen": 1852047360
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044632323232323236,
+      "loss": 2.8367,
+      "theoretical_loss": 3.452600521459652,
+      "tokens_seen": 1852112896
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004463030303030303,
+      "loss": 2.6774,
+      "theoretical_loss": 3.4525901810323263,
+      "tokens_seen": 1852178432
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004462828282828283,
+      "loss": 2.6519,
+      "theoretical_loss": 3.4525798410733133,
+      "tokens_seen": 1852243968
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044626262626262623,
+      "loss": 2.5936,
+      "theoretical_loss": 3.4525695015825755,
+      "tokens_seen": 1852309504
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044624242424242425,
+      "loss": 2.6858,
+      "theoretical_loss": 3.4525591625600742,
+      "tokens_seen": 1852375040
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044622222222222226,
+      "loss": 2.692,
+      "theoretical_loss": 3.4525488240057722,
+      "tokens_seen": 1852440576
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004462020202020202,
+      "loss": 2.6024,
+      "theoretical_loss": 3.4525384859196313,
+      "tokens_seen": 1852506112
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044618181818181817,
+      "loss": 2.6869,
+      "theoretical_loss": 3.4525281483016146,
+      "tokens_seen": 1852571648
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004461616161616162,
+      "loss": 2.7454,
+      "theoretical_loss": 3.4525178111516834,
+      "tokens_seen": 1852637184
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044614141414141414,
+      "loss": 2.6114,
+      "theoretical_loss": 3.4525074744698,
+      "tokens_seen": 1852702720
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044612121212121215,
+      "loss": 2.6078,
+      "theoretical_loss": 3.4524971382559277,
+      "tokens_seen": 1852768256
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004461010101010101,
+      "loss": 2.7454,
+      "theoretical_loss": 3.4524868025100277,
+      "tokens_seen": 1852833792
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044608080808080806,
+      "loss": 2.6441,
+      "theoretical_loss": 3.4524764672320627,
+      "tokens_seen": 1852899328
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004460606060606061,
+      "loss": 2.5881,
+      "theoretical_loss": 3.4524661324219945,
+      "tokens_seen": 1852964864
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044604040404040403,
+      "loss": 2.6752,
+      "theoretical_loss": 3.452455798079786,
+      "tokens_seen": 1853030400
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044602020202020204,
+      "loss": 2.5053,
+      "theoretical_loss": 3.4524454642053994,
+      "tokens_seen": 1853095936
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000446,
+      "loss": 2.603,
+      "theoretical_loss": 3.4524351307987966,
+      "tokens_seen": 1853161472
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000445979797979798,
+      "loss": 2.8048,
+      "theoretical_loss": 3.45242479785994,
+      "tokens_seen": 1853227008
+    },
+    {
+      "epoch": 0.12,
+      "objective/train/docs_used": 1020707,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.561333656311035,
+      "objective/train/theoretical_loss": 3.452414465388792,
+      "objective/train/tokens_used": 212151776,
+      "theoretical_loss": 3.452414465388792,
+      "tokens_seen": 1853292544
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044595959595959597,
+      "loss": 2.6155,
+      "theoretical_loss": 3.452414465388792,
+      "tokens_seen": 1853292544
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004459393939393939,
+      "loss": 2.8532,
+      "theoretical_loss": 3.452404133385315,
+      "tokens_seen": 1853358080
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044591919191919193,
+      "loss": 2.5471,
+      "theoretical_loss": 3.4523938018494706,
+      "tokens_seen": 1853423616
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004458989898989899,
+      "loss": 2.6863,
+      "theoretical_loss": 3.452383470781222,
+      "tokens_seen": 1853489152
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004458787878787879,
+      "loss": 2.5466,
+      "theoretical_loss": 3.4523731401805313,
+      "tokens_seen": 1853554688
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044585858585858586,
+      "loss": 2.7701,
+      "theoretical_loss": 3.4523628100473607,
+      "tokens_seen": 1853620224
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004458383838383838,
+      "loss": 2.6979,
+      "theoretical_loss": 3.4523524803816716,
+      "tokens_seen": 1853685760
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044581818181818183,
+      "loss": 2.8536,
+      "theoretical_loss": 3.4523421511834282,
+      "tokens_seen": 1853751296
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044579797979797984,
+      "loss": 2.5863,
+      "theoretical_loss": 3.4523318224525914,
+      "tokens_seen": 1853816832
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004457777777777778,
+      "loss": 2.8932,
+      "theoretical_loss": 3.452321494189124,
+      "tokens_seen": 1853882368
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044575757575757575,
+      "loss": 2.893,
+      "theoretical_loss": 3.452311166392988,
+      "tokens_seen": 1853947904
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004457373737373737,
+      "loss": 2.5877,
+      "theoretical_loss": 3.4523008390641463,
+      "tokens_seen": 1854013440
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004457171717171717,
+      "loss": 2.5738,
+      "theoretical_loss": 3.452290512202561,
+      "tokens_seen": 1854078976
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044569696969696973,
+      "loss": 2.6439,
+      "theoretical_loss": 3.4522801858081946,
+      "tokens_seen": 1854144512
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004456767676767677,
+      "loss": 2.6565,
+      "theoretical_loss": 3.4522698598810084,
+      "tokens_seen": 1854210048
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044565656565656564,
+      "loss": 2.4392,
+      "theoretical_loss": 3.452259534420967,
+      "tokens_seen": 1854275584
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044563636363636366,
+      "loss": 2.4127,
+      "theoretical_loss": 3.4522492094280297,
+      "tokens_seen": 1854341120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044561616161616167,
+      "loss": 2.7188,
+      "theoretical_loss": 3.4522388849021617,
+      "tokens_seen": 1854406656
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004455959595959596,
+      "loss": 2.5797,
+      "theoretical_loss": 3.4522285608433236,
+      "tokens_seen": 1854472192
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004455757575757576,
+      "loss": 2.8985,
+      "theoretical_loss": 3.4522182372514783,
+      "tokens_seen": 1854537728
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044555555555555554,
+      "loss": 2.5284,
+      "theoretical_loss": 3.4522079141265882,
+      "tokens_seen": 1854603264
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044553535353535355,
+      "loss": 2.6146,
+      "theoretical_loss": 3.4521975914686163,
+      "tokens_seen": 1854668800
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044551515151515156,
+      "loss": 2.6305,
+      "theoretical_loss": 3.4521872692775237,
+      "tokens_seen": 1854734336
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004454949494949495,
+      "loss": 2.566,
+      "theoretical_loss": 3.4521769475532738,
+      "tokens_seen": 1854799872
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004454747474747475,
+      "loss": 2.7986,
+      "theoretical_loss": 3.4521666262958286,
+      "tokens_seen": 1854865408
+    },
+    {
+      "epoch": 0.12,
+      "objective/train/docs_used": 1021895,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6850180625915527,
+      "objective/train/theoretical_loss": 3.4521563055051505,
+      "objective/train/tokens_used": 213790176,
+      "theoretical_loss": 3.4521563055051505,
+      "tokens_seen": 1854930944
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044545454545454543,
+      "loss": 2.5554,
+      "theoretical_loss": 3.4521563055051505,
+      "tokens_seen": 1854930944
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004454343434343435,
+      "loss": 2.6896,
+      "theoretical_loss": 3.452145985181202,
+      "tokens_seen": 1854996480
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044541414141414145,
+      "loss": 2.8255,
+      "theoretical_loss": 3.4521356653239454,
+      "tokens_seen": 1855062016
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004453939393939394,
+      "loss": 2.5766,
+      "theoretical_loss": 3.452125345933343,
+      "tokens_seen": 1855127552
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044537373737373737,
+      "loss": 2.8282,
+      "theoretical_loss": 3.452115027009358,
+      "tokens_seen": 1855193088
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004453535353535353,
+      "loss": 2.7015,
+      "theoretical_loss": 3.4521047085519516,
+      "tokens_seen": 1855258624
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004453333333333334,
+      "loss": 2.7837,
+      "theoretical_loss": 3.4520943905610872,
+      "tokens_seen": 1855324160
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044531313131313135,
+      "loss": 2.5069,
+      "theoretical_loss": 3.4520840730367266,
+      "tokens_seen": 1855389696
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004452929292929293,
+      "loss": 2.5115,
+      "theoretical_loss": 3.452073755978833,
+      "tokens_seen": 1855455232
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044527272727272726,
+      "loss": 2.906,
+      "theoretical_loss": 3.4520634393873673,
+      "tokens_seen": 1855520768
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004452525252525252,
+      "loss": 2.5688,
+      "theoretical_loss": 3.452053123262294,
+      "tokens_seen": 1855586304
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004452323232323233,
+      "loss": 2.5582,
+      "theoretical_loss": 3.452042807603574,
+      "tokens_seen": 1855651840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044521212121212124,
+      "loss": 2.632,
+      "theoretical_loss": 3.45203249241117,
+      "tokens_seen": 1855717376
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004451919191919192,
+      "loss": 2.7484,
+      "theoretical_loss": 3.452022177685045,
+      "tokens_seen": 1855782912
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044517171717171715,
+      "loss": 2.7794,
+      "theoretical_loss": 3.4520118634251613,
+      "tokens_seen": 1855848448
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044515151515151516,
+      "loss": 2.4757,
+      "theoretical_loss": 3.452001549631481,
+      "tokens_seen": 1855913984
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004451313131313132,
+      "loss": 2.7872,
+      "theoretical_loss": 3.451991236303967,
+      "tokens_seen": 1855979520
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044511111111111113,
+      "loss": 2.7724,
+      "theoretical_loss": 3.4519809234425813,
+      "tokens_seen": 1856045056
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004450909090909091,
+      "loss": 2.5745,
+      "theoretical_loss": 3.4519706110472868,
+      "tokens_seen": 1856110592
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044507070707070704,
+      "loss": 2.6076,
+      "theoretical_loss": 3.451960299118046,
+      "tokens_seen": 1856176128
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044505050505050506,
+      "loss": 2.7614,
+      "theoretical_loss": 3.451949987654821,
+      "tokens_seen": 1856241664
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044503030303030307,
+      "loss": 2.6492,
+      "theoretical_loss": 3.4519396766575747,
+      "tokens_seen": 1856307200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000445010101010101,
+      "loss": 2.5089,
+      "theoretical_loss": 3.451929366126269,
+      "tokens_seen": 1856372736
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000444989898989899,
+      "loss": 2.8286,
+      "theoretical_loss": 3.4519190560608672,
+      "tokens_seen": 1856438272
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000444969696969697,
+      "loss": 2.7507,
+      "theoretical_loss": 3.451908746461331,
+      "tokens_seen": 1856503808
+    },
+    {
+      "epoch": 0.12,
+      "objective/train/docs_used": 1022391,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.0687434673309326,
+      "objective/train/theoretical_loss": 3.4518984373276234,
+      "objective/train/tokens_used": 215428576,
+      "theoretical_loss": 3.4518984373276234,
+      "tokens_seen": 1856569344
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044494949494949495,
+      "loss": 2.7867,
+      "theoretical_loss": 3.4518984373276234,
+      "tokens_seen": 1856569344
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044492929292929296,
+      "loss": 2.713,
+      "theoretical_loss": 3.4518881286597067,
+      "tokens_seen": 1856634880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004449090909090909,
+      "loss": 2.8357,
+      "theoretical_loss": 3.4518778204575433,
+      "tokens_seen": 1856700416
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044488888888888887,
+      "loss": 2.7542,
+      "theoretical_loss": 3.4518675127210967,
+      "tokens_seen": 1856765952
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004448686868686869,
+      "loss": 2.4508,
+      "theoretical_loss": 3.4518572054503283,
+      "tokens_seen": 1856831488
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044484848484848484,
+      "loss": 2.7486,
+      "theoretical_loss": 3.4518468986452007,
+      "tokens_seen": 1856897024
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044482828282828285,
+      "loss": 2.5418,
+      "theoretical_loss": 3.4518365923056766,
+      "tokens_seen": 1856962560
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004448080808080808,
+      "loss": 2.4934,
+      "theoretical_loss": 3.4518262864317193,
+      "tokens_seen": 1857028096
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004447878787878788,
+      "loss": 2.6511,
+      "theoretical_loss": 3.45181598102329,
+      "tokens_seen": 1857093632
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004447676767676768,
+      "loss": 2.6574,
+      "theoretical_loss": 3.4518056760803524,
+      "tokens_seen": 1857159168
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044474747474747473,
+      "loss": 2.5949,
+      "theoretical_loss": 3.451795371602868,
+      "tokens_seen": 1857224704
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044472727272727274,
+      "loss": 2.5193,
+      "theoretical_loss": 3.4517850675908006,
+      "tokens_seen": 1857290240
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004447070707070707,
+      "loss": 2.7867,
+      "theoretical_loss": 3.451774764044112,
+      "tokens_seen": 1857355776
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004446868686868687,
+      "loss": 2.755,
+      "theoretical_loss": 3.4517644609627647,
+      "tokens_seen": 1857421312
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044466666666666667,
+      "loss": 2.6344,
+      "theoretical_loss": 3.4517541583467213,
+      "tokens_seen": 1857486848
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004446464646464646,
+      "loss": 2.667,
+      "theoretical_loss": 3.451743856195945,
+      "tokens_seen": 1857552384
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044462626262626264,
+      "loss": 2.8186,
+      "theoretical_loss": 3.4517335545103967,
+      "tokens_seen": 1857617920
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044460606060606065,
+      "loss": 2.3931,
+      "theoretical_loss": 3.4517232532900413,
+      "tokens_seen": 1857683456
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004445858585858586,
+      "loss": 2.6651,
+      "theoretical_loss": 3.45171295253484,
+      "tokens_seen": 1857748992
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044456565656565656,
+      "loss": 2.6855,
+      "theoretical_loss": 3.451702652244755,
+      "tokens_seen": 1857814528
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004445454545454545,
+      "loss": 2.609,
+      "theoretical_loss": 3.45169235241975,
+      "tokens_seen": 1857880064
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044452525252525253,
+      "loss": 2.716,
+      "theoretical_loss": 3.4516820530597867,
+      "tokens_seen": 1857945600
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044450505050505054,
+      "loss": 2.572,
+      "theoretical_loss": 3.4516717541648285,
+      "tokens_seen": 1858011136
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004444848484848485,
+      "loss": 2.4947,
+      "theoretical_loss": 3.4516614557348375,
+      "tokens_seen": 1858076672
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044446464646464645,
+      "loss": 2.7446,
+      "theoretical_loss": 3.451651157769777,
+      "tokens_seen": 1858142208
+    },
+    {
+      "epoch": 0.12,
+      "objective/train/docs_used": 1023648,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7927091121673584,
+      "objective/train/theoretical_loss": 3.451640860269608,
+      "objective/train/tokens_used": 217066976,
+      "theoretical_loss": 3.451640860269608,
+      "tokens_seen": 1858207744
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004444444444444444,
+      "loss": 2.4124,
+      "theoretical_loss": 3.451640860269608,
+      "tokens_seen": 1858207744
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004444242424242425,
+      "loss": 2.5427,
+      "theoretical_loss": 3.4516305632342945,
+      "tokens_seen": 1858273280
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044440404040404043,
+      "loss": 2.5259,
+      "theoretical_loss": 3.4516202666637987,
+      "tokens_seen": 1858338816
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004443838383838384,
+      "loss": 2.6627,
+      "theoretical_loss": 3.451609970558083,
+      "tokens_seen": 1858404352
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044436363636363635,
+      "loss": 2.7357,
+      "theoretical_loss": 3.4515996749171114,
+      "tokens_seen": 1858469888
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004443434343434343,
+      "loss": 2.7667,
+      "theoretical_loss": 3.4515893797408443,
+      "tokens_seen": 1858535424
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044432323232323237,
+      "loss": 2.5378,
+      "theoretical_loss": 3.4515790850292465,
+      "tokens_seen": 1858600960
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004443030303030303,
+      "loss": 2.5758,
+      "theoretical_loss": 3.4515687907822787,
+      "tokens_seen": 1858666496
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004442828282828283,
+      "loss": 2.536,
+      "theoretical_loss": 3.451558496999905,
+      "tokens_seen": 1858732032
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044426262626262624,
+      "loss": 2.6107,
+      "theoretical_loss": 3.451548203682088,
+      "tokens_seen": 1858797568
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044424242424242425,
+      "loss": 2.7001,
+      "theoretical_loss": 3.451537910828789,
+      "tokens_seen": 1858863104
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044422222222222226,
+      "loss": 2.6327,
+      "theoretical_loss": 3.4515276184399717,
+      "tokens_seen": 1858928640
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004442020202020202,
+      "loss": 2.5813,
+      "theoretical_loss": 3.451517326515599,
+      "tokens_seen": 1858994176
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004441818181818182,
+      "loss": 2.7582,
+      "theoretical_loss": 3.4515070350556334,
+      "tokens_seen": 1859059712
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044416161616161613,
+      "loss": 2.8071,
+      "theoretical_loss": 3.4514967440600364,
+      "tokens_seen": 1859125248
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044414141414141414,
+      "loss": 2.7393,
+      "theoretical_loss": 3.451486453528773,
+      "tokens_seen": 1859190784
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044412121212121215,
+      "loss": 2.5984,
+      "theoretical_loss": 3.4514761634618036,
+      "tokens_seen": 1859256320
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004441010101010101,
+      "loss": 2.734,
+      "theoretical_loss": 3.4514658738590915,
+      "tokens_seen": 1859321856
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044408080808080807,
+      "loss": 2.7378,
+      "theoretical_loss": 3.4514555847206,
+      "tokens_seen": 1859387392
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004440606060606061,
+      "loss": 2.7765,
+      "theoretical_loss": 3.451445296046292,
+      "tokens_seen": 1859452928
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044404040404040404,
+      "loss": 2.5779,
+      "theoretical_loss": 3.4514350078361296,
+      "tokens_seen": 1859518464
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044402020202020205,
+      "loss": 2.4927,
+      "theoretical_loss": 3.4514247200900754,
+      "tokens_seen": 1859584000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000444,
+      "loss": 2.6562,
+      "theoretical_loss": 3.4514144328080922,
+      "tokens_seen": 1859649536
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044397979797979796,
+      "loss": 2.9958,
+      "theoretical_loss": 3.4514041459901432,
+      "tokens_seen": 1859715072
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044395959595959597,
+      "loss": 2.7743,
+      "theoretical_loss": 3.4513938596361906,
+      "tokens_seen": 1859780608
+    },
+    {
+      "epoch": 0.12,
+      "objective/train/docs_used": 1024242,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.3011066913604736,
+      "objective/train/theoretical_loss": 3.451383573746197,
+      "objective/train/tokens_used": 218705376,
+      "theoretical_loss": 3.451383573746197,
+      "tokens_seen": 1859846144
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000443939393939394,
+      "loss": 2.4715,
+      "theoretical_loss": 3.451383573746197,
+      "tokens_seen": 1859846144
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044391919191919194,
+      "loss": 2.6938,
+      "theoretical_loss": 3.4513732883201254,
+      "tokens_seen": 1859911680
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004438989898989899,
+      "loss": 2.6091,
+      "theoretical_loss": 3.451363003357939,
+      "tokens_seen": 1859977216
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004438787878787879,
+      "loss": 2.6714,
+      "theoretical_loss": 3.4513527188596,
+      "tokens_seen": 1860042752
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044385858585858587,
+      "loss": 2.6668,
+      "theoretical_loss": 3.4513424348250705,
+      "tokens_seen": 1860108288
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004438383838383839,
+      "loss": 2.8937,
+      "theoretical_loss": 3.4513321512543147,
+      "tokens_seen": 1860173824
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044381818181818183,
+      "loss": 2.8199,
+      "theoretical_loss": 3.451321868147294,
+      "tokens_seen": 1860239360
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004437979797979798,
+      "loss": 2.6694,
+      "theoretical_loss": 3.451311585503972,
+      "tokens_seen": 1860304896
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004437777777777778,
+      "loss": 2.7428,
+      "theoretical_loss": 3.4513013033243114,
+      "tokens_seen": 1860370432
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044375757575757576,
+      "loss": 2.5192,
+      "theoretical_loss": 3.4512910216082746,
+      "tokens_seen": 1860435968
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044373737373737377,
+      "loss": 2.5591,
+      "theoretical_loss": 3.4512807403558243,
+      "tokens_seen": 1860501504
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004437171717171717,
+      "loss": 2.5545,
+      "theoretical_loss": 3.451270459566924,
+      "tokens_seen": 1860567040
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044369696969696974,
+      "loss": 2.7749,
+      "theoretical_loss": 3.451260179241536,
+      "tokens_seen": 1860632576
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004436767676767677,
+      "loss": 2.2815,
+      "theoretical_loss": 3.451249899379622,
+      "tokens_seen": 1860698112
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044365656565656565,
+      "loss": 2.8408,
+      "theoretical_loss": 3.4512396199811466,
+      "tokens_seen": 1860763648
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044363636363636366,
+      "loss": 2.5795,
+      "theoretical_loss": 3.451229341046072,
+      "tokens_seen": 1860829184
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004436161616161616,
+      "loss": 2.7172,
+      "theoretical_loss": 3.4512190625743604,
+      "tokens_seen": 1860894720
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044359595959595963,
+      "loss": 2.7811,
+      "theoretical_loss": 3.451208784565975,
+      "tokens_seen": 1860960256
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004435757575757576,
+      "loss": 2.5311,
+      "theoretical_loss": 3.4511985070208793,
+      "tokens_seen": 1861025792
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044355555555555554,
+      "loss": 2.4705,
+      "theoretical_loss": 3.4511882299390346,
+      "tokens_seen": 1861091328
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044353535353535355,
+      "loss": 2.8382,
+      "theoretical_loss": 3.4511779533204043,
+      "tokens_seen": 1861156864
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004435151515151515,
+      "loss": 2.5743,
+      "theoretical_loss": 3.4511676771649524,
+      "tokens_seen": 1861222400
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004434949494949495,
+      "loss": 2.7183,
+      "theoretical_loss": 3.4511574014726403,
+      "tokens_seen": 1861287936
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004434747474747475,
+      "loss": 2.7241,
+      "theoretical_loss": 3.4511471262434306,
+      "tokens_seen": 1861353472
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044345454545454544,
+      "loss": 2.7197,
+      "theoretical_loss": 3.4511368514772878,
+      "tokens_seen": 1861419008
+    },
+    {
+      "epoch": 0.12,
+      "objective/train/docs_used": 1025485,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.8361258506774902,
+      "objective/train/theoretical_loss": 3.451126577174173,
+      "objective/train/tokens_used": 220343776,
+      "theoretical_loss": 3.451126577174173,
+      "tokens_seen": 1861484544
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044343434343434345,
+      "loss": 2.6971,
+      "theoretical_loss": 3.451126577174173,
+      "tokens_seen": 1861484544
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044341414141414146,
+      "loss": 2.5431,
+      "theoretical_loss": 3.4511163033340497,
+      "tokens_seen": 1861550080
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004433939393939394,
+      "loss": 2.3856,
+      "theoretical_loss": 3.451106029956881,
+      "tokens_seen": 1861615616
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044337373737373737,
+      "loss": 2.4283,
+      "theoretical_loss": 3.45109575704263,
+      "tokens_seen": 1861681152
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044335353535353533,
+      "loss": 2.6594,
+      "theoretical_loss": 3.451085484591258,
+      "tokens_seen": 1861746688
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044333333333333334,
+      "loss": 2.663,
+      "theoretical_loss": 3.4510752126027295,
+      "tokens_seen": 1861812224
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044331313131313135,
+      "loss": 2.6829,
+      "theoretical_loss": 3.451064941077007,
+      "tokens_seen": 1861877760
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004432929292929293,
+      "loss": 2.6757,
+      "theoretical_loss": 3.451054670014053,
+      "tokens_seen": 1861943296
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044327272727272726,
+      "loss": 2.665,
+      "theoretical_loss": 3.4510443994138305,
+      "tokens_seen": 1862008832
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004432525252525252,
+      "loss": 2.6872,
+      "theoretical_loss": 3.451034129276302,
+      "tokens_seen": 1862074368
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004432323232323233,
+      "loss": 2.6956,
+      "theoretical_loss": 3.451023859601431,
+      "tokens_seen": 1862139904
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044321212121212124,
+      "loss": 2.6309,
+      "theoretical_loss": 3.45101359038918,
+      "tokens_seen": 1862205440
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004431919191919192,
+      "loss": 2.7008,
+      "theoretical_loss": 3.4510033216395124,
+      "tokens_seen": 1862270976
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044317171717171716,
+      "loss": 2.6083,
+      "theoretical_loss": 3.4509930533523905,
+      "tokens_seen": 1862336512
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004431515151515151,
+      "loss": 2.6467,
+      "theoretical_loss": 3.450982785527777,
+      "tokens_seen": 1862402048
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004431313131313132,
+      "loss": 2.9646,
+      "theoretical_loss": 3.4509725181656354,
+      "tokens_seen": 1862467584
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044311111111111114,
+      "loss": 2.6964,
+      "theoretical_loss": 3.4509622512659286,
+      "tokens_seen": 1862533120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004430909090909091,
+      "loss": 2.5257,
+      "theoretical_loss": 3.4509519848286185,
+      "tokens_seen": 1862598656
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044307070707070705,
+      "loss": 2.5954,
+      "theoretical_loss": 3.4509417188536693,
+      "tokens_seen": 1862664192
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044305050505050506,
+      "loss": 2.6717,
+      "theoretical_loss": 3.4509314533410436,
+      "tokens_seen": 1862729728
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044303030303030307,
+      "loss": 2.5898,
+      "theoretical_loss": 3.450921188290704,
+      "tokens_seen": 1862795264
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044301010101010103,
+      "loss": 2.6666,
+      "theoretical_loss": 3.450910923702613,
+      "tokens_seen": 1862860800
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000442989898989899,
+      "loss": 2.7131,
+      "theoretical_loss": 3.4509006595767344,
+      "tokens_seen": 1862926336
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044296969696969694,
+      "loss": 2.4907,
+      "theoretical_loss": 3.4508903959130306,
+      "tokens_seen": 1862991872
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044294949494949495,
+      "loss": 2.5954,
+      "theoretical_loss": 3.4508801327114647,
+      "tokens_seen": 1863057408
+    },
+    {
+      "epoch": 0.12,
+      "objective/train/docs_used": 1026014,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.185640573501587,
+      "objective/train/theoretical_loss": 3.450869869972,
+      "objective/train/tokens_used": 221982176,
+      "theoretical_loss": 3.450869869972,
+      "tokens_seen": 1863122944
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044292929292929296,
+      "loss": 2.6933,
+      "theoretical_loss": 3.450869869972,
+      "tokens_seen": 1863122944
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004429090909090909,
+      "loss": 2.906,
+      "theoretical_loss": 3.450859607694599,
+      "tokens_seen": 1863188480
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004428888888888889,
+      "loss": 2.5886,
+      "theoretical_loss": 3.4508493458792246,
+      "tokens_seen": 1863254016
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004428686868686869,
+      "loss": 2.6838,
+      "theoretical_loss": 3.4508390845258394,
+      "tokens_seen": 1863319552
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044284848484848485,
+      "loss": 2.7344,
+      "theoretical_loss": 3.4508288236344073,
+      "tokens_seen": 1863385088
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044282828282828286,
+      "loss": 2.7642,
+      "theoretical_loss": 3.4508185632048907,
+      "tokens_seen": 1863450624
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004428080808080808,
+      "loss": 2.5856,
+      "theoretical_loss": 3.4508083032372525,
+      "tokens_seen": 1863516160
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044278787878787877,
+      "loss": 2.4,
+      "theoretical_loss": 3.4507980437314556,
+      "tokens_seen": 1863581696
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004427676767676768,
+      "loss": 2.8011,
+      "theoretical_loss": 3.4507877846874635,
+      "tokens_seen": 1863647232
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044274747474747474,
+      "loss": 2.6666,
+      "theoretical_loss": 3.450777526105239,
+      "tokens_seen": 1863712768
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044272727272727275,
+      "loss": 2.6065,
+      "theoretical_loss": 3.4507672679847445,
+      "tokens_seen": 1863778304
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004427070707070707,
+      "loss": 2.651,
+      "theoretical_loss": 3.450757010325944,
+      "tokens_seen": 1863843840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004426868686868687,
+      "loss": 2.8708,
+      "theoretical_loss": 3.450746753128799,
+      "tokens_seen": 1863909376
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004426666666666667,
+      "loss": 2.6971,
+      "theoretical_loss": 3.450736496393274,
+      "tokens_seen": 1863974912
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044264646464646463,
+      "loss": 2.8966,
+      "theoretical_loss": 3.4507262401193315,
+      "tokens_seen": 1864040448
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044262626262626264,
+      "loss": 2.7598,
+      "theoretical_loss": 3.4507159843069344,
+      "tokens_seen": 1864105984
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004426060606060606,
+      "loss": 2.6357,
+      "theoretical_loss": 3.450705728956045,
+      "tokens_seen": 1864171520
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004425858585858586,
+      "loss": 2.6469,
+      "theoretical_loss": 3.4506954740666274,
+      "tokens_seen": 1864237056
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044256565656565657,
+      "loss": 2.7603,
+      "theoretical_loss": 3.450685219638644,
+      "tokens_seen": 1864302592
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004425454545454545,
+      "loss": 2.5525,
+      "theoretical_loss": 3.450674965672059,
+      "tokens_seen": 1864368128
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044252525252525254,
+      "loss": 2.4534,
+      "theoretical_loss": 3.4506647121668332,
+      "tokens_seen": 1864433664
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044250505050505055,
+      "loss": 2.7848,
+      "theoretical_loss": 3.450654459122932,
+      "tokens_seen": 1864499200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004424848484848485,
+      "loss": 2.7757,
+      "theoretical_loss": 3.450644206540316,
+      "tokens_seen": 1864564736
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044246464646464646,
+      "loss": 2.3675,
+      "theoretical_loss": 3.4506339544189504,
+      "tokens_seen": 1864630272
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004424444444444444,
+      "loss": 2.9488,
+      "theoretical_loss": 3.450623702758797,
+      "tokens_seen": 1864695808
+    },
+    {
+      "epoch": 0.12,
+      "objective/train/docs_used": 1027424,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.689354419708252,
+      "objective/train/theoretical_loss": 3.4506134515598195,
+      "objective/train/tokens_used": 223620576,
+      "theoretical_loss": 3.4506134515598195,
+      "tokens_seen": 1864761344
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044242424242424243,
+      "loss": 2.8144,
+      "theoretical_loss": 3.4506134515598195,
+      "tokens_seen": 1864761344
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044240404040404044,
+      "loss": 2.7622,
+      "theoretical_loss": 3.450603200821981,
+      "tokens_seen": 1864826880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004423838383838384,
+      "loss": 2.4991,
+      "theoretical_loss": 3.4505929505452437,
+      "tokens_seen": 1864892416
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044236363636363635,
+      "loss": 2.7188,
+      "theoretical_loss": 3.450582700729571,
+      "tokens_seen": 1864957952
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004423434343434343,
+      "loss": 2.4162,
+      "theoretical_loss": 3.4505724513749265,
+      "tokens_seen": 1865023488
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004423232323232324,
+      "loss": 2.6481,
+      "theoretical_loss": 3.4505622024812728,
+      "tokens_seen": 1865089024
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044230303030303033,
+      "loss": 2.8747,
+      "theoretical_loss": 3.450551954048573,
+      "tokens_seen": 1865154560
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004422828282828283,
+      "loss": 2.652,
+      "theoretical_loss": 3.4505417060767902,
+      "tokens_seen": 1865220096
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044226262626262625,
+      "loss": 2.5749,
+      "theoretical_loss": 3.4505314585658877,
+      "tokens_seen": 1865285632
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004422424242424242,
+      "loss": 2.5441,
+      "theoretical_loss": 3.450521211515828,
+      "tokens_seen": 1865351168
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044222222222222227,
+      "loss": 2.6244,
+      "theoretical_loss": 3.4505109649265755,
+      "tokens_seen": 1865416704
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004422020202020202,
+      "loss": 2.8392,
+      "theoretical_loss": 3.4505007187980916,
+      "tokens_seen": 1865482240
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004421818181818182,
+      "loss": 2.7035,
+      "theoretical_loss": 3.4504904731303405,
+      "tokens_seen": 1865547776
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044216161616161614,
+      "loss": 2.767,
+      "theoretical_loss": 3.4504802279232845,
+      "tokens_seen": 1865613312
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004421414141414142,
+      "loss": 2.6047,
+      "theoretical_loss": 3.4504699831768875,
+      "tokens_seen": 1865678848
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044212121212121216,
+      "loss": 2.3905,
+      "theoretical_loss": 3.4504597388911122,
+      "tokens_seen": 1865744384
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004421010101010101,
+      "loss": 2.7433,
+      "theoretical_loss": 3.4504494950659215,
+      "tokens_seen": 1865809920
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004420808080808081,
+      "loss": 2.7318,
+      "theoretical_loss": 3.4504392517012796,
+      "tokens_seen": 1865875456
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044206060606060603,
+      "loss": 2.7193,
+      "theoretical_loss": 3.450429008797148,
+      "tokens_seen": 1865940992
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004420404040404041,
+      "loss": 2.6802,
+      "theoretical_loss": 3.450418766353491,
+      "tokens_seen": 1866006528
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044202020202020205,
+      "loss": 2.5807,
+      "theoretical_loss": 3.450408524370271,
+      "tokens_seen": 1866072064
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000442,
+      "loss": 2.5472,
+      "theoretical_loss": 3.450398282847452,
+      "tokens_seen": 1866137600
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044197979797979797,
+      "loss": 2.7356,
+      "theoretical_loss": 3.4503880417849966,
+      "tokens_seen": 1866203136
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0004419595959595959,
+      "loss": 2.7341,
+      "theoretical_loss": 3.450377801182868,
+      "tokens_seen": 1866268672
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000441939393939394,
+      "loss": 2.8592,
+      "theoretical_loss": 3.4503675610410287,
+      "tokens_seen": 1866334208
+    },
+    {
+      "epoch": 0.12,
+      "objective/train/docs_used": 1028516,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.410780429840088,
+      "objective/train/theoretical_loss": 3.4503573213594434,
+      "objective/train/tokens_used": 225258976,
+      "theoretical_loss": 3.4503573213594434,
+      "tokens_seen": 1866399744
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00044191919191919195,
+      "loss": 2.5768,
+      "theoretical_loss": 3.4503573213594434,
+      "tokens_seen": 1866399744
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004418989898989899,
+      "loss": 2.6665,
+      "theoretical_loss": 3.450347082138074,
+      "tokens_seen": 1866465280
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044187878787878786,
+      "loss": 2.7986,
+      "theoretical_loss": 3.4503368433768835,
+      "tokens_seen": 1866530816
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044185858585858587,
+      "loss": 2.6747,
+      "theoretical_loss": 3.450326605075836,
+      "tokens_seen": 1866596352
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004418383838383839,
+      "loss": 2.6103,
+      "theoretical_loss": 3.450316367234894,
+      "tokens_seen": 1866661888
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044181818181818184,
+      "loss": 2.3946,
+      "theoretical_loss": 3.450306129854021,
+      "tokens_seen": 1866727424
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004417979797979798,
+      "loss": 2.7839,
+      "theoretical_loss": 3.45029589293318,
+      "tokens_seen": 1866792960
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044177777777777775,
+      "loss": 2.5326,
+      "theoretical_loss": 3.4502856564723343,
+      "tokens_seen": 1866858496
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044175757575757576,
+      "loss": 2.6544,
+      "theoretical_loss": 3.450275420471447,
+      "tokens_seen": 1866924032
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004417373737373738,
+      "loss": 2.5638,
+      "theoretical_loss": 3.4502651849304815,
+      "tokens_seen": 1866989568
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044171717171717173,
+      "loss": 2.8576,
+      "theoretical_loss": 3.4502549498494,
+      "tokens_seen": 1867055104
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004416969696969697,
+      "loss": 2.5272,
+      "theoretical_loss": 3.4502447152281674,
+      "tokens_seen": 1867120640
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004416767676767677,
+      "loss": 2.7956,
+      "theoretical_loss": 3.4502344810667456,
+      "tokens_seen": 1867186176
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044165656565656566,
+      "loss": 2.7912,
+      "theoretical_loss": 3.4502242473650986,
+      "tokens_seen": 1867251712
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044163636363636367,
+      "loss": 2.6048,
+      "theoretical_loss": 3.4502140141231887,
+      "tokens_seen": 1867317248
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004416161616161616,
+      "loss": 2.652,
+      "theoretical_loss": 3.4502037813409796,
+      "tokens_seen": 1867382784
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004415959595959596,
+      "loss": 3.0165,
+      "theoretical_loss": 3.450193549018435,
+      "tokens_seen": 1867448320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004415757575757576,
+      "loss": 2.4882,
+      "theoretical_loss": 3.4501833171555174,
+      "tokens_seen": 1867513856
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044155555555555555,
+      "loss": 2.6933,
+      "theoretical_loss": 3.4501730857521906,
+      "tokens_seen": 1867579392
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044153535353535356,
+      "loss": 2.5284,
+      "theoretical_loss": 3.4501628548084167,
+      "tokens_seen": 1867644928
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004415151515151515,
+      "loss": 2.6121,
+      "theoretical_loss": 3.4501526243241605,
+      "tokens_seen": 1867710464
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044149494949494953,
+      "loss": 2.6236,
+      "theoretical_loss": 3.4501423942993843,
+      "tokens_seen": 1867776000
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004414747474747475,
+      "loss": 2.724,
+      "theoretical_loss": 3.4501321647340513,
+      "tokens_seen": 1867841536
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044145454545454544,
+      "loss": 2.5005,
+      "theoretical_loss": 3.4501219356281254,
+      "tokens_seen": 1867907072
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044143434343434345,
+      "loss": 2.7259,
+      "theoretical_loss": 3.450111706981569,
+      "tokens_seen": 1867972608
+    },
+    {
+      "epoch": 0.13,
+      "objective/train/docs_used": 1029117,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.526183605194092,
+      "objective/train/theoretical_loss": 3.4501014787943456,
+      "objective/train/tokens_used": 226897376,
+      "theoretical_loss": 3.4501014787943456,
+      "tokens_seen": 1868038144
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004414141414141414,
+      "loss": 2.4437,
+      "theoretical_loss": 3.4501014787943456,
+      "tokens_seen": 1868038144
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004413939393939394,
+      "loss": 2.4661,
+      "theoretical_loss": 3.4500912510664192,
+      "tokens_seen": 1868103680
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004413737373737374,
+      "loss": 2.5983,
+      "theoretical_loss": 3.450081023797752,
+      "tokens_seen": 1868169216
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044135353535353533,
+      "loss": 2.4983,
+      "theoretical_loss": 3.4500707969883084,
+      "tokens_seen": 1868234752
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044133333333333335,
+      "loss": 2.5441,
+      "theoretical_loss": 3.4500605706380503,
+      "tokens_seen": 1868300288
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044131313131313136,
+      "loss": 2.7335,
+      "theoretical_loss": 3.450050344746942,
+      "tokens_seen": 1868365824
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004412929292929293,
+      "loss": 2.6975,
+      "theoretical_loss": 3.450040119314947,
+      "tokens_seen": 1868431360
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044127272727272727,
+      "loss": 2.7856,
+      "theoretical_loss": 3.450029894342028,
+      "tokens_seen": 1868496896
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004412525252525252,
+      "loss": 2.7327,
+      "theoretical_loss": 3.450019669828148,
+      "tokens_seen": 1868562432
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044123232323232324,
+      "loss": 2.394,
+      "theoretical_loss": 3.4500094457732704,
+      "tokens_seen": 1868627968
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044121212121212125,
+      "loss": 2.8116,
+      "theoretical_loss": 3.449999222177359,
+      "tokens_seen": 1868693504
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004411919191919192,
+      "loss": 2.6899,
+      "theoretical_loss": 3.449988999040377,
+      "tokens_seen": 1868759040
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044117171717171716,
+      "loss": 2.7514,
+      "theoretical_loss": 3.449978776362288,
+      "tokens_seen": 1868824576
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004411515151515151,
+      "loss": 2.7282,
+      "theoretical_loss": 3.4499685541430543,
+      "tokens_seen": 1868890112
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004411313131313132,
+      "loss": 2.7994,
+      "theoretical_loss": 3.4499583323826397,
+      "tokens_seen": 1868955648
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044111111111111114,
+      "loss": 2.7855,
+      "theoretical_loss": 3.4499481110810075,
+      "tokens_seen": 1869021184
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004410909090909091,
+      "loss": 2.6568,
+      "theoretical_loss": 3.4499378902381217,
+      "tokens_seen": 1869086720
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044107070707070706,
+      "loss": 2.6813,
+      "theoretical_loss": 3.449927669853945,
+      "tokens_seen": 1869152256
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000441050505050505,
+      "loss": 2.5984,
+      "theoretical_loss": 3.4499174499284404,
+      "tokens_seen": 1869217792
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004410303030303031,
+      "loss": 2.7613,
+      "theoretical_loss": 3.449907230461572,
+      "tokens_seen": 1869283328
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044101010101010103,
+      "loss": 2.708,
+      "theoretical_loss": 3.4498970114533023,
+      "tokens_seen": 1869348864
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000440989898989899,
+      "loss": 2.5817,
+      "theoretical_loss": 3.4498867929035955,
+      "tokens_seen": 1869414400
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044096969696969695,
+      "loss": 2.7155,
+      "theoretical_loss": 3.4498765748124143,
+      "tokens_seen": 1869479936
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044094949494949496,
+      "loss": 2.931,
+      "theoretical_loss": 3.449866357179723,
+      "tokens_seen": 1869545472
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044092929292929297,
+      "loss": 2.7079,
+      "theoretical_loss": 3.4498561400054832,
+      "tokens_seen": 1869611008
+    },
+    {
+      "epoch": 0.13,
+      "objective/train/docs_used": 1029470,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.821355104446411,
+      "objective/train/theoretical_loss": 3.44984592328966,
+      "objective/train/tokens_used": 228535776,
+      "theoretical_loss": 3.44984592328966,
+      "tokens_seen": 1869676544
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044090909090909093,
+      "loss": 2.7606,
+      "theoretical_loss": 3.44984592328966,
+      "tokens_seen": 1869676544
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004408888888888889,
+      "loss": 2.6028,
+      "theoretical_loss": 3.449835707032216,
+      "tokens_seen": 1869742080
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044086868686868684,
+      "loss": 2.73,
+      "theoretical_loss": 3.4498254912331148,
+      "tokens_seen": 1869807616
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044084848484848485,
+      "loss": 2.5199,
+      "theoretical_loss": 3.4498152758923197,
+      "tokens_seen": 1869873152
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044082828282828286,
+      "loss": 2.6589,
+      "theoretical_loss": 3.4498050610097932,
+      "tokens_seen": 1869938688
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004408080808080808,
+      "loss": 2.7433,
+      "theoretical_loss": 3.4497948465855,
+      "tokens_seen": 1870004224
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004407878787878788,
+      "loss": 2.7544,
+      "theoretical_loss": 3.4497846326194033,
+      "tokens_seen": 1870069760
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004407676767676768,
+      "loss": 2.585,
+      "theoretical_loss": 3.449774419111466,
+      "tokens_seen": 1870135296
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044074747474747474,
+      "loss": 2.6629,
+      "theoretical_loss": 3.449764206061652,
+      "tokens_seen": 1870200832
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044072727272727276,
+      "loss": 2.7832,
+      "theoretical_loss": 3.4497539934699235,
+      "tokens_seen": 1870266368
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004407070707070707,
+      "loss": 2.7545,
+      "theoretical_loss": 3.4497437813362453,
+      "tokens_seen": 1870331904
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044068686868686867,
+      "loss": 2.8311,
+      "theoretical_loss": 3.44973356966058,
+      "tokens_seen": 1870397440
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004406666666666667,
+      "loss": 3.011,
+      "theoretical_loss": 3.449723358442892,
+      "tokens_seen": 1870462976
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044064646464646464,
+      "loss": 2.5545,
+      "theoretical_loss": 3.449713147683143,
+      "tokens_seen": 1870528512
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044062626262626265,
+      "loss": 2.8115,
+      "theoretical_loss": 3.449702937381298,
+      "tokens_seen": 1870594048
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004406060606060606,
+      "loss": 2.8322,
+      "theoretical_loss": 3.44969272753732,
+      "tokens_seen": 1870659584
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004405858585858586,
+      "loss": 2.8499,
+      "theoretical_loss": 3.449682518151172,
+      "tokens_seen": 1870725120
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004405656565656566,
+      "loss": 2.9033,
+      "theoretical_loss": 3.4496723092228176,
+      "tokens_seen": 1870790656
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044054545454545453,
+      "loss": 2.5401,
+      "theoretical_loss": 3.449662100752221,
+      "tokens_seen": 1870856192
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044052525252525254,
+      "loss": 2.7093,
+      "theoretical_loss": 3.449651892739344,
+      "tokens_seen": 1870921728
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004405050505050505,
+      "loss": 2.4943,
+      "theoretical_loss": 3.449641685184152,
+      "tokens_seen": 1870987264
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004404848484848485,
+      "loss": 2.7457,
+      "theoretical_loss": 3.449631478086607,
+      "tokens_seen": 1871052800
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044046464646464647,
+      "loss": 2.6909,
+      "theoretical_loss": 3.4496212714466727,
+      "tokens_seen": 1871118336
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004404444444444445,
+      "loss": 2.8722,
+      "theoretical_loss": 3.449611065264313,
+      "tokens_seen": 1871183872
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044042424242424243,
+      "loss": 2.7971,
+      "theoretical_loss": 3.449600859539491,
+      "tokens_seen": 1871249408
+    },
+    {
+      "epoch": 0.13,
+      "objective/train/docs_used": 1030757,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.707038402557373,
+      "objective/train/theoretical_loss": 3.4495906542721704,
+      "objective/train/tokens_used": 230174176,
+      "theoretical_loss": 3.4495906542721704,
+      "tokens_seen": 1871314944
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044040404040404044,
+      "loss": 2.6459,
+      "theoretical_loss": 3.4495906542721704,
+      "tokens_seen": 1871314944
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004403838383838384,
+      "loss": 2.5098,
+      "theoretical_loss": 3.4495804494623146,
+      "tokens_seen": 1871380480
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044036363636363636,
+      "loss": 2.7132,
+      "theoretical_loss": 3.4495702451098875,
+      "tokens_seen": 1871446016
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044034343434343437,
+      "loss": 2.6227,
+      "theoretical_loss": 3.449560041214851,
+      "tokens_seen": 1871511552
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004403232323232323,
+      "loss": 2.7804,
+      "theoretical_loss": 3.4495498377771705,
+      "tokens_seen": 1871577088
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044030303030303034,
+      "loss": 2.7748,
+      "theoretical_loss": 3.4495396347968086,
+      "tokens_seen": 1871642624
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004402828282828283,
+      "loss": 2.6117,
+      "theoretical_loss": 3.4495294322737293,
+      "tokens_seen": 1871708160
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044026262626262625,
+      "loss": 2.6279,
+      "theoretical_loss": 3.449519230207895,
+      "tokens_seen": 1871773696
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044024242424242426,
+      "loss": 2.7321,
+      "theoretical_loss": 3.4495090285992704,
+      "tokens_seen": 1871839232
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004402222222222222,
+      "loss": 2.7325,
+      "theoretical_loss": 3.4494988274478184,
+      "tokens_seen": 1871904768
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044020202020202023,
+      "loss": 2.7222,
+      "theoretical_loss": 3.4494886267535025,
+      "tokens_seen": 1871970304
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004401818181818182,
+      "loss": 2.5767,
+      "theoretical_loss": 3.4494784265162863,
+      "tokens_seen": 1872035840
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044016161616161614,
+      "loss": 2.425,
+      "theoretical_loss": 3.4494682267361334,
+      "tokens_seen": 1872101376
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044014141414141416,
+      "loss": 2.5256,
+      "theoretical_loss": 3.449458027413007,
+      "tokens_seen": 1872166912
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044012121212121217,
+      "loss": 2.5125,
+      "theoretical_loss": 3.4494478285468713,
+      "tokens_seen": 1872232448
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004401010101010101,
+      "loss": 2.6882,
+      "theoretical_loss": 3.4494376301376892,
+      "tokens_seen": 1872297984
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004400808080808081,
+      "loss": 2.6736,
+      "theoretical_loss": 3.4494274321854244,
+      "tokens_seen": 1872363520
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044006060606060604,
+      "loss": 2.6651,
+      "theoretical_loss": 3.4494172346900402,
+      "tokens_seen": 1872429056
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044004040404040405,
+      "loss": 2.7091,
+      "theoretical_loss": 3.449407037651501,
+      "tokens_seen": 1872494592
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044002020202020206,
+      "loss": 2.6583,
+      "theoretical_loss": 3.4493968410697695,
+      "tokens_seen": 1872560128
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00044,
+      "loss": 2.7233,
+      "theoretical_loss": 3.449386644944809,
+      "tokens_seen": 1872625664
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043997979797979797,
+      "loss": 2.6792,
+      "theoretical_loss": 3.449376449276584,
+      "tokens_seen": 1872691200
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043995959595959593,
+      "loss": 2.6617,
+      "theoretical_loss": 3.449366254065058,
+      "tokens_seen": 1872756736
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000439939393939394,
+      "loss": 2.8347,
+      "theoretical_loss": 3.449356059310194,
+      "tokens_seen": 1872822272
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043991919191919195,
+      "loss": 2.6349,
+      "theoretical_loss": 3.4493458650119555,
+      "tokens_seen": 1872887808
+    },
+    {
+      "epoch": 0.13,
+      "objective/train/docs_used": 1031383,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7423930168151855,
+      "objective/train/theoretical_loss": 3.4493356711703065,
+      "objective/train/tokens_used": 231812576,
+      "theoretical_loss": 3.4493356711703065,
+      "tokens_seen": 1872953344
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004398989898989899,
+      "loss": 2.6487,
+      "theoretical_loss": 3.4493356711703065,
+      "tokens_seen": 1872953344
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043987878787878787,
+      "loss": 2.6297,
+      "theoretical_loss": 3.4493254777852096,
+      "tokens_seen": 1873018880
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004398585858585858,
+      "loss": 2.742,
+      "theoretical_loss": 3.44931528485663,
+      "tokens_seen": 1873084416
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004398383838383839,
+      "loss": 2.7249,
+      "theoretical_loss": 3.4493050923845305,
+      "tokens_seen": 1873149952
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043981818181818184,
+      "loss": 2.4449,
+      "theoretical_loss": 3.4492949003688747,
+      "tokens_seen": 1873215488
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004397979797979798,
+      "loss": 2.6473,
+      "theoretical_loss": 3.4492847088096257,
+      "tokens_seen": 1873281024
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043977777777777776,
+      "loss": 2.7582,
+      "theoretical_loss": 3.4492745177067476,
+      "tokens_seen": 1873346560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043975757575757577,
+      "loss": 2.7019,
+      "theoretical_loss": 3.449264327060204,
+      "tokens_seen": 1873412096
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004397373737373738,
+      "loss": 2.5911,
+      "theoretical_loss": 3.4492541368699587,
+      "tokens_seen": 1873477632
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043971717171717174,
+      "loss": 2.5838,
+      "theoretical_loss": 3.449243947135974,
+      "tokens_seen": 1873543168
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004396969696969697,
+      "loss": 2.7736,
+      "theoretical_loss": 3.4492337578582157,
+      "tokens_seen": 1873608704
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043967676767676765,
+      "loss": 2.7215,
+      "theoretical_loss": 3.449223569036646,
+      "tokens_seen": 1873674240
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043965656565656566,
+      "loss": 2.6328,
+      "theoretical_loss": 3.449213380671229,
+      "tokens_seen": 1873739776
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043963636363636367,
+      "loss": 3.0429,
+      "theoretical_loss": 3.4492031927619275,
+      "tokens_seen": 1873805312
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043961616161616163,
+      "loss": 2.6799,
+      "theoretical_loss": 3.4491930053087065,
+      "tokens_seen": 1873870848
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004395959595959596,
+      "loss": 2.6578,
+      "theoretical_loss": 3.4491828183115283,
+      "tokens_seen": 1873936384
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004395757575757576,
+      "loss": 2.5755,
+      "theoretical_loss": 3.449172631770357,
+      "tokens_seen": 1874001920
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043955555555555555,
+      "loss": 2.5907,
+      "theoretical_loss": 3.4491624456851566,
+      "tokens_seen": 1874067456
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043953535353535357,
+      "loss": 2.6247,
+      "theoretical_loss": 3.4491522600558904,
+      "tokens_seen": 1874132992
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004395151515151515,
+      "loss": 2.7296,
+      "theoretical_loss": 3.4491420748825226,
+      "tokens_seen": 1874198528
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004394949494949495,
+      "loss": 2.5939,
+      "theoretical_loss": 3.449131890165016,
+      "tokens_seen": 1874264064
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004394747474747475,
+      "loss": 2.6861,
+      "theoretical_loss": 3.449121705903335,
+      "tokens_seen": 1874329600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043945454545454545,
+      "loss": 2.6701,
+      "theoretical_loss": 3.4491115220974424,
+      "tokens_seen": 1874395136
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043943434343434346,
+      "loss": 2.4645,
+      "theoretical_loss": 3.449101338747303,
+      "tokens_seen": 1874460672
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004394141414141414,
+      "loss": 2.5077,
+      "theoretical_loss": 3.449091155852879,
+      "tokens_seen": 1874526208
+    },
+    {
+      "epoch": 0.13,
+      "objective/train/docs_used": 1032514,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.0022151470184326,
+      "objective/train/theoretical_loss": 3.4490809734141354,
+      "objective/train/tokens_used": 233450976,
+      "theoretical_loss": 3.4490809734141354,
+      "tokens_seen": 1874591744
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004393939393939394,
+      "loss": 2.7027,
+      "theoretical_loss": 3.4490809734141354,
+      "tokens_seen": 1874591744
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004393737373737374,
+      "loss": 2.6109,
+      "theoretical_loss": 3.4490707914310352,
+      "tokens_seen": 1874657280
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043935353535353534,
+      "loss": 2.7608,
+      "theoretical_loss": 3.449060609903543,
+      "tokens_seen": 1874722816
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043933333333333335,
+      "loss": 2.5745,
+      "theoretical_loss": 3.4490504288316215,
+      "tokens_seen": 1874788352
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004393131313131313,
+      "loss": 2.6831,
+      "theoretical_loss": 3.4490402482152343,
+      "tokens_seen": 1874853888
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004392929292929293,
+      "loss": 2.8604,
+      "theoretical_loss": 3.4490300680543458,
+      "tokens_seen": 1874919424
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004392727272727273,
+      "loss": 2.5409,
+      "theoretical_loss": 3.449019888348919,
+      "tokens_seen": 1874984960
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043925252525252523,
+      "loss": 2.5154,
+      "theoretical_loss": 3.4490097090989185,
+      "tokens_seen": 1875050496
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043923232323232324,
+      "loss": 2.6237,
+      "theoretical_loss": 3.4489995303043073,
+      "tokens_seen": 1875116032
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043921212121212125,
+      "loss": 2.4967,
+      "theoretical_loss": 3.448989351965049,
+      "tokens_seen": 1875181568
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004391919191919192,
+      "loss": 2.6085,
+      "theoretical_loss": 3.448979174081108,
+      "tokens_seen": 1875247104
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043917171717171717,
+      "loss": 2.7374,
+      "theoretical_loss": 3.4489689966524475,
+      "tokens_seen": 1875312640
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004391515151515151,
+      "loss": 2.3794,
+      "theoretical_loss": 3.4489588196790315,
+      "tokens_seen": 1875378176
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043913131313131314,
+      "loss": 2.6384,
+      "theoretical_loss": 3.448948643160823,
+      "tokens_seen": 1875443712
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043911111111111115,
+      "loss": 2.7276,
+      "theoretical_loss": 3.448938467097787,
+      "tokens_seen": 1875509248
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004390909090909091,
+      "loss": 2.5831,
+      "theoretical_loss": 3.448928291489886,
+      "tokens_seen": 1875574784
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043907070707070706,
+      "loss": 2.431,
+      "theoretical_loss": 3.448918116337085,
+      "tokens_seen": 1875640320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000439050505050505,
+      "loss": 2.662,
+      "theoretical_loss": 3.448907941639346,
+      "tokens_seen": 1875705856
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004390303030303031,
+      "loss": 2.4266,
+      "theoretical_loss": 3.448897767396635,
+      "tokens_seen": 1875771392
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043901010101010104,
+      "loss": 2.7484,
+      "theoretical_loss": 3.448887593608914,
+      "tokens_seen": 1875836928
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000438989898989899,
+      "loss": 2.6149,
+      "theoretical_loss": 3.448877420276147,
+      "tokens_seen": 1875902464
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043896969696969695,
+      "loss": 2.6773,
+      "theoretical_loss": 3.448867247398298,
+      "tokens_seen": 1875968000
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004389494949494949,
+      "loss": 2.6709,
+      "theoretical_loss": 3.448857074975332,
+      "tokens_seen": 1876033536
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000438929292929293,
+      "loss": 2.669,
+      "theoretical_loss": 3.448846903007211,
+      "tokens_seen": 1876099072
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043890909090909093,
+      "loss": 2.5903,
+      "theoretical_loss": 3.448836731493899,
+      "tokens_seen": 1876164608
+    },
+    {
+      "epoch": 0.13,
+      "objective/train/docs_used": 1033016,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.4885294437408447,
+      "objective/train/theoretical_loss": 3.44882656043536,
+      "objective/train/tokens_used": 235089376,
+      "theoretical_loss": 3.44882656043536,
+      "tokens_seen": 1876230144
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004388888888888889,
+      "loss": 2.6458,
+      "theoretical_loss": 3.44882656043536,
+      "tokens_seen": 1876230144
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043886868686868685,
+      "loss": 2.7158,
+      "theoretical_loss": 3.4488163898315585,
+      "tokens_seen": 1876295680
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004388484848484848,
+      "loss": 2.576,
+      "theoretical_loss": 3.4488062196824574,
+      "tokens_seen": 1876361216
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043882828282828287,
+      "loss": 2.5016,
+      "theoretical_loss": 3.448796049988021,
+      "tokens_seen": 1876426752
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004388080808080808,
+      "loss": 2.5476,
+      "theoretical_loss": 3.448785880748213,
+      "tokens_seen": 1876492288
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004387878787878788,
+      "loss": 2.7925,
+      "theoretical_loss": 3.4487757119629965,
+      "tokens_seen": 1876557824
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043876767676767674,
+      "loss": 2.7359,
+      "theoretical_loss": 3.4487655436323363,
+      "tokens_seen": 1876623360
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004387474747474748,
+      "loss": 2.5205,
+      "theoretical_loss": 3.448755375756196,
+      "tokens_seen": 1876688896
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043872727272727276,
+      "loss": 2.7716,
+      "theoretical_loss": 3.4487452083345396,
+      "tokens_seen": 1876754432
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004387070707070707,
+      "loss": 2.6872,
+      "theoretical_loss": 3.44873504136733,
+      "tokens_seen": 1876819968
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004386868686868687,
+      "loss": 2.5491,
+      "theoretical_loss": 3.4487248748545314,
+      "tokens_seen": 1876885504
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043866666666666663,
+      "loss": 2.7096,
+      "theoretical_loss": 3.448714708796108,
+      "tokens_seen": 1876951040
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004386464646464647,
+      "loss": 2.6948,
+      "theoretical_loss": 3.4487045431920236,
+      "tokens_seen": 1877016576
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043862626262626265,
+      "loss": 2.4952,
+      "theoretical_loss": 3.4486943780422417,
+      "tokens_seen": 1877082112
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004386060606060606,
+      "loss": 2.6549,
+      "theoretical_loss": 3.4486842133467266,
+      "tokens_seen": 1877147648
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043858585858585857,
+      "loss": 2.4912,
+      "theoretical_loss": 3.4486740491054415,
+      "tokens_seen": 1877213184
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004385656565656566,
+      "loss": 2.7272,
+      "theoretical_loss": 3.448663885318351,
+      "tokens_seen": 1877278720
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004385454545454546,
+      "loss": 2.9654,
+      "theoretical_loss": 3.448653721985418,
+      "tokens_seen": 1877344256
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043852525252525255,
+      "loss": 2.7963,
+      "theoretical_loss": 3.448643559106607,
+      "tokens_seen": 1877409792
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004385050505050505,
+      "loss": 2.6008,
+      "theoretical_loss": 3.448633396681882,
+      "tokens_seen": 1877475328
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043848484848484846,
+      "loss": 2.6553,
+      "theoretical_loss": 3.4486232347112065,
+      "tokens_seen": 1877540864
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043846464646464647,
+      "loss": 2.7418,
+      "theoretical_loss": 3.448613073194544,
+      "tokens_seen": 1877606400
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004384444444444445,
+      "loss": 2.9507,
+      "theoretical_loss": 3.4486029121318595,
+      "tokens_seen": 1877671936
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043842424242424244,
+      "loss": 2.7075,
+      "theoretical_loss": 3.448592751523116,
+      "tokens_seen": 1877737472
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004384040404040404,
+      "loss": 2.6087,
+      "theoretical_loss": 3.4485825913682775,
+      "tokens_seen": 1877803008
+    },
+    {
+      "epoch": 0.13,
+      "objective/train/docs_used": 1034169,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.757983922958374,
+      "objective/train/theoretical_loss": 3.448572431667308,
+      "objective/train/tokens_used": 236727776,
+      "theoretical_loss": 3.448572431667308,
+      "tokens_seen": 1877868544
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004383838383838384,
+      "loss": 2.4029,
+      "theoretical_loss": 3.448572431667308,
+      "tokens_seen": 1877868544
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043836363636363636,
+      "loss": 2.6623,
+      "theoretical_loss": 3.448562272420171,
+      "tokens_seen": 1877934080
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004383434343434344,
+      "loss": 2.6641,
+      "theoretical_loss": 3.4485521136268313,
+      "tokens_seen": 1877999616
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043832323232323233,
+      "loss": 2.6791,
+      "theoretical_loss": 3.4485419552872525,
+      "tokens_seen": 1878065152
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004383030303030303,
+      "loss": 2.5486,
+      "theoretical_loss": 3.4485317974013974,
+      "tokens_seen": 1878130688
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004382828282828283,
+      "loss": 2.7649,
+      "theoretical_loss": 3.4485216399692313,
+      "tokens_seen": 1878196224
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043826262626262626,
+      "loss": 2.7451,
+      "theoretical_loss": 3.4485114829907175,
+      "tokens_seen": 1878261760
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043824242424242427,
+      "loss": 2.7333,
+      "theoretical_loss": 3.44850132646582,
+      "tokens_seen": 1878327296
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004382222222222222,
+      "loss": 2.6926,
+      "theoretical_loss": 3.4484911703945027,
+      "tokens_seen": 1878392832
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043820202020202024,
+      "loss": 2.4957,
+      "theoretical_loss": 3.4484810147767293,
+      "tokens_seen": 1878458368
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004381818181818182,
+      "loss": 2.5296,
+      "theoretical_loss": 3.448470859612464,
+      "tokens_seen": 1878523904
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043816161616161615,
+      "loss": 2.9351,
+      "theoretical_loss": 3.4484607049016702,
+      "tokens_seen": 1878589440
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043814141414141416,
+      "loss": 2.7707,
+      "theoretical_loss": 3.448450550644313,
+      "tokens_seen": 1878654976
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004381212121212121,
+      "loss": 2.5848,
+      "theoretical_loss": 3.4484403968403554,
+      "tokens_seen": 1878720512
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043810101010101013,
+      "loss": 2.7145,
+      "theoretical_loss": 3.448430243489762,
+      "tokens_seen": 1878786048
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004380808080808081,
+      "loss": 2.7403,
+      "theoretical_loss": 3.4484200905924953,
+      "tokens_seen": 1878851584
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043806060606060604,
+      "loss": 2.778,
+      "theoretical_loss": 3.448409938148521,
+      "tokens_seen": 1878917120
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043804040404040405,
+      "loss": 2.4632,
+      "theoretical_loss": 3.448399786157802,
+      "tokens_seen": 1878982656
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043802020202020206,
+      "loss": 2.5384,
+      "theoretical_loss": 3.4483896346203027,
+      "tokens_seen": 1879048192
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000438,
+      "loss": 2.6639,
+      "theoretical_loss": 3.448379483535987,
+      "tokens_seen": 1879113728
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000437979797979798,
+      "loss": 2.6435,
+      "theoretical_loss": 3.4483693329048184,
+      "tokens_seen": 1879179264
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043795959595959593,
+      "loss": 2.61,
+      "theoretical_loss": 3.4483591827267617,
+      "tokens_seen": 1879244800
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043793939393939395,
+      "loss": 2.7294,
+      "theoretical_loss": 3.4483490330017803,
+      "tokens_seen": 1879310336
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043791919191919196,
+      "loss": 2.4736,
+      "theoretical_loss": 3.4483388837298383,
+      "tokens_seen": 1879375872
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004378989898989899,
+      "loss": 2.7058,
+      "theoretical_loss": 3.4483287349108993,
+      "tokens_seen": 1879441408
+    },
+    {
+      "epoch": 0.13,
+      "objective/train/docs_used": 1034876,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5990076065063477,
+      "objective/train/theoretical_loss": 3.448318586544928,
+      "objective/train/tokens_used": 238366176,
+      "theoretical_loss": 3.448318586544928,
+      "tokens_seen": 1879506944
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043787878787878787,
+      "loss": 2.8493,
+      "theoretical_loss": 3.448318586544928,
+      "tokens_seen": 1879506944
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043785858585858583,
+      "loss": 2.5927,
+      "theoretical_loss": 3.448308438631888,
+      "tokens_seen": 1879572480
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004378383838383839,
+      "loss": 2.7262,
+      "theoretical_loss": 3.4482982911717435,
+      "tokens_seen": 1879638016
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043781818181818185,
+      "loss": 2.8021,
+      "theoretical_loss": 3.4482881441644584,
+      "tokens_seen": 1879703552
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004377979797979798,
+      "loss": 2.8252,
+      "theoretical_loss": 3.4482779976099964,
+      "tokens_seen": 1879769088
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043777777777777776,
+      "loss": 2.7085,
+      "theoretical_loss": 3.448267851508322,
+      "tokens_seen": 1879834624
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004377575757575757,
+      "loss": 2.636,
+      "theoretical_loss": 3.448257705859399,
+      "tokens_seen": 1879900160
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004377373737373738,
+      "loss": 2.7353,
+      "theoretical_loss": 3.4482475606631913,
+      "tokens_seen": 1879965696
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043771717171717174,
+      "loss": 2.6714,
+      "theoretical_loss": 3.448237415919663,
+      "tokens_seen": 1880031232
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004376969696969697,
+      "loss": 2.5129,
+      "theoretical_loss": 3.448227271628778,
+      "tokens_seen": 1880096768
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043767676767676766,
+      "loss": 2.812,
+      "theoretical_loss": 3.4482171277905005,
+      "tokens_seen": 1880162304
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043765656565656567,
+      "loss": 2.6062,
+      "theoretical_loss": 3.448206984404795,
+      "tokens_seen": 1880227840
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004376363636363637,
+      "loss": 2.7784,
+      "theoretical_loss": 3.448196841471624,
+      "tokens_seen": 1880293376
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043761616161616164,
+      "loss": 2.6684,
+      "theoretical_loss": 3.4481866989909533,
+      "tokens_seen": 1880358912
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004375959595959596,
+      "loss": 2.5526,
+      "theoretical_loss": 3.4481765569627463,
+      "tokens_seen": 1880424448
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043757575757575755,
+      "loss": 2.494,
+      "theoretical_loss": 3.448166415386966,
+      "tokens_seen": 1880489984
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043755555555555556,
+      "loss": 2.8102,
+      "theoretical_loss": 3.448156274263578,
+      "tokens_seen": 1880555520
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043753535353535357,
+      "loss": 2.7983,
+      "theoretical_loss": 3.4481461335925463,
+      "tokens_seen": 1880621056
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043751515151515153,
+      "loss": 2.645,
+      "theoretical_loss": 3.4481359933738336,
+      "tokens_seen": 1880686592
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004374949494949495,
+      "loss": 2.5337,
+      "theoretical_loss": 3.4481258536074053,
+      "tokens_seen": 1880752128
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004374747474747475,
+      "loss": 2.8128,
+      "theoretical_loss": 3.448115714293225,
+      "tokens_seen": 1880817664
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043745454545454545,
+      "loss": 2.5904,
+      "theoretical_loss": 3.4481055754312564,
+      "tokens_seen": 1880883200
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043743434343434346,
+      "loss": 2.8262,
+      "theoretical_loss": 3.448095437021464,
+      "tokens_seen": 1880948736
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004374141414141414,
+      "loss": 2.548,
+      "theoretical_loss": 3.448085299063812,
+      "tokens_seen": 1881014272
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004373939393939394,
+      "loss": 2.2551,
+      "theoretical_loss": 3.448075161558264,
+      "tokens_seen": 1881079808
+    },
+    {
+      "epoch": 0.13,
+      "objective/train/docs_used": 1035977,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.632066488265991,
+      "objective/train/theoretical_loss": 3.448065024504784,
+      "objective/train/tokens_used": 240004576,
+      "theoretical_loss": 3.448065024504784,
+      "tokens_seen": 1881145344
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004373737373737374,
+      "loss": 2.8682,
+      "theoretical_loss": 3.448065024504784,
+      "tokens_seen": 1881145344
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043735353535353535,
+      "loss": 2.567,
+      "theoretical_loss": 3.448054887903337,
+      "tokens_seen": 1881210880
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043733333333333336,
+      "loss": 2.6562,
+      "theoretical_loss": 3.448044751753886,
+      "tokens_seen": 1881276416
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004373131313131313,
+      "loss": 2.5353,
+      "theoretical_loss": 3.448034616056396,
+      "tokens_seen": 1881341952
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004372929292929293,
+      "loss": 2.6606,
+      "theoretical_loss": 3.448024480810831,
+      "tokens_seen": 1881407488
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004372727272727273,
+      "loss": 2.7143,
+      "theoretical_loss": 3.4480143460171546,
+      "tokens_seen": 1881473024
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043725252525252524,
+      "loss": 2.723,
+      "theoretical_loss": 3.448004211675331,
+      "tokens_seen": 1881538560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043723232323232325,
+      "loss": 2.8467,
+      "theoretical_loss": 3.447994077785325,
+      "tokens_seen": 1881604096
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004372121212121212,
+      "loss": 2.727,
+      "theoretical_loss": 3.4479839443471,
+      "tokens_seen": 1881669632
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004371919191919192,
+      "loss": 2.7398,
+      "theoretical_loss": 3.4479738113606198,
+      "tokens_seen": 1881735168
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004371717171717172,
+      "loss": 2.4879,
+      "theoretical_loss": 3.4479636788258494,
+      "tokens_seen": 1881800704
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043715151515151513,
+      "loss": 2.8289,
+      "theoretical_loss": 3.4479535467427525,
+      "tokens_seen": 1881866240
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043713131313131314,
+      "loss": 2.7452,
+      "theoretical_loss": 3.4479434151112933,
+      "tokens_seen": 1881931776
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043711111111111115,
+      "loss": 2.546,
+      "theoretical_loss": 3.4479332839314365,
+      "tokens_seen": 1881997312
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004370909090909091,
+      "loss": 2.6797,
+      "theoretical_loss": 3.447923153203145,
+      "tokens_seen": 1882062848
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043707070707070707,
+      "loss": 2.7142,
+      "theoretical_loss": 3.447913022926384,
+      "tokens_seen": 1882128384
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004370505050505051,
+      "loss": 2.75,
+      "theoretical_loss": 3.4479028931011175,
+      "tokens_seen": 1882193920
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043703030303030303,
+      "loss": 2.8057,
+      "theoretical_loss": 3.4478927637273094,
+      "tokens_seen": 1882259456
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043701010101010105,
+      "loss": 2.827,
+      "theoretical_loss": 3.447882634804923,
+      "tokens_seen": 1882324992
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000436989898989899,
+      "loss": 2.4703,
+      "theoretical_loss": 3.4478725063339244,
+      "tokens_seen": 1882390528
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043696969696969696,
+      "loss": 2.749,
+      "theoretical_loss": 3.4478623783142766,
+      "tokens_seen": 1882456064
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043694949494949497,
+      "loss": 2.6755,
+      "theoretical_loss": 3.447852250745944,
+      "tokens_seen": 1882521600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043692929292929293,
+      "loss": 2.6679,
+      "theoretical_loss": 3.4478421236288908,
+      "tokens_seen": 1882587136
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00043690909090909094,
+      "loss": 2.7283,
+      "theoretical_loss": 3.447831996963081,
+      "tokens_seen": 1882652672
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0004368888888888889,
+      "loss": 2.8893,
+      "theoretical_loss": 3.447821870748479,
+      "tokens_seen": 1882718208
+    },
+    {
+      "epoch": 0.13,
+      "objective/train/docs_used": 1036495,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.311638116836548,
+      "objective/train/theoretical_loss": 3.4478117449850485,
+      "objective/train/tokens_used": 241642976,
+      "theoretical_loss": 3.4478117449850485,
+      "tokens_seen": 1882783744
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043686868686868685,
+      "loss": 2.602,
+      "theoretical_loss": 3.4478117449850485,
+      "tokens_seen": 1882783744
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043684848484848486,
+      "loss": 2.7245,
+      "theoretical_loss": 3.447801619672754,
+      "tokens_seen": 1882849280
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004368282828282829,
+      "loss": 2.6415,
+      "theoretical_loss": 3.44779149481156,
+      "tokens_seen": 1882914816
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043680808080808083,
+      "loss": 2.8219,
+      "theoretical_loss": 3.4477813704014313,
+      "tokens_seen": 1882980352
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004367878787878788,
+      "loss": 2.4011,
+      "theoretical_loss": 3.44777124644233,
+      "tokens_seen": 1883045888
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043676767676767674,
+      "loss": 2.6543,
+      "theoretical_loss": 3.4477611229342227,
+      "tokens_seen": 1883111424
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043674747474747476,
+      "loss": 2.66,
+      "theoretical_loss": 3.447750999877072,
+      "tokens_seen": 1883176960
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043672727272727277,
+      "loss": 3.0482,
+      "theoretical_loss": 3.447740877270842,
+      "tokens_seen": 1883242496
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004367070707070707,
+      "loss": 2.9646,
+      "theoretical_loss": 3.4477307551154985,
+      "tokens_seen": 1883308032
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004366868686868687,
+      "loss": 2.713,
+      "theoretical_loss": 3.4477206334110044,
+      "tokens_seen": 1883373568
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043666666666666664,
+      "loss": 2.5537,
+      "theoretical_loss": 3.4477105121573244,
+      "tokens_seen": 1883439104
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004366464646464647,
+      "loss": 2.7314,
+      "theoretical_loss": 3.447700391354423,
+      "tokens_seen": 1883504640
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043662626262626266,
+      "loss": 2.5515,
+      "theoretical_loss": 3.4476902710022634,
+      "tokens_seen": 1883570176
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004366060606060606,
+      "loss": 2.6342,
+      "theoretical_loss": 3.447680151100811,
+      "tokens_seen": 1883635712
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004365858585858586,
+      "loss": 2.5765,
+      "theoretical_loss": 3.4476700316500293,
+      "tokens_seen": 1883701248
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043656565656565653,
+      "loss": 2.624,
+      "theoretical_loss": 3.447659912649883,
+      "tokens_seen": 1883766784
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004365454545454546,
+      "loss": 2.7815,
+      "theoretical_loss": 3.447649794100336,
+      "tokens_seen": 1883832320
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043652525252525255,
+      "loss": 2.4913,
+      "theoretical_loss": 3.4476396760013532,
+      "tokens_seen": 1883897856
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004365050505050505,
+      "loss": 2.7279,
+      "theoretical_loss": 3.447629558352898,
+      "tokens_seen": 1883963392
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043648484848484847,
+      "loss": 2.5147,
+      "theoretical_loss": 3.4476194411549352,
+      "tokens_seen": 1884028928
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004364646464646465,
+      "loss": 2.669,
+      "theoretical_loss": 3.447609324407429,
+      "tokens_seen": 1884094464
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004364444444444445,
+      "loss": 2.5508,
+      "theoretical_loss": 3.4475992081103435,
+      "tokens_seen": 1884160000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043642424242424244,
+      "loss": 2.7432,
+      "theoretical_loss": 3.4475890922636436,
+      "tokens_seen": 1884225536
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004364040404040404,
+      "loss": 2.6209,
+      "theoretical_loss": 3.447578976867293,
+      "tokens_seen": 1884291072
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043638383838383836,
+      "loss": 2.4768,
+      "theoretical_loss": 3.4475688619212557,
+      "tokens_seen": 1884356608
+    },
+    {
+      "epoch": 0.14,
+      "objective/train/docs_used": 1037793,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5640203952789307,
+      "objective/train/theoretical_loss": 3.4475587474254965,
+      "objective/train/tokens_used": 243281376,
+      "theoretical_loss": 3.4475587474254965,
+      "tokens_seen": 1884422144
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043636363636363637,
+      "loss": 2.8154,
+      "theoretical_loss": 3.4475587474254965,
+      "tokens_seen": 1884422144
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004363434343434344,
+      "loss": 2.7487,
+      "theoretical_loss": 3.4475486333799794,
+      "tokens_seen": 1884487680
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043632323232323234,
+      "loss": 2.6674,
+      "theoretical_loss": 3.4475385197846693,
+      "tokens_seen": 1884553216
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004363030303030303,
+      "loss": 2.6732,
+      "theoretical_loss": 3.4475284066395293,
+      "tokens_seen": 1884618752
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004362828282828283,
+      "loss": 2.7621,
+      "theoretical_loss": 3.4475182939445252,
+      "tokens_seen": 1884684288
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043626262626262626,
+      "loss": 2.6823,
+      "theoretical_loss": 3.4475081816996207,
+      "tokens_seen": 1884749824
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004362424242424243,
+      "loss": 2.9771,
+      "theoretical_loss": 3.4474980699047792,
+      "tokens_seen": 1884815360
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043622222222222223,
+      "loss": 2.6468,
+      "theoretical_loss": 3.4474879585599667,
+      "tokens_seen": 1884880896
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004362020202020202,
+      "loss": 2.6748,
+      "theoretical_loss": 3.447477847665146,
+      "tokens_seen": 1884946432
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004361818181818182,
+      "loss": 2.6497,
+      "theoretical_loss": 3.4474677372202827,
+      "tokens_seen": 1885011968
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043616161616161616,
+      "loss": 2.4553,
+      "theoretical_loss": 3.4474576272253405,
+      "tokens_seen": 1885077504
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043614141414141417,
+      "loss": 2.7674,
+      "theoretical_loss": 3.4474475176802835,
+      "tokens_seen": 1885143040
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004361212121212121,
+      "loss": 2.7152,
+      "theoretical_loss": 3.4474374085850763,
+      "tokens_seen": 1885208576
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043610101010101013,
+      "loss": 2.5882,
+      "theoretical_loss": 3.447427299939683,
+      "tokens_seen": 1885274112
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004360808080808081,
+      "loss": 2.7256,
+      "theoretical_loss": 3.447417191744069,
+      "tokens_seen": 1885339648
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043606060606060605,
+      "loss": 2.6481,
+      "theoretical_loss": 3.4474070839981974,
+      "tokens_seen": 1885405184
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043604040404040406,
+      "loss": 2.6377,
+      "theoretical_loss": 3.447396976702033,
+      "tokens_seen": 1885470720
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000436020202020202,
+      "loss": 2.8219,
+      "theoretical_loss": 3.4473868698555403,
+      "tokens_seen": 1885536256
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000436,
+      "loss": 2.7823,
+      "theoretical_loss": 3.447376763458683,
+      "tokens_seen": 1885601792
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000435979797979798,
+      "loss": 2.7027,
+      "theoretical_loss": 3.4473666575114263,
+      "tokens_seen": 1885667328
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043595959595959594,
+      "loss": 2.7249,
+      "theoretical_loss": 3.447356552013735,
+      "tokens_seen": 1885732864
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043593939393939395,
+      "loss": 2.7471,
+      "theoretical_loss": 3.4473464469655717,
+      "tokens_seen": 1885798400
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043591919191919196,
+      "loss": 2.6487,
+      "theoretical_loss": 3.447336342366903,
+      "tokens_seen": 1885863936
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004358989898989899,
+      "loss": 2.4869,
+      "theoretical_loss": 3.4473262382176912,
+      "tokens_seen": 1885929472
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004358787878787879,
+      "loss": 2.766,
+      "theoretical_loss": 3.447316134517902,
+      "tokens_seen": 1885995008
+    },
+    {
+      "epoch": 0.14,
+      "objective/train/docs_used": 1038946,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.652374267578125,
+      "objective/train/theoretical_loss": 3.447306031267499,
+      "objective/train/tokens_used": 244919776,
+      "theoretical_loss": 3.447306031267499,
+      "tokens_seen": 1886060544
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043585858585858583,
+      "loss": 2.7403,
+      "theoretical_loss": 3.447306031267499,
+      "tokens_seen": 1886060544
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043583838383838384,
+      "loss": 2.7421,
+      "theoretical_loss": 3.4472959284664473,
+      "tokens_seen": 1886126080
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043581818181818186,
+      "loss": 2.7329,
+      "theoretical_loss": 3.4472858261147112,
+      "tokens_seen": 1886191616
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004357979797979798,
+      "loss": 2.6731,
+      "theoretical_loss": 3.447275724212255,
+      "tokens_seen": 1886257152
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043577777777777777,
+      "loss": 2.684,
+      "theoretical_loss": 3.4472656227590424,
+      "tokens_seen": 1886322688
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004357575757575757,
+      "loss": 2.4135,
+      "theoretical_loss": 3.4472555217550394,
+      "tokens_seen": 1886388224
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004357373737373738,
+      "loss": 2.6845,
+      "theoretical_loss": 3.4472454212002086,
+      "tokens_seen": 1886453760
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043571717171717175,
+      "loss": 2.8128,
+      "theoretical_loss": 3.4472353210945155,
+      "tokens_seen": 1886519296
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004356969696969697,
+      "loss": 2.6487,
+      "theoretical_loss": 3.4472252214379244,
+      "tokens_seen": 1886584832
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043567676767676766,
+      "loss": 2.8778,
+      "theoretical_loss": 3.4472151222304,
+      "tokens_seen": 1886650368
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004356565656565656,
+      "loss": 2.8099,
+      "theoretical_loss": 3.447205023471906,
+      "tokens_seen": 1886715904
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004356363636363637,
+      "loss": 2.6793,
+      "theoretical_loss": 3.4471949251624077,
+      "tokens_seen": 1886781440
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043561616161616164,
+      "loss": 2.8376,
+      "theoretical_loss": 3.4471848273018684,
+      "tokens_seen": 1886846976
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004355959595959596,
+      "loss": 2.816,
+      "theoretical_loss": 3.4471747298902535,
+      "tokens_seen": 1886912512
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043557575757575755,
+      "loss": 2.5422,
+      "theoretical_loss": 3.447164632927527,
+      "tokens_seen": 1886978048
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004355555555555555,
+      "loss": 2.7872,
+      "theoretical_loss": 3.447154536413654,
+      "tokens_seen": 1887043584
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004355353535353536,
+      "loss": 2.5499,
+      "theoretical_loss": 3.4471444403485982,
+      "tokens_seen": 1887109120
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043551515151515153,
+      "loss": 2.8599,
+      "theoretical_loss": 3.447134344732324,
+      "tokens_seen": 1887174656
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004354949494949495,
+      "loss": 2.5081,
+      "theoretical_loss": 3.447124249564797,
+      "tokens_seen": 1887240192
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043547474747474745,
+      "loss": 2.5106,
+      "theoretical_loss": 3.44711415484598,
+      "tokens_seen": 1887305728
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043545454545454546,
+      "loss": 2.8259,
+      "theoretical_loss": 3.447104060575839,
+      "tokens_seen": 1887371264
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043543434343434347,
+      "loss": 2.6445,
+      "theoretical_loss": 3.4470939667543377,
+      "tokens_seen": 1887436800
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004354141414141414,
+      "loss": 2.6661,
+      "theoretical_loss": 3.44708387338144,
+      "tokens_seen": 1887502336
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004353939393939394,
+      "loss": 2.4127,
+      "theoretical_loss": 3.447073780457112,
+      "tokens_seen": 1887567872
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043537373737373734,
+      "loss": 2.5971,
+      "theoretical_loss": 3.4470636879813172,
+      "tokens_seen": 1887633408
+    },
+    {
+      "epoch": 0.14,
+      "objective/train/docs_used": 1039523,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.781736135482788,
+      "objective/train/theoretical_loss": 3.44705359595402,
+      "objective/train/tokens_used": 246558176,
+      "theoretical_loss": 3.44705359595402,
+      "tokens_seen": 1887698944
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043535353535353535,
+      "loss": 2.7614,
+      "theoretical_loss": 3.44705359595402,
+      "tokens_seen": 1887698944
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043533333333333336,
+      "loss": 2.8807,
+      "theoretical_loss": 3.447043504375185,
+      "tokens_seen": 1887764480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004353131313131313,
+      "loss": 2.7627,
+      "theoretical_loss": 3.447033413244777,
+      "tokens_seen": 1887830016
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004352929292929293,
+      "loss": 2.763,
+      "theoretical_loss": 3.4470233225627602,
+      "tokens_seen": 1887895552
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004352727272727273,
+      "loss": 2.7779,
+      "theoretical_loss": 3.447013232329099,
+      "tokens_seen": 1887961088
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004352525252525253,
+      "loss": 2.7414,
+      "theoretical_loss": 3.4470031425437586,
+      "tokens_seen": 1888026624
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043523232323232325,
+      "loss": 2.8604,
+      "theoretical_loss": 3.446993053206703,
+      "tokens_seen": 1888092160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004352121212121212,
+      "loss": 2.4655,
+      "theoretical_loss": 3.4469829643178964,
+      "tokens_seen": 1888157696
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043519191919191917,
+      "loss": 2.7379,
+      "theoretical_loss": 3.4469728758773037,
+      "tokens_seen": 1888223232
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004351717171717172,
+      "loss": 2.8034,
+      "theoretical_loss": 3.44696278788489,
+      "tokens_seen": 1888288768
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004351515151515152,
+      "loss": 2.7015,
+      "theoretical_loss": 3.446952700340619,
+      "tokens_seen": 1888354304
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043513131313131315,
+      "loss": 2.6724,
+      "theoretical_loss": 3.446942613244455,
+      "tokens_seen": 1888419840
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004351111111111111,
+      "loss": 2.9569,
+      "theoretical_loss": 3.4469325265963637,
+      "tokens_seen": 1888485376
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004350909090909091,
+      "loss": 2.6612,
+      "theoretical_loss": 3.4469224403963086,
+      "tokens_seen": 1888550912
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043507070707070707,
+      "loss": 2.868,
+      "theoretical_loss": 3.4469123546442546,
+      "tokens_seen": 1888616448
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004350505050505051,
+      "loss": 2.6853,
+      "theoretical_loss": 3.4469022693401667,
+      "tokens_seen": 1888681984
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043503030303030304,
+      "loss": 2.698,
+      "theoretical_loss": 3.446892184484009,
+      "tokens_seen": 1888747520
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000435010101010101,
+      "loss": 2.7729,
+      "theoretical_loss": 3.4468821000757464,
+      "tokens_seen": 1888813056
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000434989898989899,
+      "loss": 2.7598,
+      "theoretical_loss": 3.4468720161153428,
+      "tokens_seen": 1888878592
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043496969696969696,
+      "loss": 2.5709,
+      "theoretical_loss": 3.446861932602763,
+      "tokens_seen": 1888944128
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000434949494949495,
+      "loss": 2.7712,
+      "theoretical_loss": 3.446851849537972,
+      "tokens_seen": 1889009664
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043492929292929293,
+      "loss": 2.7276,
+      "theoretical_loss": 3.446841766920934,
+      "tokens_seen": 1889075200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043490909090909094,
+      "loss": 2.6929,
+      "theoretical_loss": 3.446831684751614,
+      "tokens_seen": 1889140736
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004348888888888889,
+      "loss": 2.6453,
+      "theoretical_loss": 3.446821603029976,
+      "tokens_seen": 1889206272
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043486868686868686,
+      "loss": 2.5445,
+      "theoretical_loss": 3.4468115217559854,
+      "tokens_seen": 1889271808
+    },
+    {
+      "epoch": 0.14,
+      "objective/train/docs_used": 1040960,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6786797046661377,
+      "objective/train/theoretical_loss": 3.446801440929606,
+      "objective/train/tokens_used": 248196576,
+      "theoretical_loss": 3.446801440929606,
+      "tokens_seen": 1889337344
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043484848484848487,
+      "loss": 2.8896,
+      "theoretical_loss": 3.446801440929606,
+      "tokens_seen": 1889337344
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004348282828282828,
+      "loss": 2.7437,
+      "theoretical_loss": 3.4467913605508027,
+      "tokens_seen": 1889402880
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043480808080808084,
+      "loss": 2.7,
+      "theoretical_loss": 3.4467812806195397,
+      "tokens_seen": 1889468416
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004347878787878788,
+      "loss": 2.7376,
+      "theoretical_loss": 3.4467712011357827,
+      "tokens_seen": 1889533952
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043476767676767675,
+      "loss": 2.7908,
+      "theoretical_loss": 3.4467611220994954,
+      "tokens_seen": 1889599488
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043474747474747476,
+      "loss": 2.7504,
+      "theoretical_loss": 3.446751043510642,
+      "tokens_seen": 1889665024
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043472727272727277,
+      "loss": 2.4802,
+      "theoretical_loss": 3.4467409653691887,
+      "tokens_seen": 1889730560
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043470707070707073,
+      "loss": 2.7347,
+      "theoretical_loss": 3.446730887675099,
+      "tokens_seen": 1889796096
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004346868686868687,
+      "loss": 2.6914,
+      "theoretical_loss": 3.4467208104283373,
+      "tokens_seen": 1889861632
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043466666666666664,
+      "loss": 2.8039,
+      "theoretical_loss": 3.446710733628869,
+      "tokens_seen": 1889927168
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043464646464646465,
+      "loss": 2.9309,
+      "theoretical_loss": 3.446700657276658,
+      "tokens_seen": 1889992704
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043462626262626267,
+      "loss": 2.7041,
+      "theoretical_loss": 3.4466905813716697,
+      "tokens_seen": 1890058240
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004346060606060606,
+      "loss": 2.8169,
+      "theoretical_loss": 3.446680505913868,
+      "tokens_seen": 1890123776
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004345858585858586,
+      "loss": 2.67,
+      "theoretical_loss": 3.446670430903218,
+      "tokens_seen": 1890189312
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043456565656565654,
+      "loss": 2.7819,
+      "theoretical_loss": 3.4466603563396845,
+      "tokens_seen": 1890254848
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004345454545454546,
+      "loss": 2.817,
+      "theoretical_loss": 3.4466502822232323,
+      "tokens_seen": 1890320384
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043452525252525256,
+      "loss": 2.8273,
+      "theoretical_loss": 3.4466402085538252,
+      "tokens_seen": 1890385920
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004345050505050505,
+      "loss": 2.863,
+      "theoretical_loss": 3.446630135331428,
+      "tokens_seen": 1890451456
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043448484848484847,
+      "loss": 2.6555,
+      "theoretical_loss": 3.4466200625560064,
+      "tokens_seen": 1890516992
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043446464646464643,
+      "loss": 2.7843,
+      "theoretical_loss": 3.446609990227524,
+      "tokens_seen": 1890582528
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004344444444444445,
+      "loss": 2.6445,
+      "theoretical_loss": 3.446599918345946,
+      "tokens_seen": 1890648064
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043442424242424245,
+      "loss": 2.6006,
+      "theoretical_loss": 3.4465898469112366,
+      "tokens_seen": 1890713600
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004344040404040404,
+      "loss": 2.928,
+      "theoretical_loss": 3.446579775923362,
+      "tokens_seen": 1890779136
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043438383838383836,
+      "loss": 2.6915,
+      "theoretical_loss": 3.4465697053822844,
+      "tokens_seen": 1890844672
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004343636363636364,
+      "loss": 2.9646,
+      "theoretical_loss": 3.4465596352879704,
+      "tokens_seen": 1890910208
+    },
+    {
+      "epoch": 0.14,
+      "objective/train/docs_used": 1041567,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.472703456878662,
+      "objective/train/theoretical_loss": 3.446549565640384,
+      "objective/train/tokens_used": 249834976,
+      "theoretical_loss": 3.446549565640384,
+      "tokens_seen": 1890975744
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004343434343434344,
+      "loss": 2.969,
+      "theoretical_loss": 3.446549565640384,
+      "tokens_seen": 1890975744
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043432323232323234,
+      "loss": 2.5333,
+      "theoretical_loss": 3.44653949643949,
+      "tokens_seen": 1891041280
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004343030303030303,
+      "loss": 2.7445,
+      "theoretical_loss": 3.4465294276852534,
+      "tokens_seen": 1891106816
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043428282828282826,
+      "loss": 2.5441,
+      "theoretical_loss": 3.446519359377638,
+      "tokens_seen": 1891172352
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043426262626262627,
+      "loss": 2.5551,
+      "theoretical_loss": 3.44650929151661,
+      "tokens_seen": 1891237888
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004342424242424243,
+      "loss": 2.7024,
+      "theoretical_loss": 3.4464992241021326,
+      "tokens_seen": 1891303424
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043422222222222224,
+      "loss": 2.6171,
+      "theoretical_loss": 3.446489157134171,
+      "tokens_seen": 1891368960
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004342020202020202,
+      "loss": 2.7406,
+      "theoretical_loss": 3.4464790906126908,
+      "tokens_seen": 1891434496
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004341818181818182,
+      "loss": 2.6309,
+      "theoretical_loss": 3.446469024537656,
+      "tokens_seen": 1891500032
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043416161616161616,
+      "loss": 2.6551,
+      "theoretical_loss": 3.446458958909031,
+      "tokens_seen": 1891565568
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043414141414141417,
+      "loss": 2.6236,
+      "theoretical_loss": 3.4464488937267808,
+      "tokens_seen": 1891631104
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043412121212121213,
+      "loss": 2.6514,
+      "theoretical_loss": 3.4464388289908703,
+      "tokens_seen": 1891696640
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004341010101010101,
+      "loss": 2.6576,
+      "theoretical_loss": 3.4464287647012646,
+      "tokens_seen": 1891762176
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004340808080808081,
+      "loss": 2.6109,
+      "theoretical_loss": 3.4464187008579277,
+      "tokens_seen": 1891827712
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043406060606060605,
+      "loss": 2.5924,
+      "theoretical_loss": 3.446408637460825,
+      "tokens_seen": 1891893248
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043404040404040406,
+      "loss": 2.4555,
+      "theoretical_loss": 3.446398574509921,
+      "tokens_seen": 1891958784
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000434020202020202,
+      "loss": 2.7578,
+      "theoretical_loss": 3.44638851200518,
+      "tokens_seen": 1892024320
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043400000000000003,
+      "loss": 2.4286,
+      "theoretical_loss": 3.4463784499465673,
+      "tokens_seen": 1892089856
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000433979797979798,
+      "loss": 2.4674,
+      "theoretical_loss": 3.4463683883340472,
+      "tokens_seen": 1892155392
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043395959595959595,
+      "loss": 2.7483,
+      "theoretical_loss": 3.4463583271675855,
+      "tokens_seen": 1892220928
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043393939393939396,
+      "loss": 2.8933,
+      "theoretical_loss": 3.446348266447146,
+      "tokens_seen": 1892286464
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004339191919191919,
+      "loss": 2.6796,
+      "theoretical_loss": 3.446338206172694,
+      "tokens_seen": 1892352000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004338989898989899,
+      "loss": 2.6755,
+      "theoretical_loss": 3.446328146344194,
+      "tokens_seen": 1892417536
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004338787878787879,
+      "loss": 2.6965,
+      "theoretical_loss": 3.44631808696161,
+      "tokens_seen": 1892483072
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043385858585858584,
+      "loss": 2.6102,
+      "theoretical_loss": 3.4463080280249088,
+      "tokens_seen": 1892548608
+    },
+    {
+      "epoch": 0.14,
+      "objective/train/docs_used": 1042675,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5914690494537354,
+      "objective/train/theoretical_loss": 3.4462979695340534,
+      "objective/train/tokens_used": 251473376,
+      "theoretical_loss": 3.4462979695340534,
+      "tokens_seen": 1892614144
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043383838383838385,
+      "loss": 2.5992,
+      "theoretical_loss": 3.4462979695340534,
+      "tokens_seen": 1892614144
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043381818181818186,
+      "loss": 2.7988,
+      "theoretical_loss": 3.4462879114890095,
+      "tokens_seen": 1892679680
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004337979797979798,
+      "loss": 2.7278,
+      "theoretical_loss": 3.446277853889742,
+      "tokens_seen": 1892745216
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004337777777777778,
+      "loss": 2.7914,
+      "theoretical_loss": 3.4462677967362145,
+      "tokens_seen": 1892810752
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043375757575757573,
+      "loss": 2.5553,
+      "theoretical_loss": 3.4462577400283934,
+      "tokens_seen": 1892876288
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043373737373737374,
+      "loss": 2.8135,
+      "theoretical_loss": 3.4462476837662424,
+      "tokens_seen": 1892941824
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043371717171717175,
+      "loss": 2.7649,
+      "theoretical_loss": 3.446237627949727,
+      "tokens_seen": 1893007360
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004336969696969697,
+      "loss": 2.454,
+      "theoretical_loss": 3.4462275725788114,
+      "tokens_seen": 1893072896
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043367676767676767,
+      "loss": 2.9034,
+      "theoretical_loss": 3.4462175176534613,
+      "tokens_seen": 1893138432
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004336565656565656,
+      "loss": 2.8346,
+      "theoretical_loss": 3.44620746317364,
+      "tokens_seen": 1893203968
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043363636363636364,
+      "loss": 2.3368,
+      "theoretical_loss": 3.446197409139314,
+      "tokens_seen": 1893269504
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043361616161616165,
+      "loss": 2.8816,
+      "theoretical_loss": 3.446187355550448,
+      "tokens_seen": 1893335040
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004335959595959596,
+      "loss": 2.591,
+      "theoretical_loss": 3.4461773024070057,
+      "tokens_seen": 1893400576
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043357575757575756,
+      "loss": 2.8308,
+      "theoretical_loss": 3.4461672497089526,
+      "tokens_seen": 1893466112
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043355555555555557,
+      "loss": 2.7431,
+      "theoretical_loss": 3.4461571974562535,
+      "tokens_seen": 1893531648
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004335353535353536,
+      "loss": 2.6631,
+      "theoretical_loss": 3.4461471456488733,
+      "tokens_seen": 1893597184
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043351515151515154,
+      "loss": 2.6796,
+      "theoretical_loss": 3.446137094286777,
+      "tokens_seen": 1893662720
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004334949494949495,
+      "loss": 2.5672,
+      "theoretical_loss": 3.4461270433699296,
+      "tokens_seen": 1893728256
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043347474747474745,
+      "loss": 2.5745,
+      "theoretical_loss": 3.446116992898295,
+      "tokens_seen": 1893793792
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043345454545454546,
+      "loss": 2.7892,
+      "theoretical_loss": 3.4461069428718387,
+      "tokens_seen": 1893859328
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004334343434343435,
+      "loss": 2.7141,
+      "theoretical_loss": 3.4460968932905263,
+      "tokens_seen": 1893924864
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043341414141414143,
+      "loss": 2.7473,
+      "theoretical_loss": 3.4460868441543218,
+      "tokens_seen": 1893990400
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004333939393939394,
+      "loss": 2.5743,
+      "theoretical_loss": 3.44607679546319,
+      "tokens_seen": 1894055936
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043337373737373735,
+      "loss": 2.7776,
+      "theoretical_loss": 3.4460667472170963,
+      "tokens_seen": 1894121472
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004333535353535354,
+      "loss": 2.7774,
+      "theoretical_loss": 3.446056699416005,
+      "tokens_seen": 1894187008
+    },
+    {
+      "epoch": 0.14,
+      "objective/train/docs_used": 1043835,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.616513967514038,
+      "objective/train/theoretical_loss": 3.446046652059882,
+      "objective/train/tokens_used": 253111776,
+      "theoretical_loss": 3.446046652059882,
+      "tokens_seen": 1894252544
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043333333333333337,
+      "loss": 2.6525,
+      "theoretical_loss": 3.446046652059882,
+      "tokens_seen": 1894252544
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004333131313131313,
+      "loss": 2.6293,
+      "theoretical_loss": 3.4460366051486915,
+      "tokens_seen": 1894318080
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004332929292929293,
+      "loss": 2.7239,
+      "theoretical_loss": 3.446026558682398,
+      "tokens_seen": 1894383616
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043327272727272724,
+      "loss": 2.7973,
+      "theoretical_loss": 3.446016512660967,
+      "tokens_seen": 1894449152
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004332525252525253,
+      "loss": 2.6836,
+      "theoretical_loss": 3.4460064670843638,
+      "tokens_seen": 1894514688
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043323232323232326,
+      "loss": 2.7825,
+      "theoretical_loss": 3.445996421952552,
+      "tokens_seen": 1894580224
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004332121212121212,
+      "loss": 2.7624,
+      "theoretical_loss": 3.445986377265498,
+      "tokens_seen": 1894645760
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004331919191919192,
+      "loss": 2.5716,
+      "theoretical_loss": 3.445976333023166,
+      "tokens_seen": 1894711296
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004331717171717172,
+      "loss": 2.7031,
+      "theoretical_loss": 3.4459662892255203,
+      "tokens_seen": 1894776832
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004331515151515152,
+      "loss": 2.5949,
+      "theoretical_loss": 3.4459562458725275,
+      "tokens_seen": 1894842368
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043313131313131315,
+      "loss": 2.5145,
+      "theoretical_loss": 3.445946202964151,
+      "tokens_seen": 1894907904
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004331111111111111,
+      "loss": 2.6517,
+      "theoretical_loss": 3.4459361605003567,
+      "tokens_seen": 1894973440
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043309090909090907,
+      "loss": 2.5671,
+      "theoretical_loss": 3.4459261184811085,
+      "tokens_seen": 1895038976
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004330707070707071,
+      "loss": 2.6993,
+      "theoretical_loss": 3.4459160769063724,
+      "tokens_seen": 1895104512
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004330505050505051,
+      "loss": 2.6661,
+      "theoretical_loss": 3.445906035776113,
+      "tokens_seen": 1895170048
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043303030303030305,
+      "loss": 2.8961,
+      "theoretical_loss": 3.4458959950902948,
+      "tokens_seen": 1895235584
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000433010101010101,
+      "loss": 2.7048,
+      "theoretical_loss": 3.445885954848884,
+      "tokens_seen": 1895301120
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000432989898989899,
+      "loss": 2.6647,
+      "theoretical_loss": 3.445875915051844,
+      "tokens_seen": 1895366656
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043296969696969697,
+      "loss": 2.5512,
+      "theoretical_loss": 3.4458658756991407,
+      "tokens_seen": 1895432192
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000432949494949495,
+      "loss": 2.8466,
+      "theoretical_loss": 3.445855836790739,
+      "tokens_seen": 1895497728
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043292929292929294,
+      "loss": 2.7774,
+      "theoretical_loss": 3.4458457983266033,
+      "tokens_seen": 1895563264
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004329090909090909,
+      "loss": 2.8034,
+      "theoretical_loss": 3.4458357603066996,
+      "tokens_seen": 1895628800
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004328888888888889,
+      "loss": 2.6688,
+      "theoretical_loss": 3.445825722730992,
+      "tokens_seen": 1895694336
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043286868686868686,
+      "loss": 2.2792,
+      "theoretical_loss": 3.4458156855994457,
+      "tokens_seen": 1895759872
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004328484848484849,
+      "loss": 2.6926,
+      "theoretical_loss": 3.4458056489120263,
+      "tokens_seen": 1895825408
+    },
+    {
+      "epoch": 0.14,
+      "objective/train/docs_used": 1044447,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.802863836288452,
+      "objective/train/theoretical_loss": 3.4457956126686975,
+      "objective/train/tokens_used": 254750176,
+      "theoretical_loss": 3.4457956126686975,
+      "tokens_seen": 1895890944
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043282828282828283,
+      "loss": 2.6796,
+      "theoretical_loss": 3.4457956126686975,
+      "tokens_seen": 1895890944
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043280808080808084,
+      "loss": 2.7162,
+      "theoretical_loss": 3.445785576869426,
+      "tokens_seen": 1895956480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004327878787878788,
+      "loss": 2.6145,
+      "theoretical_loss": 3.445775541514175,
+      "tokens_seen": 1896022016
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043276767676767676,
+      "loss": 2.8484,
+      "theoretical_loss": 3.445765506602911,
+      "tokens_seen": 1896087552
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043274747474747477,
+      "loss": 2.6471,
+      "theoretical_loss": 3.4457554721355983,
+      "tokens_seen": 1896153088
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004327272727272727,
+      "loss": 2.6083,
+      "theoretical_loss": 3.445745438112202,
+      "tokens_seen": 1896218624
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043270707070707073,
+      "loss": 2.5806,
+      "theoretical_loss": 3.4457354045326873,
+      "tokens_seen": 1896284160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004326868686868687,
+      "loss": 2.8685,
+      "theoretical_loss": 3.4457253713970184,
+      "tokens_seen": 1896349696
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043266666666666665,
+      "loss": 2.6143,
+      "theoretical_loss": 3.4457153387051616,
+      "tokens_seen": 1896415232
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043264646464646466,
+      "loss": 2.7779,
+      "theoretical_loss": 3.4457053064570813,
+      "tokens_seen": 1896480768
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043262626262626267,
+      "loss": 2.8286,
+      "theoretical_loss": 3.4456952746527425,
+      "tokens_seen": 1896546304
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043260606060606063,
+      "loss": 2.8636,
+      "theoretical_loss": 3.44568524329211,
+      "tokens_seen": 1896611840
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004325858585858586,
+      "loss": 2.7182,
+      "theoretical_loss": 3.44567521237515,
+      "tokens_seen": 1896677376
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043256565656565654,
+      "loss": 2.7157,
+      "theoretical_loss": 3.445665181901826,
+      "tokens_seen": 1896742912
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043254545454545455,
+      "loss": 2.5607,
+      "theoretical_loss": 3.4456551518721037,
+      "tokens_seen": 1896808448
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043252525252525256,
+      "loss": 2.6054,
+      "theoretical_loss": 3.4456451222859483,
+      "tokens_seen": 1896873984
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004325050505050505,
+      "loss": 2.5935,
+      "theoretical_loss": 3.4456350931433253,
+      "tokens_seen": 1896939520
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004324848484848485,
+      "loss": 2.6931,
+      "theoretical_loss": 3.445625064444199,
+      "tokens_seen": 1897005056
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043246464646464643,
+      "loss": 2.7926,
+      "theoretical_loss": 3.445615036188534,
+      "tokens_seen": 1897070592
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004324444444444445,
+      "loss": 2.783,
+      "theoretical_loss": 3.445605008376297,
+      "tokens_seen": 1897136128
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043242424242424246,
+      "loss": 2.4811,
+      "theoretical_loss": 3.4455949810074515,
+      "tokens_seen": 1897201664
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004324040404040404,
+      "loss": 2.6182,
+      "theoretical_loss": 3.445584954081964,
+      "tokens_seen": 1897267200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043238383838383837,
+      "loss": 2.4329,
+      "theoretical_loss": 3.4455749275997976,
+      "tokens_seen": 1897332736
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004323636363636363,
+      "loss": 2.8432,
+      "theoretical_loss": 3.4455649015609193,
+      "tokens_seen": 1897398272
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004323434343434344,
+      "loss": 2.5067,
+      "theoretical_loss": 3.4455548759652936,
+      "tokens_seen": 1897463808
+    },
+    {
+      "epoch": 0.14,
+      "objective/train/docs_used": 1045650,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5606865882873535,
+      "objective/train/theoretical_loss": 3.4455448508128854,
+      "objective/train/tokens_used": 256388576,
+      "theoretical_loss": 3.4455448508128854,
+      "tokens_seen": 1897529344
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043232323232323235,
+      "loss": 2.4806,
+      "theoretical_loss": 3.4455448508128854,
+      "tokens_seen": 1897529344
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004323030303030303,
+      "loss": 2.5365,
+      "theoretical_loss": 3.4455348261036596,
+      "tokens_seen": 1897594880
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043228282828282826,
+      "loss": 2.4085,
+      "theoretical_loss": 3.445524801837582,
+      "tokens_seen": 1897660416
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004322626262626262,
+      "loss": 2.7238,
+      "theoretical_loss": 3.4455147780146174,
+      "tokens_seen": 1897725952
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004322424242424243,
+      "loss": 2.5818,
+      "theoretical_loss": 3.4455047546347304,
+      "tokens_seen": 1897791488
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043222222222222224,
+      "loss": 2.8515,
+      "theoretical_loss": 3.4454947316978863,
+      "tokens_seen": 1897857024
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004322020202020202,
+      "loss": 2.5814,
+      "theoretical_loss": 3.445484709204051,
+      "tokens_seen": 1897922560
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043218181818181816,
+      "loss": 3.0203,
+      "theoretical_loss": 3.445474687153189,
+      "tokens_seen": 1897988096
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043216161616161617,
+      "loss": 3.0408,
+      "theoretical_loss": 3.4454646655452654,
+      "tokens_seen": 1898053632
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004321414141414142,
+      "loss": 2.7143,
+      "theoretical_loss": 3.4454546443802454,
+      "tokens_seen": 1898119168
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043212121212121213,
+      "loss": 2.8371,
+      "theoretical_loss": 3.4454446236580942,
+      "tokens_seen": 1898184704
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004321010101010101,
+      "loss": 2.7068,
+      "theoretical_loss": 3.445434603378777,
+      "tokens_seen": 1898250240
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043208080808080805,
+      "loss": 2.6067,
+      "theoretical_loss": 3.445424583542258,
+      "tokens_seen": 1898315776
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043206060606060606,
+      "loss": 2.607,
+      "theoretical_loss": 3.4454145641485043,
+      "tokens_seen": 1898381312
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043204040404040407,
+      "loss": 2.4594,
+      "theoretical_loss": 3.445404545197479,
+      "tokens_seen": 1898446848
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000432020202020202,
+      "loss": 2.6032,
+      "theoretical_loss": 3.4453945266891486,
+      "tokens_seen": 1898512384
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000432,
+      "loss": 2.5731,
+      "theoretical_loss": 3.4453845086234782,
+      "tokens_seen": 1898577920
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000431979797979798,
+      "loss": 2.7252,
+      "theoretical_loss": 3.4453744910004325,
+      "tokens_seen": 1898643456
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043195959595959595,
+      "loss": 2.5493,
+      "theoretical_loss": 3.4453644738199767,
+      "tokens_seen": 1898708992
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043193939393939396,
+      "loss": 2.7383,
+      "theoretical_loss": 3.4453544570820758,
+      "tokens_seen": 1898774528
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004319191919191919,
+      "loss": 2.5631,
+      "theoretical_loss": 3.4453444407866956,
+      "tokens_seen": 1898840064
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004318989898989899,
+      "loss": 2.5864,
+      "theoretical_loss": 3.4453344249338005,
+      "tokens_seen": 1898905600
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004318787878787879,
+      "loss": 2.5916,
+      "theoretical_loss": 3.445324409523356,
+      "tokens_seen": 1898971136
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004318585858585859,
+      "loss": 2.7156,
+      "theoretical_loss": 3.4453143945553277,
+      "tokens_seen": 1899036672
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00043183838383838386,
+      "loss": 2.8104,
+      "theoretical_loss": 3.4453043800296808,
+      "tokens_seen": 1899102208
+    },
+    {
+      "epoch": 0.14,
+      "objective/train/docs_used": 1046191,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.9813382625579834,
+      "objective/train/theoretical_loss": 3.4452943659463795,
+      "objective/train/tokens_used": 258026976,
+      "theoretical_loss": 3.4452943659463795,
+      "tokens_seen": 1899167744
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0004318181818181818,
+      "loss": 2.7552,
+      "theoretical_loss": 3.4452943659463795,
+      "tokens_seen": 1899167744
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004317979797979798,
+      "loss": 2.7213,
+      "theoretical_loss": 3.4452843523053898,
+      "tokens_seen": 1899233280
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004317777777777778,
+      "loss": 2.5726,
+      "theoretical_loss": 3.445274339106677,
+      "tokens_seen": 1899298816
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004317575757575758,
+      "loss": 2.6529,
+      "theoretical_loss": 3.445264326350206,
+      "tokens_seen": 1899364352
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043173737373737375,
+      "loss": 2.6968,
+      "theoretical_loss": 3.4452543140359415,
+      "tokens_seen": 1899429888
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004317171717171717,
+      "loss": 2.6248,
+      "theoretical_loss": 3.44524430216385,
+      "tokens_seen": 1899495424
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004316969696969697,
+      "loss": 2.8309,
+      "theoretical_loss": 3.4452342907338958,
+      "tokens_seen": 1899560960
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004316767676767677,
+      "loss": 2.5161,
+      "theoretical_loss": 3.445224279746044,
+      "tokens_seen": 1899626496
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004316565656565657,
+      "loss": 2.6626,
+      "theoretical_loss": 3.4452142692002603,
+      "tokens_seen": 1899692032
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043163636363636364,
+      "loss": 2.6857,
+      "theoretical_loss": 3.44520425909651,
+      "tokens_seen": 1899757568
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043161616161616165,
+      "loss": 2.7009,
+      "theoretical_loss": 3.445194249434758,
+      "tokens_seen": 1899823104
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004315959595959596,
+      "loss": 2.4982,
+      "theoretical_loss": 3.44518424021497,
+      "tokens_seen": 1899888640
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043157575757575757,
+      "loss": 2.698,
+      "theoretical_loss": 3.4451742314371105,
+      "tokens_seen": 1899954176
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004315555555555556,
+      "loss": 2.7291,
+      "theoretical_loss": 3.4451642231011452,
+      "tokens_seen": 1900019712
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043153535353535353,
+      "loss": 2.6778,
+      "theoretical_loss": 3.4451542152070393,
+      "tokens_seen": 1900085248
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043151515151515154,
+      "loss": 2.676,
+      "theoretical_loss": 3.4451442077547583,
+      "tokens_seen": 1900150784
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004314949494949495,
+      "loss": 2.7419,
+      "theoretical_loss": 3.445134200744267,
+      "tokens_seen": 1900216320
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043147474747474746,
+      "loss": 2.5326,
+      "theoretical_loss": 3.44512419417553,
+      "tokens_seen": 1900281856
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043145454545454547,
+      "loss": 2.5492,
+      "theoretical_loss": 3.4451141880485148,
+      "tokens_seen": 1900347392
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004314343434343435,
+      "loss": 2.4373,
+      "theoretical_loss": 3.4451041823631843,
+      "tokens_seen": 1900412928
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043141414141414144,
+      "loss": 2.5494,
+      "theoretical_loss": 3.445094177119505,
+      "tokens_seen": 1900478464
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004313939393939394,
+      "loss": 2.5809,
+      "theoretical_loss": 3.4450841723174426,
+      "tokens_seen": 1900544000
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043137373737373735,
+      "loss": 2.4989,
+      "theoretical_loss": 3.4450741679569608,
+      "tokens_seen": 1900609536
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043135353535353536,
+      "loss": 2.677,
+      "theoretical_loss": 3.445064164038026,
+      "tokens_seen": 1900675072
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004313333333333334,
+      "loss": 2.6094,
+      "theoretical_loss": 3.4450541605606038,
+      "tokens_seen": 1900740608
+    },
+    {
+      "epoch": 0.15,
+      "objective/train/docs_used": 1047224,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5529582500457764,
+      "objective/train/theoretical_loss": 3.4450441575246584,
+      "objective/train/tokens_used": 259665376,
+      "theoretical_loss": 3.4450441575246584,
+      "tokens_seen": 1900806144
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043131313131313133,
+      "loss": 2.4374,
+      "theoretical_loss": 3.4450441575246584,
+      "tokens_seen": 1900806144
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004312929292929293,
+      "loss": 2.5186,
+      "theoretical_loss": 3.4450341549301564,
+      "tokens_seen": 1900871680
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043127272727272724,
+      "loss": 2.7014,
+      "theoretical_loss": 3.4450241527770613,
+      "tokens_seen": 1900937216
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004312525252525253,
+      "loss": 2.8584,
+      "theoretical_loss": 3.4450141510653403,
+      "tokens_seen": 1901002752
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043123232323232327,
+      "loss": 2.5654,
+      "theoretical_loss": 3.4450041497949577,
+      "tokens_seen": 1901068288
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004312121212121212,
+      "loss": 2.6917,
+      "theoretical_loss": 3.444994148965879,
+      "tokens_seen": 1901133824
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004311919191919192,
+      "loss": 2.5157,
+      "theoretical_loss": 3.4449841485780697,
+      "tokens_seen": 1901199360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043117171717171714,
+      "loss": 2.457,
+      "theoretical_loss": 3.444974148631495,
+      "tokens_seen": 1901264896
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004311515151515152,
+      "loss": 2.7993,
+      "theoretical_loss": 3.4449641491261196,
+      "tokens_seen": 1901330432
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043113131313131316,
+      "loss": 2.3877,
+      "theoretical_loss": 3.4449541500619096,
+      "tokens_seen": 1901395968
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004311111111111111,
+      "loss": 2.762,
+      "theoretical_loss": 3.4449441514388304,
+      "tokens_seen": 1901461504
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043109090909090907,
+      "loss": 2.5943,
+      "theoretical_loss": 3.4449341532568467,
+      "tokens_seen": 1901527040
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004310707070707071,
+      "loss": 2.8475,
+      "theoretical_loss": 3.4449241555159245,
+      "tokens_seen": 1901592576
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004310505050505051,
+      "loss": 2.5425,
+      "theoretical_loss": 3.4449141582160285,
+      "tokens_seen": 1901658112
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043103030303030305,
+      "loss": 2.6034,
+      "theoretical_loss": 3.4449041613571247,
+      "tokens_seen": 1901723648
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000431010101010101,
+      "loss": 2.6465,
+      "theoretical_loss": 3.4448941649391784,
+      "tokens_seen": 1901789184
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043098989898989897,
+      "loss": 2.8249,
+      "theoretical_loss": 3.444884168962154,
+      "tokens_seen": 1901854720
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000430969696969697,
+      "loss": 2.6515,
+      "theoretical_loss": 3.444874173426018,
+      "tokens_seen": 1901920256
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000430949494949495,
+      "loss": 2.7534,
+      "theoretical_loss": 3.444864178330735,
+      "tokens_seen": 1901985792
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043092929292929294,
+      "loss": 2.5677,
+      "theoretical_loss": 3.444854183676271,
+      "tokens_seen": 1902051328
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004309090909090909,
+      "loss": 2.5041,
+      "theoretical_loss": 3.444844189462591,
+      "tokens_seen": 1902116864
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004308888888888889,
+      "loss": 2.4229,
+      "theoretical_loss": 3.44483419568966,
+      "tokens_seen": 1902182400
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043086868686868687,
+      "loss": 2.8729,
+      "theoretical_loss": 3.4448242023574447,
+      "tokens_seen": 1902247936
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004308484848484849,
+      "loss": 2.726,
+      "theoretical_loss": 3.444814209465909,
+      "tokens_seen": 1902313472
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043082828282828284,
+      "loss": 2.6824,
+      "theoretical_loss": 3.444804217015019,
+      "tokens_seen": 1902379008
+    },
+    {
+      "epoch": 0.15,
+      "objective/train/docs_used": 1047755,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.809091329574585,
+      "objective/train/theoretical_loss": 3.44479422500474,
+      "objective/train/tokens_used": 261303776,
+      "theoretical_loss": 3.44479422500474,
+      "tokens_seen": 1902444544
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004308080808080808,
+      "loss": 2.5754,
+      "theoretical_loss": 3.44479422500474,
+      "tokens_seen": 1902444544
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004307878787878788,
+      "loss": 2.6735,
+      "theoretical_loss": 3.4447842334350374,
+      "tokens_seen": 1902510080
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043076767676767676,
+      "loss": 2.7893,
+      "theoretical_loss": 3.4447742423058765,
+      "tokens_seen": 1902575616
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043074747474747477,
+      "loss": 2.7023,
+      "theoretical_loss": 3.4447642516172223,
+      "tokens_seen": 1902641152
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043072727272727273,
+      "loss": 2.6068,
+      "theoretical_loss": 3.4447542613690416,
+      "tokens_seen": 1902706688
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043070707070707074,
+      "loss": 2.6704,
+      "theoretical_loss": 3.444744271561298,
+      "tokens_seen": 1902772224
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004306868686868687,
+      "loss": 2.5841,
+      "theoretical_loss": 3.4447342821939584,
+      "tokens_seen": 1902837760
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043066666666666665,
+      "loss": 2.5968,
+      "theoretical_loss": 3.4447242932669875,
+      "tokens_seen": 1902903296
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043064646464646467,
+      "loss": 2.4953,
+      "theoretical_loss": 3.4447143047803506,
+      "tokens_seen": 1902968832
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004306262626262626,
+      "loss": 2.7003,
+      "theoretical_loss": 3.444704316734013,
+      "tokens_seen": 1903034368
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043060606060606063,
+      "loss": 2.6531,
+      "theoretical_loss": 3.444694329127941,
+      "tokens_seen": 1903099904
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004305858585858586,
+      "loss": 2.6429,
+      "theoretical_loss": 3.4446843419621,
+      "tokens_seen": 1903165440
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043056565656565655,
+      "loss": 2.6939,
+      "theoretical_loss": 3.4446743552364545,
+      "tokens_seen": 1903230976
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043054545454545456,
+      "loss": 2.5613,
+      "theoretical_loss": 3.4446643689509697,
+      "tokens_seen": 1903296512
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043052525252525257,
+      "loss": 2.8203,
+      "theoretical_loss": 3.444654383105613,
+      "tokens_seen": 1903362048
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004305050505050505,
+      "loss": 2.6728,
+      "theoretical_loss": 3.4446443977003476,
+      "tokens_seen": 1903427584
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004304848484848485,
+      "loss": 2.6016,
+      "theoretical_loss": 3.44463441273514,
+      "tokens_seen": 1903493120
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043046464646464644,
+      "loss": 2.7691,
+      "theoretical_loss": 3.444624428209956,
+      "tokens_seen": 1903558656
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043044444444444445,
+      "loss": 2.7279,
+      "theoretical_loss": 3.4446144441247606,
+      "tokens_seen": 1903624192
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043042424242424246,
+      "loss": 2.7133,
+      "theoretical_loss": 3.4446044604795194,
+      "tokens_seen": 1903689728
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004304040404040404,
+      "loss": 2.7337,
+      "theoretical_loss": 3.4445944772741974,
+      "tokens_seen": 1903755264
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004303838383838384,
+      "loss": 2.7528,
+      "theoretical_loss": 3.444584494508761,
+      "tokens_seen": 1903820800
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043036363636363633,
+      "loss": 2.6488,
+      "theoretical_loss": 3.444574512183175,
+      "tokens_seen": 1903886336
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043034343434343434,
+      "loss": 2.686,
+      "theoretical_loss": 3.444564530297405,
+      "tokens_seen": 1903951872
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043032323232323235,
+      "loss": 2.7676,
+      "theoretical_loss": 3.444554548851416,
+      "tokens_seen": 1904017408
+    },
+    {
+      "epoch": 0.15,
+      "objective/train/docs_used": 1048883,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5599634647369385,
+      "objective/train/theoretical_loss": 3.4445445678451745,
+      "objective/train/tokens_used": 262942176,
+      "theoretical_loss": 3.4445445678451745,
+      "tokens_seen": 1904082944
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004303030303030303,
+      "loss": 2.8532,
+      "theoretical_loss": 3.4445445678451745,
+      "tokens_seen": 1904082944
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043028282828282827,
+      "loss": 2.8837,
+      "theoretical_loss": 3.4445345872786453,
+      "tokens_seen": 1904148480
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004302626262626262,
+      "loss": 2.6033,
+      "theoretical_loss": 3.444524607151794,
+      "tokens_seen": 1904214016
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004302424242424243,
+      "loss": 2.7305,
+      "theoretical_loss": 3.444514627464587,
+      "tokens_seen": 1904279552
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043022222222222225,
+      "loss": 2.7995,
+      "theoretical_loss": 3.444504648216988,
+      "tokens_seen": 1904345088
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004302020202020202,
+      "loss": 2.8029,
+      "theoretical_loss": 3.4444946694089635,
+      "tokens_seen": 1904410624
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043018181818181816,
+      "loss": 2.6914,
+      "theoretical_loss": 3.4444846910404796,
+      "tokens_seen": 1904476160
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043016161616161617,
+      "loss": 2.6082,
+      "theoretical_loss": 3.444474713111501,
+      "tokens_seen": 1904541696
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004301414141414142,
+      "loss": 2.8255,
+      "theoretical_loss": 3.4444647356219935,
+      "tokens_seen": 1904607232
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043012121212121214,
+      "loss": 2.5401,
+      "theoretical_loss": 3.444454758571922,
+      "tokens_seen": 1904672768
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004301010101010101,
+      "loss": 2.7175,
+      "theoretical_loss": 3.4444447819612534,
+      "tokens_seen": 1904738304
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043008080808080805,
+      "loss": 2.5902,
+      "theoretical_loss": 3.444434805789952,
+      "tokens_seen": 1904803840
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004300606060606061,
+      "loss": 2.8099,
+      "theoretical_loss": 3.444424830057984,
+      "tokens_seen": 1904869376
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004300404040404041,
+      "loss": 2.6037,
+      "theoretical_loss": 3.444414854765314,
+      "tokens_seen": 1904934912
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043002020202020203,
+      "loss": 2.7313,
+      "theoretical_loss": 3.444404879911909,
+      "tokens_seen": 1905000448
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00043,
+      "loss": 2.8642,
+      "theoretical_loss": 3.4443949054977336,
+      "tokens_seen": 1905065984
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042997979797979795,
+      "loss": 2.777,
+      "theoretical_loss": 3.4443849315227535,
+      "tokens_seen": 1905131520
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000429959595959596,
+      "loss": 2.5088,
+      "theoretical_loss": 3.444374957986934,
+      "tokens_seen": 1905197056
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042993939393939397,
+      "loss": 2.6678,
+      "theoretical_loss": 3.444364984890241,
+      "tokens_seen": 1905262592
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004299191919191919,
+      "loss": 2.7726,
+      "theoretical_loss": 3.4443550122326405,
+      "tokens_seen": 1905328128
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004298989898989899,
+      "loss": 2.6621,
+      "theoretical_loss": 3.444345040014097,
+      "tokens_seen": 1905393664
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004298787878787879,
+      "loss": 2.8235,
+      "theoretical_loss": 3.444335068234577,
+      "tokens_seen": 1905459200
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004298585858585859,
+      "loss": 2.807,
+      "theoretical_loss": 3.4443250968940458,
+      "tokens_seen": 1905524736
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042983838383838386,
+      "loss": 2.785,
+      "theoretical_loss": 3.4443151259924685,
+      "tokens_seen": 1905590272
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004298181818181818,
+      "loss": 2.69,
+      "theoretical_loss": 3.444305155529811,
+      "tokens_seen": 1905655808
+    },
+    {
+      "epoch": 0.15,
+      "objective/train/docs_used": 1049481,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5614824295043945,
+      "objective/train/theoretical_loss": 3.4442951855060393,
+      "objective/train/tokens_used": 264580576,
+      "theoretical_loss": 3.4442951855060393,
+      "tokens_seen": 1905721344
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004297979797979798,
+      "loss": 2.5091,
+      "theoretical_loss": 3.4442951855060393,
+      "tokens_seen": 1905721344
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004297777777777778,
+      "loss": 2.7097,
+      "theoretical_loss": 3.4442852159211186,
+      "tokens_seen": 1905786880
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004297575757575758,
+      "loss": 2.5279,
+      "theoretical_loss": 3.444275246775015,
+      "tokens_seen": 1905852416
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042973737373737375,
+      "loss": 2.6529,
+      "theoretical_loss": 3.4442652780676926,
+      "tokens_seen": 1905917952
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004297171717171717,
+      "loss": 2.6415,
+      "theoretical_loss": 3.444255309799119,
+      "tokens_seen": 1905983488
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004296969696969697,
+      "loss": 2.7135,
+      "theoretical_loss": 3.444245341969258,
+      "tokens_seen": 1906049024
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004296767676767677,
+      "loss": 2.7133,
+      "theoretical_loss": 3.4442353745780765,
+      "tokens_seen": 1906114560
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004296565656565657,
+      "loss": 2.5431,
+      "theoretical_loss": 3.44422540762554,
+      "tokens_seen": 1906180096
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042963636363636365,
+      "loss": 2.6903,
+      "theoretical_loss": 3.444215441111613,
+      "tokens_seen": 1906245632
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004296161616161616,
+      "loss": 2.6747,
+      "theoretical_loss": 3.444205475036262,
+      "tokens_seen": 1906311168
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004295959595959596,
+      "loss": 2.6883,
+      "theoretical_loss": 3.444195509399453,
+      "tokens_seen": 1906376704
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042957575757575757,
+      "loss": 2.5868,
+      "theoretical_loss": 3.444185544201151,
+      "tokens_seen": 1906442240
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004295555555555556,
+      "loss": 2.8323,
+      "theoretical_loss": 3.444175579441322,
+      "tokens_seen": 1906507776
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042953535353535354,
+      "loss": 2.766,
+      "theoretical_loss": 3.444165615119931,
+      "tokens_seen": 1906573312
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042951515151515155,
+      "loss": 2.665,
+      "theoretical_loss": 3.4441556512369442,
+      "tokens_seen": 1906638848
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004294949494949495,
+      "loss": 2.7014,
+      "theoretical_loss": 3.4441456877923273,
+      "tokens_seen": 1906704384
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042947474747474746,
+      "loss": 2.6946,
+      "theoretical_loss": 3.4441357247860456,
+      "tokens_seen": 1906769920
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004294545454545455,
+      "loss": 2.4496,
+      "theoretical_loss": 3.4441257622180648,
+      "tokens_seen": 1906835456
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042943434343434343,
+      "loss": 2.6594,
+      "theoretical_loss": 3.4441158000883507,
+      "tokens_seen": 1906900992
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042941414141414144,
+      "loss": 2.3002,
+      "theoretical_loss": 3.444105838396869,
+      "tokens_seen": 1906966528
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004293939393939394,
+      "loss": 2.5308,
+      "theoretical_loss": 3.4440958771435852,
+      "tokens_seen": 1907032064
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042937373737373736,
+      "loss": 2.5182,
+      "theoretical_loss": 3.444085916328465,
+      "tokens_seen": 1907097600
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042935353535353537,
+      "loss": 2.5564,
+      "theoretical_loss": 3.444075955951474,
+      "tokens_seen": 1907163136
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004293333333333334,
+      "loss": 2.5707,
+      "theoretical_loss": 3.4440659960125783,
+      "tokens_seen": 1907228672
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042931313131313134,
+      "loss": 2.6122,
+      "theoretical_loss": 3.4440560365117427,
+      "tokens_seen": 1907294208
+    },
+    {
+      "epoch": 0.15,
+      "objective/train/docs_used": 1050597,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.497748374938965,
+      "objective/train/theoretical_loss": 3.444046077448934,
+      "objective/train/tokens_used": 266218976,
+      "theoretical_loss": 3.444046077448934,
+      "tokens_seen": 1907359744
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004292929292929293,
+      "loss": 2.6615,
+      "theoretical_loss": 3.444046077448934,
+      "tokens_seen": 1907359744
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042927272727272725,
+      "loss": 2.7089,
+      "theoretical_loss": 3.4440361188241173,
+      "tokens_seen": 1907425280
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042925252525252526,
+      "loss": 2.6821,
+      "theoretical_loss": 3.444026160637258,
+      "tokens_seen": 1907490816
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042923232323232327,
+      "loss": 2.6433,
+      "theoretical_loss": 3.4440162028883226,
+      "tokens_seen": 1907556352
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042921212121212123,
+      "loss": 2.7893,
+      "theoretical_loss": 3.444006245577276,
+      "tokens_seen": 1907621888
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004291919191919192,
+      "loss": 2.4906,
+      "theoretical_loss": 3.443996288704084,
+      "tokens_seen": 1907687424
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042917171717171714,
+      "loss": 2.745,
+      "theoretical_loss": 3.443986332268713,
+      "tokens_seen": 1907752960
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004291515151515152,
+      "loss": 2.4249,
+      "theoretical_loss": 3.443976376271128,
+      "tokens_seen": 1907818496
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042913131313131316,
+      "loss": 2.7227,
+      "theoretical_loss": 3.4439664207112948,
+      "tokens_seen": 1907884032
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004291111111111111,
+      "loss": 2.762,
+      "theoretical_loss": 3.4439564655891797,
+      "tokens_seen": 1907949568
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004290909090909091,
+      "loss": 2.7586,
+      "theoretical_loss": 3.4439465109047473,
+      "tokens_seen": 1908015104
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042907070707070703,
+      "loss": 2.616,
+      "theoretical_loss": 3.4439365566579645,
+      "tokens_seen": 1908080640
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004290505050505051,
+      "loss": 2.6937,
+      "theoretical_loss": 3.4439266028487965,
+      "tokens_seen": 1908146176
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042903030303030306,
+      "loss": 2.6258,
+      "theoretical_loss": 3.4439166494772087,
+      "tokens_seen": 1908211712
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000429010101010101,
+      "loss": 2.7516,
+      "theoretical_loss": 3.4439066965431677,
+      "tokens_seen": 1908277248
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042898989898989897,
+      "loss": 2.7227,
+      "theoretical_loss": 3.4438967440466386,
+      "tokens_seen": 1908342784
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042896969696969693,
+      "loss": 2.4991,
+      "theoretical_loss": 3.443886791987587,
+      "tokens_seen": 1908408320
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000428949494949495,
+      "loss": 2.5172,
+      "theoretical_loss": 3.443876840365979,
+      "tokens_seen": 1908473856
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042892929292929295,
+      "loss": 2.6511,
+      "theoretical_loss": 3.4438668891817805,
+      "tokens_seen": 1908539392
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004289090909090909,
+      "loss": 2.6616,
+      "theoretical_loss": 3.443856938434957,
+      "tokens_seen": 1908604928
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042888888888888886,
+      "loss": 2.7852,
+      "theoretical_loss": 3.4438469881254745,
+      "tokens_seen": 1908670464
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004288686868686869,
+      "loss": 2.6391,
+      "theoretical_loss": 3.443837038253298,
+      "tokens_seen": 1908736000
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004288484848484849,
+      "loss": 2.6146,
+      "theoretical_loss": 3.443827088818394,
+      "tokens_seen": 1908801536
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042882828282828284,
+      "loss": 2.6307,
+      "theoretical_loss": 3.4438171398207285,
+      "tokens_seen": 1908867072
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004288080808080808,
+      "loss": 2.4644,
+      "theoretical_loss": 3.443807191260267,
+      "tokens_seen": 1908932608
+    },
+    {
+      "epoch": 0.15,
+      "objective/train/docs_used": 1051477,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.080481767654419,
+      "objective/train/theoretical_loss": 3.4437972431369746,
+      "objective/train/tokens_used": 267857376,
+      "theoretical_loss": 3.4437972431369746,
+      "tokens_seen": 1908998144
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042878787878787876,
+      "loss": 2.712,
+      "theoretical_loss": 3.4437972431369746,
+      "tokens_seen": 1908998144
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042876767676767677,
+      "loss": 2.772,
+      "theoretical_loss": 3.4437872954508175,
+      "tokens_seen": 1909063680
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004287474747474748,
+      "loss": 2.712,
+      "theoretical_loss": 3.443777348201762,
+      "tokens_seen": 1909129216
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042872727272727274,
+      "loss": 2.7066,
+      "theoretical_loss": 3.443767401389773,
+      "tokens_seen": 1909194752
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004287070707070707,
+      "loss": 2.7274,
+      "theoretical_loss": 3.4437574550148176,
+      "tokens_seen": 1909260288
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004286868686868687,
+      "loss": 2.6661,
+      "theoretical_loss": 3.4437475090768603,
+      "tokens_seen": 1909325824
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042866666666666666,
+      "loss": 2.663,
+      "theoretical_loss": 3.4437375635758674,
+      "tokens_seen": 1909391360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042864646464646467,
+      "loss": 2.6362,
+      "theoretical_loss": 3.4437276185118053,
+      "tokens_seen": 1909456896
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042862626262626263,
+      "loss": 2.514,
+      "theoretical_loss": 3.4437176738846382,
+      "tokens_seen": 1909522432
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004286060606060606,
+      "loss": 2.5079,
+      "theoretical_loss": 3.4437077296943337,
+      "tokens_seen": 1909587968
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004285858585858586,
+      "loss": 2.801,
+      "theoretical_loss": 3.4436977859408566,
+      "tokens_seen": 1909653504
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042856565656565655,
+      "loss": 2.6743,
+      "theoretical_loss": 3.443687842624173,
+      "tokens_seen": 1909719040
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042854545454545456,
+      "loss": 2.6999,
+      "theoretical_loss": 3.443677899744249,
+      "tokens_seen": 1909784576
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004285252525252525,
+      "loss": 2.603,
+      "theoretical_loss": 3.4436679573010496,
+      "tokens_seen": 1909850112
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042850505050505053,
+      "loss": 2.7413,
+      "theoretical_loss": 3.4436580152945413,
+      "tokens_seen": 1909915648
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004284848484848485,
+      "loss": 2.3256,
+      "theoretical_loss": 3.44364807372469,
+      "tokens_seen": 1909981184
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042846464646464645,
+      "loss": 2.6853,
+      "theoretical_loss": 3.4436381325914613,
+      "tokens_seen": 1910046720
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042844444444444446,
+      "loss": 2.4406,
+      "theoretical_loss": 3.443628191894821,
+      "tokens_seen": 1910112256
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004284242424242424,
+      "loss": 2.7554,
+      "theoretical_loss": 3.443618251634735,
+      "tokens_seen": 1910177792
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004284040404040404,
+      "loss": 2.4338,
+      "theoretical_loss": 3.443608311811169,
+      "tokens_seen": 1910243328
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004283838383838384,
+      "loss": 2.5513,
+      "theoretical_loss": 3.443598372424089,
+      "tokens_seen": 1910308864
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004283636363636364,
+      "loss": 2.9312,
+      "theoretical_loss": 3.443588433473461,
+      "tokens_seen": 1910374400
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042834343434343435,
+      "loss": 2.6278,
+      "theoretical_loss": 3.443578494959251,
+      "tokens_seen": 1910439936
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042832323232323236,
+      "loss": 2.8923,
+      "theoretical_loss": 3.4435685568814245,
+      "tokens_seen": 1910505472
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004283030303030303,
+      "loss": 2.7032,
+      "theoretical_loss": 3.4435586192399477,
+      "tokens_seen": 1910571008
+    },
+    {
+      "epoch": 0.15,
+      "objective/train/docs_used": 1051849,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.255452871322632,
+      "objective/train/theoretical_loss": 3.443548682034786,
+      "objective/train/tokens_used": 269495776,
+      "theoretical_loss": 3.443548682034786,
+      "tokens_seen": 1910636544
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004282828282828283,
+      "loss": 2.7182,
+      "theoretical_loss": 3.443548682034786,
+      "tokens_seen": 1910636544
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004282626262626263,
+      "loss": 2.5879,
+      "theoretical_loss": 3.443538745265905,
+      "tokens_seen": 1910702080
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042824242424242424,
+      "loss": 2.5467,
+      "theoretical_loss": 3.443528808933272,
+      "tokens_seen": 1910767616
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042822222222222225,
+      "loss": 2.8346,
+      "theoretical_loss": 3.443518873036852,
+      "tokens_seen": 1910833152
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004282020202020202,
+      "loss": 2.3565,
+      "theoretical_loss": 3.44350893757661,
+      "tokens_seen": 1910898688
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042818181818181817,
+      "loss": 2.7003,
+      "theoretical_loss": 3.443499002552514,
+      "tokens_seen": 1910964224
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004281616161616162,
+      "loss": 2.3673,
+      "theoretical_loss": 3.443489067964528,
+      "tokens_seen": 1911029760
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004281414141414142,
+      "loss": 2.8019,
+      "theoretical_loss": 3.4434791338126187,
+      "tokens_seen": 1911095296
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042812121212121215,
+      "loss": 2.7316,
+      "theoretical_loss": 3.443469200096752,
+      "tokens_seen": 1911160832
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004281010101010101,
+      "loss": 2.8181,
+      "theoretical_loss": 3.4434592668168933,
+      "tokens_seen": 1911226368
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042808080808080806,
+      "loss": 2.7831,
+      "theoretical_loss": 3.4434493339730095,
+      "tokens_seen": 1911291904
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042806060606060607,
+      "loss": 2.8366,
+      "theoretical_loss": 3.4434394015650653,
+      "tokens_seen": 1911357440
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004280404040404041,
+      "loss": 2.5423,
+      "theoretical_loss": 3.4434294695930276,
+      "tokens_seen": 1911422976
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042802020202020204,
+      "loss": 2.5892,
+      "theoretical_loss": 3.443419538056862,
+      "tokens_seen": 1911488512
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000428,
+      "loss": 2.5842,
+      "theoretical_loss": 3.4434096069565348,
+      "tokens_seen": 1911554048
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042797979797979795,
+      "loss": 2.7545,
+      "theoretical_loss": 3.4433996762920107,
+      "tokens_seen": 1911619584
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000427959595959596,
+      "loss": 2.591,
+      "theoretical_loss": 3.443389746063257,
+      "tokens_seen": 1911685120
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000427939393939394,
+      "loss": 2.8032,
+      "theoretical_loss": 3.443379816270239,
+      "tokens_seen": 1911750656
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042791919191919193,
+      "loss": 2.6334,
+      "theoretical_loss": 3.4433698869129232,
+      "tokens_seen": 1911816192
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004278989898989899,
+      "loss": 2.7227,
+      "theoretical_loss": 3.443359957991275,
+      "tokens_seen": 1911881728
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042787878787878784,
+      "loss": 2.7052,
+      "theoretical_loss": 3.4433500295052597,
+      "tokens_seen": 1911947264
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004278585858585859,
+      "loss": 2.5298,
+      "theoretical_loss": 3.4433401014548446,
+      "tokens_seen": 1912012800
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042783838383838387,
+      "loss": 2.5337,
+      "theoretical_loss": 3.4433301738399953,
+      "tokens_seen": 1912078336
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004278181818181818,
+      "loss": 2.5673,
+      "theoretical_loss": 3.4433202466606767,
+      "tokens_seen": 1912143872
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004277979797979798,
+      "loss": 2.8075,
+      "theoretical_loss": 3.4433103199168564,
+      "tokens_seen": 1912209408
+    },
+    {
+      "epoch": 0.15,
+      "objective/train/docs_used": 1053170,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.4347217082977295,
+      "objective/train/theoretical_loss": 3.4433003936084994,
+      "objective/train/tokens_used": 271134176,
+      "theoretical_loss": 3.4433003936084994,
+      "tokens_seen": 1912274944
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004277777777777778,
+      "loss": 2.6555,
+      "theoretical_loss": 3.4433003936084994,
+      "tokens_seen": 1912274944
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004277575757575758,
+      "loss": 2.5505,
+      "theoretical_loss": 3.443290467735572,
+      "tokens_seen": 1912340480
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042773737373737376,
+      "loss": 2.7639,
+      "theoretical_loss": 3.44328054229804,
+      "tokens_seen": 1912406016
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004277171717171717,
+      "loss": 2.8434,
+      "theoretical_loss": 3.4432706172958687,
+      "tokens_seen": 1912471552
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004276969696969697,
+      "loss": 2.7693,
+      "theoretical_loss": 3.4432606927290257,
+      "tokens_seen": 1912537088
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004276767676767677,
+      "loss": 2.601,
+      "theoretical_loss": 3.4432507685974754,
+      "tokens_seen": 1912602624
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004276565656565657,
+      "loss": 2.3856,
+      "theoretical_loss": 3.443240844901185,
+      "tokens_seen": 1912668160
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042763636363636365,
+      "loss": 2.6203,
+      "theoretical_loss": 3.4432309216401196,
+      "tokens_seen": 1912733696
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004276161616161616,
+      "loss": 2.762,
+      "theoretical_loss": 3.4432209988142457,
+      "tokens_seen": 1912799232
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004275959595959596,
+      "loss": 2.5514,
+      "theoretical_loss": 3.4432110764235295,
+      "tokens_seen": 1912864768
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004275757575757576,
+      "loss": 2.6362,
+      "theoretical_loss": 3.4432011544679364,
+      "tokens_seen": 1912930304
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004275555555555556,
+      "loss": 2.6411,
+      "theoretical_loss": 3.443191232947433,
+      "tokens_seen": 1912995840
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042753535353535354,
+      "loss": 2.7044,
+      "theoretical_loss": 3.4431813118619847,
+      "tokens_seen": 1913061376
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004275151515151515,
+      "loss": 2.5536,
+      "theoretical_loss": 3.443171391211558,
+      "tokens_seen": 1913126912
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004274949494949495,
+      "loss": 2.7948,
+      "theoretical_loss": 3.4431614709961185,
+      "tokens_seen": 1913192448
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042747474747474747,
+      "loss": 2.7015,
+      "theoretical_loss": 3.4431515512156325,
+      "tokens_seen": 1913257984
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004274545454545455,
+      "loss": 2.7094,
+      "theoretical_loss": 3.4431416318700663,
+      "tokens_seen": 1913323520
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042743434343434344,
+      "loss": 2.7221,
+      "theoretical_loss": 3.443131712959386,
+      "tokens_seen": 1913389056
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042741414141414145,
+      "loss": 2.7083,
+      "theoretical_loss": 3.443121794483557,
+      "tokens_seen": 1913454592
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004273939393939394,
+      "loss": 2.5024,
+      "theoretical_loss": 3.4431118764425452,
+      "tokens_seen": 1913520128
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042737373737373736,
+      "loss": 2.5784,
+      "theoretical_loss": 3.4431019588363174,
+      "tokens_seen": 1913585664
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004273535353535354,
+      "loss": 2.864,
+      "theoretical_loss": 3.4430920416648396,
+      "tokens_seen": 1913651200
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042733333333333333,
+      "loss": 2.8075,
+      "theoretical_loss": 3.443082124928077,
+      "tokens_seen": 1913716736
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042731313131313134,
+      "loss": 2.7673,
+      "theoretical_loss": 3.443072208625997,
+      "tokens_seen": 1913782272
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004272929292929293,
+      "loss": 2.662,
+      "theoretical_loss": 3.4430622927585643,
+      "tokens_seen": 1913847808
+    },
+    {
+      "epoch": 0.15,
+      "objective/train/docs_used": 1053822,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.910274028778076,
+      "objective/train/theoretical_loss": 3.443052377325746,
+      "objective/train/tokens_used": 272772576,
+      "theoretical_loss": 3.443052377325746,
+      "tokens_seen": 1913913344
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042727272727272726,
+      "loss": 2.6898,
+      "theoretical_loss": 3.443052377325746,
+      "tokens_seen": 1913913344
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042725252525252527,
+      "loss": 2.5891,
+      "theoretical_loss": 3.4430424623275075,
+      "tokens_seen": 1913978880
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004272323232323233,
+      "loss": 2.6262,
+      "theoretical_loss": 3.4430325477638153,
+      "tokens_seen": 1914044416
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042721212121212123,
+      "loss": 2.5299,
+      "theoretical_loss": 3.443022633634635,
+      "tokens_seen": 1914109952
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004271919191919192,
+      "loss": 2.6919,
+      "theoretical_loss": 3.4430127199399334,
+      "tokens_seen": 1914175488
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042717171717171715,
+      "loss": 2.5389,
+      "theoretical_loss": 3.443002806679676,
+      "tokens_seen": 1914241024
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042715151515151516,
+      "loss": 2.4234,
+      "theoretical_loss": 3.442992893853829,
+      "tokens_seen": 1914306560
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042713131313131317,
+      "loss": 2.7233,
+      "theoretical_loss": 3.4429829814623583,
+      "tokens_seen": 1914372096
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004271111111111111,
+      "loss": 2.4909,
+      "theoretical_loss": 3.4429730695052303,
+      "tokens_seen": 1914437632
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004270909090909091,
+      "loss": 2.7478,
+      "theoretical_loss": 3.442963157982412,
+      "tokens_seen": 1914503168
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042707070707070704,
+      "loss": 2.5509,
+      "theoretical_loss": 3.4429532468938673,
+      "tokens_seen": 1914568704
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042705050505050505,
+      "loss": 2.525,
+      "theoretical_loss": 3.4429433362395643,
+      "tokens_seen": 1914634240
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042703030303030306,
+      "loss": 2.7176,
+      "theoretical_loss": 3.442933426019468,
+      "tokens_seen": 1914699776
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000427010101010101,
+      "loss": 2.7536,
+      "theoretical_loss": 3.442923516233545,
+      "tokens_seen": 1914765312
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000426989898989899,
+      "loss": 2.6505,
+      "theoretical_loss": 3.4429136068817616,
+      "tokens_seen": 1914830848
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042696969696969693,
+      "loss": 2.54,
+      "theoretical_loss": 3.4429036979640832,
+      "tokens_seen": 1914896384
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000426949494949495,
+      "loss": 2.7148,
+      "theoretical_loss": 3.4428937894804763,
+      "tokens_seen": 1914961920
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042692929292929296,
+      "loss": 2.4606,
+      "theoretical_loss": 3.442883881430908,
+      "tokens_seen": 1915027456
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004269090909090909,
+      "loss": 2.6427,
+      "theoretical_loss": 3.4428739738153427,
+      "tokens_seen": 1915092992
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042688888888888887,
+      "loss": 2.5241,
+      "theoretical_loss": 3.442864066633747,
+      "tokens_seen": 1915158528
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004268686868686868,
+      "loss": 2.7025,
+      "theoretical_loss": 3.442854159886088,
+      "tokens_seen": 1915224064
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004268484848484849,
+      "loss": 2.8231,
+      "theoretical_loss": 3.442844253572331,
+      "tokens_seen": 1915289600
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042682828282828285,
+      "loss": 2.6401,
+      "theoretical_loss": 3.4428343476924432,
+      "tokens_seen": 1915355136
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0004268080808080808,
+      "loss": 2.7003,
+      "theoretical_loss": 3.4428244422463887,
+      "tokens_seen": 1915420672
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042678787878787876,
+      "loss": 2.4902,
+      "theoretical_loss": 3.4428145372341357,
+      "tokens_seen": 1915486208
+    },
+    {
+      "epoch": 0.15,
+      "objective/train/docs_used": 1054940,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.356700897216797,
+      "objective/train/theoretical_loss": 3.4428046326556494,
+      "objective/train/tokens_used": 274410976,
+      "theoretical_loss": 3.4428046326556494,
+      "tokens_seen": 1915551744
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00042676767676767677,
+      "loss": 2.4488,
+      "theoretical_loss": 3.4428046326556494,
+      "tokens_seen": 1915551744
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004267474747474748,
+      "loss": 2.7953,
+      "theoretical_loss": 3.442794728510896,
+      "tokens_seen": 1915617280
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042672727272727274,
+      "loss": 2.6184,
+      "theoretical_loss": 3.4427848247998423,
+      "tokens_seen": 1915682816
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004267070707070707,
+      "loss": 2.4235,
+      "theoretical_loss": 3.4427749215224535,
+      "tokens_seen": 1915748352
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042668686868686865,
+      "loss": 2.7641,
+      "theoretical_loss": 3.4427650186786964,
+      "tokens_seen": 1915813888
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004266666666666667,
+      "loss": 2.7819,
+      "theoretical_loss": 3.442755116268537,
+      "tokens_seen": 1915879424
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004266464646464647,
+      "loss": 2.6544,
+      "theoretical_loss": 3.4427452142919415,
+      "tokens_seen": 1915944960
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042662626262626263,
+      "loss": 2.6319,
+      "theoretical_loss": 3.442735312748876,
+      "tokens_seen": 1916010496
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004266060606060606,
+      "loss": 2.7644,
+      "theoretical_loss": 3.442725411639307,
+      "tokens_seen": 1916076032
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004265858585858586,
+      "loss": 2.6702,
+      "theoretical_loss": 3.4427155109632004,
+      "tokens_seen": 1916141568
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004265656565656566,
+      "loss": 2.5994,
+      "theoretical_loss": 3.4427056107205223,
+      "tokens_seen": 1916207104
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042654545454545457,
+      "loss": 2.7542,
+      "theoretical_loss": 3.4426957109112393,
+      "tokens_seen": 1916272640
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004265252525252525,
+      "loss": 2.4553,
+      "theoretical_loss": 3.442685811535317,
+      "tokens_seen": 1916338176
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004265050505050505,
+      "loss": 2.6332,
+      "theoretical_loss": 3.4426759125927227,
+      "tokens_seen": 1916403712
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004264848484848485,
+      "loss": 2.6213,
+      "theoretical_loss": 3.4426660140834215,
+      "tokens_seen": 1916469248
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004264646464646465,
+      "loss": 2.6587,
+      "theoretical_loss": 3.44265611600738,
+      "tokens_seen": 1916534784
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042644444444444446,
+      "loss": 2.6661,
+      "theoretical_loss": 3.442646218364565,
+      "tokens_seen": 1916600320
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004264242424242424,
+      "loss": 2.4961,
+      "theoretical_loss": 3.442636321154942,
+      "tokens_seen": 1916665856
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042640404040404043,
+      "loss": 2.3362,
+      "theoretical_loss": 3.442626424378477,
+      "tokens_seen": 1916731392
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004263838383838384,
+      "loss": 2.6343,
+      "theoretical_loss": 3.4426165280351366,
+      "tokens_seen": 1916796928
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004263636363636364,
+      "loss": 2.6596,
+      "theoretical_loss": 3.4426066321248876,
+      "tokens_seen": 1916862464
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042634343434343435,
+      "loss": 2.8295,
+      "theoretical_loss": 3.4425967366476957,
+      "tokens_seen": 1916928000
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004263232323232323,
+      "loss": 2.6982,
+      "theoretical_loss": 3.442586841603527,
+      "tokens_seen": 1916993536
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004263030303030303,
+      "loss": 2.4981,
+      "theoretical_loss": 3.4425769469923475,
+      "tokens_seen": 1917059072
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004262828282828283,
+      "loss": 2.7179,
+      "theoretical_loss": 3.442567052814124,
+      "tokens_seen": 1917124608
+    },
+    {
+      "epoch": 0.16,
+      "objective/train/docs_used": 1055508,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.873598575592041,
+      "objective/train/theoretical_loss": 3.4425571590688233,
+      "objective/train/tokens_used": 276049376,
+      "theoretical_loss": 3.4425571590688233,
+      "tokens_seen": 1917190144
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004262626262626263,
+      "loss": 2.8046,
+      "theoretical_loss": 3.4425571590688233,
+      "tokens_seen": 1917190144
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042624242424242425,
+      "loss": 2.6275,
+      "theoretical_loss": 3.442547265756411,
+      "tokens_seen": 1917255680
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042622222222222226,
+      "loss": 2.5438,
+      "theoretical_loss": 3.4425373728768527,
+      "tokens_seen": 1917321216
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004262020202020202,
+      "loss": 2.5688,
+      "theoretical_loss": 3.4425274804301154,
+      "tokens_seen": 1917386752
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042618181818181817,
+      "loss": 2.5858,
+      "theoretical_loss": 3.442517588416165,
+      "tokens_seen": 1917452288
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004261616161616162,
+      "loss": 2.3179,
+      "theoretical_loss": 3.442507696834969,
+      "tokens_seen": 1917517824
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042614141414141414,
+      "loss": 2.6822,
+      "theoretical_loss": 3.442497805686492,
+      "tokens_seen": 1917583360
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042612121212121215,
+      "loss": 2.8473,
+      "theoretical_loss": 3.4424879149707017,
+      "tokens_seen": 1917648896
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004261010101010101,
+      "loss": 2.7661,
+      "theoretical_loss": 3.442478024687563,
+      "tokens_seen": 1917714432
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042608080808080806,
+      "loss": 2.6435,
+      "theoretical_loss": 3.442468134837043,
+      "tokens_seen": 1917779968
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004260606060606061,
+      "loss": 2.5611,
+      "theoretical_loss": 3.4424582454191084,
+      "tokens_seen": 1917845504
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004260404040404041,
+      "loss": 2.9118,
+      "theoretical_loss": 3.4424483564337245,
+      "tokens_seen": 1917911040
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042602020202020204,
+      "loss": 2.5781,
+      "theoretical_loss": 3.4424384678808586,
+      "tokens_seen": 1917976576
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000426,
+      "loss": 2.6566,
+      "theoretical_loss": 3.4424285797604757,
+      "tokens_seen": 1918042112
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042597979797979796,
+      "loss": 2.5477,
+      "theoretical_loss": 3.4424186920725437,
+      "tokens_seen": 1918107648
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042595959595959597,
+      "loss": 2.6297,
+      "theoretical_loss": 3.4424088048170276,
+      "tokens_seen": 1918173184
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000425939393939394,
+      "loss": 2.5819,
+      "theoretical_loss": 3.4423989179938945,
+      "tokens_seen": 1918238720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042591919191919194,
+      "loss": 2.8178,
+      "theoretical_loss": 3.4423890316031107,
+      "tokens_seen": 1918304256
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004258989898989899,
+      "loss": 2.5821,
+      "theoretical_loss": 3.442379145644642,
+      "tokens_seen": 1918369792
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042587878787878785,
+      "loss": 2.6713,
+      "theoretical_loss": 3.442369260118455,
+      "tokens_seen": 1918435328
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004258585858585859,
+      "loss": 2.6253,
+      "theoretical_loss": 3.442359375024516,
+      "tokens_seen": 1918500864
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042583838383838387,
+      "loss": 2.7187,
+      "theoretical_loss": 3.4423494903627914,
+      "tokens_seen": 1918566400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042581818181818183,
+      "loss": 2.6976,
+      "theoretical_loss": 3.442339606133247,
+      "tokens_seen": 1918631936
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004257979797979798,
+      "loss": 2.7475,
+      "theoretical_loss": 3.4423297223358507,
+      "tokens_seen": 1918697472
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042577777777777774,
+      "loss": 2.7081,
+      "theoretical_loss": 3.442319838970567,
+      "tokens_seen": 1918763008
+    },
+    {
+      "epoch": 0.16,
+      "objective/train/docs_used": 1056802,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.889559268951416,
+      "objective/train/theoretical_loss": 3.4423099560373633,
+      "objective/train/tokens_used": 277687776,
+      "theoretical_loss": 3.4423099560373633,
+      "tokens_seen": 1918828544
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004257575757575758,
+      "loss": 2.8355,
+      "theoretical_loss": 3.4423099560373633,
+      "tokens_seen": 1918828544
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042573737373737377,
+      "loss": 2.7127,
+      "theoretical_loss": 3.442300073536206,
+      "tokens_seen": 1918894080
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004257171717171717,
+      "loss": 2.5629,
+      "theoretical_loss": 3.4422901914670607,
+      "tokens_seen": 1918959616
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004256969696969697,
+      "loss": 2.5977,
+      "theoretical_loss": 3.4422803098298944,
+      "tokens_seen": 1919025152
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042567676767676764,
+      "loss": 2.4683,
+      "theoretical_loss": 3.4422704286246733,
+      "tokens_seen": 1919090688
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004256565656565657,
+      "loss": 2.502,
+      "theoretical_loss": 3.442260547851364,
+      "tokens_seen": 1919156224
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042563636363636366,
+      "loss": 2.6394,
+      "theoretical_loss": 3.4422506675099322,
+      "tokens_seen": 1919221760
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004256161616161616,
+      "loss": 2.5128,
+      "theoretical_loss": 3.442240787600345,
+      "tokens_seen": 1919287296
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042559595959595957,
+      "loss": 2.5185,
+      "theoretical_loss": 3.4422309081225686,
+      "tokens_seen": 1919352832
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004255757575757576,
+      "loss": 2.6894,
+      "theoretical_loss": 3.4422210290765687,
+      "tokens_seen": 1919418368
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004255555555555556,
+      "loss": 2.5627,
+      "theoretical_loss": 3.442211150462313,
+      "tokens_seen": 1919483904
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042553535353535355,
+      "loss": 2.8682,
+      "theoretical_loss": 3.4422012722797666,
+      "tokens_seen": 1919549440
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004255151515151515,
+      "loss": 2.5446,
+      "theoretical_loss": 3.442191394528897,
+      "tokens_seen": 1919614976
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042549494949494946,
+      "loss": 2.6554,
+      "theoretical_loss": 3.4421815172096695,
+      "tokens_seen": 1919680512
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004254747474747475,
+      "loss": 2.808,
+      "theoretical_loss": 3.4421716403220515,
+      "tokens_seen": 1919746048
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004254545454545455,
+      "loss": 2.5093,
+      "theoretical_loss": 3.4421617638660083,
+      "tokens_seen": 1919811584
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042543434343434344,
+      "loss": 2.6735,
+      "theoretical_loss": 3.442151887841508,
+      "tokens_seen": 1919877120
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004254141414141414,
+      "loss": 2.6459,
+      "theoretical_loss": 3.442142012248515,
+      "tokens_seen": 1919942656
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004253939393939394,
+      "loss": 2.4776,
+      "theoretical_loss": 3.442132137086997,
+      "tokens_seen": 1920008192
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042537373737373737,
+      "loss": 2.5164,
+      "theoretical_loss": 3.4421222623569205,
+      "tokens_seen": 1920073728
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004253535353535354,
+      "loss": 2.7869,
+      "theoretical_loss": 3.442112388058251,
+      "tokens_seen": 1920139264
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042533333333333334,
+      "loss": 2.5343,
+      "theoretical_loss": 3.442102514190956,
+      "tokens_seen": 1920204800
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004253131313131313,
+      "loss": 2.6066,
+      "theoretical_loss": 3.4420926407550008,
+      "tokens_seen": 1920270336
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004252929292929293,
+      "loss": 2.7356,
+      "theoretical_loss": 3.442082767750353,
+      "tokens_seen": 1920335872
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042527272727272726,
+      "loss": 2.6266,
+      "theoretical_loss": 3.4420728951769783,
+      "tokens_seen": 1920401408
+    },
+    {
+      "epoch": 0.16,
+      "objective/train/docs_used": 1058271,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.8692092895507812,
+      "objective/train/theoretical_loss": 3.4420630230348435,
+      "objective/train/tokens_used": 279326176,
+      "theoretical_loss": 3.4420630230348435,
+      "tokens_seen": 1920466944
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042525252525252527,
+      "loss": 2.789,
+      "theoretical_loss": 3.4420630230348435,
+      "tokens_seen": 1920466944
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042523232323232323,
+      "loss": 2.5816,
+      "theoretical_loss": 3.4420531513239148,
+      "tokens_seen": 1920532480
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042521212121212124,
+      "loss": 2.4545,
+      "theoretical_loss": 3.4420432800441585,
+      "tokens_seen": 1920598016
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004251919191919192,
+      "loss": 2.5497,
+      "theoretical_loss": 3.4420334091955413,
+      "tokens_seen": 1920663552
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042517171717171715,
+      "loss": 2.4662,
+      "theoretical_loss": 3.44202353877803,
+      "tokens_seen": 1920729088
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042515151515151516,
+      "loss": 2.581,
+      "theoretical_loss": 3.4420136687915903,
+      "tokens_seen": 1920794624
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004251313131313131,
+      "loss": 2.5163,
+      "theoretical_loss": 3.442003799236189,
+      "tokens_seen": 1920860160
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042511111111111113,
+      "loss": 2.6854,
+      "theoretical_loss": 3.441993930111793,
+      "tokens_seen": 1920925696
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004250909090909091,
+      "loss": 2.6518,
+      "theoretical_loss": 3.4419840614183688,
+      "tokens_seen": 1920991232
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042507070707070705,
+      "loss": 2.7454,
+      "theoretical_loss": 3.4419741931558816,
+      "tokens_seen": 1921056768
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042505050505050506,
+      "loss": 2.4956,
+      "theoretical_loss": 3.4419643253242995,
+      "tokens_seen": 1921122304
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042503030303030307,
+      "loss": 2.7089,
+      "theoretical_loss": 3.441954457923588,
+      "tokens_seen": 1921187840
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000425010101010101,
+      "loss": 2.9059,
+      "theoretical_loss": 3.441944590953714,
+      "tokens_seen": 1921253376
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000424989898989899,
+      "loss": 2.7142,
+      "theoretical_loss": 3.4419347244146437,
+      "tokens_seen": 1921318912
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000424969696969697,
+      "loss": 2.66,
+      "theoretical_loss": 3.4419248583063435,
+      "tokens_seen": 1921384448
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042494949494949495,
+      "loss": 2.5657,
+      "theoretical_loss": 3.4419149926287806,
+      "tokens_seen": 1921449984
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042492929292929296,
+      "loss": 2.5889,
+      "theoretical_loss": 3.441905127381921,
+      "tokens_seen": 1921515520
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004249090909090909,
+      "loss": 2.8425,
+      "theoretical_loss": 3.4418952625657306,
+      "tokens_seen": 1921581056
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004248888888888889,
+      "loss": 2.642,
+      "theoretical_loss": 3.4418853981801774,
+      "tokens_seen": 1921646592
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004248686868686869,
+      "loss": 2.6522,
+      "theoretical_loss": 3.4418755342252267,
+      "tokens_seen": 1921712128
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004248484848484849,
+      "loss": 2.6508,
+      "theoretical_loss": 3.441865670700846,
+      "tokens_seen": 1921777664
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042482828282828285,
+      "loss": 2.7685,
+      "theoretical_loss": 3.4418558076070003,
+      "tokens_seen": 1921843200
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004248080808080808,
+      "loss": 2.5812,
+      "theoretical_loss": 3.4418459449436574,
+      "tokens_seen": 1921908736
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042478787878787877,
+      "loss": 2.419,
+      "theoretical_loss": 3.4418360827107835,
+      "tokens_seen": 1921974272
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004247676767676768,
+      "loss": 2.5664,
+      "theoretical_loss": 3.4418262209083457,
+      "tokens_seen": 1922039808
+    },
+    {
+      "epoch": 0.16,
+      "objective/train/docs_used": 1058853,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.736966609954834,
+      "objective/train/theoretical_loss": 3.4418163595363094,
+      "objective/train/tokens_used": 280964576,
+      "theoretical_loss": 3.4418163595363094,
+      "tokens_seen": 1922105344
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004247474747474748,
+      "loss": 2.7834,
+      "theoretical_loss": 3.4418163595363094,
+      "tokens_seen": 1922105344
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042472727272727275,
+      "loss": 2.6645,
+      "theoretical_loss": 3.4418064985946417,
+      "tokens_seen": 1922170880
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004247070707070707,
+      "loss": 2.6125,
+      "theoretical_loss": 3.441796638083309,
+      "tokens_seen": 1922236416
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042468686868686866,
+      "loss": 2.6596,
+      "theoretical_loss": 3.441786778002278,
+      "tokens_seen": 1922301952
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004246666666666667,
+      "loss": 2.631,
+      "theoretical_loss": 3.4417769183515157,
+      "tokens_seen": 1922367488
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004246464646464647,
+      "loss": 2.5757,
+      "theoretical_loss": 3.4417670591309877,
+      "tokens_seen": 1922433024
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042462626262626264,
+      "loss": 2.5161,
+      "theoretical_loss": 3.441757200340662,
+      "tokens_seen": 1922498560
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004246060606060606,
+      "loss": 2.6606,
+      "theoretical_loss": 3.4417473419805034,
+      "tokens_seen": 1922564096
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042458585858585855,
+      "loss": 2.7274,
+      "theoretical_loss": 3.4417374840504795,
+      "tokens_seen": 1922629632
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004245656565656566,
+      "loss": 2.6871,
+      "theoretical_loss": 3.4417276265505565,
+      "tokens_seen": 1922695168
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004245454545454546,
+      "loss": 2.8311,
+      "theoretical_loss": 3.441717769480701,
+      "tokens_seen": 1922760704
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042452525252525253,
+      "loss": 2.5027,
+      "theoretical_loss": 3.4417079128408803,
+      "tokens_seen": 1922826240
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004245050505050505,
+      "loss": 2.7313,
+      "theoretical_loss": 3.4416980566310604,
+      "tokens_seen": 1922891776
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004244848484848485,
+      "loss": 2.6135,
+      "theoretical_loss": 3.4416882008512073,
+      "tokens_seen": 1922957312
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004244646464646465,
+      "loss": 2.6602,
+      "theoretical_loss": 3.4416783455012885,
+      "tokens_seen": 1923022848
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042444444444444447,
+      "loss": 2.6453,
+      "theoretical_loss": 3.4416684905812707,
+      "tokens_seen": 1923088384
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004244242424242424,
+      "loss": 2.57,
+      "theoretical_loss": 3.4416586360911197,
+      "tokens_seen": 1923153920
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004244040404040404,
+      "loss": 2.7243,
+      "theoretical_loss": 3.441648782030802,
+      "tokens_seen": 1923219456
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004243838383838384,
+      "loss": 2.8261,
+      "theoretical_loss": 3.4416389284002853,
+      "tokens_seen": 1923284992
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004243636363636364,
+      "loss": 2.4296,
+      "theoretical_loss": 3.4416290751995353,
+      "tokens_seen": 1923350528
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042434343434343436,
+      "loss": 2.5943,
+      "theoretical_loss": 3.4416192224285194,
+      "tokens_seen": 1923416064
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004243232323232323,
+      "loss": 2.6942,
+      "theoretical_loss": 3.4416093700872032,
+      "tokens_seen": 1923481600
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042430303030303033,
+      "loss": 2.6438,
+      "theoretical_loss": 3.441599518175554,
+      "tokens_seen": 1923547136
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004242828282828283,
+      "loss": 2.5493,
+      "theoretical_loss": 3.441589666693538,
+      "tokens_seen": 1923612672
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004242626262626263,
+      "loss": 2.6233,
+      "theoretical_loss": 3.4415798156411226,
+      "tokens_seen": 1923678208
+    },
+    {
+      "epoch": 0.16,
+      "objective/train/docs_used": 1059748,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.226703643798828,
+      "objective/train/theoretical_loss": 3.4415699650182736,
+      "objective/train/tokens_used": 282602976,
+      "theoretical_loss": 3.4415699650182736,
+      "tokens_seen": 1923743744
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042424242424242425,
+      "loss": 2.5937,
+      "theoretical_loss": 3.4415699650182736,
+      "tokens_seen": 1923743744
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004242222222222222,
+      "loss": 2.6444,
+      "theoretical_loss": 3.4415601148249584,
+      "tokens_seen": 1923809280
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004242020202020202,
+      "loss": 2.6106,
+      "theoretical_loss": 3.4415502650611427,
+      "tokens_seen": 1923874816
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004241818181818182,
+      "loss": 2.6631,
+      "theoretical_loss": 3.441540415726794,
+      "tokens_seen": 1923940352
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004241616161616162,
+      "loss": 2.6376,
+      "theoretical_loss": 3.4415305668218785,
+      "tokens_seen": 1924005888
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042414141414141415,
+      "loss": 2.7537,
+      "theoretical_loss": 3.4415207183463625,
+      "tokens_seen": 1924071424
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042412121212121216,
+      "loss": 2.858,
+      "theoretical_loss": 3.4415108703002137,
+      "tokens_seen": 1924136960
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004241010101010101,
+      "loss": 2.7537,
+      "theoretical_loss": 3.441501022683398,
+      "tokens_seen": 1924202496
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042408080808080807,
+      "loss": 2.7596,
+      "theoretical_loss": 3.441491175495882,
+      "tokens_seen": 1924268032
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004240606060606061,
+      "loss": 2.5014,
+      "theoretical_loss": 3.4414813287376327,
+      "tokens_seen": 1924333568
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042404040404040404,
+      "loss": 2.6273,
+      "theoretical_loss": 3.4414714824086166,
+      "tokens_seen": 1924399104
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042402020202020205,
+      "loss": 2.4565,
+      "theoretical_loss": 3.4414616365088007,
+      "tokens_seen": 1924464640
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000424,
+      "loss": 2.5366,
+      "theoretical_loss": 3.4414517910381512,
+      "tokens_seen": 1924530176
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042397979797979796,
+      "loss": 2.5527,
+      "theoretical_loss": 3.441441945996635,
+      "tokens_seen": 1924595712
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000423959595959596,
+      "loss": 2.6158,
+      "theoretical_loss": 3.441432101384218,
+      "tokens_seen": 1924661248
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042393939393939393,
+      "loss": 2.5519,
+      "theoretical_loss": 3.441422257200869,
+      "tokens_seen": 1924726784
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042391919191919194,
+      "loss": 2.6452,
+      "theoretical_loss": 3.4414124134465522,
+      "tokens_seen": 1924792320
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004238989898989899,
+      "loss": 2.4951,
+      "theoretical_loss": 3.4414025701212365,
+      "tokens_seen": 1924857856
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042387878787878786,
+      "loss": 2.6911,
+      "theoretical_loss": 3.4413927272248865,
+      "tokens_seen": 1924923392
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042385858585858587,
+      "loss": 2.6374,
+      "theoretical_loss": 3.441382884757471,
+      "tokens_seen": 1924988928
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004238383838383839,
+      "loss": 2.995,
+      "theoretical_loss": 3.4413730427189546,
+      "tokens_seen": 1925054464
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042381818181818183,
+      "loss": 2.9582,
+      "theoretical_loss": 3.441363201109305,
+      "tokens_seen": 1925120000
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004237979797979798,
+      "loss": 2.7288,
+      "theoretical_loss": 3.44135335992849,
+      "tokens_seen": 1925185536
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042377777777777775,
+      "loss": 2.5508,
+      "theoretical_loss": 3.4413435191764745,
+      "tokens_seen": 1925251072
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042375757575757576,
+      "loss": 2.5517,
+      "theoretical_loss": 3.4413336788532263,
+      "tokens_seen": 1925316608
+    },
+    {
+      "epoch": 0.16,
+      "objective/train/docs_used": 1060184,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.955732583999634,
+      "objective/train/theoretical_loss": 3.441323838958712,
+      "objective/train/tokens_used": 284241376,
+      "theoretical_loss": 3.441323838958712,
+      "tokens_seen": 1925382144
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042373737373737377,
+      "loss": 2.7418,
+      "theoretical_loss": 3.441323838958712,
+      "tokens_seen": 1925382144
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042371717171717173,
+      "loss": 2.5483,
+      "theoretical_loss": 3.4413139994928974,
+      "tokens_seen": 1925447680
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004236969696969697,
+      "loss": 2.646,
+      "theoretical_loss": 3.4413041604557506,
+      "tokens_seen": 1925513216
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042367676767676764,
+      "loss": 2.7451,
+      "theoretical_loss": 3.441294321847238,
+      "tokens_seen": 1925578752
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004236565656565657,
+      "loss": 2.6215,
+      "theoretical_loss": 3.4412844836673253,
+      "tokens_seen": 1925644288
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042363636363636366,
+      "loss": 2.6979,
+      "theoretical_loss": 3.44127464591598,
+      "tokens_seen": 1925709824
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004236161616161616,
+      "loss": 2.5959,
+      "theoretical_loss": 3.4412648085931696,
+      "tokens_seen": 1925775360
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004235959595959596,
+      "loss": 2.5962,
+      "theoretical_loss": 3.44125497169886,
+      "tokens_seen": 1925840896
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042357575757575753,
+      "loss": 2.7386,
+      "theoretical_loss": 3.4412451352330176,
+      "tokens_seen": 1925906432
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004235555555555556,
+      "loss": 2.6526,
+      "theoretical_loss": 3.4412352991956094,
+      "tokens_seen": 1925971968
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042353535353535356,
+      "loss": 2.552,
+      "theoretical_loss": 3.441225463586603,
+      "tokens_seen": 1926037504
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004235151515151515,
+      "loss": 2.6995,
+      "theoretical_loss": 3.4412156284059643,
+      "tokens_seen": 1926103040
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042349494949494947,
+      "loss": 2.7203,
+      "theoretical_loss": 3.44120579365366,
+      "tokens_seen": 1926168576
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004234747474747475,
+      "loss": 2.6077,
+      "theoretical_loss": 3.441195959329658,
+      "tokens_seen": 1926234112
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004234545454545455,
+      "loss": 2.7539,
+      "theoretical_loss": 3.4411861254339238,
+      "tokens_seen": 1926299648
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042343434343434345,
+      "loss": 2.5654,
+      "theoretical_loss": 3.4411762919664244,
+      "tokens_seen": 1926365184
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004234141414141414,
+      "loss": 2.6057,
+      "theoretical_loss": 3.441166458927127,
+      "tokens_seen": 1926430720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042339393939393936,
+      "loss": 2.713,
+      "theoretical_loss": 3.4411566263159985,
+      "tokens_seen": 1926496256
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004233737373737374,
+      "loss": 2.6969,
+      "theoretical_loss": 3.441146794133005,
+      "tokens_seen": 1926561792
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004233535353535354,
+      "loss": 2.7568,
+      "theoretical_loss": 3.4411369623781134,
+      "tokens_seen": 1926627328
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042333333333333334,
+      "loss": 2.5438,
+      "theoretical_loss": 3.4411271310512914,
+      "tokens_seen": 1926692864
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004233131313131313,
+      "loss": 2.6664,
+      "theoretical_loss": 3.441117300152505,
+      "tokens_seen": 1926758400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004232929292929293,
+      "loss": 2.6309,
+      "theoretical_loss": 3.441107469681721,
+      "tokens_seen": 1926823936
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004232727272727273,
+      "loss": 2.4625,
+      "theoretical_loss": 3.441097639638907,
+      "tokens_seen": 1926889472
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004232525252525253,
+      "loss": 2.6909,
+      "theoretical_loss": 3.4410878100240287,
+      "tokens_seen": 1926955008
+    },
+    {
+      "epoch": 0.16,
+      "objective/train/docs_used": 1061402,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.8198060989379883,
+      "objective/train/theoretical_loss": 3.4410779808370533,
+      "objective/train/tokens_used": 285879776,
+      "theoretical_loss": 3.4410779808370533,
+      "tokens_seen": 1927020544
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042323232323232323,
+      "loss": 2.7919,
+      "theoretical_loss": 3.4410779808370533,
+      "tokens_seen": 1927020544
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004232121212121212,
+      "loss": 2.6017,
+      "theoretical_loss": 3.441068152077948,
+      "tokens_seen": 1927086080
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004231919191919192,
+      "loss": 2.7291,
+      "theoretical_loss": 3.4410583237466796,
+      "tokens_seen": 1927151616
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004231717171717172,
+      "loss": 2.5611,
+      "theoretical_loss": 3.441048495843215,
+      "tokens_seen": 1927217152
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042315151515151517,
+      "loss": 2.5936,
+      "theoretical_loss": 3.4410386683675203,
+      "tokens_seen": 1927282688
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004231313131313131,
+      "loss": 2.8128,
+      "theoretical_loss": 3.4410288413195627,
+      "tokens_seen": 1927348224
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042311111111111114,
+      "loss": 2.6127,
+      "theoretical_loss": 3.441019014699309,
+      "tokens_seen": 1927413760
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004230909090909091,
+      "loss": 2.662,
+      "theoretical_loss": 3.4410091885067264,
+      "tokens_seen": 1927479296
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004230707070707071,
+      "loss": 2.8147,
+      "theoretical_loss": 3.440999362741781,
+      "tokens_seen": 1927544832
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042305050505050506,
+      "loss": 2.7461,
+      "theoretical_loss": 3.440989537404441,
+      "tokens_seen": 1927610368
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000423030303030303,
+      "loss": 2.6908,
+      "theoretical_loss": 3.4409797124946726,
+      "tokens_seen": 1927675904
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042301010101010103,
+      "loss": 2.7072,
+      "theoretical_loss": 3.4409698880124413,
+      "tokens_seen": 1927741440
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000422989898989899,
+      "loss": 2.3188,
+      "theoretical_loss": 3.440960063957716,
+      "tokens_seen": 1927806976
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000422969696969697,
+      "loss": 2.7596,
+      "theoretical_loss": 3.4409502403304626,
+      "tokens_seen": 1927872512
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042294949494949496,
+      "loss": 2.6218,
+      "theoretical_loss": 3.4409404171306477,
+      "tokens_seen": 1927938048
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042292929292929297,
+      "loss": 2.5852,
+      "theoretical_loss": 3.440930594358239,
+      "tokens_seen": 1928003584
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004229090909090909,
+      "loss": 2.6241,
+      "theoretical_loss": 3.440920772013203,
+      "tokens_seen": 1928069120
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004228888888888889,
+      "loss": 2.7613,
+      "theoretical_loss": 3.440910950095506,
+      "tokens_seen": 1928134656
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004228686868686869,
+      "loss": 2.7854,
+      "theoretical_loss": 3.440901128605116,
+      "tokens_seen": 1928200192
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042284848484848485,
+      "loss": 2.5871,
+      "theoretical_loss": 3.4408913075419987,
+      "tokens_seen": 1928265728
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042282828282828286,
+      "loss": 2.6012,
+      "theoretical_loss": 3.4408814869061217,
+      "tokens_seen": 1928331264
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004228080808080808,
+      "loss": 2.663,
+      "theoretical_loss": 3.440871666697452,
+      "tokens_seen": 1928396800
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042278787878787877,
+      "loss": 2.6573,
+      "theoretical_loss": 3.440861846915956,
+      "tokens_seen": 1928462336
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004227676767676768,
+      "loss": 2.5695,
+      "theoretical_loss": 3.4408520275616006,
+      "tokens_seen": 1928527872
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004227474747474748,
+      "loss": 2.8241,
+      "theoretical_loss": 3.4408422086343533,
+      "tokens_seen": 1928593408
+    },
+    {
+      "epoch": 0.16,
+      "objective/train/docs_used": 1062041,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.024728298187256,
+      "objective/train/theoretical_loss": 3.4408323901341813,
+      "objective/train/tokens_used": 287518176,
+      "theoretical_loss": 3.4408323901341813,
+      "tokens_seen": 1928658944
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042272727272727275,
+      "loss": 2.6987,
+      "theoretical_loss": 3.4408323901341813,
+      "tokens_seen": 1928658944
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004227070707070707,
+      "loss": 2.6473,
+      "theoretical_loss": 3.44082257206105,
+      "tokens_seen": 1928724480
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042268686868686867,
+      "loss": 2.7776,
+      "theoretical_loss": 3.440812754414927,
+      "tokens_seen": 1928790016
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004226666666666667,
+      "loss": 2.7868,
+      "theoretical_loss": 3.4408029371957802,
+      "tokens_seen": 1928855552
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004226464646464647,
+      "loss": 2.6503,
+      "theoretical_loss": 3.4407931204035753,
+      "tokens_seen": 1928921088
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042262626262626264,
+      "loss": 2.5474,
+      "theoretical_loss": 3.44078330403828,
+      "tokens_seen": 1928986624
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004226060606060606,
+      "loss": 2.6479,
+      "theoretical_loss": 3.440773488099861,
+      "tokens_seen": 1929052160
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042258585858585856,
+      "loss": 2.6409,
+      "theoretical_loss": 3.4407636725882846,
+      "tokens_seen": 1929117696
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004225656565656566,
+      "loss": 2.5776,
+      "theoretical_loss": 3.440753857503519,
+      "tokens_seen": 1929183232
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004225454545454546,
+      "loss": 2.6831,
+      "theoretical_loss": 3.44074404284553,
+      "tokens_seen": 1929248768
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042252525252525254,
+      "loss": 2.7039,
+      "theoretical_loss": 3.440734228614285,
+      "tokens_seen": 1929314304
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004225050505050505,
+      "loss": 2.5256,
+      "theoretical_loss": 3.4407244148097504,
+      "tokens_seen": 1929379840
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042248484848484845,
+      "loss": 2.8189,
+      "theoretical_loss": 3.4407146014318943,
+      "tokens_seen": 1929445376
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004224646464646465,
+      "loss": 2.7982,
+      "theoretical_loss": 3.4407047884806827,
+      "tokens_seen": 1929510912
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004224444444444445,
+      "loss": 2.5438,
+      "theoretical_loss": 3.4406949759560836,
+      "tokens_seen": 1929576448
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042242424242424243,
+      "loss": 2.5654,
+      "theoretical_loss": 3.4406851638580624,
+      "tokens_seen": 1929641984
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004224040404040404,
+      "loss": 2.5339,
+      "theoretical_loss": 3.440675352186587,
+      "tokens_seen": 1929707520
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042238383838383834,
+      "loss": 2.3803,
+      "theoretical_loss": 3.440665540941625,
+      "tokens_seen": 1929773056
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004223636363636364,
+      "loss": 2.5308,
+      "theoretical_loss": 3.4406557301231424,
+      "tokens_seen": 1929838592
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042234343434343437,
+      "loss": 2.5711,
+      "theoretical_loss": 3.4406459197311063,
+      "tokens_seen": 1929904128
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004223232323232323,
+      "loss": 2.6102,
+      "theoretical_loss": 3.440636109765484,
+      "tokens_seen": 1929969664
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004223030303030303,
+      "loss": 2.5921,
+      "theoretical_loss": 3.440626300226242,
+      "tokens_seen": 1930035200
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004222828282828283,
+      "loss": 2.5174,
+      "theoretical_loss": 3.440616491113348,
+      "tokens_seen": 1930100736
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004222626262626263,
+      "loss": 2.7388,
+      "theoretical_loss": 3.4406066824267683,
+      "tokens_seen": 1930166272
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042224242424242426,
+      "loss": 2.7482,
+      "theoretical_loss": 3.44059687416647,
+      "tokens_seen": 1930231808
+    },
+    {
+      "epoch": 0.16,
+      "objective/train/docs_used": 1063329,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.848512649536133,
+      "objective/train/theoretical_loss": 3.440587066332421,
+      "objective/train/tokens_used": 289156576,
+      "theoretical_loss": 3.440587066332421,
+      "tokens_seen": 1930297344
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004222222222222222,
+      "loss": 2.5179,
+      "theoretical_loss": 3.440587066332421,
+      "tokens_seen": 1930297344
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042220202020202017,
+      "loss": 2.451,
+      "theoretical_loss": 3.4405772589245873,
+      "tokens_seen": 1930362880
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004221818181818182,
+      "loss": 2.9025,
+      "theoretical_loss": 3.440567451942936,
+      "tokens_seen": 1930428416
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004221616161616162,
+      "loss": 2.5348,
+      "theoretical_loss": 3.440557645387435,
+      "tokens_seen": 1930493952
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042214141414141415,
+      "loss": 2.5781,
+      "theoretical_loss": 3.44054783925805,
+      "tokens_seen": 1930559488
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004221212121212121,
+      "loss": 2.7627,
+      "theoretical_loss": 3.4405380335547493,
+      "tokens_seen": 1930625024
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004221010101010101,
+      "loss": 2.5685,
+      "theoretical_loss": 3.4405282282774987,
+      "tokens_seen": 1930690560
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004220808080808081,
+      "loss": 2.5545,
+      "theoretical_loss": 3.4405184234262656,
+      "tokens_seen": 1930756096
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004220606060606061,
+      "loss": 2.519,
+      "theoretical_loss": 3.440508619001018,
+      "tokens_seen": 1930821632
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042204040404040404,
+      "loss": 2.6168,
+      "theoretical_loss": 3.440498815001722,
+      "tokens_seen": 1930887168
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000422020202020202,
+      "loss": 2.513,
+      "theoretical_loss": 3.440489011428345,
+      "tokens_seen": 1930952704
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000422,
+      "loss": 2.6129,
+      "theoretical_loss": 3.440479208280853,
+      "tokens_seen": 1931018240
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042197979797979797,
+      "loss": 2.8086,
+      "theoretical_loss": 3.440469405559215,
+      "tokens_seen": 1931083776
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000421959595959596,
+      "loss": 2.7246,
+      "theoretical_loss": 3.4404596032633963,
+      "tokens_seen": 1931149312
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042193939393939394,
+      "loss": 2.5076,
+      "theoretical_loss": 3.440449801393364,
+      "tokens_seen": 1931214848
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042191919191919195,
+      "loss": 2.6279,
+      "theoretical_loss": 3.440439999949087,
+      "tokens_seen": 1931280384
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004218989898989899,
+      "loss": 2.7267,
+      "theoretical_loss": 3.440430198930531,
+      "tokens_seen": 1931345920
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042187878787878786,
+      "loss": 2.6061,
+      "theoretical_loss": 3.440420398337663,
+      "tokens_seen": 1931411456
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042185858585858587,
+      "loss": 2.6032,
+      "theoretical_loss": 3.44041059817045,
+      "tokens_seen": 1931476992
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042183838383838383,
+      "loss": 2.4926,
+      "theoretical_loss": 3.4404007984288594,
+      "tokens_seen": 1931542528
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042181818181818184,
+      "loss": 2.5209,
+      "theoretical_loss": 3.440390999112858,
+      "tokens_seen": 1931608064
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004217979797979798,
+      "loss": 2.4727,
+      "theoretical_loss": 3.4403812002224137,
+      "tokens_seen": 1931673600
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042177777777777775,
+      "loss": 2.8118,
+      "theoretical_loss": 3.440371401757493,
+      "tokens_seen": 1931739136
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00042175757575757577,
+      "loss": 2.6462,
+      "theoretical_loss": 3.4403616037180624,
+      "tokens_seen": 1931804672
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0004217373737373738,
+      "loss": 2.7645,
+      "theoretical_loss": 3.44035180610409,
+      "tokens_seen": 1931870208
+    },
+    {
+      "epoch": 0.16,
+      "objective/train/docs_used": 1064052,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.8039486408233643,
+      "objective/train/theoretical_loss": 3.440342008915542,
+      "objective/train/tokens_used": 290794976,
+      "theoretical_loss": 3.440342008915542,
+      "tokens_seen": 1931935744
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042171717171717173,
+      "loss": 2.663,
+      "theoretical_loss": 3.440342008915542,
+      "tokens_seen": 1931935744
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004216969696969697,
+      "loss": 2.5882,
+      "theoretical_loss": 3.440332212152386,
+      "tokens_seen": 1932001280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042167676767676765,
+      "loss": 2.931,
+      "theoretical_loss": 3.440322415814589,
+      "tokens_seen": 1932066816
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042165656565656566,
+      "loss": 2.5881,
+      "theoretical_loss": 3.4403126199021186,
+      "tokens_seen": 1932132352
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042163636363636367,
+      "loss": 2.6762,
+      "theoretical_loss": 3.4403028244149416,
+      "tokens_seen": 1932197888
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004216161616161616,
+      "loss": 2.4638,
+      "theoretical_loss": 3.4402930293530245,
+      "tokens_seen": 1932263424
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004215959595959596,
+      "loss": 2.5095,
+      "theoretical_loss": 3.440283234716335,
+      "tokens_seen": 1932328960
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042157575757575754,
+      "loss": 2.6447,
+      "theoretical_loss": 3.44027344050484,
+      "tokens_seen": 1932394496
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004215555555555556,
+      "loss": 2.6329,
+      "theoretical_loss": 3.440263646718507,
+      "tokens_seen": 1932460032
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042153535353535356,
+      "loss": 2.497,
+      "theoretical_loss": 3.4402538533573024,
+      "tokens_seen": 1932525568
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004215151515151515,
+      "loss": 2.6166,
+      "theoretical_loss": 3.440244060421194,
+      "tokens_seen": 1932591104
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004214949494949495,
+      "loss": 2.5755,
+      "theoretical_loss": 3.440234267910149,
+      "tokens_seen": 1932656640
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004214747474747475,
+      "loss": 2.7671,
+      "theoretical_loss": 3.4402244758241336,
+      "tokens_seen": 1932722176
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004214545454545455,
+      "loss": 2.7073,
+      "theoretical_loss": 3.4402146841631165,
+      "tokens_seen": 1932787712
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042143434343434345,
+      "loss": 2.6975,
+      "theoretical_loss": 3.4402048929270634,
+      "tokens_seen": 1932853248
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004214141414141414,
+      "loss": 2.7931,
+      "theoretical_loss": 3.4401951021159416,
+      "tokens_seen": 1932918784
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042139393939393937,
+      "loss": 2.6141,
+      "theoretical_loss": 3.4401853117297194,
+      "tokens_seen": 1932984320
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042137373737373743,
+      "loss": 2.6676,
+      "theoretical_loss": 3.4401755217683627,
+      "tokens_seen": 1933049856
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004213535353535354,
+      "loss": 2.5838,
+      "theoretical_loss": 3.440165732231839,
+      "tokens_seen": 1933115392
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042133333333333335,
+      "loss": 2.7085,
+      "theoretical_loss": 3.440155943120116,
+      "tokens_seen": 1933180928
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004213131313131313,
+      "loss": 2.6783,
+      "theoretical_loss": 3.440146154433161,
+      "tokens_seen": 1933246464
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042129292929292926,
+      "loss": 2.6153,
+      "theoretical_loss": 3.4401363661709397,
+      "tokens_seen": 1933312000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004212727272727273,
+      "loss": 2.6217,
+      "theoretical_loss": 3.4401265783334205,
+      "tokens_seen": 1933377536
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004212525252525253,
+      "loss": 2.6967,
+      "theoretical_loss": 3.4401167909205705,
+      "tokens_seen": 1933443072
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042123232323232324,
+      "loss": 2.7279,
+      "theoretical_loss": 3.440107003932357,
+      "tokens_seen": 1933508608
+    },
+    {
+      "epoch": 0.17,
+      "objective/train/docs_used": 1065201,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.028951406478882,
+      "objective/train/theoretical_loss": 3.440097217368746,
+      "objective/train/tokens_used": 292433376,
+      "theoretical_loss": 3.440097217368746,
+      "tokens_seen": 1933574144
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004212121212121212,
+      "loss": 2.831,
+      "theoretical_loss": 3.440097217368746,
+      "tokens_seen": 1933574144
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004211919191919192,
+      "loss": 2.6147,
+      "theoretical_loss": 3.4400874312297063,
+      "tokens_seen": 1933639680
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004211717171717172,
+      "loss": 2.6807,
+      "theoretical_loss": 3.4400776455152036,
+      "tokens_seen": 1933705216
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004211515151515152,
+      "loss": 2.5114,
+      "theoretical_loss": 3.4400678602252066,
+      "tokens_seen": 1933770752
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042113131313131313,
+      "loss": 2.3262,
+      "theoretical_loss": 3.4400580753596817,
+      "tokens_seen": 1933836288
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004211111111111111,
+      "loss": 2.5457,
+      "theoretical_loss": 3.440048290918596,
+      "tokens_seen": 1933901824
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004210909090909091,
+      "loss": 2.5826,
+      "theoretical_loss": 3.4400385069019164,
+      "tokens_seen": 1933967360
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004210707070707071,
+      "loss": 2.4071,
+      "theoretical_loss": 3.4400287233096107,
+      "tokens_seen": 1934032896
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042105050505050507,
+      "loss": 2.6196,
+      "theoretical_loss": 3.4400189401416466,
+      "tokens_seen": 1934098432
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000421030303030303,
+      "loss": 2.5962,
+      "theoretical_loss": 3.4400091573979896,
+      "tokens_seen": 1934163968
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042101010101010104,
+      "loss": 2.669,
+      "theoretical_loss": 3.439999375078609,
+      "tokens_seen": 1934229504
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000420989898989899,
+      "loss": 2.4876,
+      "theoretical_loss": 3.439989593183471,
+      "tokens_seen": 1934295040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000420969696969697,
+      "loss": 2.4545,
+      "theoretical_loss": 3.4399798117125426,
+      "tokens_seen": 1934360576
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042094949494949496,
+      "loss": 2.6715,
+      "theoretical_loss": 3.439970030665791,
+      "tokens_seen": 1934426112
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004209292929292929,
+      "loss": 2.3414,
+      "theoretical_loss": 3.439960250043184,
+      "tokens_seen": 1934491648
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042090909090909093,
+      "loss": 2.5181,
+      "theoretical_loss": 3.4399504698446886,
+      "tokens_seen": 1934557184
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004208888888888889,
+      "loss": 2.4947,
+      "theoretical_loss": 3.439940690070272,
+      "tokens_seen": 1934622720
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004208686868686869,
+      "loss": 2.554,
+      "theoretical_loss": 3.4399309107199016,
+      "tokens_seen": 1934688256
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042084848484848485,
+      "loss": 2.7057,
+      "theoretical_loss": 3.439921131793544,
+      "tokens_seen": 1934753792
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042082828282828286,
+      "loss": 2.5841,
+      "theoretical_loss": 3.439911353291168,
+      "tokens_seen": 1934819328
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004208080808080808,
+      "loss": 2.5806,
+      "theoretical_loss": 3.4399015752127386,
+      "tokens_seen": 1934884864
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004207878787878788,
+      "loss": 2.7166,
+      "theoretical_loss": 3.439891797558225,
+      "tokens_seen": 1934950400
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004207676767676768,
+      "loss": 2.9193,
+      "theoretical_loss": 3.4398820203275937,
+      "tokens_seen": 1935015936
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042074747474747475,
+      "loss": 2.5482,
+      "theoretical_loss": 3.439872243520812,
+      "tokens_seen": 1935081472
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042072727272727276,
+      "loss": 2.641,
+      "theoretical_loss": 3.4398624671378473,
+      "tokens_seen": 1935147008
+    },
+    {
+      "epoch": 0.17,
+      "objective/train/docs_used": 1066155,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.908733367919922,
+      "objective/train/theoretical_loss": 3.4398526911786664,
+      "objective/train/tokens_used": 294071776,
+      "theoretical_loss": 3.4398526911786664,
+      "tokens_seen": 1935212544
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004207070707070707,
+      "loss": 2.8247,
+      "theoretical_loss": 3.4398526911786664,
+      "tokens_seen": 1935212544
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042068686868686867,
+      "loss": 2.7459,
+      "theoretical_loss": 3.4398429156432373,
+      "tokens_seen": 1935278080
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004206666666666667,
+      "loss": 2.5958,
+      "theoretical_loss": 3.439833140531527,
+      "tokens_seen": 1935343616
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042064646464646464,
+      "loss": 2.6365,
+      "theoretical_loss": 3.4398233658435027,
+      "tokens_seen": 1935409152
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042062626262626265,
+      "loss": 2.5277,
+      "theoretical_loss": 3.4398135915791315,
+      "tokens_seen": 1935474688
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004206060606060606,
+      "loss": 2.5039,
+      "theoretical_loss": 3.4398038177383805,
+      "tokens_seen": 1935540224
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042058585858585856,
+      "loss": 2.6533,
+      "theoretical_loss": 3.439794044321218,
+      "tokens_seen": 1935605760
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004205656565656566,
+      "loss": 2.7235,
+      "theoretical_loss": 3.4397842713276106,
+      "tokens_seen": 1935671296
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004205454545454546,
+      "loss": 2.5335,
+      "theoretical_loss": 3.439774498757526,
+      "tokens_seen": 1935736832
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042052525252525254,
+      "loss": 2.4197,
+      "theoretical_loss": 3.439764726610931,
+      "tokens_seen": 1935802368
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004205050505050505,
+      "loss": 2.6907,
+      "theoretical_loss": 3.4397549548877926,
+      "tokens_seen": 1935867904
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042048484848484846,
+      "loss": 2.481,
+      "theoretical_loss": 3.4397451835880792,
+      "tokens_seen": 1935933440
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042046464646464647,
+      "loss": 2.4428,
+      "theoretical_loss": 3.439735412711757,
+      "tokens_seen": 1935998976
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004204444444444445,
+      "loss": 2.7433,
+      "theoretical_loss": 3.4397256422587947,
+      "tokens_seen": 1936064512
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042042424242424244,
+      "loss": 2.5442,
+      "theoretical_loss": 3.4397158722291583,
+      "tokens_seen": 1936130048
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004204040404040404,
+      "loss": 2.4129,
+      "theoretical_loss": 3.4397061026228157,
+      "tokens_seen": 1936195584
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042038383838383835,
+      "loss": 2.5154,
+      "theoretical_loss": 3.439696333439734,
+      "tokens_seen": 1936261120
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004203636363636364,
+      "loss": 2.4758,
+      "theoretical_loss": 3.439686564679881,
+      "tokens_seen": 1936326656
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042034343434343437,
+      "loss": 2.7579,
+      "theoretical_loss": 3.4396767963432238,
+      "tokens_seen": 1936392192
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042032323232323233,
+      "loss": 2.5475,
+      "theoretical_loss": 3.4396670284297297,
+      "tokens_seen": 1936457728
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004203030303030303,
+      "loss": 2.5622,
+      "theoretical_loss": 3.4396572609393656,
+      "tokens_seen": 1936523264
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042028282828282824,
+      "loss": 2.6701,
+      "theoretical_loss": 3.4396474938720996,
+      "tokens_seen": 1936588800
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004202626262626263,
+      "loss": 2.8353,
+      "theoretical_loss": 3.439637727227899,
+      "tokens_seen": 1936654336
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042024242424242426,
+      "loss": 2.7227,
+      "theoretical_loss": 3.439627961006731,
+      "tokens_seen": 1936719872
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004202222222222222,
+      "loss": 2.5612,
+      "theoretical_loss": 3.4396181952085625,
+      "tokens_seen": 1936785408
+    },
+    {
+      "epoch": 0.17,
+      "objective/train/docs_used": 1066685,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.676300048828125,
+      "objective/train/theoretical_loss": 3.4396084298333616,
+      "objective/train/tokens_used": 295710176,
+      "theoretical_loss": 3.4396084298333616,
+      "tokens_seen": 1936850944
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004202020202020202,
+      "loss": 2.6837,
+      "theoretical_loss": 3.4396084298333616,
+      "tokens_seen": 1936850944
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004201818181818182,
+      "loss": 2.4526,
+      "theoretical_loss": 3.4395986648810943,
+      "tokens_seen": 1936916480
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004201616161616162,
+      "loss": 2.698,
+      "theoretical_loss": 3.43958890035173,
+      "tokens_seen": 1936982016
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042014141414141416,
+      "loss": 2.6901,
+      "theoretical_loss": 3.439579136245235,
+      "tokens_seen": 1937047552
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004201212121212121,
+      "loss": 2.6405,
+      "theoretical_loss": 3.439569372561577,
+      "tokens_seen": 1937113088
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042010101010101007,
+      "loss": 2.7245,
+      "theoretical_loss": 3.439559609300722,
+      "tokens_seen": 1937178624
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004200808080808081,
+      "loss": 2.7824,
+      "theoretical_loss": 3.439549846462639,
+      "tokens_seen": 1937244160
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004200606060606061,
+      "loss": 2.3995,
+      "theoretical_loss": 3.4395400840472954,
+      "tokens_seen": 1937309696
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042004040404040405,
+      "loss": 2.558,
+      "theoretical_loss": 3.439530322054658,
+      "tokens_seen": 1937375232
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000420020202020202,
+      "loss": 2.6701,
+      "theoretical_loss": 3.439520560484694,
+      "tokens_seen": 1937440768
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00042,
+      "loss": 2.4783,
+      "theoretical_loss": 3.4395107993373712,
+      "tokens_seen": 1937506304
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000419979797979798,
+      "loss": 2.6693,
+      "theoretical_loss": 3.439501038612657,
+      "tokens_seen": 1937571840
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000419959595959596,
+      "loss": 2.5367,
+      "theoretical_loss": 3.4394912783105185,
+      "tokens_seen": 1937637376
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041993939393939394,
+      "loss": 2.6768,
+      "theoretical_loss": 3.4394815184309238,
+      "tokens_seen": 1937702912
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004199191919191919,
+      "loss": 2.794,
+      "theoretical_loss": 3.4394717589738395,
+      "tokens_seen": 1937768448
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004198989898989899,
+      "loss": 2.5629,
+      "theoretical_loss": 3.4394619999392333,
+      "tokens_seen": 1937833984
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041987878787878787,
+      "loss": 2.8272,
+      "theoretical_loss": 3.4394522413270727,
+      "tokens_seen": 1937899520
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004198585858585859,
+      "loss": 2.8337,
+      "theoretical_loss": 3.4394424831373254,
+      "tokens_seen": 1937965056
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041983838383838383,
+      "loss": 2.5516,
+      "theoretical_loss": 3.439432725369958,
+      "tokens_seen": 1938030592
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041981818181818185,
+      "loss": 2.689,
+      "theoretical_loss": 3.439422968024939,
+      "tokens_seen": 1938096128
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004197979797979798,
+      "loss": 2.6582,
+      "theoretical_loss": 3.4394132111022353,
+      "tokens_seen": 1938161664
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004197777777777778,
+      "loss": 2.449,
+      "theoretical_loss": 3.4394034546018144,
+      "tokens_seen": 1938227200
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004197777777777778,
+      "loss": 2.9064,
+      "theoretical_loss": 3.4393936985236433,
+      "tokens_seen": 1938292736
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041975757575757577,
+      "loss": 2.8677,
+      "theoretical_loss": 3.439383942867691,
+      "tokens_seen": 1938358272
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041973737373737373,
+      "loss": 2.6937,
+      "theoretical_loss": 3.4393741876339226,
+      "tokens_seen": 1938423808
+    },
+    {
+      "epoch": 0.17,
+      "objective/train/docs_used": 1067842,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.121251344680786,
+      "objective/train/theoretical_loss": 3.439364432822307,
+      "objective/train/tokens_used": 297348576,
+      "theoretical_loss": 3.439364432822307,
+      "tokens_seen": 1938489344
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041971717171717174,
+      "loss": 2.7767,
+      "theoretical_loss": 3.439364432822307,
+      "tokens_seen": 1938489344
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004196969696969697,
+      "loss": 2.7315,
+      "theoretical_loss": 3.439354678432812,
+      "tokens_seen": 1938554880
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004196767676767677,
+      "loss": 2.7844,
+      "theoretical_loss": 3.439344924465404,
+      "tokens_seen": 1938620416
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041965656565656566,
+      "loss": 2.6795,
+      "theoretical_loss": 3.4393351709200513,
+      "tokens_seen": 1938685952
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004196363636363637,
+      "loss": 2.6743,
+      "theoretical_loss": 3.439325417796721,
+      "tokens_seen": 1938751488
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041961616161616163,
+      "loss": 2.605,
+      "theoretical_loss": 3.4393156650953802,
+      "tokens_seen": 1938817024
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004195959595959596,
+      "loss": 2.6641,
+      "theoretical_loss": 3.439305912815997,
+      "tokens_seen": 1938882560
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004195757575757576,
+      "loss": 2.8231,
+      "theoretical_loss": 3.4392961609585386,
+      "tokens_seen": 1938948096
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041955555555555556,
+      "loss": 2.7356,
+      "theoretical_loss": 3.439286409522973,
+      "tokens_seen": 1939013632
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041953535353535357,
+      "loss": 2.6914,
+      "theoretical_loss": 3.439276658509267,
+      "tokens_seen": 1939079168
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004195151515151515,
+      "loss": 2.6307,
+      "theoretical_loss": 3.4392669079173883,
+      "tokens_seen": 1939144704
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004194949494949495,
+      "loss": 2.339,
+      "theoretical_loss": 3.4392571577473046,
+      "tokens_seen": 1939210240
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004194747474747475,
+      "loss": 2.614,
+      "theoretical_loss": 3.439247407998983,
+      "tokens_seen": 1939275776
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004194545454545455,
+      "loss": 2.5329,
+      "theoretical_loss": 3.4392376586723916,
+      "tokens_seen": 1939341312
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041943434343434346,
+      "loss": 2.789,
+      "theoretical_loss": 3.439227909767497,
+      "tokens_seen": 1939406848
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004194141414141414,
+      "loss": 2.7102,
+      "theoretical_loss": 3.439218161284268,
+      "tokens_seen": 1939472384
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004193939393939394,
+      "loss": 2.7777,
+      "theoretical_loss": 3.439208413222671,
+      "tokens_seen": 1939537920
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004193737373737374,
+      "loss": 2.593,
+      "theoretical_loss": 3.4391986655826736,
+      "tokens_seen": 1939603456
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004193535353535354,
+      "loss": 2.662,
+      "theoretical_loss": 3.439188918364244,
+      "tokens_seen": 1939668992
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041933333333333335,
+      "loss": 2.8318,
+      "theoretical_loss": 3.439179171567349,
+      "tokens_seen": 1939734528
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004193131313131313,
+      "loss": 2.8346,
+      "theoretical_loss": 3.439169425191957,
+      "tokens_seen": 1939800064
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041929292929292927,
+      "loss": 2.4422,
+      "theoretical_loss": 3.439159679238035,
+      "tokens_seen": 1939865600
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041927272727272733,
+      "loss": 2.5032,
+      "theoretical_loss": 3.4391499337055498,
+      "tokens_seen": 1939931136
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004192525252525253,
+      "loss": 2.6996,
+      "theoretical_loss": 3.4391401885944703,
+      "tokens_seen": 1939996672
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041923232323232325,
+      "loss": 2.7708,
+      "theoretical_loss": 3.439130443904763,
+      "tokens_seen": 1940062208
+    },
+    {
+      "epoch": 0.17,
+      "objective/train/docs_used": 1068281,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.4002110958099365,
+      "objective/train/theoretical_loss": 3.4391206996363968,
+      "objective/train/tokens_used": 298986976,
+      "theoretical_loss": 3.4391206996363968,
+      "tokens_seen": 1940127744
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004192121212121212,
+      "loss": 2.7074,
+      "theoretical_loss": 3.4391206996363968,
+      "tokens_seen": 1940127744
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041919191919191916,
+      "loss": 2.7498,
+      "theoretical_loss": 3.439110955789338,
+      "tokens_seen": 1940193280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004191717171717172,
+      "loss": 2.583,
+      "theoretical_loss": 3.4391012123635534,
+      "tokens_seen": 1940258816
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004191515151515152,
+      "loss": 2.6829,
+      "theoretical_loss": 3.4390914693590124,
+      "tokens_seen": 1940324352
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041913131313131314,
+      "loss": 2.6073,
+      "theoretical_loss": 3.439081726775682,
+      "tokens_seen": 1940389888
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004191111111111111,
+      "loss": 2.8032,
+      "theoretical_loss": 3.4390719846135296,
+      "tokens_seen": 1940455424
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041909090909090905,
+      "loss": 2.6993,
+      "theoretical_loss": 3.4390622428725224,
+      "tokens_seen": 1940520960
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004190707070707071,
+      "loss": 2.5315,
+      "theoretical_loss": 3.439052501552628,
+      "tokens_seen": 1940586496
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004190505050505051,
+      "loss": 2.6999,
+      "theoretical_loss": 3.4390427606538148,
+      "tokens_seen": 1940652032
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041903030303030303,
+      "loss": 2.6889,
+      "theoretical_loss": 3.4390330201760495,
+      "tokens_seen": 1940717568
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000419010101010101,
+      "loss": 2.6122,
+      "theoretical_loss": 3.439023280119301,
+      "tokens_seen": 1940783104
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000418989898989899,
+      "loss": 2.7696,
+      "theoretical_loss": 3.439013540483535,
+      "tokens_seen": 1940848640
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000418969696969697,
+      "loss": 2.644,
+      "theoretical_loss": 3.4390038012687203,
+      "tokens_seen": 1940914176
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041894949494949497,
+      "loss": 2.4996,
+      "theoretical_loss": 3.4389940624748236,
+      "tokens_seen": 1940979712
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004189292929292929,
+      "loss": 2.6993,
+      "theoretical_loss": 3.438984324101814,
+      "tokens_seen": 1941045248
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004189090909090909,
+      "loss": 2.4784,
+      "theoretical_loss": 3.4389745861496577,
+      "tokens_seen": 1941110784
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004188888888888889,
+      "loss": 2.6716,
+      "theoretical_loss": 3.438964848618323,
+      "tokens_seen": 1941176320
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004188686868686869,
+      "loss": 2.7038,
+      "theoretical_loss": 3.438955111507777,
+      "tokens_seen": 1941241856
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041884848484848486,
+      "loss": 2.4977,
+      "theoretical_loss": 3.438945374817988,
+      "tokens_seen": 1941307392
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004188282828282828,
+      "loss": 2.8805,
+      "theoretical_loss": 3.4389356385489234,
+      "tokens_seen": 1941372928
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041880808080808083,
+      "loss": 2.6442,
+      "theoretical_loss": 3.43892590270055,
+      "tokens_seen": 1941438464
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004187878787878788,
+      "loss": 2.3895,
+      "theoretical_loss": 3.4389161672728368,
+      "tokens_seen": 1941504000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004187676767676768,
+      "loss": 2.9731,
+      "theoretical_loss": 3.4389064322657505,
+      "tokens_seen": 1941569536
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041874747474747475,
+      "loss": 2.5621,
+      "theoretical_loss": 3.438896697679259,
+      "tokens_seen": 1941635072
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004187272727272727,
+      "loss": 2.6273,
+      "theoretical_loss": 3.4388869635133297,
+      "tokens_seen": 1941700608
+    },
+    {
+      "epoch": 0.17,
+      "objective/train/docs_used": 1069559,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.2460758686065674,
+      "objective/train/theoretical_loss": 3.4388772297679306,
+      "objective/train/tokens_used": 300625376,
+      "theoretical_loss": 3.4388772297679306,
+      "tokens_seen": 1941766144
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004187070707070707,
+      "loss": 2.6064,
+      "theoretical_loss": 3.4388772297679306,
+      "tokens_seen": 1941766144
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004186868686868687,
+      "loss": 2.589,
+      "theoretical_loss": 3.438867496443029,
+      "tokens_seen": 1941831680
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004186666666666667,
+      "loss": 2.6783,
+      "theoretical_loss": 3.438857763538593,
+      "tokens_seen": 1941897216
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041864646464646464,
+      "loss": 2.7044,
+      "theoretical_loss": 3.4388480310545892,
+      "tokens_seen": 1941962752
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041862626262626266,
+      "loss": 2.364,
+      "theoretical_loss": 3.4388382989909863,
+      "tokens_seen": 1942028288
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004186060606060606,
+      "loss": 2.6038,
+      "theoretical_loss": 3.438828567347752,
+      "tokens_seen": 1942093824
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041858585858585857,
+      "loss": 2.5249,
+      "theoretical_loss": 3.4388188361248533,
+      "tokens_seen": 1942159360
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004185656565656566,
+      "loss": 2.7292,
+      "theoretical_loss": 3.4388091053222585,
+      "tokens_seen": 1942224896
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041854545454545454,
+      "loss": 2.9845,
+      "theoretical_loss": 3.4387993749399346,
+      "tokens_seen": 1942290432
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041852525252525255,
+      "loss": 2.6296,
+      "theoretical_loss": 3.4387896449778497,
+      "tokens_seen": 1942355968
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004185050505050505,
+      "loss": 2.507,
+      "theoretical_loss": 3.4387799154359717,
+      "tokens_seen": 1942421504
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041848484848484846,
+      "loss": 2.5466,
+      "theoretical_loss": 3.4387701863142675,
+      "tokens_seen": 1942487040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004184646464646465,
+      "loss": 2.6209,
+      "theoretical_loss": 3.4387604576127053,
+      "tokens_seen": 1942552576
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004184444444444445,
+      "loss": 2.6257,
+      "theoretical_loss": 3.438750729331253,
+      "tokens_seen": 1942618112
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041842424242424244,
+      "loss": 2.5931,
+      "theoretical_loss": 3.438741001469878,
+      "tokens_seen": 1942683648
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004184040404040404,
+      "loss": 2.6329,
+      "theoretical_loss": 3.438731274028547,
+      "tokens_seen": 1942749184
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041838383838383835,
+      "loss": 2.6169,
+      "theoretical_loss": 3.43872154700723,
+      "tokens_seen": 1942814720
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041836363636363637,
+      "loss": 2.7174,
+      "theoretical_loss": 3.438711820405893,
+      "tokens_seen": 1942880256
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004183434343434344,
+      "loss": 2.7735,
+      "theoretical_loss": 3.438702094224504,
+      "tokens_seen": 1942945792
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041832323232323233,
+      "loss": 2.5786,
+      "theoretical_loss": 3.4386923684630304,
+      "tokens_seen": 1943011328
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004183030303030303,
+      "loss": 2.4048,
+      "theoretical_loss": 3.438682643121441,
+      "tokens_seen": 1943076864
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041828282828282825,
+      "loss": 2.4826,
+      "theoretical_loss": 3.4386729181997024,
+      "tokens_seen": 1943142400
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004182626262626263,
+      "loss": 2.6934,
+      "theoretical_loss": 3.4386631936977827,
+      "tokens_seen": 1943207936
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041824242424242427,
+      "loss": 2.691,
+      "theoretical_loss": 3.43865346961565,
+      "tokens_seen": 1943273472
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004182222222222222,
+      "loss": 2.6558,
+      "theoretical_loss": 3.4386437459532715,
+      "tokens_seen": 1943339008
+    },
+    {
+      "epoch": 0.17,
+      "objective/train/docs_used": 1070184,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5214946269989014,
+      "objective/train/theoretical_loss": 3.4386340227106147,
+      "objective/train/tokens_used": 302263776,
+      "theoretical_loss": 3.4386340227106147,
+      "tokens_seen": 1943404544
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004182020202020202,
+      "loss": 2.5596,
+      "theoretical_loss": 3.4386340227106147,
+      "tokens_seen": 1943404544
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041818181818181814,
+      "loss": 2.5195,
+      "theoretical_loss": 3.438624299887648,
+      "tokens_seen": 1943470080
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004181616161616162,
+      "loss": 2.7051,
+      "theoretical_loss": 3.438614577484339,
+      "tokens_seen": 1943535616
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041814141414141416,
+      "loss": 2.7595,
+      "theoretical_loss": 3.438604855500655,
+      "tokens_seen": 1943601152
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004181212121212121,
+      "loss": 2.9422,
+      "theoretical_loss": 3.4385951339365644,
+      "tokens_seen": 1943666688
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004181010101010101,
+      "loss": 2.6183,
+      "theoretical_loss": 3.438585412792034,
+      "tokens_seen": 1943732224
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041808080808080814,
+      "loss": 2.8163,
+      "theoretical_loss": 3.4385756920670327,
+      "tokens_seen": 1943797760
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004180606060606061,
+      "loss": 2.7602,
+      "theoretical_loss": 3.438565971761528,
+      "tokens_seen": 1943863296
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041804040404040406,
+      "loss": 2.8253,
+      "theoretical_loss": 3.4385562518754864,
+      "tokens_seen": 1943928832
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000418020202020202,
+      "loss": 2.3846,
+      "theoretical_loss": 3.4385465324088766,
+      "tokens_seen": 1943994368
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041799999999999997,
+      "loss": 2.5301,
+      "theoretical_loss": 3.438536813361667,
+      "tokens_seen": 1944059904
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041797979797979803,
+      "loss": 2.6371,
+      "theoretical_loss": 3.4385270947338245,
+      "tokens_seen": 1944125440
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000417959595959596,
+      "loss": 2.5773,
+      "theoretical_loss": 3.4385173765253167,
+      "tokens_seen": 1944190976
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041793939393939395,
+      "loss": 2.4892,
+      "theoretical_loss": 3.4385076587361123,
+      "tokens_seen": 1944256512
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004179191919191919,
+      "loss": 2.7892,
+      "theoretical_loss": 3.4384979413661787,
+      "tokens_seen": 1944322048
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004178989898989899,
+      "loss": 2.5332,
+      "theoretical_loss": 3.438488224415483,
+      "tokens_seen": 1944387584
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004178787878787879,
+      "loss": 2.636,
+      "theoretical_loss": 3.4384785078839935,
+      "tokens_seen": 1944453120
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004178585858585859,
+      "loss": 2.6614,
+      "theoretical_loss": 3.438468791771678,
+      "tokens_seen": 1944518656
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041783838383838384,
+      "loss": 2.7253,
+      "theoretical_loss": 3.4384590760785043,
+      "tokens_seen": 1944584192
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004178181818181818,
+      "loss": 2.7251,
+      "theoretical_loss": 3.4384493608044404,
+      "tokens_seen": 1944649728
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004177979797979798,
+      "loss": 2.6503,
+      "theoretical_loss": 3.4384396459494537,
+      "tokens_seen": 1944715264
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004177777777777778,
+      "loss": 2.4993,
+      "theoretical_loss": 3.4384299315135123,
+      "tokens_seen": 1944780800
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004177575757575758,
+      "loss": 2.7246,
+      "theoretical_loss": 3.438420217496584,
+      "tokens_seen": 1944846336
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041773737373737373,
+      "loss": 2.3471,
+      "theoretical_loss": 3.438410503898636,
+      "tokens_seen": 1944911872
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041771717171717174,
+      "loss": 2.6327,
+      "theoretical_loss": 3.438400790719637,
+      "tokens_seen": 1944977408
+    },
+    {
+      "epoch": 0.17,
+      "objective/train/docs_used": 1071164,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.9253249168395996,
+      "objective/train/theoretical_loss": 3.4383910779595537,
+      "objective/train/tokens_used": 303902176,
+      "theoretical_loss": 3.4383910779595537,
+      "tokens_seen": 1945042944
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004176969696969697,
+      "loss": 2.591,
+      "theoretical_loss": 3.4383910779595537,
+      "tokens_seen": 1945042944
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004176767676767677,
+      "loss": 2.5417,
+      "theoretical_loss": 3.4383813656183557,
+      "tokens_seen": 1945108480
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041765656565656567,
+      "loss": 2.5524,
+      "theoretical_loss": 3.438371653696009,
+      "tokens_seen": 1945174016
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004176363636363636,
+      "loss": 2.686,
+      "theoretical_loss": 3.4383619421924823,
+      "tokens_seen": 1945239552
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041761616161616164,
+      "loss": 2.652,
+      "theoretical_loss": 3.4383522311077432,
+      "tokens_seen": 1945305088
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004175959595959596,
+      "loss": 2.5249,
+      "theoretical_loss": 3.43834252044176,
+      "tokens_seen": 1945370624
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004175757575757576,
+      "loss": 2.7325,
+      "theoretical_loss": 3.4383328101945,
+      "tokens_seen": 1945436160
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041755555555555556,
+      "loss": 2.6777,
+      "theoretical_loss": 3.438323100365931,
+      "tokens_seen": 1945501696
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041753535353535357,
+      "loss": 2.8154,
+      "theoretical_loss": 3.4383133909560213,
+      "tokens_seen": 1945567232
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041751515151515153,
+      "loss": 2.6971,
+      "theoretical_loss": 3.4383036819647383,
+      "tokens_seen": 1945632768
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004174949494949495,
+      "loss": 2.6609,
+      "theoretical_loss": 3.4382939733920503,
+      "tokens_seen": 1945698304
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004174747474747475,
+      "loss": 2.6654,
+      "theoretical_loss": 3.4382842652379244,
+      "tokens_seen": 1945763840
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041745454545454545,
+      "loss": 2.6757,
+      "theoretical_loss": 3.4382745575023295,
+      "tokens_seen": 1945829376
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041743434343434347,
+      "loss": 2.8648,
+      "theoretical_loss": 3.4382648501852326,
+      "tokens_seen": 1945894912
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004174141414141414,
+      "loss": 2.7462,
+      "theoretical_loss": 3.4382551432866024,
+      "tokens_seen": 1945960448
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004173939393939394,
+      "loss": 2.5422,
+      "theoretical_loss": 3.438245436806406,
+      "tokens_seen": 1946025984
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004173737373737374,
+      "loss": 2.5824,
+      "theoretical_loss": 3.438235730744611,
+      "tokens_seen": 1946091520
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041735353535353535,
+      "loss": 2.5466,
+      "theoretical_loss": 3.4382260251011862,
+      "tokens_seen": 1946157056
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041733333333333336,
+      "loss": 2.5127,
+      "theoretical_loss": 3.4382163198760995,
+      "tokens_seen": 1946222592
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004173131313131313,
+      "loss": 2.6517,
+      "theoretical_loss": 3.4382066150693182,
+      "tokens_seen": 1946288128
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041729292929292927,
+      "loss": 2.6034,
+      "theoretical_loss": 3.43819691068081,
+      "tokens_seen": 1946353664
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004172727272727273,
+      "loss": 2.6726,
+      "theoretical_loss": 3.438187206710543,
+      "tokens_seen": 1946419200
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004172525252525253,
+      "loss": 2.6317,
+      "theoretical_loss": 3.4381775031584856,
+      "tokens_seen": 1946484736
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041723232323232325,
+      "loss": 2.7176,
+      "theoretical_loss": 3.4381678000246056,
+      "tokens_seen": 1946550272
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004172121212121212,
+      "loss": 2.5587,
+      "theoretical_loss": 3.43815809730887,
+      "tokens_seen": 1946615808
+    },
+    {
+      "epoch": 0.17,
+      "objective/train/docs_used": 1071829,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5584614276885986,
+      "objective/train/theoretical_loss": 3.4381483950112477,
+      "objective/train/tokens_used": 305540576,
+      "theoretical_loss": 3.4381483950112477,
+      "tokens_seen": 1946681344
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041719191919191916,
+      "loss": 2.5635,
+      "theoretical_loss": 3.4381483950112477,
+      "tokens_seen": 1946681344
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004171717171717172,
+      "loss": 2.6877,
+      "theoretical_loss": 3.438138693131706,
+      "tokens_seen": 1946746880
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004171515151515152,
+      "loss": 2.7448,
+      "theoretical_loss": 3.4381289916702134,
+      "tokens_seen": 1946812416
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041713131313131314,
+      "loss": 2.5441,
+      "theoretical_loss": 3.4381192906267373,
+      "tokens_seen": 1946877952
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004171111111111111,
+      "loss": 2.8343,
+      "theoretical_loss": 3.438109590001246,
+      "tokens_seen": 1946943488
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041709090909090906,
+      "loss": 2.7814,
+      "theoretical_loss": 3.4380998897937065,
+      "tokens_seen": 1947009024
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004170707070707071,
+      "loss": 2.5301,
+      "theoretical_loss": 3.438090190004088,
+      "tokens_seen": 1947074560
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004170505050505051,
+      "loss": 2.6603,
+      "theoretical_loss": 3.438080490632358,
+      "tokens_seen": 1947140096
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041703030303030304,
+      "loss": 2.4052,
+      "theoretical_loss": 3.438070791678484,
+      "tokens_seen": 1947205632
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000417010101010101,
+      "loss": 2.5384,
+      "theoretical_loss": 3.4380610931424345,
+      "tokens_seen": 1947271168
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041698989898989895,
+      "loss": 2.5251,
+      "theoretical_loss": 3.4380513950241767,
+      "tokens_seen": 1947336704
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000416969696969697,
+      "loss": 2.754,
+      "theoretical_loss": 3.438041697323679,
+      "tokens_seen": 1947402240
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041694949494949497,
+      "loss": 2.5165,
+      "theoretical_loss": 3.43803200004091,
+      "tokens_seen": 1947467776
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041692929292929293,
+      "loss": 2.6118,
+      "theoretical_loss": 3.4380223031758366,
+      "tokens_seen": 1947533312
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004169090909090909,
+      "loss": 2.6706,
+      "theoretical_loss": 3.438012606728427,
+      "tokens_seen": 1947598848
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004168888888888889,
+      "loss": 2.5127,
+      "theoretical_loss": 3.43800291069865,
+      "tokens_seen": 1947664384
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004168686868686869,
+      "loss": 2.6097,
+      "theoretical_loss": 3.4379932150864723,
+      "tokens_seen": 1947729920
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041684848484848486,
+      "loss": 2.7995,
+      "theoretical_loss": 3.4379835198918625,
+      "tokens_seen": 1947795456
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004168282828282828,
+      "loss": 2.5767,
+      "theoretical_loss": 3.4379738251147884,
+      "tokens_seen": 1947860992
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004168080808080808,
+      "loss": 2.7866,
+      "theoretical_loss": 3.4379641307552182,
+      "tokens_seen": 1947926528
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004167878787878788,
+      "loss": 2.3767,
+      "theoretical_loss": 3.43795443681312,
+      "tokens_seen": 1947992064
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004167676767676768,
+      "loss": 2.6098,
+      "theoretical_loss": 3.4379447432884613,
+      "tokens_seen": 1948057600
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00041674747474747476,
+      "loss": 2.5764,
+      "theoretical_loss": 3.4379350501812107,
+      "tokens_seen": 1948123136
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004167272727272727,
+      "loss": 2.7819,
+      "theoretical_loss": 3.437925357491335,
+      "tokens_seen": 1948188672
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004167070707070707,
+      "loss": 2.6237,
+      "theoretical_loss": 3.437915665218804,
+      "tokens_seen": 1948254208
+    },
+    {
+      "epoch": 0.17,
+      "objective/train/docs_used": 1073161,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.831909656524658,
+      "objective/train/theoretical_loss": 3.437905973363584,
+      "objective/train/tokens_used": 307178976,
+      "theoretical_loss": 3.437905973363584,
+      "tokens_seen": 1948319744
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0004166868686868687,
+      "loss": 2.6736,
+      "theoretical_loss": 3.437905973363584,
+      "tokens_seen": 1948319744
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004166666666666667,
+      "loss": 2.6268,
+      "theoretical_loss": 3.437896281925644,
+      "tokens_seen": 1948385280
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041664646464646465,
+      "loss": 2.8293,
+      "theoretical_loss": 3.4378865909049514,
+      "tokens_seen": 1948450816
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004166262626262626,
+      "loss": 2.5125,
+      "theoretical_loss": 3.4378769003014744,
+      "tokens_seen": 1948516352
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004166060606060606,
+      "loss": 2.7312,
+      "theoretical_loss": 3.437867210115181,
+      "tokens_seen": 1948581888
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004165858585858586,
+      "loss": 2.5568,
+      "theoretical_loss": 3.43785752034604,
+      "tokens_seen": 1948647424
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004165656565656566,
+      "loss": 2.9045,
+      "theoretical_loss": 3.437847830994018,
+      "tokens_seen": 1948712960
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041654545454545454,
+      "loss": 2.6776,
+      "theoretical_loss": 3.4378381420590842,
+      "tokens_seen": 1948778496
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041652525252525255,
+      "loss": 2.5809,
+      "theoretical_loss": 3.437828453541206,
+      "tokens_seen": 1948844032
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004165050505050505,
+      "loss": 2.5214,
+      "theoretical_loss": 3.4378187654403516,
+      "tokens_seen": 1948909568
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041648484848484847,
+      "loss": 2.5708,
+      "theoretical_loss": 3.437809077756489,
+      "tokens_seen": 1948975104
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004164646464646465,
+      "loss": 2.7088,
+      "theoretical_loss": 3.437799390489586,
+      "tokens_seen": 1949040640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041644444444444444,
+      "loss": 2.5165,
+      "theoretical_loss": 3.437789703639611,
+      "tokens_seen": 1949106176
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041642424242424245,
+      "loss": 2.6073,
+      "theoretical_loss": 3.4377800172065323,
+      "tokens_seen": 1949171712
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004164040404040404,
+      "loss": 2.5349,
+      "theoretical_loss": 3.4377703311903174,
+      "tokens_seen": 1949237248
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041638383838383836,
+      "loss": 2.5828,
+      "theoretical_loss": 3.437760645590934,
+      "tokens_seen": 1949302784
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041636363636363637,
+      "loss": 2.6568,
+      "theoretical_loss": 3.437750960408351,
+      "tokens_seen": 1949368320
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004163434343434344,
+      "loss": 2.7027,
+      "theoretical_loss": 3.4377412756425363,
+      "tokens_seen": 1949433856
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041632323232323234,
+      "loss": 2.5977,
+      "theoretical_loss": 3.437731591293457,
+      "tokens_seen": 1949499392
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004163030303030303,
+      "loss": 2.5681,
+      "theoretical_loss": 3.4377219073610825,
+      "tokens_seen": 1949564928
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004162828282828283,
+      "loss": 2.4887,
+      "theoretical_loss": 3.4377122238453803,
+      "tokens_seen": 1949630464
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041626262626262626,
+      "loss": 2.7247,
+      "theoretical_loss": 3.437702540746318,
+      "tokens_seen": 1949696000
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004162424242424243,
+      "loss": 2.6263,
+      "theoretical_loss": 3.4376928580638646,
+      "tokens_seen": 1949761536
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041622222222222223,
+      "loss": 2.7817,
+      "theoretical_loss": 3.4376831757979875,
+      "tokens_seen": 1949827072
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004162020202020202,
+      "loss": 2.6438,
+      "theoretical_loss": 3.437673493948655,
+      "tokens_seen": 1949892608
+    },
+    {
+      "epoch": 0.18,
+      "objective/train/docs_used": 1073760,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.301966428756714,
+      "objective/train/theoretical_loss": 3.437663812515835,
+      "objective/train/tokens_used": 308817376,
+      "theoretical_loss": 3.437663812515835,
+      "tokens_seen": 1949958144
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004161818181818182,
+      "loss": 2.5007,
+      "theoretical_loss": 3.437663812515835,
+      "tokens_seen": 1949958144
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004161616161616162,
+      "loss": 2.8909,
+      "theoretical_loss": 3.437654131499496,
+      "tokens_seen": 1950023680
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041614141414141417,
+      "loss": 2.6493,
+      "theoretical_loss": 3.4376444508996054,
+      "tokens_seen": 1950089216
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004161212121212121,
+      "loss": 2.5106,
+      "theoretical_loss": 3.437634770716132,
+      "tokens_seen": 1950154752
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004161010101010101,
+      "loss": 2.4883,
+      "theoretical_loss": 3.4376250909490436,
+      "tokens_seen": 1950220288
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004160808080808081,
+      "loss": 2.6714,
+      "theoretical_loss": 3.4376154115983084,
+      "tokens_seen": 1950285824
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004160606060606061,
+      "loss": 2.4813,
+      "theoretical_loss": 3.4376057326638945,
+      "tokens_seen": 1950351360
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041604040404040406,
+      "loss": 2.7367,
+      "theoretical_loss": 3.4375960541457697,
+      "tokens_seen": 1950416896
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000416020202020202,
+      "loss": 2.6954,
+      "theoretical_loss": 3.437586376043902,
+      "tokens_seen": 1950482432
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000416,
+      "loss": 2.605,
+      "theoretical_loss": 3.43757669835826,
+      "tokens_seen": 1950547968
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041597979797979804,
+      "loss": 2.552,
+      "theoretical_loss": 3.437567021088812,
+      "tokens_seen": 1950613504
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000415959595959596,
+      "loss": 2.6863,
+      "theoretical_loss": 3.4375573442355254,
+      "tokens_seen": 1950679040
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041593939393939395,
+      "loss": 2.6037,
+      "theoretical_loss": 3.4375476677983685,
+      "tokens_seen": 1950744576
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004159191919191919,
+      "loss": 2.7154,
+      "theoretical_loss": 3.43753799177731,
+      "tokens_seen": 1950810112
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041589898989898987,
+      "loss": 2.7183,
+      "theoretical_loss": 3.437528316172317,
+      "tokens_seen": 1950875648
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041587878787878793,
+      "loss": 2.7948,
+      "theoretical_loss": 3.437518640983359,
+      "tokens_seen": 1950941184
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004158585858585859,
+      "loss": 2.5989,
+      "theoretical_loss": 3.4375089662104035,
+      "tokens_seen": 1951006720
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041583838383838385,
+      "loss": 2.6351,
+      "theoretical_loss": 3.4374992918534177,
+      "tokens_seen": 1951072256
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004158181818181818,
+      "loss": 2.4972,
+      "theoretical_loss": 3.4374896179123713,
+      "tokens_seen": 1951137792
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041579797979797976,
+      "loss": 2.5836,
+      "theoretical_loss": 3.4374799443872317,
+      "tokens_seen": 1951203328
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004157777777777778,
+      "loss": 2.4592,
+      "theoretical_loss": 3.4374702712779666,
+      "tokens_seen": 1951268864
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004157575757575758,
+      "loss": 2.6278,
+      "theoretical_loss": 3.4374605985845452,
+      "tokens_seen": 1951334400
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041573737373737374,
+      "loss": 2.7606,
+      "theoretical_loss": 3.4374509263069353,
+      "tokens_seen": 1951399936
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004157171717171717,
+      "loss": 2.5732,
+      "theoretical_loss": 3.437441254445104,
+      "tokens_seen": 1951465472
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004156969696969697,
+      "loss": 2.6677,
+      "theoretical_loss": 3.4374315829990207,
+      "tokens_seen": 1951531008
+    },
+    {
+      "epoch": 0.18,
+      "objective/train/docs_used": 1074960,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.260453701019287,
+      "objective/train/theoretical_loss": 3.437421911968653,
+      "objective/train/tokens_used": 310455776,
+      "theoretical_loss": 3.437421911968653,
+      "tokens_seen": 1951596544
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004156767676767677,
+      "loss": 2.4721,
+      "theoretical_loss": 3.437421911968653,
+      "tokens_seen": 1951596544
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004156565656565657,
+      "loss": 2.5506,
+      "theoretical_loss": 3.43741224135397,
+      "tokens_seen": 1951662080
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041563636363636363,
+      "loss": 2.6751,
+      "theoretical_loss": 3.4374025711549385,
+      "tokens_seen": 1951727616
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004156161616161616,
+      "loss": 2.5199,
+      "theoretical_loss": 3.4373929013715276,
+      "tokens_seen": 1951793152
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004155959595959596,
+      "loss": 2.6086,
+      "theoretical_loss": 3.437383232003705,
+      "tokens_seen": 1951858688
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004155757575757576,
+      "loss": 2.5168,
+      "theoretical_loss": 3.4373735630514393,
+      "tokens_seen": 1951924224
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041555555555555557,
+      "loss": 2.6447,
+      "theoretical_loss": 3.4373638945146983,
+      "tokens_seen": 1951989760
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004155353535353535,
+      "loss": 2.4685,
+      "theoretical_loss": 3.4373542263934507,
+      "tokens_seen": 1952055296
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041551515151515154,
+      "loss": 2.5438,
+      "theoretical_loss": 3.4373445586876636,
+      "tokens_seen": 1952120832
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004154949494949495,
+      "loss": 2.5168,
+      "theoretical_loss": 3.4373348913973065,
+      "tokens_seen": 1952186368
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004154747474747475,
+      "loss": 2.5002,
+      "theoretical_loss": 3.4373252245223473,
+      "tokens_seen": 1952251904
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041545454545454546,
+      "loss": 2.6449,
+      "theoretical_loss": 3.4373155580627532,
+      "tokens_seen": 1952317440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004154343434343434,
+      "loss": 2.5862,
+      "theoretical_loss": 3.437305892018494,
+      "tokens_seen": 1952382976
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041541414141414143,
+      "loss": 2.6386,
+      "theoretical_loss": 3.437296226389537,
+      "tokens_seen": 1952448512
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004153939393939394,
+      "loss": 2.448,
+      "theoretical_loss": 3.43728656117585,
+      "tokens_seen": 1952514048
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004153737373737374,
+      "loss": 2.7955,
+      "theoretical_loss": 3.4372768963774023,
+      "tokens_seen": 1952579584
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041535353535353535,
+      "loss": 2.5514,
+      "theoretical_loss": 3.437267231994161,
+      "tokens_seen": 1952645120
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041533333333333336,
+      "loss": 2.7479,
+      "theoretical_loss": 3.4372575680260953,
+      "tokens_seen": 1952710656
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004153131313131313,
+      "loss": 2.6988,
+      "theoretical_loss": 3.437247904473173,
+      "tokens_seen": 1952776192
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004152929292929293,
+      "loss": 2.716,
+      "theoretical_loss": 3.437238241335362,
+      "tokens_seen": 1952841728
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004152727272727273,
+      "loss": 2.7452,
+      "theoretical_loss": 3.4372285786126318,
+      "tokens_seen": 1952907264
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041525252525252525,
+      "loss": 2.5785,
+      "theoretical_loss": 3.437218916304949,
+      "tokens_seen": 1952972800
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041523232323232326,
+      "loss": 2.5327,
+      "theoretical_loss": 3.4372092544122825,
+      "tokens_seen": 1953038336
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004152121212121212,
+      "loss": 2.6263,
+      "theoretical_loss": 3.4371995929346006,
+      "tokens_seen": 1953103872
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041519191919191917,
+      "loss": 2.5292,
+      "theoretical_loss": 3.437189931871872,
+      "tokens_seen": 1953169408
+    },
+    {
+      "epoch": 0.18,
+      "objective/train/docs_used": 1075533,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5056207180023193,
+      "objective/train/theoretical_loss": 3.437180271224064,
+      "objective/train/tokens_used": 312094176,
+      "theoretical_loss": 3.437180271224064,
+      "tokens_seen": 1953234944
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004151717171717172,
+      "loss": 2.6481,
+      "theoretical_loss": 3.437180271224064,
+      "tokens_seen": 1953234944
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004151515151515152,
+      "loss": 2.801,
+      "theoretical_loss": 3.437170610991146,
+      "tokens_seen": 1953300480
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041513131313131315,
+      "loss": 2.3028,
+      "theoretical_loss": 3.4371609511730856,
+      "tokens_seen": 1953366016
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004151111111111111,
+      "loss": 2.6984,
+      "theoretical_loss": 3.437151291769851,
+      "tokens_seen": 1953431552
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041509090909090906,
+      "loss": 2.6107,
+      "theoretical_loss": 3.437141632781411,
+      "tokens_seen": 1953497088
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004150707070707071,
+      "loss": 2.8986,
+      "theoretical_loss": 3.4371319742077326,
+      "tokens_seen": 1953562624
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004150505050505051,
+      "loss": 2.4985,
+      "theoretical_loss": 3.4371223160487854,
+      "tokens_seen": 1953628160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041503030303030304,
+      "loss": 2.4593,
+      "theoretical_loss": 3.437112658304537,
+      "tokens_seen": 1953693696
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000415010101010101,
+      "loss": 2.7836,
+      "theoretical_loss": 3.4371030009749566,
+      "tokens_seen": 1953759232
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041498989898989896,
+      "loss": 2.5408,
+      "theoretical_loss": 3.437093344060011,
+      "tokens_seen": 1953824768
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000414969696969697,
+      "loss": 2.7091,
+      "theoretical_loss": 3.4370836875596695,
+      "tokens_seen": 1953890304
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000414949494949495,
+      "loss": 2.745,
+      "theoretical_loss": 3.437074031473901,
+      "tokens_seen": 1953955840
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041492929292929293,
+      "loss": 2.6009,
+      "theoretical_loss": 3.437064375802672,
+      "tokens_seen": 1954021376
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004149090909090909,
+      "loss": 2.3948,
+      "theoretical_loss": 3.4370547205459516,
+      "tokens_seen": 1954086912
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041488888888888885,
+      "loss": 2.5832,
+      "theoretical_loss": 3.4370450657037086,
+      "tokens_seen": 1954152448
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004148686868686869,
+      "loss": 2.6297,
+      "theoretical_loss": 3.4370354112759114,
+      "tokens_seen": 1954217984
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041484848484848487,
+      "loss": 2.5004,
+      "theoretical_loss": 3.4370257572625276,
+      "tokens_seen": 1954283520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041482828282828283,
+      "loss": 2.7232,
+      "theoretical_loss": 3.4370161036635256,
+      "tokens_seen": 1954349056
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004148080808080808,
+      "loss": 2.7519,
+      "theoretical_loss": 3.437006450478875,
+      "tokens_seen": 1954414592
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004147878787878788,
+      "loss": 2.6824,
+      "theoretical_loss": 3.4369967977085416,
+      "tokens_seen": 1954480128
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004147676767676768,
+      "loss": 2.5209,
+      "theoretical_loss": 3.436987145352496,
+      "tokens_seen": 1954545664
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041474747474747476,
+      "loss": 2.6794,
+      "theoretical_loss": 3.436977493410706,
+      "tokens_seen": 1954611200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004147272727272727,
+      "loss": 2.577,
+      "theoretical_loss": 3.436967841883139,
+      "tokens_seen": 1954676736
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004147070707070707,
+      "loss": 2.6933,
+      "theoretical_loss": 3.436958190769764,
+      "tokens_seen": 1954742272
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004146868686868687,
+      "loss": 2.7894,
+      "theoretical_loss": 3.4369485400705497,
+      "tokens_seen": 1954807808
+    },
+    {
+      "epoch": 0.18,
+      "objective/train/docs_used": 1076714,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.991243839263916,
+      "objective/train/theoretical_loss": 3.4369388897854636,
+      "objective/train/tokens_used": 313732576,
+      "theoretical_loss": 3.4369388897854636,
+      "tokens_seen": 1954873344
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004146666666666667,
+      "loss": 2.5591,
+      "theoretical_loss": 3.4369388897854636,
+      "tokens_seen": 1954873344
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041464646464646466,
+      "loss": 2.6222,
+      "theoretical_loss": 3.436929239914475,
+      "tokens_seen": 1954938880
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004146262626262626,
+      "loss": 2.4555,
+      "theoretical_loss": 3.436919590457552,
+      "tokens_seen": 1955004416
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004146060606060606,
+      "loss": 2.699,
+      "theoretical_loss": 3.436909941414662,
+      "tokens_seen": 1955069952
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041458585858585863,
+      "loss": 2.587,
+      "theoretical_loss": 3.4369002927857744,
+      "tokens_seen": 1955135488
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004145656565656566,
+      "loss": 2.4276,
+      "theoretical_loss": 3.436890644570857,
+      "tokens_seen": 1955201024
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041454545454545455,
+      "loss": 2.6416,
+      "theoretical_loss": 3.436880996769879,
+      "tokens_seen": 1955266560
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004145252525252525,
+      "loss": 2.5871,
+      "theoretical_loss": 3.436871349382808,
+      "tokens_seen": 1955332096
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004145050505050505,
+      "loss": 2.6925,
+      "theoretical_loss": 3.436861702409612,
+      "tokens_seen": 1955397632
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041448484848484853,
+      "loss": 2.7467,
+      "theoretical_loss": 3.4368520558502604,
+      "tokens_seen": 1955463168
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004144646464646465,
+      "loss": 2.5199,
+      "theoretical_loss": 3.4368424097047208,
+      "tokens_seen": 1955528704
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041444444444444444,
+      "loss": 2.6255,
+      "theoretical_loss": 3.436832763972962,
+      "tokens_seen": 1955594240
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041442424242424245,
+      "loss": 2.7099,
+      "theoretical_loss": 3.4368231186549525,
+      "tokens_seen": 1955659776
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004144040404040404,
+      "loss": 2.6867,
+      "theoretical_loss": 3.43681347375066,
+      "tokens_seen": 1955725312
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004143838383838384,
+      "loss": 2.7098,
+      "theoretical_loss": 3.436803829260054,
+      "tokens_seen": 1955790848
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004143636363636364,
+      "loss": 2.3875,
+      "theoretical_loss": 3.4367941851831016,
+      "tokens_seen": 1955856384
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041434343434343433,
+      "loss": 2.6654,
+      "theoretical_loss": 3.4367845415197715,
+      "tokens_seen": 1955921920
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041432323232323235,
+      "loss": 2.6225,
+      "theoretical_loss": 3.4367748982700332,
+      "tokens_seen": 1955987456
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004143030303030303,
+      "loss": 2.6428,
+      "theoretical_loss": 3.436765255433854,
+      "tokens_seen": 1956052992
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004142828282828283,
+      "loss": 2.6854,
+      "theoretical_loss": 3.436755613011203,
+      "tokens_seen": 1956118528
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041426262626262627,
+      "loss": 2.4692,
+      "theoretical_loss": 3.436745971002048,
+      "tokens_seen": 1956184064
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004142424242424243,
+      "loss": 2.7986,
+      "theoretical_loss": 3.4367363294063575,
+      "tokens_seen": 1956249600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041422222222222224,
+      "loss": 2.6346,
+      "theoretical_loss": 3.4367266882241,
+      "tokens_seen": 1956315136
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004142020202020202,
+      "loss": 2.8425,
+      "theoretical_loss": 3.436717047455245,
+      "tokens_seen": 1956380672
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004141818181818182,
+      "loss": 2.5537,
+      "theoretical_loss": 3.436707407099759,
+      "tokens_seen": 1956446208
+    },
+    {
+      "epoch": 0.18,
+      "objective/train/docs_used": 1077107,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.886265754699707,
+      "objective/train/theoretical_loss": 3.4366977671576118,
+      "objective/train/tokens_used": 315370976,
+      "theoretical_loss": 3.4366977671576118,
+      "tokens_seen": 1956511744
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041416161616161616,
+      "loss": 2.6059,
+      "theoretical_loss": 3.4366977671576118,
+      "tokens_seen": 1956511744
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004141414141414142,
+      "loss": 2.7731,
+      "theoretical_loss": 3.436688127628771,
+      "tokens_seen": 1956577280
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041412121212121213,
+      "loss": 2.5947,
+      "theoretical_loss": 3.4366784885132056,
+      "tokens_seen": 1956642816
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004141010101010101,
+      "loss": 2.571,
+      "theoretical_loss": 3.436668849810884,
+      "tokens_seen": 1956708352
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004140808080808081,
+      "loss": 2.4155,
+      "theoretical_loss": 3.436659211521774,
+      "tokens_seen": 1956773888
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041406060606060606,
+      "loss": 2.5718,
+      "theoretical_loss": 3.4366495736458456,
+      "tokens_seen": 1956839424
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041404040404040407,
+      "loss": 2.4914,
+      "theoretical_loss": 3.4366399361830657,
+      "tokens_seen": 1956904960
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000414020202020202,
+      "loss": 2.7574,
+      "theoretical_loss": 3.436630299133403,
+      "tokens_seen": 1956970496
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000414,
+      "loss": 2.5857,
+      "theoretical_loss": 3.4366206624968267,
+      "tokens_seen": 1957036032
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000413979797979798,
+      "loss": 2.6283,
+      "theoretical_loss": 3.4366110262733045,
+      "tokens_seen": 1957101568
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000413959595959596,
+      "loss": 2.6388,
+      "theoretical_loss": 3.4366013904628048,
+      "tokens_seen": 1957167104
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041393939393939396,
+      "loss": 2.4807,
+      "theoretical_loss": 3.4365917550652973,
+      "tokens_seen": 1957232640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004139191919191919,
+      "loss": 2.5695,
+      "theoretical_loss": 3.4365821200807494,
+      "tokens_seen": 1957298176
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041389898989898987,
+      "loss": 2.7659,
+      "theoretical_loss": 3.4365724855091293,
+      "tokens_seen": 1957363712
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004138787878787879,
+      "loss": 2.6308,
+      "theoretical_loss": 3.4365628513504065,
+      "tokens_seen": 1957429248
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004138585858585859,
+      "loss": 2.6777,
+      "theoretical_loss": 3.4365532176045486,
+      "tokens_seen": 1957494784
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041383838383838385,
+      "loss": 2.4554,
+      "theoretical_loss": 3.436543584271524,
+      "tokens_seen": 1957560320
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004138181818181818,
+      "loss": 2.8151,
+      "theoretical_loss": 3.4365339513513025,
+      "tokens_seen": 1957625856
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041379797979797977,
+      "loss": 2.473,
+      "theoretical_loss": 3.4365243188438512,
+      "tokens_seen": 1957691392
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041377777777777783,
+      "loss": 2.5284,
+      "theoretical_loss": 3.436514686749139,
+      "tokens_seen": 1957756928
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004137575757575758,
+      "loss": 2.5062,
+      "theoretical_loss": 3.4365050550671348,
+      "tokens_seen": 1957822464
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041373737373737374,
+      "loss": 2.6967,
+      "theoretical_loss": 3.4364954237978065,
+      "tokens_seen": 1957888000
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004137171717171717,
+      "loss": 2.6102,
+      "theoretical_loss": 3.436485792941123,
+      "tokens_seen": 1957953536
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041369696969696966,
+      "loss": 2.6571,
+      "theoretical_loss": 3.436476162497053,
+      "tokens_seen": 1958019072
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004136767676767677,
+      "loss": 2.9256,
+      "theoretical_loss": 3.4364665324655643,
+      "tokens_seen": 1958084608
+    },
+    {
+      "epoch": 0.18,
+      "objective/train/docs_used": 1078288,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.1963400840759277,
+      "objective/train/theoretical_loss": 3.4364569028466265,
+      "objective/train/tokens_used": 317009376,
+      "theoretical_loss": 3.4364569028466265,
+      "tokens_seen": 1958150144
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004136565656565657,
+      "loss": 2.6203,
+      "theoretical_loss": 3.4364569028466265,
+      "tokens_seen": 1958150144
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041363636363636364,
+      "loss": 2.7172,
+      "theoretical_loss": 3.4364472736402067,
+      "tokens_seen": 1958215680
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004136161616161616,
+      "loss": 2.5027,
+      "theoretical_loss": 3.436437644846275,
+      "tokens_seen": 1958281216
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004135959595959596,
+      "loss": 2.6738,
+      "theoretical_loss": 3.4364280164647987,
+      "tokens_seen": 1958346752
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004135757575757576,
+      "loss": 2.5267,
+      "theoretical_loss": 3.4364183884957464,
+      "tokens_seen": 1958412288
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004135555555555556,
+      "loss": 2.4939,
+      "theoretical_loss": 3.4364087609390874,
+      "tokens_seen": 1958477824
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041353535353535353,
+      "loss": 2.5811,
+      "theoretical_loss": 3.43639913379479,
+      "tokens_seen": 1958543360
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004135151515151515,
+      "loss": 2.3717,
+      "theoretical_loss": 3.436389507062822,
+      "tokens_seen": 1958608896
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004134949494949495,
+      "loss": 2.5317,
+      "theoretical_loss": 3.4363798807431527,
+      "tokens_seen": 1958674432
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004134747474747475,
+      "loss": 2.5381,
+      "theoretical_loss": 3.436370254835751,
+      "tokens_seen": 1958739968
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041345454545454547,
+      "loss": 2.5417,
+      "theoretical_loss": 3.4363606293405837,
+      "tokens_seen": 1958805504
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004134343434343434,
+      "loss": 2.6023,
+      "theoretical_loss": 3.4363510042576215,
+      "tokens_seen": 1958871040
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041341414141414143,
+      "loss": 2.8136,
+      "theoretical_loss": 3.436341379586832,
+      "tokens_seen": 1958936576
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004133939393939394,
+      "loss": 2.6125,
+      "theoretical_loss": 3.4363317553281836,
+      "tokens_seen": 1959002112
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004133737373737374,
+      "loss": 2.5957,
+      "theoretical_loss": 3.436322131481645,
+      "tokens_seen": 1959067648
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041335353535353536,
+      "loss": 2.6129,
+      "theoretical_loss": 3.436312508047185,
+      "tokens_seen": 1959133184
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004133333333333333,
+      "loss": 2.5397,
+      "theoretical_loss": 3.4363028850247717,
+      "tokens_seen": 1959198720
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004133131313131313,
+      "loss": 2.409,
+      "theoretical_loss": 3.436293262414374,
+      "tokens_seen": 1959264256
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004132929292929293,
+      "loss": 2.4528,
+      "theoretical_loss": 3.4362836402159607,
+      "tokens_seen": 1959329792
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004132727272727273,
+      "loss": 2.6568,
+      "theoretical_loss": 3.4362740184294998,
+      "tokens_seen": 1959395328
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041325252525252525,
+      "loss": 2.6499,
+      "theoretical_loss": 3.43626439705496,
+      "tokens_seen": 1959460864
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041323232323232326,
+      "loss": 2.3524,
+      "theoretical_loss": 3.4362547760923103,
+      "tokens_seen": 1959526400
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004132121212121212,
+      "loss": 2.5487,
+      "theoretical_loss": 3.4362451555415197,
+      "tokens_seen": 1959591936
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004131919191919192,
+      "loss": 2.5426,
+      "theoretical_loss": 3.436235535402555,
+      "tokens_seen": 1959657472
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004131717171717172,
+      "loss": 2.3804,
+      "theoretical_loss": 3.4362259156753874,
+      "tokens_seen": 1959723008
+    },
+    {
+      "epoch": 0.18,
+      "objective/train/docs_used": 1079604,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.561680793762207,
+      "objective/train/theoretical_loss": 3.4362162963599827,
+      "objective/train/tokens_used": 318647776,
+      "theoretical_loss": 3.4362162963599827,
+      "tokens_seen": 1959788544
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041315151515151514,
+      "loss": 2.7004,
+      "theoretical_loss": 3.4362162963599827,
+      "tokens_seen": 1959788544
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041313131313131315,
+      "loss": 2.4122,
+      "theoretical_loss": 3.436206677456312,
+      "tokens_seen": 1959854080
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004131111111111111,
+      "loss": 2.7408,
+      "theoretical_loss": 3.436197058964342,
+      "tokens_seen": 1959919616
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041309090909090907,
+      "loss": 2.7353,
+      "theoretical_loss": 3.4361874408840425,
+      "tokens_seen": 1959985152
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004130707070707071,
+      "loss": 2.7172,
+      "theoretical_loss": 3.4361778232153815,
+      "tokens_seen": 1960050688
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004130505050505051,
+      "loss": 2.5049,
+      "theoretical_loss": 3.436168205958328,
+      "tokens_seen": 1960116224
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041303030303030305,
+      "loss": 2.602,
+      "theoretical_loss": 3.4361585891128508,
+      "tokens_seen": 1960181760
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000413010101010101,
+      "loss": 2.6146,
+      "theoretical_loss": 3.436148972678917,
+      "tokens_seen": 1960247296
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041298989898989896,
+      "loss": 2.7808,
+      "theoretical_loss": 3.436139356656498,
+      "tokens_seen": 1960312832
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041296969696969697,
+      "loss": 2.6078,
+      "theoretical_loss": 3.43612974104556,
+      "tokens_seen": 1960378368
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000412949494949495,
+      "loss": 2.7435,
+      "theoretical_loss": 3.436120125846072,
+      "tokens_seen": 1960443904
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041292929292929294,
+      "loss": 2.5543,
+      "theoretical_loss": 3.436110511058004,
+      "tokens_seen": 1960509440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004129090909090909,
+      "loss": 2.8017,
+      "theoretical_loss": 3.4361008966813236,
+      "tokens_seen": 1960574976
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004128888888888889,
+      "loss": 2.7139,
+      "theoretical_loss": 3.4360912827159997,
+      "tokens_seen": 1960640512
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004128686868686869,
+      "loss": 2.7691,
+      "theoretical_loss": 3.4360816691620006,
+      "tokens_seen": 1960706048
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004128484848484849,
+      "loss": 2.558,
+      "theoretical_loss": 3.436072056019295,
+      "tokens_seen": 1960771584
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041282828282828283,
+      "loss": 2.5107,
+      "theoretical_loss": 3.4360624432878524,
+      "tokens_seen": 1960837120
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004128080808080808,
+      "loss": 2.4678,
+      "theoretical_loss": 3.4360528309676406,
+      "tokens_seen": 1960902656
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004127878787878788,
+      "loss": 2.6761,
+      "theoretical_loss": 3.4360432190586288,
+      "tokens_seen": 1960968192
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004127676767676768,
+      "loss": 2.6746,
+      "theoretical_loss": 3.436033607560785,
+      "tokens_seen": 1961033728
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041274747474747477,
+      "loss": 2.8905,
+      "theoretical_loss": 3.4360239964740784,
+      "tokens_seen": 1961099264
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004127272727272727,
+      "loss": 2.3687,
+      "theoretical_loss": 3.4360143857984777,
+      "tokens_seen": 1961164800
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004127070707070707,
+      "loss": 2.7308,
+      "theoretical_loss": 3.4360047755339513,
+      "tokens_seen": 1961230336
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041268686868686875,
+      "loss": 2.6371,
+      "theoretical_loss": 3.4359951656804677,
+      "tokens_seen": 1961295872
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004126666666666667,
+      "loss": 2.6793,
+      "theoretical_loss": 3.4359855562379957,
+      "tokens_seen": 1961361408
+    },
+    {
+      "epoch": 0.18,
+      "objective/train/docs_used": 1080156,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.839082717895508,
+      "objective/train/theoretical_loss": 3.435975947206505,
+      "objective/train/tokens_used": 320286176,
+      "theoretical_loss": 3.435975947206505,
+      "tokens_seen": 1961426944
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041264646464646466,
+      "loss": 2.7623,
+      "theoretical_loss": 3.435975947206505,
+      "tokens_seen": 1961426944
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004126262626262626,
+      "loss": 2.7329,
+      "theoretical_loss": 3.435966338585963,
+      "tokens_seen": 1961492480
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004126060606060606,
+      "loss": 2.5897,
+      "theoretical_loss": 3.4359567303763385,
+      "tokens_seen": 1961558016
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041258585858585864,
+      "loss": 2.4654,
+      "theoretical_loss": 3.435947122577601,
+      "tokens_seen": 1961623552
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004125656565656566,
+      "loss": 2.7386,
+      "theoretical_loss": 3.4359375151897185,
+      "tokens_seen": 1961689088
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041254545454545455,
+      "loss": 2.6012,
+      "theoretical_loss": 3.43592790821266,
+      "tokens_seen": 1961754624
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004125252525252525,
+      "loss": 2.6477,
+      "theoretical_loss": 3.4359183016463946,
+      "tokens_seen": 1961820160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041250505050505047,
+      "loss": 2.8662,
+      "theoretical_loss": 3.43590869549089,
+      "tokens_seen": 1961885696
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041248484848484853,
+      "loss": 2.455,
+      "theoretical_loss": 3.435899089746116,
+      "tokens_seen": 1961951232
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004124646464646465,
+      "loss": 2.8418,
+      "theoretical_loss": 3.4358894844120407,
+      "tokens_seen": 1962016768
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041244444444444445,
+      "loss": 2.666,
+      "theoretical_loss": 3.435879879488633,
+      "tokens_seen": 1962082304
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004124242424242424,
+      "loss": 2.6833,
+      "theoretical_loss": 3.435870274975861,
+      "tokens_seen": 1962147840
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004124040404040404,
+      "loss": 2.5037,
+      "theoretical_loss": 3.4358606708736943,
+      "tokens_seen": 1962213376
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004123838383838384,
+      "loss": 2.5875,
+      "theoretical_loss": 3.435851067182101,
+      "tokens_seen": 1962278912
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004123636363636364,
+      "loss": 2.5752,
+      "theoretical_loss": 3.435841463901051,
+      "tokens_seen": 1962344448
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041234343434343434,
+      "loss": 2.7171,
+      "theoretical_loss": 3.435831861030512,
+      "tokens_seen": 1962409984
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004123232323232323,
+      "loss": 2.5322,
+      "theoretical_loss": 3.4358222585704525,
+      "tokens_seen": 1962475520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004123030303030303,
+      "loss": 2.621,
+      "theoretical_loss": 3.4358126565208416,
+      "tokens_seen": 1962541056
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004122828282828283,
+      "loss": 2.6025,
+      "theoretical_loss": 3.4358030548816485,
+      "tokens_seen": 1962606592
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004122626262626263,
+      "loss": 2.6889,
+      "theoretical_loss": 3.435793453652842,
+      "tokens_seen": 1962672128
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041224242424242423,
+      "loss": 2.5616,
+      "theoretical_loss": 3.4357838528343896,
+      "tokens_seen": 1962737664
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041222222222222224,
+      "loss": 2.445,
+      "theoretical_loss": 3.4357742524262616,
+      "tokens_seen": 1962803200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004122020202020202,
+      "loss": 2.5863,
+      "theoretical_loss": 3.4357646524284258,
+      "tokens_seen": 1962868736
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004121818181818182,
+      "loss": 2.5758,
+      "theoretical_loss": 3.435755052840851,
+      "tokens_seen": 1962934272
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041216161616161617,
+      "loss": 2.7752,
+      "theoretical_loss": 3.4357454536635066,
+      "tokens_seen": 1962999808
+    },
+    {
+      "epoch": 0.18,
+      "objective/train/docs_used": 1081069,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6743087768554688,
+      "objective/train/theoretical_loss": 3.435735854896361,
+      "objective/train/tokens_used": 321924576,
+      "theoretical_loss": 3.435735854896361,
+      "tokens_seen": 1963065344
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004121414141414141,
+      "loss": 2.397,
+      "theoretical_loss": 3.435735854896361,
+      "tokens_seen": 1963065344
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041212121212121214,
+      "loss": 2.7075,
+      "theoretical_loss": 3.4357262565393825,
+      "tokens_seen": 1963130880
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004121010101010101,
+      "loss": 2.6683,
+      "theoretical_loss": 3.4357166585925407,
+      "tokens_seen": 1963196416
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004120808080808081,
+      "loss": 2.4466,
+      "theoretical_loss": 3.4357070610558043,
+      "tokens_seen": 1963261952
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041206060606060606,
+      "loss": 2.7529,
+      "theoretical_loss": 3.4356974639291415,
+      "tokens_seen": 1963327488
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041204040404040407,
+      "loss": 2.6604,
+      "theoretical_loss": 3.4356878672125215,
+      "tokens_seen": 1963393024
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041202020202020203,
+      "loss": 2.4729,
+      "theoretical_loss": 3.435678270905913,
+      "tokens_seen": 1963458560
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000412,
+      "loss": 2.9114,
+      "theoretical_loss": 3.4356686750092846,
+      "tokens_seen": 1963524096
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000411979797979798,
+      "loss": 2.6102,
+      "theoretical_loss": 3.4356590795226056,
+      "tokens_seen": 1963589632
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041195959595959595,
+      "loss": 2.5086,
+      "theoretical_loss": 3.4356494844458445,
+      "tokens_seen": 1963655168
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041193939393939396,
+      "loss": 2.7648,
+      "theoretical_loss": 3.43563988977897,
+      "tokens_seen": 1963720704
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004119191919191919,
+      "loss": 2.4557,
+      "theoretical_loss": 3.435630295521951,
+      "tokens_seen": 1963786240
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004118989898989899,
+      "loss": 2.5652,
+      "theoretical_loss": 3.4356207016747566,
+      "tokens_seen": 1963851776
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004118787878787879,
+      "loss": 2.5733,
+      "theoretical_loss": 3.4356111082373553,
+      "tokens_seen": 1963917312
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004118585858585859,
+      "loss": 2.6456,
+      "theoretical_loss": 3.435601515209716,
+      "tokens_seen": 1963982848
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041183838383838386,
+      "loss": 2.5374,
+      "theoretical_loss": 3.435591922591807,
+      "tokens_seen": 1964048384
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004118181818181818,
+      "loss": 2.4767,
+      "theoretical_loss": 3.435582330383598,
+      "tokens_seen": 1964113920
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041179797979797977,
+      "loss": 2.737,
+      "theoretical_loss": 3.4355727385850576,
+      "tokens_seen": 1964179456
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004117777777777778,
+      "loss": 2.4674,
+      "theoretical_loss": 3.435563147196155,
+      "tokens_seen": 1964244992
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004117575757575758,
+      "loss": 2.4722,
+      "theoretical_loss": 3.435553556216858,
+      "tokens_seen": 1964310528
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041173737373737375,
+      "loss": 2.5974,
+      "theoretical_loss": 3.4355439656471356,
+      "tokens_seen": 1964376064
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004117171717171717,
+      "loss": 2.8329,
+      "theoretical_loss": 3.4355343754869576,
+      "tokens_seen": 1964441600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041169696969696966,
+      "loss": 2.654,
+      "theoretical_loss": 3.435524785736292,
+      "tokens_seen": 1964507136
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041167676767676773,
+      "loss": 2.7161,
+      "theoretical_loss": 3.4355151963951083,
+      "tokens_seen": 1964572672
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0004116565656565657,
+      "loss": 2.5793,
+      "theoretical_loss": 3.4355056074633747,
+      "tokens_seen": 1964638208
+    },
+    {
+      "epoch": 0.18,
+      "objective/train/docs_used": 1081685,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.539858341217041,
+      "objective/train/theoretical_loss": 3.4354960189410604,
+      "objective/train/tokens_used": 323562976,
+      "theoretical_loss": 3.4354960189410604,
+      "tokens_seen": 1964703744
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00041163636363636364,
+      "loss": 2.7547,
+      "theoretical_loss": 3.4354960189410604,
+      "tokens_seen": 1964703744
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004116161616161616,
+      "loss": 2.8963,
+      "theoretical_loss": 3.435486430828134,
+      "tokens_seen": 1964769280
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041159595959595956,
+      "loss": 2.6418,
+      "theoretical_loss": 3.4354768431245652,
+      "tokens_seen": 1964834816
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004115757575757576,
+      "loss": 2.8092,
+      "theoretical_loss": 3.435467255830322,
+      "tokens_seen": 1964900352
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004115555555555556,
+      "loss": 2.5482,
+      "theoretical_loss": 3.4354576689453733,
+      "tokens_seen": 1964965888
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041153535353535354,
+      "loss": 2.6656,
+      "theoretical_loss": 3.4354480824696885,
+      "tokens_seen": 1965031424
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004115151515151515,
+      "loss": 2.71,
+      "theoretical_loss": 3.4354384964032363,
+      "tokens_seen": 1965096960
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004114949494949495,
+      "loss": 2.7688,
+      "theoretical_loss": 3.4354289107459848,
+      "tokens_seen": 1965162496
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004114747474747475,
+      "loss": 2.3955,
+      "theoretical_loss": 3.435419325497904,
+      "tokens_seen": 1965228032
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041145454545454547,
+      "loss": 2.6943,
+      "theoretical_loss": 3.4354097406589625,
+      "tokens_seen": 1965293568
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041143434343434343,
+      "loss": 2.6809,
+      "theoretical_loss": 3.4354001562291288,
+      "tokens_seen": 1965359104
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004114141414141414,
+      "loss": 2.6625,
+      "theoretical_loss": 3.435390572208372,
+      "tokens_seen": 1965424640
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004113939393939394,
+      "loss": 2.6743,
+      "theoretical_loss": 3.4353809885966613,
+      "tokens_seen": 1965490176
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004113737373737374,
+      "loss": 2.4488,
+      "theoretical_loss": 3.4353714053939655,
+      "tokens_seen": 1965555712
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041135353535353536,
+      "loss": 2.6777,
+      "theoretical_loss": 3.4353618226002527,
+      "tokens_seen": 1965621248
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004113333333333333,
+      "loss": 2.6546,
+      "theoretical_loss": 3.435352240215493,
+      "tokens_seen": 1965686784
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041131313131313133,
+      "loss": 2.5162,
+      "theoretical_loss": 3.4353426582396542,
+      "tokens_seen": 1965752320
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004112929292929293,
+      "loss": 2.5366,
+      "theoretical_loss": 3.4353330766727064,
+      "tokens_seen": 1965817856
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004112727272727273,
+      "loss": 2.6484,
+      "theoretical_loss": 3.4353234955146177,
+      "tokens_seen": 1965883392
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041125252525252526,
+      "loss": 2.574,
+      "theoretical_loss": 3.4353139147653575,
+      "tokens_seen": 1965948928
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004112323232323232,
+      "loss": 2.3125,
+      "theoretical_loss": 3.4353043344248944,
+      "tokens_seen": 1966014464
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004112121212121212,
+      "loss": 2.5873,
+      "theoretical_loss": 3.4352947544931967,
+      "tokens_seen": 1966080000
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041119191919191924,
+      "loss": 2.6623,
+      "theoretical_loss": 3.435285174970235,
+      "tokens_seen": 1966145536
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004111717171717172,
+      "loss": 2.7218,
+      "theoretical_loss": 3.435275595855977,
+      "tokens_seen": 1966211072
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041115151515151515,
+      "loss": 2.4976,
+      "theoretical_loss": 3.4352660171503913,
+      "tokens_seen": 1966276608
+    },
+    {
+      "epoch": 0.19,
+      "objective/train/docs_used": 1082856,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7312498092651367,
+      "objective/train/theoretical_loss": 3.4352564388534477,
+      "objective/train/tokens_used": 325201376,
+      "theoretical_loss": 3.4352564388534477,
+      "tokens_seen": 1966342144
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041113131313131316,
+      "loss": 2.6919,
+      "theoretical_loss": 3.4352564388534477,
+      "tokens_seen": 1966342144
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004111111111111111,
+      "loss": 2.5451,
+      "theoretical_loss": 3.4352468609651154,
+      "tokens_seen": 1966407680
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041109090909090913,
+      "loss": 2.4027,
+      "theoretical_loss": 3.435237283485362,
+      "tokens_seen": 1966473216
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004110707070707071,
+      "loss": 2.6824,
+      "theoretical_loss": 3.4352277064141585,
+      "tokens_seen": 1966538752
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041105050505050504,
+      "loss": 2.4505,
+      "theoretical_loss": 3.4352181297514717,
+      "tokens_seen": 1966604288
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041103030303030305,
+      "loss": 2.4184,
+      "theoretical_loss": 3.4352085534972723,
+      "tokens_seen": 1966669824
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000411010101010101,
+      "loss": 2.5168,
+      "theoretical_loss": 3.4351989776515275,
+      "tokens_seen": 1966735360
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000410989898989899,
+      "loss": 2.5689,
+      "theoretical_loss": 3.435189402214208,
+      "tokens_seen": 1966800896
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000410969696969697,
+      "loss": 3.0605,
+      "theoretical_loss": 3.435179827185282,
+      "tokens_seen": 1966866432
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000410949494949495,
+      "loss": 2.6067,
+      "theoretical_loss": 3.4351702525647183,
+      "tokens_seen": 1966931968
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041092929292929295,
+      "loss": 2.7054,
+      "theoretical_loss": 3.435160678352486,
+      "tokens_seen": 1966997504
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004109090909090909,
+      "loss": 2.3814,
+      "theoretical_loss": 3.4351511045485545,
+      "tokens_seen": 1967063040
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004108888888888889,
+      "loss": 2.8035,
+      "theoretical_loss": 3.435141531152892,
+      "tokens_seen": 1967128576
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041086868686868687,
+      "loss": 2.5896,
+      "theoretical_loss": 3.4351319581654685,
+      "tokens_seen": 1967194112
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004108484848484849,
+      "loss": 2.9124,
+      "theoretical_loss": 3.4351223855862525,
+      "tokens_seen": 1967259648
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041082828282828284,
+      "loss": 2.7392,
+      "theoretical_loss": 3.4351128134152122,
+      "tokens_seen": 1967325184
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004108080808080808,
+      "loss": 2.5456,
+      "theoretical_loss": 3.435103241652318,
+      "tokens_seen": 1967390720
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004107878787878788,
+      "loss": 2.6759,
+      "theoretical_loss": 3.4350936702975376,
+      "tokens_seen": 1967456256
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041076767676767676,
+      "loss": 2.6334,
+      "theoretical_loss": 3.435084099350841,
+      "tokens_seen": 1967521792
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004107474747474748,
+      "loss": 2.409,
+      "theoretical_loss": 3.435074528812197,
+      "tokens_seen": 1967587328
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041072727272727273,
+      "loss": 2.4961,
+      "theoretical_loss": 3.4350649586815742,
+      "tokens_seen": 1967652864
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004107070707070707,
+      "loss": 2.6618,
+      "theoretical_loss": 3.435055388958942,
+      "tokens_seen": 1967718400
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004106868686868687,
+      "loss": 2.6335,
+      "theoretical_loss": 3.4350458196442695,
+      "tokens_seen": 1967783936
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004106666666666667,
+      "loss": 2.6031,
+      "theoretical_loss": 3.435036250737525,
+      "tokens_seen": 1967849472
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041064646464646467,
+      "loss": 2.2975,
+      "theoretical_loss": 3.4350266822386786,
+      "tokens_seen": 1967915008
+    },
+    {
+      "epoch": 0.19,
+      "objective/train/docs_used": 1083372,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.8083508014678955,
+      "objective/train/theoretical_loss": 3.4350171141476986,
+      "objective/train/tokens_used": 326839776,
+      "theoretical_loss": 3.4350171141476986,
+      "tokens_seen": 1967980544
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004106262626262626,
+      "loss": 2.7665,
+      "theoretical_loss": 3.4350171141476986,
+      "tokens_seen": 1967980544
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004106060606060606,
+      "loss": 2.6247,
+      "theoretical_loss": 3.435007546464554,
+      "tokens_seen": 1968046080
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004105858585858586,
+      "loss": 2.6699,
+      "theoretical_loss": 3.4349979791892142,
+      "tokens_seen": 1968111616
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004105656565656566,
+      "loss": 2.6583,
+      "theoretical_loss": 3.434988412321648,
+      "tokens_seen": 1968177152
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041054545454545456,
+      "loss": 2.6518,
+      "theoretical_loss": 3.4349788458618247,
+      "tokens_seen": 1968242688
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004105252525252525,
+      "loss": 2.5485,
+      "theoretical_loss": 3.434969279809713,
+      "tokens_seen": 1968308224
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004105050505050505,
+      "loss": 2.3527,
+      "theoretical_loss": 3.434959714165282,
+      "tokens_seen": 1968373760
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041048484848484854,
+      "loss": 2.6313,
+      "theoretical_loss": 3.4349501489285013,
+      "tokens_seen": 1968439296
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004104646464646465,
+      "loss": 2.6005,
+      "theoretical_loss": 3.4349405840993392,
+      "tokens_seen": 1968504832
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041044444444444445,
+      "loss": 2.6,
+      "theoretical_loss": 3.434931019677765,
+      "tokens_seen": 1968570368
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004104242424242424,
+      "loss": 2.7116,
+      "theoretical_loss": 3.434921455663748,
+      "tokens_seen": 1968635904
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041040404040404037,
+      "loss": 2.48,
+      "theoretical_loss": 3.4349118920572574,
+      "tokens_seen": 1968701440
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041038383838383843,
+      "loss": 2.661,
+      "theoretical_loss": 3.4349023288582616,
+      "tokens_seen": 1968766976
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004103636363636364,
+      "loss": 2.5923,
+      "theoretical_loss": 3.4348927660667306,
+      "tokens_seen": 1968832512
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041034343434343435,
+      "loss": 2.3861,
+      "theoretical_loss": 3.4348832036826322,
+      "tokens_seen": 1968898048
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004103232323232323,
+      "loss": 2.492,
+      "theoretical_loss": 3.434873641705937,
+      "tokens_seen": 1968963584
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004103030303030303,
+      "loss": 2.7685,
+      "theoretical_loss": 3.434864080136612,
+      "tokens_seen": 1969029120
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004102828282828283,
+      "loss": 2.8386,
+      "theoretical_loss": 3.434854518974629,
+      "tokens_seen": 1969094656
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004102626262626263,
+      "loss": 2.8557,
+      "theoretical_loss": 3.434844958219955,
+      "tokens_seen": 1969160192
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041024242424242424,
+      "loss": 2.5892,
+      "theoretical_loss": 3.4348353978725603,
+      "tokens_seen": 1969225728
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004102222222222222,
+      "loss": 2.5535,
+      "theoretical_loss": 3.434825837932413,
+      "tokens_seen": 1969291264
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004102020202020202,
+      "loss": 2.66,
+      "theoretical_loss": 3.4348162783994827,
+      "tokens_seen": 1969356800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004101818181818182,
+      "loss": 2.5879,
+      "theoretical_loss": 3.4348067192737384,
+      "tokens_seen": 1969422336
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004101616161616162,
+      "loss": 2.7818,
+      "theoretical_loss": 3.4347971605551493,
+      "tokens_seen": 1969487872
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041014141414141413,
+      "loss": 2.5091,
+      "theoretical_loss": 3.4347876022436843,
+      "tokens_seen": 1969553408
+    },
+    {
+      "epoch": 0.19,
+      "objective/train/docs_used": 1084308,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.9995901584625244,
+      "objective/train/theoretical_loss": 3.434778044339313,
+      "objective/train/tokens_used": 328478176,
+      "theoretical_loss": 3.434778044339313,
+      "tokens_seen": 1969618944
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041012121212121214,
+      "loss": 2.6201,
+      "theoretical_loss": 3.434778044339313,
+      "tokens_seen": 1969618944
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004101010101010101,
+      "loss": 2.7089,
+      "theoretical_loss": 3.434768486842005,
+      "tokens_seen": 1969684480
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004100808080808081,
+      "loss": 2.6555,
+      "theoretical_loss": 3.4347589297517276,
+      "tokens_seen": 1969750016
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041006060606060607,
+      "loss": 2.3727,
+      "theoretical_loss": 3.434749373068451,
+      "tokens_seen": 1969815552
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000410040404040404,
+      "loss": 2.6942,
+      "theoretical_loss": 3.4347398167921446,
+      "tokens_seen": 1969881088
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041002020202020203,
+      "loss": 2.5807,
+      "theoretical_loss": 3.4347302609227772,
+      "tokens_seen": 1969946624
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00041,
+      "loss": 2.6507,
+      "theoretical_loss": 3.434720705460318,
+      "tokens_seen": 1970012160
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000409979797979798,
+      "loss": 2.5104,
+      "theoretical_loss": 3.434711150404736,
+      "tokens_seen": 1970077696
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040995959595959596,
+      "loss": 2.57,
+      "theoretical_loss": 3.4347015957560005,
+      "tokens_seen": 1970143232
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040993939393939397,
+      "loss": 2.3079,
+      "theoretical_loss": 3.4346920415140803,
+      "tokens_seen": 1970208768
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004099191919191919,
+      "loss": 2.6171,
+      "theoretical_loss": 3.4346824876789452,
+      "tokens_seen": 1970274304
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004098989898989899,
+      "loss": 2.5322,
+      "theoretical_loss": 3.434672934250564,
+      "tokens_seen": 1970339840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004098787878787879,
+      "loss": 2.6325,
+      "theoretical_loss": 3.434663381228906,
+      "tokens_seen": 1970405376
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040985858585858585,
+      "loss": 2.6111,
+      "theoretical_loss": 3.434653828613939,
+      "tokens_seen": 1970470912
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040983838383838386,
+      "loss": 2.6792,
+      "theoretical_loss": 3.434644276405635,
+      "tokens_seen": 1970536448
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004098181818181818,
+      "loss": 2.6454,
+      "theoretical_loss": 3.434634724603961,
+      "tokens_seen": 1970601984
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004097979797979798,
+      "loss": 2.408,
+      "theoretical_loss": 3.434625173208886,
+      "tokens_seen": 1970667520
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004097777777777778,
+      "loss": 2.5299,
+      "theoretical_loss": 3.43461562222038,
+      "tokens_seen": 1970733056
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004097575757575758,
+      "loss": 2.701,
+      "theoretical_loss": 3.4346060716384126,
+      "tokens_seen": 1970798592
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040973737373737376,
+      "loss": 2.6637,
+      "theoretical_loss": 3.434596521462952,
+      "tokens_seen": 1970864128
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004097171717171717,
+      "loss": 2.4817,
+      "theoretical_loss": 3.434586971693968,
+      "tokens_seen": 1970929664
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040969696969696967,
+      "loss": 2.6051,
+      "theoretical_loss": 3.4345774223314294,
+      "tokens_seen": 1970995200
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004096767676767677,
+      "loss": 2.4224,
+      "theoretical_loss": 3.434567873375306,
+      "tokens_seen": 1971060736
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004096565656565657,
+      "loss": 2.8169,
+      "theoretical_loss": 3.4345583248255656,
+      "tokens_seen": 1971126272
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040963636363636365,
+      "loss": 2.439,
+      "theoretical_loss": 3.4345487766821794,
+      "tokens_seen": 1971191808
+    },
+    {
+      "epoch": 0.19,
+      "objective/train/docs_used": 1085454,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7234983444213867,
+      "objective/train/theoretical_loss": 3.434539228945115,
+      "objective/train/tokens_used": 330116576,
+      "theoretical_loss": 3.434539228945115,
+      "tokens_seen": 1971257344
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004096161616161616,
+      "loss": 2.4934,
+      "theoretical_loss": 3.434539228945115,
+      "tokens_seen": 1971257344
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040959595959595956,
+      "loss": 2.5968,
+      "theoretical_loss": 3.4345296816143422,
+      "tokens_seen": 1971322880
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040957575757575763,
+      "loss": 2.8397,
+      "theoretical_loss": 3.43452013468983,
+      "tokens_seen": 1971388416
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004095555555555556,
+      "loss": 2.6118,
+      "theoretical_loss": 3.4345105881715483,
+      "tokens_seen": 1971453952
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040953535353535354,
+      "loss": 2.785,
+      "theoretical_loss": 3.4345010420594653,
+      "tokens_seen": 1971519488
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004095151515151515,
+      "loss": 2.6511,
+      "theoretical_loss": 3.4344914963535507,
+      "tokens_seen": 1971585024
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040949494949494945,
+      "loss": 2.5009,
+      "theoretical_loss": 3.434481951053774,
+      "tokens_seen": 1971650560
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004094747474747475,
+      "loss": 2.7502,
+      "theoretical_loss": 3.4344724061601037,
+      "tokens_seen": 1971716096
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004094545454545455,
+      "loss": 2.4277,
+      "theoretical_loss": 3.4344628616725097,
+      "tokens_seen": 1971781632
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040943434343434343,
+      "loss": 2.6531,
+      "theoretical_loss": 3.434453317590961,
+      "tokens_seen": 1971847168
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004094141414141414,
+      "loss": 2.5949,
+      "theoretical_loss": 3.4344437739154268,
+      "tokens_seen": 1971912704
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040939393939393946,
+      "loss": 2.7611,
+      "theoretical_loss": 3.434434230645876,
+      "tokens_seen": 1971978240
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004093737373737374,
+      "loss": 2.6381,
+      "theoretical_loss": 3.434424687782279,
+      "tokens_seen": 1972043776
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040935353535353537,
+      "loss": 2.4472,
+      "theoretical_loss": 3.4344151453246035,
+      "tokens_seen": 1972109312
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004093333333333333,
+      "loss": 2.448,
+      "theoretical_loss": 3.4344056032728196,
+      "tokens_seen": 1972174848
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004093131313131313,
+      "loss": 2.6905,
+      "theoretical_loss": 3.4343960616268965,
+      "tokens_seen": 1972240384
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040929292929292935,
+      "loss": 2.8451,
+      "theoretical_loss": 3.4343865203868034,
+      "tokens_seen": 1972305920
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004092727272727273,
+      "loss": 2.4575,
+      "theoretical_loss": 3.4343769795525096,
+      "tokens_seen": 1972371456
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040925252525252526,
+      "loss": 2.8152,
+      "theoretical_loss": 3.4343674391239842,
+      "tokens_seen": 1972436992
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004092323232323232,
+      "loss": 2.4402,
+      "theoretical_loss": 3.4343578991011965,
+      "tokens_seen": 1972502528
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004092121212121212,
+      "loss": 2.4436,
+      "theoretical_loss": 3.4343483594841158,
+      "tokens_seen": 1972568064
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040919191919191924,
+      "loss": 2.845,
+      "theoretical_loss": 3.4343388202727114,
+      "tokens_seen": 1972633600
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004091717171717172,
+      "loss": 2.7507,
+      "theoretical_loss": 3.4343292814669524,
+      "tokens_seen": 1972699136
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040915151515151515,
+      "loss": 2.6888,
+      "theoretical_loss": 3.4343197430668084,
+      "tokens_seen": 1972764672
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004091313131313131,
+      "loss": 2.6904,
+      "theoretical_loss": 3.4343102050722485,
+      "tokens_seen": 1972830208
+    },
+    {
+      "epoch": 0.19,
+      "objective/train/docs_used": 1086100,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.58170747756958,
+      "objective/train/theoretical_loss": 3.434300667483242,
+      "objective/train/tokens_used": 331754976,
+      "theoretical_loss": 3.434300667483242,
+      "tokens_seen": 1972895744
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004091111111111111,
+      "loss": 2.5805,
+      "theoretical_loss": 3.434300667483242,
+      "tokens_seen": 1972895744
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040909090909090913,
+      "loss": 2.5536,
+      "theoretical_loss": 3.434291130299758,
+      "tokens_seen": 1972961280
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004090707070707071,
+      "loss": 2.4578,
+      "theoretical_loss": 3.4342815935217663,
+      "tokens_seen": 1973026816
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040905050505050505,
+      "loss": 2.7724,
+      "theoretical_loss": 3.4342720571492356,
+      "tokens_seen": 1973092352
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000409030303030303,
+      "loss": 2.7268,
+      "theoretical_loss": 3.434262521182135,
+      "tokens_seen": 1973157888
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000409010101010101,
+      "loss": 2.3653,
+      "theoretical_loss": 3.434252985620435,
+      "tokens_seen": 1973223424
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000408989898989899,
+      "loss": 2.8847,
+      "theoretical_loss": 3.4342434504641037,
+      "tokens_seen": 1973288960
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000408969696969697,
+      "loss": 2.5669,
+      "theoretical_loss": 3.4342339157131114,
+      "tokens_seen": 1973354496
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040894949494949494,
+      "loss": 2.6521,
+      "theoretical_loss": 3.434224381367426,
+      "tokens_seen": 1973420032
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040892929292929295,
+      "loss": 2.7036,
+      "theoretical_loss": 3.434214847427018,
+      "tokens_seen": 1973485568
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004089090909090909,
+      "loss": 2.5247,
+      "theoretical_loss": 3.434205313891857,
+      "tokens_seen": 1973551104
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004088888888888889,
+      "loss": 2.5833,
+      "theoretical_loss": 3.434195780761911,
+      "tokens_seen": 1973616640
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004088686868686869,
+      "loss": 2.6482,
+      "theoretical_loss": 3.4341862480371503,
+      "tokens_seen": 1973682176
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040884848484848483,
+      "loss": 2.8507,
+      "theoretical_loss": 3.434176715717544,
+      "tokens_seen": 1973747712
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040882828282828284,
+      "loss": 2.4306,
+      "theoretical_loss": 3.434167183803061,
+      "tokens_seen": 1973813248
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004088080808080808,
+      "loss": 2.6021,
+      "theoretical_loss": 3.4341576522936714,
+      "tokens_seen": 1973878784
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004087878787878788,
+      "loss": 2.7193,
+      "theoretical_loss": 3.4341481211893443,
+      "tokens_seen": 1973944320
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040876767676767677,
+      "loss": 2.6304,
+      "theoretical_loss": 3.4341385904900483,
+      "tokens_seen": 1974009856
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004087474747474748,
+      "loss": 2.5909,
+      "theoretical_loss": 3.4341290601957537,
+      "tokens_seen": 1974075392
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040872727272727274,
+      "loss": 2.6612,
+      "theoretical_loss": 3.4341195303064294,
+      "tokens_seen": 1974140928
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004087070707070707,
+      "loss": 2.5381,
+      "theoretical_loss": 3.4341100008220447,
+      "tokens_seen": 1974206464
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004086868686868687,
+      "loss": 2.8058,
+      "theoretical_loss": 3.4341004717425694,
+      "tokens_seen": 1974272000
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040866666666666666,
+      "loss": 2.653,
+      "theoretical_loss": 3.4340909430679725,
+      "tokens_seen": 1974337536
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040864646464646467,
+      "loss": 2.6647,
+      "theoretical_loss": 3.4340814147982233,
+      "tokens_seen": 1974403072
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040862626262626263,
+      "loss": 2.6635,
+      "theoretical_loss": 3.4340718869332907,
+      "tokens_seen": 1974468608
+    },
+    {
+      "epoch": 0.19,
+      "objective/train/docs_used": 1087285,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.9132044315338135,
+      "objective/train/theoretical_loss": 3.4340623594731454,
+      "objective/train/tokens_used": 333393376,
+      "theoretical_loss": 3.4340623594731454,
+      "tokens_seen": 1974534144
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004086060606060606,
+      "loss": 2.8076,
+      "theoretical_loss": 3.4340623594731454,
+      "tokens_seen": 1974534144
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004085858585858586,
+      "loss": 2.4489,
+      "theoretical_loss": 3.4340528324177555,
+      "tokens_seen": 1974599680
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004085656565656566,
+      "loss": 2.5364,
+      "theoretical_loss": 3.4340433057670907,
+      "tokens_seen": 1974665216
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040854545454545457,
+      "loss": 2.5742,
+      "theoretical_loss": 3.434033779521121,
+      "tokens_seen": 1974730752
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004085252525252525,
+      "loss": 2.5656,
+      "theoretical_loss": 3.434024253679815,
+      "tokens_seen": 1974796288
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004085050505050505,
+      "loss": 2.7908,
+      "theoretical_loss": 3.4340147282431426,
+      "tokens_seen": 1974861824
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004084848484848485,
+      "loss": 2.8222,
+      "theoretical_loss": 3.434005203211073,
+      "tokens_seen": 1974927360
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004084646464646465,
+      "loss": 2.6662,
+      "theoretical_loss": 3.4339956785835755,
+      "tokens_seen": 1974992896
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040844444444444446,
+      "loss": 2.673,
+      "theoretical_loss": 3.4339861543606194,
+      "tokens_seen": 1975058432
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004084242424242424,
+      "loss": 2.5198,
+      "theoretical_loss": 3.4339766305421744,
+      "tokens_seen": 1975123968
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040840404040404037,
+      "loss": 2.5674,
+      "theoretical_loss": 3.4339671071282094,
+      "tokens_seen": 1975189504
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040838383838383844,
+      "loss": 2.503,
+      "theoretical_loss": 3.433957584118695,
+      "tokens_seen": 1975255040
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004083636363636364,
+      "loss": 2.6536,
+      "theoretical_loss": 3.4339480615135987,
+      "tokens_seen": 1975320576
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040834343434343435,
+      "loss": 2.5693,
+      "theoretical_loss": 3.4339385393128916,
+      "tokens_seen": 1975386112
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004083232323232323,
+      "loss": 2.4123,
+      "theoretical_loss": 3.4339290175165424,
+      "tokens_seen": 1975451648
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040830303030303026,
+      "loss": 2.7539,
+      "theoretical_loss": 3.4339194961245205,
+      "tokens_seen": 1975517184
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040828282828282833,
+      "loss": 2.8219,
+      "theoretical_loss": 3.4339099751367956,
+      "tokens_seen": 1975582720
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004082626262626263,
+      "loss": 2.4702,
+      "theoretical_loss": 3.4339004545533367,
+      "tokens_seen": 1975648256
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040824242424242424,
+      "loss": 2.5782,
+      "theoretical_loss": 3.433890934374114,
+      "tokens_seen": 1975713792
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004082222222222222,
+      "loss": 2.6308,
+      "theoretical_loss": 3.433881414599096,
+      "tokens_seen": 1975779328
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004082020202020202,
+      "loss": 2.5198,
+      "theoretical_loss": 3.4338718952282523,
+      "tokens_seen": 1975844864
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004081818181818182,
+      "loss": 2.5395,
+      "theoretical_loss": 3.4338623762615526,
+      "tokens_seen": 1975910400
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004081616161616162,
+      "loss": 2.5631,
+      "theoretical_loss": 3.4338528576989664,
+      "tokens_seen": 1975975936
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040814141414141414,
+      "loss": 2.5006,
+      "theoretical_loss": 3.433843339540463,
+      "tokens_seen": 1976041472
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004081212121212121,
+      "loss": 2.6091,
+      "theoretical_loss": 3.4338338217860116,
+      "tokens_seen": 1976107008
+    },
+    {
+      "epoch": 0.19,
+      "objective/train/docs_used": 1088636,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.3727774620056152,
+      "objective/train/theoretical_loss": 3.4338243044355825,
+      "objective/train/tokens_used": 335031776,
+      "theoretical_loss": 3.4338243044355825,
+      "tokens_seen": 1976172544
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004081010101010101,
+      "loss": 2.5955,
+      "theoretical_loss": 3.4338243044355825,
+      "tokens_seen": 1976172544
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004080808080808081,
+      "loss": 2.664,
+      "theoretical_loss": 3.433814787489144,
+      "tokens_seen": 1976238080
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040806060606060607,
+      "loss": 2.7374,
+      "theoretical_loss": 3.4338052709466664,
+      "tokens_seen": 1976303616
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040804040404040403,
+      "loss": 2.6428,
+      "theoretical_loss": 3.433795754808119,
+      "tokens_seen": 1976369152
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040802020202020204,
+      "loss": 2.6984,
+      "theoretical_loss": 3.4337862390734717,
+      "tokens_seen": 1976434688
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000408,
+      "loss": 2.7755,
+      "theoretical_loss": 3.4337767237426924,
+      "tokens_seen": 1976500224
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000407979797979798,
+      "loss": 2.5779,
+      "theoretical_loss": 3.433767208815752,
+      "tokens_seen": 1976565760
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040795959595959596,
+      "loss": 2.5811,
+      "theoretical_loss": 3.43375769429262,
+      "tokens_seen": 1976631296
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004079393939393939,
+      "loss": 2.634,
+      "theoretical_loss": 3.4337481801732648,
+      "tokens_seen": 1976696832
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040791919191919193,
+      "loss": 2.7384,
+      "theoretical_loss": 3.4337386664576566,
+      "tokens_seen": 1976762368
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004078989898989899,
+      "loss": 2.517,
+      "theoretical_loss": 3.433729153145765,
+      "tokens_seen": 1976827904
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004078787878787879,
+      "loss": 2.4324,
+      "theoretical_loss": 3.433719640237559,
+      "tokens_seen": 1976893440
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040785858585858586,
+      "loss": 2.6609,
+      "theoretical_loss": 3.4337101277330087,
+      "tokens_seen": 1976958976
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040783838383838387,
+      "loss": 2.5041,
+      "theoretical_loss": 3.433700615632083,
+      "tokens_seen": 1977024512
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004078181818181818,
+      "loss": 2.4608,
+      "theoretical_loss": 3.433691103934752,
+      "tokens_seen": 1977090048
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004077979797979798,
+      "loss": 2.6694,
+      "theoretical_loss": 3.4336815926409843,
+      "tokens_seen": 1977155584
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004077777777777778,
+      "loss": 2.6661,
+      "theoretical_loss": 3.4336720817507507,
+      "tokens_seen": 1977221120
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040775757575757575,
+      "loss": 2.6757,
+      "theoretical_loss": 3.433662571264019,
+      "tokens_seen": 1977286656
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040773737373737376,
+      "loss": 2.6244,
+      "theoretical_loss": 3.4336530611807605,
+      "tokens_seen": 1977352192
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004077171717171717,
+      "loss": 2.7615,
+      "theoretical_loss": 3.4336435515009436,
+      "tokens_seen": 1977417728
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040769696969696973,
+      "loss": 2.8321,
+      "theoretical_loss": 3.433634042224538,
+      "tokens_seen": 1977483264
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004076767676767677,
+      "loss": 2.6079,
+      "theoretical_loss": 3.433624533351513,
+      "tokens_seen": 1977548800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004076565656565657,
+      "loss": 2.5141,
+      "theoretical_loss": 3.4336150248818393,
+      "tokens_seen": 1977614336
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040763636363636365,
+      "loss": 2.5957,
+      "theoretical_loss": 3.4336055168154846,
+      "tokens_seen": 1977679872
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004076161616161616,
+      "loss": 2.5912,
+      "theoretical_loss": 3.43359600915242,
+      "tokens_seen": 1977745408
+    },
+    {
+      "epoch": 0.19,
+      "objective/train/docs_used": 1089194,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.324873208999634,
+      "objective/train/theoretical_loss": 3.4335865018926146,
+      "objective/train/tokens_used": 336670176,
+      "theoretical_loss": 3.4335865018926146,
+      "tokens_seen": 1977810944
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004075959595959596,
+      "loss": 2.7609,
+      "theoretical_loss": 3.4335865018926146,
+      "tokens_seen": 1977810944
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004075757575757576,
+      "loss": 2.6968,
+      "theoretical_loss": 3.433576995036037,
+      "tokens_seen": 1977876480
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004075555555555556,
+      "loss": 2.7448,
+      "theoretical_loss": 3.433567488582658,
+      "tokens_seen": 1977942016
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040753535353535355,
+      "loss": 2.5191,
+      "theoretical_loss": 3.433557982532447,
+      "tokens_seen": 1978007552
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004075151515151515,
+      "loss": 2.62,
+      "theoretical_loss": 3.433548476885372,
+      "tokens_seen": 1978073088
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004074949494949495,
+      "loss": 2.632,
+      "theoretical_loss": 3.433538971641405,
+      "tokens_seen": 1978138624
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040747474747474747,
+      "loss": 2.6667,
+      "theoretical_loss": 3.433529466800514,
+      "tokens_seen": 1978204160
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004074545454545455,
+      "loss": 2.6104,
+      "theoretical_loss": 3.433519962362668,
+      "tokens_seen": 1978269696
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040743434343434344,
+      "loss": 2.6996,
+      "theoretical_loss": 3.433510458327838,
+      "tokens_seen": 1978335232
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004074141414141414,
+      "loss": 2.6062,
+      "theoretical_loss": 3.433500954695993,
+      "tokens_seen": 1978400768
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004073939393939394,
+      "loss": 2.5305,
+      "theoretical_loss": 3.4334914514671024,
+      "tokens_seen": 1978466304
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004073737373737374,
+      "loss": 2.5134,
+      "theoretical_loss": 3.433481948641136,
+      "tokens_seen": 1978531840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004073535353535354,
+      "loss": 2.7429,
+      "theoretical_loss": 3.433472446218063,
+      "tokens_seen": 1978597376
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040733333333333333,
+      "loss": 2.6326,
+      "theoretical_loss": 3.4334629441978537,
+      "tokens_seen": 1978662912
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004073131313131313,
+      "loss": 2.7723,
+      "theoretical_loss": 3.4334534425804764,
+      "tokens_seen": 1978728448
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004072929292929293,
+      "loss": 2.8548,
+      "theoretical_loss": 3.433443941365902,
+      "tokens_seen": 1978793984
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004072727272727273,
+      "loss": 2.289,
+      "theoretical_loss": 3.4334344405541,
+      "tokens_seen": 1978859520
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040725252525252527,
+      "loss": 2.4712,
+      "theoretical_loss": 3.4334249401450387,
+      "tokens_seen": 1978925056
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004072323232323232,
+      "loss": 2.5277,
+      "theoretical_loss": 3.433415440138689,
+      "tokens_seen": 1978990592
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004072121212121212,
+      "loss": 2.5607,
+      "theoretical_loss": 3.43340594053502,
+      "tokens_seen": 1979056128
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040719191919191925,
+      "loss": 2.6904,
+      "theoretical_loss": 3.4333964413340015,
+      "tokens_seen": 1979121664
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004071717171717172,
+      "loss": 2.6751,
+      "theoretical_loss": 3.4333869425356025,
+      "tokens_seen": 1979187200
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040715151515151516,
+      "loss": 2.2723,
+      "theoretical_loss": 3.4333774441397935,
+      "tokens_seen": 1979252736
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004071313131313131,
+      "loss": 2.6722,
+      "theoretical_loss": 3.4333679461465434,
+      "tokens_seen": 1979318272
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004071111111111111,
+      "loss": 2.723,
+      "theoretical_loss": 3.433358448555822,
+      "tokens_seen": 1979383808
+    },
+    {
+      "epoch": 0.19,
+      "objective/train/docs_used": 1090532,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.0842976570129395,
+      "objective/train/theoretical_loss": 3.4333489513675994,
+      "objective/train/tokens_used": 338308576,
+      "theoretical_loss": 3.4333489513675994,
+      "tokens_seen": 1979449344
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040709090909090914,
+      "loss": 2.5824,
+      "theoretical_loss": 3.4333489513675994,
+      "tokens_seen": 1979449344
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004070707070707071,
+      "loss": 2.7518,
+      "theoretical_loss": 3.4333394545818443,
+      "tokens_seen": 1979514880
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040705050505050505,
+      "loss": 2.7846,
+      "theoretical_loss": 3.4333299581985273,
+      "tokens_seen": 1979580416
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000407030303030303,
+      "loss": 2.6402,
+      "theoretical_loss": 3.433320462217617,
+      "tokens_seen": 1979645952
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000407010101010101,
+      "loss": 2.6241,
+      "theoretical_loss": 3.433310966639084,
+      "tokens_seen": 1979711488
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040698989898989903,
+      "loss": 2.7387,
+      "theoretical_loss": 3.4333014714628973,
+      "tokens_seen": 1979777024
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000406969696969697,
+      "loss": 2.5149,
+      "theoretical_loss": 3.433291976689027,
+      "tokens_seen": 1979842560
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040694949494949495,
+      "loss": 2.7728,
+      "theoretical_loss": 3.433282482317442,
+      "tokens_seen": 1979908096
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004069292929292929,
+      "loss": 2.5955,
+      "theoretical_loss": 3.4332729883481123,
+      "tokens_seen": 1979973632
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004069090909090909,
+      "loss": 2.6294,
+      "theoretical_loss": 3.4332634947810083,
+      "tokens_seen": 1980039168
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004068888888888889,
+      "loss": 2.6673,
+      "theoretical_loss": 3.4332540016160986,
+      "tokens_seen": 1980104704
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004068686868686869,
+      "loss": 2.6137,
+      "theoretical_loss": 3.4332445088533534,
+      "tokens_seen": 1980170240
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040684848484848484,
+      "loss": 2.6255,
+      "theoretical_loss": 3.433235016492742,
+      "tokens_seen": 1980235776
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040682828282828285,
+      "loss": 2.3272,
+      "theoretical_loss": 3.433225524534235,
+      "tokens_seen": 1980301312
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004068080808080808,
+      "loss": 2.7693,
+      "theoretical_loss": 3.4332160329778003,
+      "tokens_seen": 1980366848
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004067878787878788,
+      "loss": 2.7496,
+      "theoretical_loss": 3.433206541823409,
+      "tokens_seen": 1980432384
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004067676767676768,
+      "loss": 2.5695,
+      "theoretical_loss": 3.4331970510710303,
+      "tokens_seen": 1980497920
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040674747474747473,
+      "loss": 2.5576,
+      "theoretical_loss": 3.433187560720634,
+      "tokens_seen": 1980563456
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040672727272727274,
+      "loss": 2.7761,
+      "theoretical_loss": 3.4331780707721897,
+      "tokens_seen": 1980628992
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004067070707070707,
+      "loss": 2.7222,
+      "theoretical_loss": 3.433168581225667,
+      "tokens_seen": 1980694528
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004066868686868687,
+      "loss": 2.6214,
+      "theoretical_loss": 3.433159092081036,
+      "tokens_seen": 1980760064
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040666666666666667,
+      "loss": 2.6567,
+      "theoretical_loss": 3.433149603338266,
+      "tokens_seen": 1980825600
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004066464646464647,
+      "loss": 2.5001,
+      "theoretical_loss": 3.433140114997326,
+      "tokens_seen": 1980891136
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00040662626262626264,
+      "loss": 2.5997,
+      "theoretical_loss": 3.433130627058187,
+      "tokens_seen": 1980956672
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0004066060606060606,
+      "loss": 2.5609,
+      "theoretical_loss": 3.433121139520818,
+      "tokens_seen": 1981022208
+    },
+    {
+      "epoch": 0.19,
+      "objective/train/docs_used": 1091139,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.3586647510528564,
+      "objective/train/theoretical_loss": 3.433111652385189,
+      "objective/train/tokens_used": 339946976,
+      "theoretical_loss": 3.433111652385189,
+      "tokens_seen": 1981087744
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004065858585858586,
+      "loss": 2.4624,
+      "theoretical_loss": 3.433111652385189,
+      "tokens_seen": 1981087744
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040656565656565656,
+      "loss": 2.6518,
+      "theoretical_loss": 3.4331021656512695,
+      "tokens_seen": 1981153280
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040654545454545457,
+      "loss": 2.4553,
+      "theoretical_loss": 3.433092679319029,
+      "tokens_seen": 1981218816
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040652525252525253,
+      "loss": 2.6482,
+      "theoretical_loss": 3.4330831933884367,
+      "tokens_seen": 1981284352
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004065050505050505,
+      "loss": 2.6432,
+      "theoretical_loss": 3.4330737078594638,
+      "tokens_seen": 1981349888
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004064848484848485,
+      "loss": 2.4884,
+      "theoretical_loss": 3.4330642227320793,
+      "tokens_seen": 1981415424
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004064646464646465,
+      "loss": 2.748,
+      "theoretical_loss": 3.433054738006253,
+      "tokens_seen": 1981480960
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040644444444444446,
+      "loss": 2.5247,
+      "theoretical_loss": 3.433045253681954,
+      "tokens_seen": 1981546496
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004064242424242424,
+      "loss": 2.5808,
+      "theoretical_loss": 3.4330357697591527,
+      "tokens_seen": 1981612032
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004064040404040404,
+      "loss": 2.5666,
+      "theoretical_loss": 3.4330262862378182,
+      "tokens_seen": 1981677568
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004063838383838384,
+      "loss": 2.5371,
+      "theoretical_loss": 3.4330168031179213,
+      "tokens_seen": 1981743104
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004063636363636364,
+      "loss": 2.6204,
+      "theoretical_loss": 3.433007320399431,
+      "tokens_seen": 1981808640
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040634343434343436,
+      "loss": 2.6785,
+      "theoretical_loss": 3.4329978380823167,
+      "tokens_seen": 1981874176
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004063232323232323,
+      "loss": 2.4334,
+      "theoretical_loss": 3.4329883561665486,
+      "tokens_seen": 1981939712
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040630303030303027,
+      "loss": 2.8279,
+      "theoretical_loss": 3.432978874652097,
+      "tokens_seen": 1982005248
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040628282828282834,
+      "loss": 2.5103,
+      "theoretical_loss": 3.4329693935389303,
+      "tokens_seen": 1982070784
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004062626262626263,
+      "loss": 2.7507,
+      "theoretical_loss": 3.4329599128270196,
+      "tokens_seen": 1982136320
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040624242424242425,
+      "loss": 2.8613,
+      "theoretical_loss": 3.432950432516334,
+      "tokens_seen": 1982201856
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004062222222222222,
+      "loss": 2.3773,
+      "theoretical_loss": 3.4329409526068426,
+      "tokens_seen": 1982267392
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040620202020202016,
+      "loss": 2.5538,
+      "theoretical_loss": 3.4329314730985163,
+      "tokens_seen": 1982332928
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040618181818181823,
+      "loss": 2.7964,
+      "theoretical_loss": 3.4329219939913247,
+      "tokens_seen": 1982398464
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004061616161616162,
+      "loss": 2.7069,
+      "theoretical_loss": 3.4329125152852367,
+      "tokens_seen": 1982464000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040614141414141414,
+      "loss": 2.6791,
+      "theoretical_loss": 3.432903036980223,
+      "tokens_seen": 1982529536
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004061212121212121,
+      "loss": 2.5445,
+      "theoretical_loss": 3.4328935590762533,
+      "tokens_seen": 1982595072
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040610101010101006,
+      "loss": 2.6502,
+      "theoretical_loss": 3.4328840815732966,
+      "tokens_seen": 1982660608
+    },
+    {
+      "epoch": 0.2,
+      "objective/train/docs_used": 1091959,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.3474152088165283,
+      "objective/train/theoretical_loss": 3.4328746044713236,
+      "objective/train/tokens_used": 341585376,
+      "theoretical_loss": 3.4328746044713236,
+      "tokens_seen": 1982726144
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004060808080808081,
+      "loss": 2.7832,
+      "theoretical_loss": 3.4328746044713236,
+      "tokens_seen": 1982726144
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004060606060606061,
+      "loss": 2.7179,
+      "theoretical_loss": 3.4328651277703033,
+      "tokens_seen": 1982791680
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040604040404040403,
+      "loss": 2.835,
+      "theoretical_loss": 3.4328556514702058,
+      "tokens_seen": 1982857216
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000406020202020202,
+      "loss": 2.4766,
+      "theoretical_loss": 3.432846175571001,
+      "tokens_seen": 1982922752
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040600000000000006,
+      "loss": 2.4752,
+      "theoretical_loss": 3.432836700072659,
+      "tokens_seen": 1982988288
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000405979797979798,
+      "loss": 2.816,
+      "theoretical_loss": 3.4328272249751492,
+      "tokens_seen": 1983053824
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040595959595959597,
+      "loss": 2.5177,
+      "theoretical_loss": 3.432817750278441,
+      "tokens_seen": 1983119360
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040593939393939393,
+      "loss": 2.6334,
+      "theoretical_loss": 3.432808275982505,
+      "tokens_seen": 1983184896
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004059191919191919,
+      "loss": 2.5028,
+      "theoretical_loss": 3.4327988020873104,
+      "tokens_seen": 1983250432
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040589898989898995,
+      "loss": 2.5369,
+      "theoretical_loss": 3.4327893285928273,
+      "tokens_seen": 1983315968
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004058787878787879,
+      "loss": 2.7305,
+      "theoretical_loss": 3.4327798554990254,
+      "tokens_seen": 1983381504
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040585858585858586,
+      "loss": 2.7252,
+      "theoretical_loss": 3.432770382805874,
+      "tokens_seen": 1983447040
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004058383838383838,
+      "loss": 2.6841,
+      "theoretical_loss": 3.4327609105133443,
+      "tokens_seen": 1983512576
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040581818181818183,
+      "loss": 2.7614,
+      "theoretical_loss": 3.432751438621405,
+      "tokens_seen": 1983578112
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040579797979797984,
+      "loss": 2.857,
+      "theoretical_loss": 3.4327419671300263,
+      "tokens_seen": 1983643648
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004057777777777778,
+      "loss": 2.7226,
+      "theoretical_loss": 3.4327324960391783,
+      "tokens_seen": 1983709184
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040575757575757576,
+      "loss": 2.3924,
+      "theoretical_loss": 3.43272302534883,
+      "tokens_seen": 1983774720
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004057373737373737,
+      "loss": 2.7689,
+      "theoretical_loss": 3.4327135550589514,
+      "tokens_seen": 1983840256
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004057171717171717,
+      "loss": 2.5132,
+      "theoretical_loss": 3.4327040851695134,
+      "tokens_seen": 1983905792
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040569696969696973,
+      "loss": 2.5889,
+      "theoretical_loss": 3.432694615680485,
+      "tokens_seen": 1983971328
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004056767676767677,
+      "loss": 2.7441,
+      "theoretical_loss": 3.4326851465918358,
+      "tokens_seen": 1984036864
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040565656565656565,
+      "loss": 2.6829,
+      "theoretical_loss": 3.4326756779035357,
+      "tokens_seen": 1984102400
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040563636363636366,
+      "loss": 2.701,
+      "theoretical_loss": 3.4326662096155554,
+      "tokens_seen": 1984167936
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004056161616161616,
+      "loss": 2.8375,
+      "theoretical_loss": 3.4326567417278637,
+      "tokens_seen": 1984233472
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040559595959595963,
+      "loss": 2.5529,
+      "theoretical_loss": 3.4326472742404315,
+      "tokens_seen": 1984299008
+    },
+    {
+      "epoch": 0.2,
+      "objective/train/docs_used": 1092471,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.014146327972412,
+      "objective/train/theoretical_loss": 3.432637807153228,
+      "objective/train/tokens_used": 343223776,
+      "theoretical_loss": 3.432637807153228,
+      "tokens_seen": 1984364544
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004055757575757576,
+      "loss": 2.7457,
+      "theoretical_loss": 3.432637807153228,
+      "tokens_seen": 1984364544
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040555555555555554,
+      "loss": 2.7626,
+      "theoretical_loss": 3.4326283404662226,
+      "tokens_seen": 1984430080
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040553535353535355,
+      "loss": 2.7682,
+      "theoretical_loss": 3.4326188741793864,
+      "tokens_seen": 1984495616
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004055151515151515,
+      "loss": 2.6067,
+      "theoretical_loss": 3.432609408292689,
+      "tokens_seen": 1984561152
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004054949494949495,
+      "loss": 2.6628,
+      "theoretical_loss": 3.4325999428060987,
+      "tokens_seen": 1984626688
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004054747474747475,
+      "loss": 2.8009,
+      "theoretical_loss": 3.432590477719587,
+      "tokens_seen": 1984692224
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004054545454545455,
+      "loss": 2.7039,
+      "theoretical_loss": 3.432581013033124,
+      "tokens_seen": 1984757760
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040543434343434344,
+      "loss": 2.6553,
+      "theoretical_loss": 3.4325715487466777,
+      "tokens_seen": 1984823296
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004054141414141414,
+      "loss": 2.5869,
+      "theoretical_loss": 3.43256208486022,
+      "tokens_seen": 1984888832
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004053939393939394,
+      "loss": 2.6068,
+      "theoretical_loss": 3.43255262137372,
+      "tokens_seen": 1984954368
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040537373737373737,
+      "loss": 2.5967,
+      "theoretical_loss": 3.4325431582871473,
+      "tokens_seen": 1985019904
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004053535353535354,
+      "loss": 2.4943,
+      "theoretical_loss": 3.432533695600472,
+      "tokens_seen": 1985085440
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040533333333333334,
+      "loss": 2.9349,
+      "theoretical_loss": 3.4325242333136643,
+      "tokens_seen": 1985150976
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004053131313131313,
+      "loss": 2.7874,
+      "theoretical_loss": 3.4325147714266944,
+      "tokens_seen": 1985216512
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004052929292929293,
+      "loss": 2.7921,
+      "theoretical_loss": 3.432505309939531,
+      "tokens_seen": 1985282048
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004052727272727273,
+      "loss": 2.419,
+      "theoretical_loss": 3.432495848852145,
+      "tokens_seen": 1985347584
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004052525252525253,
+      "loss": 2.7875,
+      "theoretical_loss": 3.432486388164506,
+      "tokens_seen": 1985413120
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040523232323232323,
+      "loss": 2.7653,
+      "theoretical_loss": 3.4324769278765834,
+      "tokens_seen": 1985478656
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004052121212121212,
+      "loss": 2.9169,
+      "theoretical_loss": 3.432467467988348,
+      "tokens_seen": 1985544192
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004051919191919192,
+      "loss": 2.6095,
+      "theoretical_loss": 3.4324580084997693,
+      "tokens_seen": 1985609728
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004051717171717172,
+      "loss": 2.718,
+      "theoretical_loss": 3.4324485494108177,
+      "tokens_seen": 1985675264
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040515151515151517,
+      "loss": 2.6026,
+      "theoretical_loss": 3.4324390907214624,
+      "tokens_seen": 1985740800
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004051313131313131,
+      "loss": 2.6342,
+      "theoretical_loss": 3.4324296324316736,
+      "tokens_seen": 1985806336
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004051111111111111,
+      "loss": 2.5302,
+      "theoretical_loss": 3.432420174541422,
+      "tokens_seen": 1985871872
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040509090909090915,
+      "loss": 2.8093,
+      "theoretical_loss": 3.4324107170506757,
+      "tokens_seen": 1985937408
+    },
+    {
+      "epoch": 0.2,
+      "objective/train/docs_used": 1093318,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.581864595413208,
+      "objective/train/theoretical_loss": 3.4324012599594065,
+      "objective/train/tokens_used": 344862176,
+      "theoretical_loss": 3.4324012599594065,
+      "tokens_seen": 1986002944
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004050707070707071,
+      "loss": 2.4938,
+      "theoretical_loss": 3.4324012599594065,
+      "tokens_seen": 1986002944
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040505050505050506,
+      "loss": 2.6114,
+      "theoretical_loss": 3.432391803267583,
+      "tokens_seen": 1986068480
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000405030303030303,
+      "loss": 2.6796,
+      "theoretical_loss": 3.4323823469751766,
+      "tokens_seen": 1986134016
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040501010101010097,
+      "loss": 2.6682,
+      "theoretical_loss": 3.432372891082156,
+      "tokens_seen": 1986199552
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040498989898989904,
+      "loss": 2.6571,
+      "theoretical_loss": 3.4323634355884916,
+      "tokens_seen": 1986265088
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000404969696969697,
+      "loss": 2.5284,
+      "theoretical_loss": 3.4323539804941534,
+      "tokens_seen": 1986330624
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040494949494949495,
+      "loss": 2.685,
+      "theoretical_loss": 3.432344525799111,
+      "tokens_seen": 1986396160
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004049292929292929,
+      "loss": 2.7383,
+      "theoretical_loss": 3.432335071503335,
+      "tokens_seen": 1986461696
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004049090909090909,
+      "loss": 2.6443,
+      "theoretical_loss": 3.4323256176067947,
+      "tokens_seen": 1986527232
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040488888888888893,
+      "loss": 2.7833,
+      "theoretical_loss": 3.4323161641094604,
+      "tokens_seen": 1986592768
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004048686868686869,
+      "loss": 2.7787,
+      "theoretical_loss": 3.4323067110113024,
+      "tokens_seen": 1986658304
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040484848484848484,
+      "loss": 2.7694,
+      "theoretical_loss": 3.4322972583122904,
+      "tokens_seen": 1986723840
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004048282828282828,
+      "loss": 2.6806,
+      "theoretical_loss": 3.4322878060123942,
+      "tokens_seen": 1986789376
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004048080808080808,
+      "loss": 2.8119,
+      "theoretical_loss": 3.4322783541115838,
+      "tokens_seen": 1986854912
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004047878787878788,
+      "loss": 2.59,
+      "theoretical_loss": 3.432268902609829,
+      "tokens_seen": 1986920448
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004047676767676768,
+      "loss": 2.721,
+      "theoretical_loss": 3.4322594515071003,
+      "tokens_seen": 1986985984
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040474747474747474,
+      "loss": 2.5832,
+      "theoretical_loss": 3.432250000803368,
+      "tokens_seen": 1987051520
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040472727272727275,
+      "loss": 2.7105,
+      "theoretical_loss": 3.4322405504986007,
+      "tokens_seen": 1987117056
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004047070707070707,
+      "loss": 2.779,
+      "theoretical_loss": 3.4322311005927704,
+      "tokens_seen": 1987182592
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004046868686868687,
+      "loss": 2.8089,
+      "theoretical_loss": 3.432221651085845,
+      "tokens_seen": 1987248128
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040466666666666667,
+      "loss": 2.6261,
+      "theoretical_loss": 3.432212201977796,
+      "tokens_seen": 1987313664
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040464646464646463,
+      "loss": 2.5074,
+      "theoretical_loss": 3.4322027532685926,
+      "tokens_seen": 1987379200
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040462626262626264,
+      "loss": 2.4476,
+      "theoretical_loss": 3.432193304958205,
+      "tokens_seen": 1987444736
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004046060606060606,
+      "loss": 2.7534,
+      "theoretical_loss": 3.4321838570466037,
+      "tokens_seen": 1987510272
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004045858585858586,
+      "loss": 2.6793,
+      "theoretical_loss": 3.432174409533758,
+      "tokens_seen": 1987575808
+    },
+    {
+      "epoch": 0.2,
+      "objective/train/docs_used": 1094373,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.851362943649292,
+      "objective/train/theoretical_loss": 3.4321649624196384,
+      "objective/train/tokens_used": 346500576,
+      "theoretical_loss": 3.4321649624196384,
+      "tokens_seen": 1987641344
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040456565656565657,
+      "loss": 2.5657,
+      "theoretical_loss": 3.4321649624196384,
+      "tokens_seen": 1987641344
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004045454545454546,
+      "loss": 2.3815,
+      "theoretical_loss": 3.432155515704215,
+      "tokens_seen": 1987706880
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040452525252525253,
+      "loss": 2.7743,
+      "theoretical_loss": 3.432146069387458,
+      "tokens_seen": 1987772416
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004045050505050505,
+      "loss": 2.7407,
+      "theoretical_loss": 3.4321366234693365,
+      "tokens_seen": 1987837952
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004044848484848485,
+      "loss": 2.5287,
+      "theoretical_loss": 3.432127177949821,
+      "tokens_seen": 1987903488
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040446464646464646,
+      "loss": 2.6323,
+      "theoretical_loss": 3.432117732828882,
+      "tokens_seen": 1987969024
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040444444444444447,
+      "loss": 2.6108,
+      "theoretical_loss": 3.4321082881064884,
+      "tokens_seen": 1988034560
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004044242424242424,
+      "loss": 2.4885,
+      "theoretical_loss": 3.4320988437826117,
+      "tokens_seen": 1988100096
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004044040404040404,
+      "loss": 2.5306,
+      "theoretical_loss": 3.4320893998572215,
+      "tokens_seen": 1988165632
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004043838383838384,
+      "loss": 2.5742,
+      "theoretical_loss": 3.432079956330287,
+      "tokens_seen": 1988231168
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040436363636363635,
+      "loss": 2.7162,
+      "theoretical_loss": 3.4320705132017792,
+      "tokens_seen": 1988296704
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040434343434343436,
+      "loss": 2.7681,
+      "theoretical_loss": 3.4320610704716676,
+      "tokens_seen": 1988362240
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004043232323232323,
+      "loss": 2.8476,
+      "theoretical_loss": 3.4320516281399227,
+      "tokens_seen": 1988427776
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040430303030303033,
+      "loss": 2.7233,
+      "theoretical_loss": 3.4320421862065147,
+      "tokens_seen": 1988493312
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004042828282828283,
+      "loss": 2.949,
+      "theoretical_loss": 3.432032744671413,
+      "tokens_seen": 1988558848
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004042626262626263,
+      "loss": 2.7295,
+      "theoretical_loss": 3.4320233035345877,
+      "tokens_seen": 1988624384
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040424242424242425,
+      "loss": 2.7111,
+      "theoretical_loss": 3.4320138627960097,
+      "tokens_seen": 1988689920
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004042222222222222,
+      "loss": 2.7302,
+      "theoretical_loss": 3.4320044224556483,
+      "tokens_seen": 1988755456
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004042020202020202,
+      "loss": 2.7109,
+      "theoretical_loss": 3.4319949825134737,
+      "tokens_seen": 1988820992
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004041818181818182,
+      "loss": 2.9131,
+      "theoretical_loss": 3.4319855429694566,
+      "tokens_seen": 1988886528
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004041616161616162,
+      "loss": 2.6445,
+      "theoretical_loss": 3.4319761038235663,
+      "tokens_seen": 1988952064
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040414141414141415,
+      "loss": 2.6379,
+      "theoretical_loss": 3.431966665075773,
+      "tokens_seen": 1989017600
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004041212121212121,
+      "loss": 2.6443,
+      "theoretical_loss": 3.4319572267260474,
+      "tokens_seen": 1989083136
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004041010101010101,
+      "loss": 2.6639,
+      "theoretical_loss": 3.431947788774359,
+      "tokens_seen": 1989148672
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004040808080808081,
+      "loss": 2.9648,
+      "theoretical_loss": 3.431938351220678,
+      "tokens_seen": 1989214208
+    },
+    {
+      "debugging/Self-BLEU-5": 0.4028481195939984,
+      "debugging/distinct-1-grams": 0.8260221595150598,
+      "debugging/distinct-2-grams": 0.9811327641385946,
+      "debugging/entropy-1-grams": 5.420510909354855,
+      "debugging/entropy-2-grams": 5.998582774683988,
+      "debugging/length": 472.57142857142856,
+      "debugging/num_segments": 7,
+      "epoch": 0.2,
+      "objective/train/docs_used": 1094923,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.8087246417999268,
+      "objective/train/theoretical_loss": 3.431928914064975,
+      "objective/train/tokens_used": 348138976,
+      "theoretical_loss": 3.431928914064975,
+      "tokens_seen": 1989279744
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004040606060606061,
+      "loss": 2.7047,
+      "theoretical_loss": 3.431928914064975,
+      "tokens_seen": 1989279744
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040404040404040404,
+      "loss": 2.6149,
+      "theoretical_loss": 3.4319194773072192,
+      "tokens_seen": 1989345280
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000404020202020202,
+      "loss": 2.6278,
+      "theoretical_loss": 3.4319100409473813,
+      "tokens_seen": 1989410816
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000404,
+      "loss": 2.6593,
+      "theoretical_loss": 3.4319006049854313,
+      "tokens_seen": 1989476352
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000403979797979798,
+      "loss": 2.6405,
+      "theoretical_loss": 3.4318911694213394,
+      "tokens_seen": 1989541888
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000403959595959596,
+      "loss": 2.745,
+      "theoretical_loss": 3.431881734255076,
+      "tokens_seen": 1989607424
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040393939393939393,
+      "loss": 2.841,
+      "theoretical_loss": 3.431872299486611,
+      "tokens_seen": 1989672960
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004039191919191919,
+      "loss": 2.6359,
+      "theoretical_loss": 3.4318628651159138,
+      "tokens_seen": 1989738496
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040389898989898995,
+      "loss": 2.7627,
+      "theoretical_loss": 3.4318534311429554,
+      "tokens_seen": 1989804032
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004038787878787879,
+      "loss": 2.4968,
+      "theoretical_loss": 3.4318439975677055,
+      "tokens_seen": 1989869568
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040385858585858587,
+      "loss": 2.5069,
+      "theoretical_loss": 3.431834564390135,
+      "tokens_seen": 1989935104
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004038383838383838,
+      "loss": 2.7199,
+      "theoretical_loss": 3.4318251316102124,
+      "tokens_seen": 1990000640
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004038181818181818,
+      "loss": 2.3488,
+      "theoretical_loss": 3.4318156992279096,
+      "tokens_seen": 1990066176
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040379797979797985,
+      "loss": 2.7606,
+      "theoretical_loss": 3.4318062672431964,
+      "tokens_seen": 1990131712
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004037777777777778,
+      "loss": 2.6489,
+      "theoretical_loss": 3.431796835656042,
+      "tokens_seen": 1990197248
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040375757575757576,
+      "loss": 2.8979,
+      "theoretical_loss": 3.431787404466417,
+      "tokens_seen": 1990262784
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004037373737373737,
+      "loss": 2.6212,
+      "theoretical_loss": 3.431777973674292,
+      "tokens_seen": 1990328320
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040371717171717173,
+      "loss": 2.8688,
+      "theoretical_loss": 3.431768543279637,
+      "tokens_seen": 1990393856
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040369696969696974,
+      "loss": 2.858,
+      "theoretical_loss": 3.431759113282422,
+      "tokens_seen": 1990459392
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004036767676767677,
+      "loss": 2.5182,
+      "theoretical_loss": 3.431749683682617,
+      "tokens_seen": 1990524928
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040365656565656565,
+      "loss": 2.5937,
+      "theoretical_loss": 3.4317402544801925,
+      "tokens_seen": 1990590464
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004036363636363636,
+      "loss": 2.6897,
+      "theoretical_loss": 3.4317308256751184,
+      "tokens_seen": 1990656000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004036161616161616,
+      "loss": 2.7109,
+      "theoretical_loss": 3.4317213972673652,
+      "tokens_seen": 1990721536
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040359595959595963,
+      "loss": 2.7822,
+      "theoretical_loss": 3.4317119692569023,
+      "tokens_seen": 1990787072
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004035757575757576,
+      "loss": 2.7954,
+      "theoretical_loss": 3.431702541643701,
+      "tokens_seen": 1990852608
+    },
+    {
+      "epoch": 0.2,
+      "objective/train/docs_used": 1096094,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.850614547729492,
+      "objective/train/theoretical_loss": 3.431693114427731,
+      "objective/train/tokens_used": 349777376,
+      "theoretical_loss": 3.431693114427731,
+      "tokens_seen": 1990918144
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040355555555555555,
+      "loss": 2.7385,
+      "theoretical_loss": 3.431693114427731,
+      "tokens_seen": 1990918144
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040353535353535356,
+      "loss": 2.9012,
+      "theoretical_loss": 3.4316836876089623,
+      "tokens_seen": 1990983680
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004035151515151515,
+      "loss": 2.6501,
+      "theoretical_loss": 3.431674261187365,
+      "tokens_seen": 1991049216
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004034949494949495,
+      "loss": 2.6113,
+      "theoretical_loss": 3.4316648351629095,
+      "tokens_seen": 1991114752
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004034747474747475,
+      "loss": 2.6352,
+      "theoretical_loss": 3.4316554095355665,
+      "tokens_seen": 1991180288
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040345454545454544,
+      "loss": 2.5908,
+      "theoretical_loss": 3.4316459843053053,
+      "tokens_seen": 1991245824
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040343434343434345,
+      "loss": 2.466,
+      "theoretical_loss": 3.4316365594720963,
+      "tokens_seen": 1991311360
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004034141414141414,
+      "loss": 2.9378,
+      "theoretical_loss": 3.43162713503591,
+      "tokens_seen": 1991376896
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004033939393939394,
+      "loss": 2.8731,
+      "theoretical_loss": 3.431617710996717,
+      "tokens_seen": 1991442432
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004033737373737374,
+      "loss": 2.5793,
+      "theoretical_loss": 3.431608287354487,
+      "tokens_seen": 1991507968
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004033535353535354,
+      "loss": 2.8316,
+      "theoretical_loss": 3.4315988641091906,
+      "tokens_seen": 1991573504
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040333333333333334,
+      "loss": 2.7059,
+      "theoretical_loss": 3.4315894412607966,
+      "tokens_seen": 1991639040
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004033131313131313,
+      "loss": 2.9928,
+      "theoretical_loss": 3.431580018809277,
+      "tokens_seen": 1991704576
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004032929292929293,
+      "loss": 2.709,
+      "theoretical_loss": 3.431570596754601,
+      "tokens_seen": 1991770112
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040327272727272727,
+      "loss": 2.7009,
+      "theoretical_loss": 3.4315611750967396,
+      "tokens_seen": 1991835648
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004032525252525253,
+      "loss": 2.4751,
+      "theoretical_loss": 3.431551753835662,
+      "tokens_seen": 1991901184
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040323232323232324,
+      "loss": 2.5752,
+      "theoretical_loss": 3.4315423329713397,
+      "tokens_seen": 1991966720
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004032121212121212,
+      "loss": 2.6393,
+      "theoretical_loss": 3.4315329125037417,
+      "tokens_seen": 1992032256
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004031919191919192,
+      "loss": 2.8421,
+      "theoretical_loss": 3.4315234924328393,
+      "tokens_seen": 1992097792
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004031717171717172,
+      "loss": 2.65,
+      "theoretical_loss": 3.431514072758602,
+      "tokens_seen": 1992163328
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040315151515151517,
+      "loss": 2.6166,
+      "theoretical_loss": 3.4315046534810003,
+      "tokens_seen": 1992228864
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040313131313131313,
+      "loss": 2.6158,
+      "theoretical_loss": 3.4314952346000047,
+      "tokens_seen": 1992294400
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004031111111111111,
+      "loss": 2.7504,
+      "theoretical_loss": 3.4314858161155852,
+      "tokens_seen": 1992359936
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004030909090909091,
+      "loss": 2.5385,
+      "theoretical_loss": 3.431476398027712,
+      "tokens_seen": 1992425472
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004030707070707071,
+      "loss": 2.4957,
+      "theoretical_loss": 3.431466980336355,
+      "tokens_seen": 1992491008
+    },
+    {
+      "epoch": 0.2,
+      "objective/train/docs_used": 1096583,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7095284461975098,
+      "objective/train/theoretical_loss": 3.4314575630414854,
+      "objective/train/tokens_used": 351415776,
+      "theoretical_loss": 3.4314575630414854,
+      "tokens_seen": 1992556544
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040305050505050506,
+      "loss": 2.6214,
+      "theoretical_loss": 3.4314575630414854,
+      "tokens_seen": 1992556544
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000403030303030303,
+      "loss": 2.807,
+      "theoretical_loss": 3.4314481461430724,
+      "tokens_seen": 1992622080
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000403010101010101,
+      "loss": 2.634,
+      "theoretical_loss": 3.4314387296410875,
+      "tokens_seen": 1992687616
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040298989898989904,
+      "loss": 2.7925,
+      "theoretical_loss": 3.4314293135355003,
+      "tokens_seen": 1992753152
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000402969696969697,
+      "loss": 2.4985,
+      "theoretical_loss": 3.431419897826281,
+      "tokens_seen": 1992818688
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040294949494949496,
+      "loss": 2.8183,
+      "theoretical_loss": 3.4314104825133995,
+      "tokens_seen": 1992884224
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004029292929292929,
+      "loss": 2.7924,
+      "theoretical_loss": 3.431401067596827,
+      "tokens_seen": 1992949760
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040290909090909087,
+      "loss": 2.6961,
+      "theoretical_loss": 3.431391653076533,
+      "tokens_seen": 1993015296
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040288888888888894,
+      "loss": 2.4038,
+      "theoretical_loss": 3.431382238952488,
+      "tokens_seen": 1993080832
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004028686868686869,
+      "loss": 2.7434,
+      "theoretical_loss": 3.4313728252246634,
+      "tokens_seen": 1993146368
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040284848484848485,
+      "loss": 2.718,
+      "theoretical_loss": 3.4313634118930274,
+      "tokens_seen": 1993211904
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004028282828282828,
+      "loss": 2.7766,
+      "theoretical_loss": 3.431353998957552,
+      "tokens_seen": 1993277440
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040280808080808076,
+      "loss": 2.7098,
+      "theoretical_loss": 3.4313445864182066,
+      "tokens_seen": 1993342976
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040278787878787883,
+      "loss": 2.7335,
+      "theoretical_loss": 3.4313351742749623,
+      "tokens_seen": 1993408512
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004027676767676768,
+      "loss": 2.6668,
+      "theoretical_loss": 3.4313257625277886,
+      "tokens_seen": 1993474048
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040274747474747474,
+      "loss": 2.5202,
+      "theoretical_loss": 3.4313163511766565,
+      "tokens_seen": 1993539584
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004027272727272727,
+      "loss": 2.7599,
+      "theoretical_loss": 3.4313069402215355,
+      "tokens_seen": 1993605120
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004027070707070707,
+      "loss": 2.6586,
+      "theoretical_loss": 3.431297529662397,
+      "tokens_seen": 1993670656
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004026868686868687,
+      "loss": 2.8594,
+      "theoretical_loss": 3.43128811949921,
+      "tokens_seen": 1993736192
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004026666666666667,
+      "loss": 2.6204,
+      "theoretical_loss": 3.431278709731946,
+      "tokens_seen": 1993801728
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040264646464646464,
+      "loss": 2.6484,
+      "theoretical_loss": 3.431269300360575,
+      "tokens_seen": 1993867264
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004026262626262626,
+      "loss": 2.6423,
+      "theoretical_loss": 3.431259891385067,
+      "tokens_seen": 1993932800
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004026060606060606,
+      "loss": 2.6703,
+      "theoretical_loss": 3.431250482805392,
+      "tokens_seen": 1993998336
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004025858585858586,
+      "loss": 2.5785,
+      "theoretical_loss": 3.4312410746215214,
+      "tokens_seen": 1994063872
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040256565656565657,
+      "loss": 2.655,
+      "theoretical_loss": 3.4312316668334253,
+      "tokens_seen": 1994129408
+    },
+    {
+      "epoch": 0.2,
+      "objective/train/docs_used": 1097915,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.4799556732177734,
+      "objective/train/theoretical_loss": 3.431222259441073,
+      "objective/train/tokens_used": 353054176,
+      "theoretical_loss": 3.431222259441073,
+      "tokens_seen": 1994194944
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040254545454545453,
+      "loss": 2.6861,
+      "theoretical_loss": 3.431222259441073,
+      "tokens_seen": 1994194944
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040252525252525254,
+      "loss": 2.6419,
+      "theoretical_loss": 3.4312128524444363,
+      "tokens_seen": 1994260480
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040250505050505055,
+      "loss": 2.548,
+      "theoretical_loss": 3.431203445843485,
+      "tokens_seen": 1994326016
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004024848484848485,
+      "loss": 2.5519,
+      "theoretical_loss": 3.4311940396381884,
+      "tokens_seen": 1994391552
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040246464646464646,
+      "loss": 2.6708,
+      "theoretical_loss": 3.4311846338285186,
+      "tokens_seen": 1994457088
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004024444444444444,
+      "loss": 2.6083,
+      "theoretical_loss": 3.4311752284144443,
+      "tokens_seen": 1994522624
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040242424242424243,
+      "loss": 2.5466,
+      "theoretical_loss": 3.4311658233959372,
+      "tokens_seen": 1994588160
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040240404040404044,
+      "loss": 2.5841,
+      "theoretical_loss": 3.4311564187729675,
+      "tokens_seen": 1994653696
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004023838383838384,
+      "loss": 2.6176,
+      "theoretical_loss": 3.431147014545505,
+      "tokens_seen": 1994719232
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040236363636363636,
+      "loss": 2.6555,
+      "theoretical_loss": 3.4311376107135203,
+      "tokens_seen": 1994784768
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040234343434343437,
+      "loss": 2.681,
+      "theoretical_loss": 3.4311282072769833,
+      "tokens_seen": 1994850304
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004023232323232323,
+      "loss": 2.5542,
+      "theoretical_loss": 3.4311188042358656,
+      "tokens_seen": 1994915840
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040230303030303034,
+      "loss": 2.8682,
+      "theoretical_loss": 3.4311094015901364,
+      "tokens_seen": 1994981376
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004022828282828283,
+      "loss": 2.7279,
+      "theoretical_loss": 3.431099999339767,
+      "tokens_seen": 1995046912
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040226262626262625,
+      "loss": 2.8612,
+      "theoretical_loss": 3.431090597484727,
+      "tokens_seen": 1995112448
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040224242424242426,
+      "loss": 2.6292,
+      "theoretical_loss": 3.4310811960249867,
+      "tokens_seen": 1995177984
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004022222222222222,
+      "loss": 2.5724,
+      "theoretical_loss": 3.431071794960517,
+      "tokens_seen": 1995243520
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040220202020202023,
+      "loss": 2.6812,
+      "theoretical_loss": 3.4310623942912892,
+      "tokens_seen": 1995309056
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004021818181818182,
+      "loss": 2.9322,
+      "theoretical_loss": 3.431052994017272,
+      "tokens_seen": 1995374592
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004021616161616162,
+      "loss": 2.7268,
+      "theoretical_loss": 3.431043594138436,
+      "tokens_seen": 1995440128
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040214141414141415,
+      "loss": 2.7061,
+      "theoretical_loss": 3.431034194654753,
+      "tokens_seen": 1995505664
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004021212121212121,
+      "loss": 2.7505,
+      "theoretical_loss": 3.4310247955661923,
+      "tokens_seen": 1995571200
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004021010101010101,
+      "loss": 2.5533,
+      "theoretical_loss": 3.4310153968727244,
+      "tokens_seen": 1995636736
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004020808080808081,
+      "loss": 2.5434,
+      "theoretical_loss": 3.43100599857432,
+      "tokens_seen": 1995702272
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004020606060606061,
+      "loss": 2.7473,
+      "theoretical_loss": 3.4309966006709494,
+      "tokens_seen": 1995767808
+    },
+    {
+      "epoch": 0.2,
+      "objective/train/docs_used": 1098508,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5174620151519775,
+      "objective/train/theoretical_loss": 3.4309872031625828,
+      "objective/train/tokens_used": 354692576,
+      "theoretical_loss": 3.4309872031625828,
+      "tokens_seen": 1995833344
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040204040404040405,
+      "loss": 2.7295,
+      "theoretical_loss": 3.4309872031625828,
+      "tokens_seen": 1995833344
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000402020202020202,
+      "loss": 2.3933,
+      "theoretical_loss": 3.4309778060491913,
+      "tokens_seen": 1995898880
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000402,
+      "loss": 2.7622,
+      "theoretical_loss": 3.4309684093307444,
+      "tokens_seen": 1995964416
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000401979797979798,
+      "loss": 2.8072,
+      "theoretical_loss": 3.430959013007213,
+      "tokens_seen": 1996029952
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000401959595959596,
+      "loss": 2.4624,
+      "theoretical_loss": 3.430949617078568,
+      "tokens_seen": 1996095488
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040193939393939394,
+      "loss": 2.6096,
+      "theoretical_loss": 3.4309402215447795,
+      "tokens_seen": 1996161024
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004019191919191919,
+      "loss": 2.8545,
+      "theoretical_loss": 3.430930826405817,
+      "tokens_seen": 1996226560
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004018989898989899,
+      "loss": 2.6149,
+      "theoretical_loss": 3.4309214316616528,
+      "tokens_seen": 1996292096
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004018787878787879,
+      "loss": 2.621,
+      "theoretical_loss": 3.4309120373122557,
+      "tokens_seen": 1996357632
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004018585858585859,
+      "loss": 2.8016,
+      "theoretical_loss": 3.4309026433575966,
+      "tokens_seen": 1996423168
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040183838383838383,
+      "loss": 2.7612,
+      "theoretical_loss": 3.4308932497976468,
+      "tokens_seen": 1996488704
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004018181818181818,
+      "loss": 2.5173,
+      "theoretical_loss": 3.4308838566323754,
+      "tokens_seen": 1996554240
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040179797979797985,
+      "loss": 2.5817,
+      "theoretical_loss": 3.4308744638617537,
+      "tokens_seen": 1996619776
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004017777777777778,
+      "loss": 2.5235,
+      "theoretical_loss": 3.4308650714857523,
+      "tokens_seen": 1996685312
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040175757575757577,
+      "loss": 2.5699,
+      "theoretical_loss": 3.4308556795043414,
+      "tokens_seen": 1996750848
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004017373737373737,
+      "loss": 2.7443,
+      "theoretical_loss": 3.4308462879174915,
+      "tokens_seen": 1996816384
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004017171717171717,
+      "loss": 2.5264,
+      "theoretical_loss": 3.4308368967251726,
+      "tokens_seen": 1996881920
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040169696969696975,
+      "loss": 2.647,
+      "theoretical_loss": 3.4308275059273563,
+      "tokens_seen": 1996947456
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004016767676767677,
+      "loss": 2.6705,
+      "theoretical_loss": 3.430818115524012,
+      "tokens_seen": 1997012992
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040165656565656566,
+      "loss": 2.7531,
+      "theoretical_loss": 3.430808725515111,
+      "tokens_seen": 1997078528
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004016363636363636,
+      "loss": 2.7024,
+      "theoretical_loss": 3.430799335900623,
+      "tokens_seen": 1997144064
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040161616161616163,
+      "loss": 2.7327,
+      "theoretical_loss": 3.4307899466805187,
+      "tokens_seen": 1997209600
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040159595959595964,
+      "loss": 2.746,
+      "theoretical_loss": 3.430780557854769,
+      "tokens_seen": 1997275136
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004015757575757576,
+      "loss": 2.7495,
+      "theoretical_loss": 3.4307711694233443,
+      "tokens_seen": 1997340672
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00040155555555555555,
+      "loss": 2.608,
+      "theoretical_loss": 3.430761781386215,
+      "tokens_seen": 1997406208
+    },
+    {
+      "epoch": 0.2,
+      "objective/train/docs_used": 1099394,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.8477187156677246,
+      "objective/train/theoretical_loss": 3.430752393743351,
+      "objective/train/tokens_used": 356330976,
+      "theoretical_loss": 3.430752393743351,
+      "tokens_seen": 1997471744
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0004015353535353535,
+      "loss": 2.63,
+      "theoretical_loss": 3.430752393743351,
+      "tokens_seen": 1997471744
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004015151515151515,
+      "loss": 2.8339,
+      "theoretical_loss": 3.430743006494724,
+      "tokens_seen": 1997537280
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040149494949494953,
+      "loss": 2.9765,
+      "theoretical_loss": 3.4307336196403035,
+      "tokens_seen": 1997602816
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004014747474747475,
+      "loss": 2.9089,
+      "theoretical_loss": 3.4307242331800607,
+      "tokens_seen": 1997668352
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040145454545454545,
+      "loss": 2.7895,
+      "theoretical_loss": 3.4307148471139657,
+      "tokens_seen": 1997733888
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040143434343434346,
+      "loss": 2.7117,
+      "theoretical_loss": 3.430705461441989,
+      "tokens_seen": 1997799424
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004014141414141414,
+      "loss": 2.6922,
+      "theoretical_loss": 3.4306960761641014,
+      "tokens_seen": 1997864960
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004013939393939394,
+      "loss": 2.6537,
+      "theoretical_loss": 3.4306866912802736,
+      "tokens_seen": 1997930496
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004013737373737374,
+      "loss": 2.6435,
+      "theoretical_loss": 3.430677306790475,
+      "tokens_seen": 1997996032
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040135353535353534,
+      "loss": 2.7962,
+      "theoretical_loss": 3.4306679226946777,
+      "tokens_seen": 1998061568
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040133333333333335,
+      "loss": 2.5895,
+      "theoretical_loss": 3.430658538992851,
+      "tokens_seen": 1998127104
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004013131313131313,
+      "loss": 2.7134,
+      "theoretical_loss": 3.430649155684966,
+      "tokens_seen": 1998192640
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004012929292929293,
+      "loss": 2.6309,
+      "theoretical_loss": 3.4306397727709936,
+      "tokens_seen": 1998258176
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004012727272727273,
+      "loss": 2.573,
+      "theoretical_loss": 3.4306303902509034,
+      "tokens_seen": 1998323712
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004012525252525253,
+      "loss": 2.8616,
+      "theoretical_loss": 3.4306210081246666,
+      "tokens_seen": 1998389248
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040123232323232324,
+      "loss": 2.587,
+      "theoretical_loss": 3.430611626392254,
+      "tokens_seen": 1998454784
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004012121212121212,
+      "loss": 2.7992,
+      "theoretical_loss": 3.4306022450536355,
+      "tokens_seen": 1998520320
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004011919191919192,
+      "loss": 2.5505,
+      "theoretical_loss": 3.4305928641087817,
+      "tokens_seen": 1998585856
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040117171717171717,
+      "loss": 2.7951,
+      "theoretical_loss": 3.4305834835576636,
+      "tokens_seen": 1998651392
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004011515151515152,
+      "loss": 2.6681,
+      "theoretical_loss": 3.430574103400252,
+      "tokens_seen": 1998716928
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040113131313131313,
+      "loss": 2.5223,
+      "theoretical_loss": 3.4305647236365164,
+      "tokens_seen": 1998782464
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004011111111111111,
+      "loss": 2.7467,
+      "theoretical_loss": 3.430555344266428,
+      "tokens_seen": 1998848000
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004010909090909091,
+      "loss": 2.7558,
+      "theoretical_loss": 3.4305459652899577,
+      "tokens_seen": 1998913536
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040107070707070706,
+      "loss": 2.6164,
+      "theoretical_loss": 3.4305365867070754,
+      "tokens_seen": 1998979072
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040105050505050507,
+      "loss": 2.878,
+      "theoretical_loss": 3.430527208517752,
+      "tokens_seen": 1999044608
+    },
+    {
+      "epoch": 0.21,
+      "objective/train/docs_used": 1100643,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.998358726501465,
+      "objective/train/theoretical_loss": 3.4305178307219584,
+      "objective/train/tokens_used": 357969376,
+      "theoretical_loss": 3.4305178307219584,
+      "tokens_seen": 1999110144
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000401030303030303,
+      "loss": 2.8925,
+      "theoretical_loss": 3.4305178307219584,
+      "tokens_seen": 1999110144
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000401010101010101,
+      "loss": 2.8018,
+      "theoretical_loss": 3.430508453319665,
+      "tokens_seen": 1999175680
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000400989898989899,
+      "loss": 2.7074,
+      "theoretical_loss": 3.430499076310842,
+      "tokens_seen": 1999241216
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000400969696969697,
+      "loss": 2.6598,
+      "theoretical_loss": 3.4304896996954604,
+      "tokens_seen": 1999306752
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040094949494949496,
+      "loss": 2.555,
+      "theoretical_loss": 3.4304803234734904,
+      "tokens_seen": 1999372288
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004009292929292929,
+      "loss": 2.7593,
+      "theoretical_loss": 3.430470947644903,
+      "tokens_seen": 1999437824
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004009090909090909,
+      "loss": 3.0023,
+      "theoretical_loss": 3.4304615722096687,
+      "tokens_seen": 1999503360
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004008888888888889,
+      "loss": 2.7062,
+      "theoretical_loss": 3.4304521971677584,
+      "tokens_seen": 1999568896
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004008686868686869,
+      "loss": 2.711,
+      "theoretical_loss": 3.430442822519142,
+      "tokens_seen": 1999634432
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040084848484848486,
+      "loss": 2.6887,
+      "theoretical_loss": 3.4304334482637904,
+      "tokens_seen": 1999699968
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004008282828282828,
+      "loss": 2.6289,
+      "theoretical_loss": 3.4304240744016745,
+      "tokens_seen": 1999765504
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004008080808080808,
+      "loss": 2.6162,
+      "theoretical_loss": 3.430414700932765,
+      "tokens_seen": 1999831040
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040078787878787883,
+      "loss": 2.7507,
+      "theoretical_loss": 3.4304053278570317,
+      "tokens_seen": 1999896576
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004007676767676768,
+      "loss": 2.6243,
+      "theoretical_loss": 3.430395955174446,
+      "tokens_seen": 1999962112
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040074747474747475,
+      "loss": 2.6822,
+      "theoretical_loss": 3.4303865828849784,
+      "tokens_seen": 2000027648
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004007272727272727,
+      "loss": 2.5374,
+      "theoretical_loss": 3.4303772109885995,
+      "tokens_seen": 2000093184
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004007070707070707,
+      "loss": 2.8277,
+      "theoretical_loss": 3.43036783948528,
+      "tokens_seen": 2000158720
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040068686868686873,
+      "loss": 2.9846,
+      "theoretical_loss": 3.43035846837499,
+      "tokens_seen": 2000224256
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004006666666666667,
+      "loss": 2.718,
+      "theoretical_loss": 3.4303490976577007,
+      "tokens_seen": 2000289792
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040064646464646464,
+      "loss": 2.4615,
+      "theoretical_loss": 3.430339727333383,
+      "tokens_seen": 2000355328
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004006262626262626,
+      "loss": 2.7228,
+      "theoretical_loss": 3.4303303574020063,
+      "tokens_seen": 2000420864
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040060606060606066,
+      "loss": 2.5237,
+      "theoretical_loss": 3.4303209878635426,
+      "tokens_seen": 2000486400
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004005858585858586,
+      "loss": 2.866,
+      "theoretical_loss": 3.430311618717962,
+      "tokens_seen": 2000551936
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004005656565656566,
+      "loss": 2.6097,
+      "theoretical_loss": 3.4303022499652354,
+      "tokens_seen": 2000617472
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040054545454545453,
+      "loss": 2.7682,
+      "theoretical_loss": 3.4302928816053324,
+      "tokens_seen": 2000683008
+    },
+    {
+      "epoch": 0.21,
+      "objective/train/docs_used": 1101270,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.287954092025757,
+      "objective/train/theoretical_loss": 3.4302835136382255,
+      "objective/train/tokens_used": 359607776,
+      "theoretical_loss": 3.4302835136382255,
+      "tokens_seen": 2000748544
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004005252525252525,
+      "loss": 2.6407,
+      "theoretical_loss": 3.4302835136382255,
+      "tokens_seen": 2000748544
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040050505050505056,
+      "loss": 2.6081,
+      "theoretical_loss": 3.4302741460638844,
+      "tokens_seen": 2000814080
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004004848484848485,
+      "loss": 2.9468,
+      "theoretical_loss": 3.430264778882279,
+      "tokens_seen": 2000879616
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040046464646464647,
+      "loss": 2.6506,
+      "theoretical_loss": 3.430255412093381,
+      "tokens_seen": 2000945152
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004004444444444444,
+      "loss": 2.7632,
+      "theoretical_loss": 3.430246045697161,
+      "tokens_seen": 2001010688
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040042424242424244,
+      "loss": 2.7678,
+      "theoretical_loss": 3.4302366796935893,
+      "tokens_seen": 2001076224
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040040404040404045,
+      "loss": 2.7699,
+      "theoretical_loss": 3.4302273140826367,
+      "tokens_seen": 2001141760
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004003838383838384,
+      "loss": 2.8298,
+      "theoretical_loss": 3.430217948864274,
+      "tokens_seen": 2001207296
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040036363636363636,
+      "loss": 2.6612,
+      "theoretical_loss": 3.430208584038472,
+      "tokens_seen": 2001272832
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004003434343434343,
+      "loss": 2.684,
+      "theoretical_loss": 3.430199219605201,
+      "tokens_seen": 2001338368
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040032323232323233,
+      "loss": 2.6532,
+      "theoretical_loss": 3.430189855564432,
+      "tokens_seen": 2001403904
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040030303030303034,
+      "loss": 2.8199,
+      "theoretical_loss": 3.4301804919161354,
+      "tokens_seen": 2001469440
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004002828282828283,
+      "loss": 2.7311,
+      "theoretical_loss": 3.430171128660283,
+      "tokens_seen": 2001534976
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040026262626262625,
+      "loss": 2.6164,
+      "theoretical_loss": 3.4301617657968437,
+      "tokens_seen": 2001600512
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040024242424242427,
+      "loss": 2.7514,
+      "theoretical_loss": 3.4301524033257893,
+      "tokens_seen": 2001666048
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004002222222222222,
+      "loss": 2.7445,
+      "theoretical_loss": 3.4301430412470904,
+      "tokens_seen": 2001731584
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040020202020202023,
+      "loss": 2.5826,
+      "theoretical_loss": 3.4301336795607176,
+      "tokens_seen": 2001797120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004001818181818182,
+      "loss": 2.6789,
+      "theoretical_loss": 3.4301243182666417,
+      "tokens_seen": 2001862656
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040016161616161615,
+      "loss": 2.6448,
+      "theoretical_loss": 3.4301149573648333,
+      "tokens_seen": 2001928192
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040014141414141416,
+      "loss": 2.6389,
+      "theoretical_loss": 3.430105596855263,
+      "tokens_seen": 2001993728
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004001212121212121,
+      "loss": 2.4755,
+      "theoretical_loss": 3.430096236737902,
+      "tokens_seen": 2002059264
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004001010101010101,
+      "loss": 2.7416,
+      "theoretical_loss": 3.4300868770127213,
+      "tokens_seen": 2002124800
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004000808080808081,
+      "loss": 2.7288,
+      "theoretical_loss": 3.43007751767969,
+      "tokens_seen": 2002190336
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004000606060606061,
+      "loss": 2.7819,
+      "theoretical_loss": 3.4300681587387807,
+      "tokens_seen": 2002255872
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00040004040404040405,
+      "loss": 2.7553,
+      "theoretical_loss": 3.430058800189963,
+      "tokens_seen": 2002321408
+    },
+    {
+      "epoch": 0.21,
+      "objective/train/docs_used": 1102668,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.553368330001831,
+      "objective/train/theoretical_loss": 3.430049442033208,
+      "objective/train/tokens_used": 361246176,
+      "theoretical_loss": 3.430049442033208,
+      "tokens_seen": 2002386944
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000400020202020202,
+      "loss": 2.6432,
+      "theoretical_loss": 3.430049442033208,
+      "tokens_seen": 2002386944
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0004,
+      "loss": 2.7453,
+      "theoretical_loss": 3.4300400842684864,
+      "tokens_seen": 2002452480
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000399979797979798,
+      "loss": 2.5433,
+      "theoretical_loss": 3.4300307268957693,
+      "tokens_seen": 2002518016
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000399959595959596,
+      "loss": 2.645,
+      "theoretical_loss": 3.430021369915027,
+      "tokens_seen": 2002583552
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039993939393939394,
+      "loss": 2.7866,
+      "theoretical_loss": 3.4300120133262304,
+      "tokens_seen": 2002649088
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003999191919191919,
+      "loss": 2.7484,
+      "theoretical_loss": 3.43000265712935,
+      "tokens_seen": 2002714624
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003998989898989899,
+      "loss": 2.4926,
+      "theoretical_loss": 3.429993301324357,
+      "tokens_seen": 2002780160
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003998787878787879,
+      "loss": 2.6038,
+      "theoretical_loss": 3.4299839459112222,
+      "tokens_seen": 2002845696
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003998585858585859,
+      "loss": 2.7284,
+      "theoretical_loss": 3.4299745908899157,
+      "tokens_seen": 2002911232
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039983838383838384,
+      "loss": 2.8709,
+      "theoretical_loss": 3.4299652362604087,
+      "tokens_seen": 2002976768
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003998181818181818,
+      "loss": 2.6466,
+      "theoretical_loss": 3.4299558820226723,
+      "tokens_seen": 2003042304
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003997979797979798,
+      "loss": 2.5868,
+      "theoretical_loss": 3.429946528176677,
+      "tokens_seen": 2003107840
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003997777777777778,
+      "loss": 2.6806,
+      "theoretical_loss": 3.4299371747223937,
+      "tokens_seen": 2003173376
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039975757575757577,
+      "loss": 2.7211,
+      "theoretical_loss": 3.429927821659793,
+      "tokens_seen": 2003238912
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039973737373737373,
+      "loss": 2.6575,
+      "theoretical_loss": 3.429918468988845,
+      "tokens_seen": 2003304448
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003997171717171717,
+      "loss": 2.3557,
+      "theoretical_loss": 3.4299091167095215,
+      "tokens_seen": 2003369984
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039969696969696975,
+      "loss": 2.5517,
+      "theoretical_loss": 3.4298997648217933,
+      "tokens_seen": 2003435520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003996767676767677,
+      "loss": 2.876,
+      "theoretical_loss": 3.42989041332563,
+      "tokens_seen": 2003501056
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039965656565656567,
+      "loss": 2.6888,
+      "theoretical_loss": 3.4298810622210043,
+      "tokens_seen": 2003566592
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003996363636363636,
+      "loss": 2.7384,
+      "theoretical_loss": 3.4298717115078854,
+      "tokens_seen": 2003632128
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003996161616161616,
+      "loss": 2.4843,
+      "theoretical_loss": 3.429862361186245,
+      "tokens_seen": 2003697664
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039959595959595964,
+      "loss": 2.6384,
+      "theoretical_loss": 3.4298530112560535,
+      "tokens_seen": 2003763200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003995757575757576,
+      "loss": 2.7286,
+      "theoretical_loss": 3.429843661717282,
+      "tokens_seen": 2003828736
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039955555555555556,
+      "loss": 2.8028,
+      "theoretical_loss": 3.4298343125699002,
+      "tokens_seen": 2003894272
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003995353535353535,
+      "loss": 2.7432,
+      "theoretical_loss": 3.4298249638138807,
+      "tokens_seen": 2003959808
+    },
+    {
+      "epoch": 0.21,
+      "objective/train/docs_used": 1103172,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.3909668922424316,
+      "objective/train/theoretical_loss": 3.4298156154491934,
+      "objective/train/tokens_used": 362884576,
+      "theoretical_loss": 3.4298156154491934,
+      "tokens_seen": 2004025344
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039951515151515147,
+      "loss": 2.5196,
+      "theoretical_loss": 3.4298156154491934,
+      "tokens_seen": 2004025344
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039949494949494954,
+      "loss": 2.7479,
+      "theoretical_loss": 3.4298062674758087,
+      "tokens_seen": 2004090880
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003994747474747475,
+      "loss": 2.5021,
+      "theoretical_loss": 3.429796919893698,
+      "tokens_seen": 2004156416
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039945454545454545,
+      "loss": 2.791,
+      "theoretical_loss": 3.4297875727028324,
+      "tokens_seen": 2004221952
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003994343434343434,
+      "loss": 2.8064,
+      "theoretical_loss": 3.429778225903182,
+      "tokens_seen": 2004287488
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003994141414141414,
+      "loss": 2.7233,
+      "theoretical_loss": 3.429768879494718,
+      "tokens_seen": 2004353024
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039939393939393943,
+      "loss": 2.6588,
+      "theoretical_loss": 3.4297595334774114,
+      "tokens_seen": 2004418560
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003993737373737374,
+      "loss": 2.6098,
+      "theoretical_loss": 3.4297501878512326,
+      "tokens_seen": 2004484096
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039935353535353534,
+      "loss": 2.7278,
+      "theoretical_loss": 3.429740842616153,
+      "tokens_seen": 2004549632
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003993333333333333,
+      "loss": 2.5958,
+      "theoretical_loss": 3.429731497772143,
+      "tokens_seen": 2004615168
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003993131313131313,
+      "loss": 2.8843,
+      "theoretical_loss": 3.4297221533191737,
+      "tokens_seen": 2004680704
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003992929292929293,
+      "loss": 2.6212,
+      "theoretical_loss": 3.429712809257216,
+      "tokens_seen": 2004746240
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003992727272727273,
+      "loss": 2.7094,
+      "theoretical_loss": 3.4297034655862406,
+      "tokens_seen": 2004811776
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039925252525252524,
+      "loss": 2.7256,
+      "theoretical_loss": 3.429694122306218,
+      "tokens_seen": 2004877312
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039923232323232325,
+      "loss": 2.8226,
+      "theoretical_loss": 3.4296847794171197,
+      "tokens_seen": 2004942848
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003992121212121212,
+      "loss": 2.7651,
+      "theoretical_loss": 3.4296754369189166,
+      "tokens_seen": 2005008384
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003991919191919192,
+      "loss": 2.7283,
+      "theoretical_loss": 3.4296660948115782,
+      "tokens_seen": 2005073920
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039917171717171717,
+      "loss": 2.6783,
+      "theoretical_loss": 3.4296567530950774,
+      "tokens_seen": 2005139456
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039915151515151513,
+      "loss": 2.7694,
+      "theoretical_loss": 3.429647411769384,
+      "tokens_seen": 2005204992
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039913131313131314,
+      "loss": 2.855,
+      "theoretical_loss": 3.4296380708344687,
+      "tokens_seen": 2005270528
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039911111111111115,
+      "loss": 2.6749,
+      "theoretical_loss": 3.429628730290303,
+      "tokens_seen": 2005336064
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003990909090909091,
+      "loss": 2.4842,
+      "theoretical_loss": 3.4296193901368577,
+      "tokens_seen": 2005401600
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039907070707070706,
+      "loss": 2.9729,
+      "theoretical_loss": 3.429610050374103,
+      "tokens_seen": 2005467136
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003990505050505051,
+      "loss": 2.7454,
+      "theoretical_loss": 3.4296007110020104,
+      "tokens_seen": 2005532672
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039903030303030303,
+      "loss": 2.6091,
+      "theoretical_loss": 3.4295913720205506,
+      "tokens_seen": 2005598208
+    },
+    {
+      "epoch": 0.21,
+      "objective/train/docs_used": 1104281,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.4382286071777344,
+      "objective/train/theoretical_loss": 3.4295820334296945,
+      "objective/train/tokens_used": 364522976,
+      "theoretical_loss": 3.4295820334296945,
+      "tokens_seen": 2005663744
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039901010101010104,
+      "loss": 2.7794,
+      "theoretical_loss": 3.4295820334296945,
+      "tokens_seen": 2005663744
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000398989898989899,
+      "loss": 2.5314,
+      "theoretical_loss": 3.429572695229413,
+      "tokens_seen": 2005729280
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039896969696969696,
+      "loss": 2.5465,
+      "theoretical_loss": 3.429563357419677,
+      "tokens_seen": 2005794816
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039894949494949497,
+      "loss": 2.6578,
+      "theoretical_loss": 3.4295540200004577,
+      "tokens_seen": 2005860352
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003989292929292929,
+      "loss": 2.7723,
+      "theoretical_loss": 3.4295446829717258,
+      "tokens_seen": 2005925888
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039890909090909094,
+      "loss": 2.5896,
+      "theoretical_loss": 3.429535346333452,
+      "tokens_seen": 2005991424
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003988888888888889,
+      "loss": 2.4996,
+      "theoretical_loss": 3.429526010085607,
+      "tokens_seen": 2006056960
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003988686868686869,
+      "loss": 2.8004,
+      "theoretical_loss": 3.4295166742281626,
+      "tokens_seen": 2006122496
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039884848484848486,
+      "loss": 2.5072,
+      "theoretical_loss": 3.4295073387610895,
+      "tokens_seen": 2006188032
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003988282828282828,
+      "loss": 2.6983,
+      "theoretical_loss": 3.4294980036843583,
+      "tokens_seen": 2006253568
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039880808080808083,
+      "loss": 2.9369,
+      "theoretical_loss": 3.4294886689979394,
+      "tokens_seen": 2006319104
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003987878787878788,
+      "loss": 2.5722,
+      "theoretical_loss": 3.4294793347018047,
+      "tokens_seen": 2006384640
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003987676767676768,
+      "loss": 2.586,
+      "theoretical_loss": 3.4294700007959245,
+      "tokens_seen": 2006450176
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039874747474747475,
+      "loss": 2.8116,
+      "theoretical_loss": 3.4294606672802703,
+      "tokens_seen": 2006515712
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003987272727272727,
+      "loss": 2.7575,
+      "theoretical_loss": 3.4294513341548125,
+      "tokens_seen": 2006581248
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003987070707070707,
+      "loss": 2.8497,
+      "theoretical_loss": 3.4294420014195226,
+      "tokens_seen": 2006646784
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039868686868686873,
+      "loss": 2.8441,
+      "theoretical_loss": 3.4294326690743713,
+      "tokens_seen": 2006712320
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003986666666666667,
+      "loss": 2.559,
+      "theoretical_loss": 3.429423337119329,
+      "tokens_seen": 2006777856
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039864646464646465,
+      "loss": 2.4556,
+      "theoretical_loss": 3.429414005554367,
+      "tokens_seen": 2006843392
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003986262626262626,
+      "loss": 2.5888,
+      "theoretical_loss": 3.429404674379457,
+      "tokens_seen": 2006908928
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003986060606060606,
+      "loss": 2.6556,
+      "theoretical_loss": 3.429395343594569,
+      "tokens_seen": 2006974464
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003985858585858586,
+      "loss": 2.6575,
+      "theoretical_loss": 3.4293860131996743,
+      "tokens_seen": 2007040000
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003985656565656566,
+      "loss": 2.597,
+      "theoretical_loss": 3.429376683194744,
+      "tokens_seen": 2007105536
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039854545454545454,
+      "loss": 2.6176,
+      "theoretical_loss": 3.4293673535797486,
+      "tokens_seen": 2007171072
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003985252525252525,
+      "loss": 2.7297,
+      "theoretical_loss": 3.4293580243546598,
+      "tokens_seen": 2007236608
+    },
+    {
+      "epoch": 0.21,
+      "objective/train/docs_used": 1105495,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7771756649017334,
+      "objective/train/theoretical_loss": 3.4293486955194483,
+      "objective/train/tokens_used": 366161376,
+      "theoretical_loss": 3.4293486955194483,
+      "tokens_seen": 2007302144
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039850505050505056,
+      "loss": 2.7023,
+      "theoretical_loss": 3.4293486955194483,
+      "tokens_seen": 2007302144
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003984848484848485,
+      "loss": 2.7546,
+      "theoretical_loss": 3.4293393670740846,
+      "tokens_seen": 2007367680
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003984646464646465,
+      "loss": 2.6322,
+      "theoretical_loss": 3.42933003901854,
+      "tokens_seen": 2007433216
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039844444444444443,
+      "loss": 2.806,
+      "theoretical_loss": 3.429320711352786,
+      "tokens_seen": 2007498752
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003984242424242424,
+      "loss": 2.5904,
+      "theoretical_loss": 3.429311384076793,
+      "tokens_seen": 2007564288
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039840404040404045,
+      "loss": 2.8624,
+      "theoretical_loss": 3.4293020571905317,
+      "tokens_seen": 2007629824
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003983838383838384,
+      "loss": 2.6415,
+      "theoretical_loss": 3.429292730693974,
+      "tokens_seen": 2007695360
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039836363636363637,
+      "loss": 2.5293,
+      "theoretical_loss": 3.42928340458709,
+      "tokens_seen": 2007760896
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003983434343434343,
+      "loss": 2.5056,
+      "theoretical_loss": 3.4292740788698515,
+      "tokens_seen": 2007826432
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039832323232323234,
+      "loss": 2.6948,
+      "theoretical_loss": 3.4292647535422285,
+      "tokens_seen": 2007891968
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039830303030303035,
+      "loss": 2.602,
+      "theoretical_loss": 3.4292554286041934,
+      "tokens_seen": 2007957504
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003982828282828283,
+      "loss": 2.6045,
+      "theoretical_loss": 3.429246104055716,
+      "tokens_seen": 2008023040
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039826262626262626,
+      "loss": 2.8573,
+      "theoretical_loss": 3.4292367798967676,
+      "tokens_seen": 2008088576
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003982424242424242,
+      "loss": 2.5131,
+      "theoretical_loss": 3.4292274561273195,
+      "tokens_seen": 2008154112
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039822222222222223,
+      "loss": 2.6732,
+      "theoretical_loss": 3.4292181327473426,
+      "tokens_seen": 2008219648
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039820202020202024,
+      "loss": 2.591,
+      "theoretical_loss": 3.429208809756808,
+      "tokens_seen": 2008285184
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003981818181818182,
+      "loss": 2.5137,
+      "theoretical_loss": 3.429199487155686,
+      "tokens_seen": 2008350720
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039816161616161615,
+      "loss": 2.5927,
+      "theoretical_loss": 3.4291901649439493,
+      "tokens_seen": 2008416256
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039814141414141416,
+      "loss": 2.8069,
+      "theoretical_loss": 3.429180843121567,
+      "tokens_seen": 2008481792
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003981212121212121,
+      "loss": 2.6222,
+      "theoretical_loss": 3.4291715216885112,
+      "tokens_seen": 2008547328
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039810101010101013,
+      "loss": 2.7161,
+      "theoretical_loss": 3.429162200644753,
+      "tokens_seen": 2008612864
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003980808080808081,
+      "loss": 2.7619,
+      "theoretical_loss": 3.429152879990263,
+      "tokens_seen": 2008678400
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039806060606060605,
+      "loss": 2.697,
+      "theoretical_loss": 3.429143559725013,
+      "tokens_seen": 2008743936
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039804040404040406,
+      "loss": 2.5508,
+      "theoretical_loss": 3.429134239848973,
+      "tokens_seen": 2008809472
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000398020202020202,
+      "loss": 2.6974,
+      "theoretical_loss": 3.4291249203621144,
+      "tokens_seen": 2008875008
+    },
+    {
+      "epoch": 0.21,
+      "objective/train/docs_used": 1105995,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.0905230045318604,
+      "objective/train/theoretical_loss": 3.4291156012644084,
+      "objective/train/tokens_used": 367799776,
+      "theoretical_loss": 3.4291156012644084,
+      "tokens_seen": 2008940544
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000398,
+      "loss": 2.7571,
+      "theoretical_loss": 3.4291156012644084,
+      "tokens_seen": 2008940544
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000397979797979798,
+      "loss": 2.4411,
+      "theoretical_loss": 3.429106282555826,
+      "tokens_seen": 2009006080
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000397959595959596,
+      "loss": 2.5121,
+      "theoretical_loss": 3.4290969642363387,
+      "tokens_seen": 2009071616
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039793939393939395,
+      "loss": 2.521,
+      "theoretical_loss": 3.429087646305917,
+      "tokens_seen": 2009137152
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003979191919191919,
+      "loss": 2.6003,
+      "theoretical_loss": 3.4290783287645317,
+      "tokens_seen": 2009202688
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003978989898989899,
+      "loss": 2.8013,
+      "theoretical_loss": 3.429069011612155,
+      "tokens_seen": 2009268224
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003978787878787879,
+      "loss": 2.8218,
+      "theoretical_loss": 3.429059694848757,
+      "tokens_seen": 2009333760
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003978585858585859,
+      "loss": 2.5237,
+      "theoretical_loss": 3.4290503784743085,
+      "tokens_seen": 2009399296
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039783838383838384,
+      "loss": 2.6163,
+      "theoretical_loss": 3.429041062488782,
+      "tokens_seen": 2009464832
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003978181818181818,
+      "loss": 2.7317,
+      "theoretical_loss": 3.429031746892147,
+      "tokens_seen": 2009530368
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003977979797979798,
+      "loss": 2.5755,
+      "theoretical_loss": 3.429022431684375,
+      "tokens_seen": 2009595904
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039777777777777777,
+      "loss": 2.5838,
+      "theoretical_loss": 3.4290131168654385,
+      "tokens_seen": 2009661440
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003977575757575758,
+      "loss": 2.6078,
+      "theoretical_loss": 3.4290038024353064,
+      "tokens_seen": 2009726976
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039773737373737373,
+      "loss": 2.7659,
+      "theoretical_loss": 3.4289944883939514,
+      "tokens_seen": 2009792512
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003977171717171717,
+      "loss": 2.5651,
+      "theoretical_loss": 3.428985174741344,
+      "tokens_seen": 2009858048
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003976969696969697,
+      "loss": 2.3688,
+      "theoretical_loss": 3.428975861477455,
+      "tokens_seen": 2009923584
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003976767676767677,
+      "loss": 2.7842,
+      "theoretical_loss": 3.4289665486022565,
+      "tokens_seen": 2009989120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039765656565656567,
+      "loss": 2.6497,
+      "theoretical_loss": 3.4289572361157186,
+      "tokens_seen": 2010054656
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039763636363636363,
+      "loss": 2.5598,
+      "theoretical_loss": 3.428947924017813,
+      "tokens_seen": 2010120192
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003976161616161616,
+      "loss": 2.7141,
+      "theoretical_loss": 3.42893861230851,
+      "tokens_seen": 2010185728
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003975959595959596,
+      "loss": 2.6488,
+      "theoretical_loss": 3.428929300987782,
+      "tokens_seen": 2010251264
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003975757575757576,
+      "loss": 2.6941,
+      "theoretical_loss": 3.428919990055599,
+      "tokens_seen": 2010316800
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039755555555555556,
+      "loss": 2.6565,
+      "theoretical_loss": 3.428910679511932,
+      "tokens_seen": 2010382336
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003975353535353535,
+      "loss": 2.5045,
+      "theoretical_loss": 3.428901369356754,
+      "tokens_seen": 2010447872
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003975151515151515,
+      "loss": 2.8396,
+      "theoretical_loss": 3.4288920595900336,
+      "tokens_seen": 2010513408
+    },
+    {
+      "epoch": 0.21,
+      "objective/train/docs_used": 1107334,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7171061038970947,
+      "objective/train/theoretical_loss": 3.4288827502117436,
+      "objective/train/tokens_used": 369438176,
+      "theoretical_loss": 3.4288827502117436,
+      "tokens_seen": 2010578944
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039749494949494954,
+      "loss": 2.6364,
+      "theoretical_loss": 3.4288827502117436,
+      "tokens_seen": 2010578944
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003974747474747475,
+      "loss": 2.616,
+      "theoretical_loss": 3.4288734412218544,
+      "tokens_seen": 2010644480
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039745454545454546,
+      "loss": 2.9022,
+      "theoretical_loss": 3.4288641326203377,
+      "tokens_seen": 2010710016
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003974343434343434,
+      "loss": 2.6037,
+      "theoretical_loss": 3.428854824407164,
+      "tokens_seen": 2010775552
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039741414141414137,
+      "loss": 2.5749,
+      "theoretical_loss": 3.428845516582305,
+      "tokens_seen": 2010841088
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039739393939393944,
+      "loss": 2.8528,
+      "theoretical_loss": 3.4288362091457314,
+      "tokens_seen": 2010906624
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003973737373737374,
+      "loss": 2.7218,
+      "theoretical_loss": 3.428826902097415,
+      "tokens_seen": 2010972160
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039735353535353535,
+      "loss": 2.8239,
+      "theoretical_loss": 3.428817595437326,
+      "tokens_seen": 2011037696
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003973333333333333,
+      "loss": 2.7298,
+      "theoretical_loss": 3.4288082891654366,
+      "tokens_seen": 2011103232
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039731313131313137,
+      "loss": 2.6982,
+      "theoretical_loss": 3.428798983281717,
+      "tokens_seen": 2011168768
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039729292929292933,
+      "loss": 2.8055,
+      "theoretical_loss": 3.4287896777861384,
+      "tokens_seen": 2011234304
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003972727272727273,
+      "loss": 2.5813,
+      "theoretical_loss": 3.4287803726786734,
+      "tokens_seen": 2011299840
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039725252525252524,
+      "loss": 2.6005,
+      "theoretical_loss": 3.4287710679592913,
+      "tokens_seen": 2011365376
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003972323232323232,
+      "loss": 2.8153,
+      "theoretical_loss": 3.428761763627964,
+      "tokens_seen": 2011430912
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039721212121212126,
+      "loss": 2.2978,
+      "theoretical_loss": 3.4287524596846635,
+      "tokens_seen": 2011496448
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003971919191919192,
+      "loss": 2.5187,
+      "theoretical_loss": 3.42874315612936,
+      "tokens_seen": 2011561984
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003971717171717172,
+      "loss": 2.6217,
+      "theoretical_loss": 3.4287338529620244,
+      "tokens_seen": 2011627520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039715151515151513,
+      "loss": 2.6869,
+      "theoretical_loss": 3.428724550182629,
+      "tokens_seen": 2011693056
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039713131313131315,
+      "loss": 2.6536,
+      "theoretical_loss": 3.4287152477911436,
+      "tokens_seen": 2011758592
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039711111111111116,
+      "loss": 2.6725,
+      "theoretical_loss": 3.4287059457875406,
+      "tokens_seen": 2011824128
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003970909090909091,
+      "loss": 2.7913,
+      "theoretical_loss": 3.428696644171791,
+      "tokens_seen": 2011889664
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039707070707070707,
+      "loss": 2.7083,
+      "theoretical_loss": 3.428687342943866,
+      "tokens_seen": 2011955200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000397050505050505,
+      "loss": 2.7214,
+      "theoretical_loss": 3.428678042103736,
+      "tokens_seen": 2012020736
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039703030303030304,
+      "loss": 2.839,
+      "theoretical_loss": 3.428668741651373,
+      "tokens_seen": 2012086272
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039701010101010105,
+      "loss": 2.6332,
+      "theoretical_loss": 3.4286594415867477,
+      "tokens_seen": 2012151808
+    },
+    {
+      "epoch": 0.21,
+      "objective/train/docs_used": 1107920,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5978739261627197,
+      "objective/train/theoretical_loss": 3.4286501419098316,
+      "objective/train/tokens_used": 371076576,
+      "theoretical_loss": 3.4286501419098316,
+      "tokens_seen": 2012217344
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000396989898989899,
+      "loss": 2.5902,
+      "theoretical_loss": 3.4286501419098316,
+      "tokens_seen": 2012217344
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039696969696969696,
+      "loss": 2.9242,
+      "theoretical_loss": 3.428640842620596,
+      "tokens_seen": 2012282880
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000396949494949495,
+      "loss": 2.8167,
+      "theoretical_loss": 3.4286315437190122,
+      "tokens_seen": 2012348416
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039692929292929293,
+      "loss": 2.5715,
+      "theoretical_loss": 3.4286222452050508,
+      "tokens_seen": 2012413952
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039690909090909094,
+      "loss": 2.6004,
+      "theoretical_loss": 3.428612947078684,
+      "tokens_seen": 2012479488
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003968888888888889,
+      "loss": 2.6496,
+      "theoretical_loss": 3.428603649339882,
+      "tokens_seen": 2012545024
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039686868686868686,
+      "loss": 2.3663,
+      "theoretical_loss": 3.4285943519886164,
+      "tokens_seen": 2012610560
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039684848484848487,
+      "loss": 2.7802,
+      "theoretical_loss": 3.4285850550248584,
+      "tokens_seen": 2012676096
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003968282828282828,
+      "loss": 2.6398,
+      "theoretical_loss": 3.42857575844858,
+      "tokens_seen": 2012741632
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039680808080808083,
+      "loss": 2.457,
+      "theoretical_loss": 3.4285664622597514,
+      "tokens_seen": 2012807168
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003967878787878788,
+      "loss": 2.5914,
+      "theoretical_loss": 3.428557166458344,
+      "tokens_seen": 2012872704
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003967676767676768,
+      "loss": 2.4352,
+      "theoretical_loss": 3.4285478710443296,
+      "tokens_seen": 2012938240
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039674747474747476,
+      "loss": 2.3116,
+      "theoretical_loss": 3.428538576017679,
+      "tokens_seen": 2013003776
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003967272727272727,
+      "loss": 2.6761,
+      "theoretical_loss": 3.4285292813783634,
+      "tokens_seen": 2013069312
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039670707070707073,
+      "loss": 2.6778,
+      "theoretical_loss": 3.4285199871263545,
+      "tokens_seen": 2013134848
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003966868686868687,
+      "loss": 2.6637,
+      "theoretical_loss": 3.4285106932616225,
+      "tokens_seen": 2013200384
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003966666666666667,
+      "loss": 2.6304,
+      "theoretical_loss": 3.4285013997841403,
+      "tokens_seen": 2013265920
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039664646464646465,
+      "loss": 2.6416,
+      "theoretical_loss": 3.428492106693878,
+      "tokens_seen": 2013331456
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003966262626262626,
+      "loss": 2.6832,
+      "theoretical_loss": 3.428482813990807,
+      "tokens_seen": 2013396992
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003966060606060606,
+      "loss": 2.7265,
+      "theoretical_loss": 3.4284735216748983,
+      "tokens_seen": 2013462528
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039658585858585863,
+      "loss": 2.2403,
+      "theoretical_loss": 3.4284642297461243,
+      "tokens_seen": 2013528064
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003965656565656566,
+      "loss": 2.5304,
+      "theoretical_loss": 3.428454938204455,
+      "tokens_seen": 2013593600
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00039654545454545454,
+      "loss": 2.6138,
+      "theoretical_loss": 3.4284456470498625,
+      "tokens_seen": 2013659136
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003965252525252525,
+      "loss": 2.2908,
+      "theoretical_loss": 3.4284363562823175,
+      "tokens_seen": 2013724672
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003965050505050505,
+      "loss": 2.5686,
+      "theoretical_loss": 3.4284270659017917,
+      "tokens_seen": 2013790208
+    },
+    {
+      "epoch": 0.21,
+      "objective/train/docs_used": 1109221,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5685839653015137,
+      "objective/train/theoretical_loss": 3.4284177759082564,
+      "objective/train/tokens_used": 372714976,
+      "theoretical_loss": 3.4284177759082564,
+      "tokens_seen": 2013855744
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003964848484848485,
+      "loss": 2.6221,
+      "theoretical_loss": 3.4284177759082564,
+      "tokens_seen": 2013855744
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003964646464646465,
+      "loss": 2.5984,
+      "theoretical_loss": 3.4284084863016826,
+      "tokens_seen": 2013921280
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039644444444444444,
+      "loss": 2.4075,
+      "theoretical_loss": 3.428399197082042,
+      "tokens_seen": 2013986816
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003964242424242424,
+      "loss": 2.5087,
+      "theoretical_loss": 3.4283899082493052,
+      "tokens_seen": 2014052352
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039640404040404046,
+      "loss": 2.5658,
+      "theoretical_loss": 3.428380619803444,
+      "tokens_seen": 2014117888
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003963838383838384,
+      "loss": 2.7391,
+      "theoretical_loss": 3.4283713317444295,
+      "tokens_seen": 2014183424
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003963636363636364,
+      "loss": 2.7968,
+      "theoretical_loss": 3.4283620440722333,
+      "tokens_seen": 2014248960
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039634343434343433,
+      "loss": 2.6103,
+      "theoretical_loss": 3.4283527567868264,
+      "tokens_seen": 2014314496
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003963232323232323,
+      "loss": 2.7116,
+      "theoretical_loss": 3.4283434698881807,
+      "tokens_seen": 2014380032
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039630303030303035,
+      "loss": 2.7277,
+      "theoretical_loss": 3.4283341833762666,
+      "tokens_seen": 2014445568
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003962828282828283,
+      "loss": 2.7381,
+      "theoretical_loss": 3.4283248972510556,
+      "tokens_seen": 2014511104
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039626262626262627,
+      "loss": 2.6238,
+      "theoretical_loss": 3.42831561151252,
+      "tokens_seen": 2014576640
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003962424242424242,
+      "loss": 2.3142,
+      "theoretical_loss": 3.42830632616063,
+      "tokens_seen": 2014642176
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003962222222222222,
+      "loss": 2.741,
+      "theoretical_loss": 3.428297041195357,
+      "tokens_seen": 2014707712
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039620202020202024,
+      "loss": 2.7352,
+      "theoretical_loss": 3.4282877566166734,
+      "tokens_seen": 2014773248
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003961818181818182,
+      "loss": 2.6948,
+      "theoretical_loss": 3.428278472424549,
+      "tokens_seen": 2014838784
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039616161616161616,
+      "loss": 2.7039,
+      "theoretical_loss": 3.4282691886189562,
+      "tokens_seen": 2014904320
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003961414141414141,
+      "loss": 2.4815,
+      "theoretical_loss": 3.4282599051998663,
+      "tokens_seen": 2014969856
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003961212121212121,
+      "loss": 2.6714,
+      "theoretical_loss": 3.4282506221672504,
+      "tokens_seen": 2015035392
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039610101010101014,
+      "loss": 2.7503,
+      "theoretical_loss": 3.4282413395210796,
+      "tokens_seen": 2015100928
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003960808080808081,
+      "loss": 2.5656,
+      "theoretical_loss": 3.4282320572613254,
+      "tokens_seen": 2015166464
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039606060606060605,
+      "loss": 2.3771,
+      "theoretical_loss": 3.428222775387959,
+      "tokens_seen": 2015232000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000396040404040404,
+      "loss": 2.5924,
+      "theoretical_loss": 3.4282134939009525,
+      "tokens_seen": 2015297536
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000396020202020202,
+      "loss": 2.5664,
+      "theoretical_loss": 3.4282042128002765,
+      "tokens_seen": 2015363072
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039600000000000003,
+      "loss": 2.6235,
+      "theoretical_loss": 3.4281949320859026,
+      "tokens_seen": 2015428608
+    },
+    {
+      "epoch": 0.22,
+      "objective/train/docs_used": 1110352,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.888545274734497,
+      "objective/train/theoretical_loss": 3.4281856517578024,
+      "objective/train/tokens_used": 374353376,
+      "theoretical_loss": 3.4281856517578024,
+      "tokens_seen": 2015494144
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000395979797979798,
+      "loss": 2.527,
+      "theoretical_loss": 3.4281856517578024,
+      "tokens_seen": 2015494144
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039595959595959594,
+      "loss": 2.8353,
+      "theoretical_loss": 3.4281763718159466,
+      "tokens_seen": 2015559680
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039593939393939396,
+      "loss": 2.3749,
+      "theoretical_loss": 3.428167092260307,
+      "tokens_seen": 2015625216
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003959191919191919,
+      "loss": 2.5688,
+      "theoretical_loss": 3.4281578130908548,
+      "tokens_seen": 2015690752
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003958989898989899,
+      "loss": 2.6914,
+      "theoretical_loss": 3.428148534307562,
+      "tokens_seen": 2015756288
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003958787878787879,
+      "loss": 2.6239,
+      "theoretical_loss": 3.4281392559103994,
+      "tokens_seen": 2015821824
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039585858585858584,
+      "loss": 2.3188,
+      "theoretical_loss": 3.428129977899338,
+      "tokens_seen": 2015887360
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039583838383838385,
+      "loss": 2.5494,
+      "theoretical_loss": 3.4281207002743503,
+      "tokens_seen": 2015952896
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003958181818181818,
+      "loss": 2.5859,
+      "theoretical_loss": 3.4281114230354066,
+      "tokens_seen": 2016018432
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003957979797979798,
+      "loss": 2.6417,
+      "theoretical_loss": 3.428102146182479,
+      "tokens_seen": 2016083968
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039577777777777777,
+      "loss": 2.3921,
+      "theoretical_loss": 3.428092869715538,
+      "tokens_seen": 2016149504
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003957575757575758,
+      "loss": 2.8298,
+      "theoretical_loss": 3.4280835936345566,
+      "tokens_seen": 2016215040
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039573737373737374,
+      "loss": 2.6659,
+      "theoretical_loss": 3.4280743179395046,
+      "tokens_seen": 2016280576
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003957171717171717,
+      "loss": 2.5375,
+      "theoretical_loss": 3.4280650426303536,
+      "tokens_seen": 2016346112
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003956969696969697,
+      "loss": 2.6967,
+      "theoretical_loss": 3.428055767707076,
+      "tokens_seen": 2016411648
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039567676767676767,
+      "loss": 2.5046,
+      "theoretical_loss": 3.428046493169643,
+      "tokens_seen": 2016477184
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003956565656565657,
+      "loss": 2.6307,
+      "theoretical_loss": 3.428037219018025,
+      "tokens_seen": 2016542720
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039563636363636363,
+      "loss": 2.6712,
+      "theoretical_loss": 3.428027945252194,
+      "tokens_seen": 2016608256
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039561616161616164,
+      "loss": 2.6101,
+      "theoretical_loss": 3.4280186718721217,
+      "tokens_seen": 2016673792
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003955959595959596,
+      "loss": 2.6178,
+      "theoretical_loss": 3.428009398877779,
+      "tokens_seen": 2016739328
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003955757575757576,
+      "loss": 2.6002,
+      "theoretical_loss": 3.428000126269138,
+      "tokens_seen": 2016804864
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039555555555555557,
+      "loss": 2.4496,
+      "theoretical_loss": 3.427990854046169,
+      "tokens_seen": 2016870400
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003955353535353535,
+      "loss": 2.6548,
+      "theoretical_loss": 3.4279815822088455,
+      "tokens_seen": 2016935936
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039551515151515154,
+      "loss": 2.4536,
+      "theoretical_loss": 3.427972310757136,
+      "tokens_seen": 2017001472
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003954949494949495,
+      "loss": 2.6407,
+      "theoretical_loss": 3.4279630396910146,
+      "tokens_seen": 2017067008
+    },
+    {
+      "epoch": 0.22,
+      "objective/train/docs_used": 1110721,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7697877883911133,
+      "objective/train/theoretical_loss": 3.4279537690104513,
+      "objective/train/tokens_used": 375991776,
+      "theoretical_loss": 3.4279537690104513,
+      "tokens_seen": 2017132544
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003954747474747475,
+      "loss": 2.5957,
+      "theoretical_loss": 3.4279537690104513,
+      "tokens_seen": 2017132544
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039545454545454546,
+      "loss": 2.5364,
+      "theoretical_loss": 3.427944498715418,
+      "tokens_seen": 2017198080
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003954343434343434,
+      "loss": 2.8179,
+      "theoretical_loss": 3.427935228805886,
+      "tokens_seen": 2017263616
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039541414141414143,
+      "loss": 2.6458,
+      "theoretical_loss": 3.4279259592818265,
+      "tokens_seen": 2017329152
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039539393939393944,
+      "loss": 2.58,
+      "theoretical_loss": 3.4279166901432117,
+      "tokens_seen": 2017394688
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003953737373737374,
+      "loss": 2.2848,
+      "theoretical_loss": 3.427907421390012,
+      "tokens_seen": 2017460224
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039535353535353535,
+      "loss": 2.5158,
+      "theoretical_loss": 3.4278981530221997,
+      "tokens_seen": 2017525760
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003953333333333333,
+      "loss": 2.5589,
+      "theoretical_loss": 3.4278888850397458,
+      "tokens_seen": 2017591296
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003953131313131313,
+      "loss": 2.5954,
+      "theoretical_loss": 3.4278796174426223,
+      "tokens_seen": 2017656832
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039529292929292933,
+      "loss": 2.7181,
+      "theoretical_loss": 3.4278703502308003,
+      "tokens_seen": 2017722368
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003952727272727273,
+      "loss": 2.5333,
+      "theoretical_loss": 3.427861083404251,
+      "tokens_seen": 2017787904
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039525252525252525,
+      "loss": 2.4276,
+      "theoretical_loss": 3.4278518169629466,
+      "tokens_seen": 2017853440
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003952323232323232,
+      "loss": 2.5602,
+      "theoretical_loss": 3.4278425509068575,
+      "tokens_seen": 2017918976
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039521212121212127,
+      "loss": 2.7124,
+      "theoretical_loss": 3.4278332852359563,
+      "tokens_seen": 2017984512
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003951919191919192,
+      "loss": 2.6338,
+      "theoretical_loss": 3.4278240199502137,
+      "tokens_seen": 2018050048
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003951717171717172,
+      "loss": 2.5743,
+      "theoretical_loss": 3.427814755049601,
+      "tokens_seen": 2018115584
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039515151515151514,
+      "loss": 2.7207,
+      "theoretical_loss": 3.4278054905340913,
+      "tokens_seen": 2018181120
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003951313131313131,
+      "loss": 2.7285,
+      "theoretical_loss": 3.4277962264036543,
+      "tokens_seen": 2018246656
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039511111111111116,
+      "loss": 2.3587,
+      "theoretical_loss": 3.427786962658262,
+      "tokens_seen": 2018312192
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003950909090909091,
+      "loss": 2.4289,
+      "theoretical_loss": 3.427777699297886,
+      "tokens_seen": 2018377728
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003950707070707071,
+      "loss": 2.6449,
+      "theoretical_loss": 3.4277684363224985,
+      "tokens_seen": 2018443264
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039505050505050503,
+      "loss": 2.3899,
+      "theoretical_loss": 3.4277591737320696,
+      "tokens_seen": 2018508800
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039503030303030304,
+      "loss": 2.6567,
+      "theoretical_loss": 3.427749911526572,
+      "tokens_seen": 2018574336
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039501010101010105,
+      "loss": 2.7126,
+      "theoretical_loss": 3.427740649705976,
+      "tokens_seen": 2018639872
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000394989898989899,
+      "loss": 2.6026,
+      "theoretical_loss": 3.4277313882702547,
+      "tokens_seen": 2018705408
+    },
+    {
+      "epoch": 0.22,
+      "objective/train/docs_used": 1111912,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.503018856048584,
+      "objective/train/theoretical_loss": 3.4277221272193783,
+      "objective/train/tokens_used": 377630176,
+      "theoretical_loss": 3.4277221272193783,
+      "tokens_seen": 2018770944
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039496969696969697,
+      "loss": 2.6544,
+      "theoretical_loss": 3.4277221272193783,
+      "tokens_seen": 2018770944
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003949494949494949,
+      "loss": 2.7159,
+      "theoretical_loss": 3.4277128665533194,
+      "tokens_seen": 2018836480
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039492929292929294,
+      "loss": 2.6576,
+      "theoretical_loss": 3.427703606272048,
+      "tokens_seen": 2018902016
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039490909090909095,
+      "loss": 2.9404,
+      "theoretical_loss": 3.427694346375537,
+      "tokens_seen": 2018967552
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003948888888888889,
+      "loss": 2.5293,
+      "theoretical_loss": 3.4276850868637574,
+      "tokens_seen": 2019033088
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039486868686868686,
+      "loss": 2.7324,
+      "theoretical_loss": 3.4276758277366803,
+      "tokens_seen": 2019098624
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039484848484848487,
+      "loss": 2.5299,
+      "theoretical_loss": 3.4276665689942782,
+      "tokens_seen": 2019164160
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039482828282828283,
+      "loss": 2.6959,
+      "theoretical_loss": 3.427657310636522,
+      "tokens_seen": 2019229696
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039480808080808084,
+      "loss": 2.5669,
+      "theoretical_loss": 3.4276480526633835,
+      "tokens_seen": 2019295232
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003947878787878788,
+      "loss": 2.7492,
+      "theoretical_loss": 3.4276387950748335,
+      "tokens_seen": 2019360768
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039476767676767675,
+      "loss": 2.3663,
+      "theoretical_loss": 3.4276295378708452,
+      "tokens_seen": 2019426304
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039474747474747476,
+      "loss": 2.4772,
+      "theoretical_loss": 3.427620281051388,
+      "tokens_seen": 2019491840
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003947272727272727,
+      "loss": 2.5573,
+      "theoretical_loss": 3.4276110246164353,
+      "tokens_seen": 2019557376
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039470707070707073,
+      "loss": 2.5873,
+      "theoretical_loss": 3.4276017685659577,
+      "tokens_seen": 2019622912
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003946868686868687,
+      "loss": 2.6352,
+      "theoretical_loss": 3.4275925128999267,
+      "tokens_seen": 2019688448
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003946666666666667,
+      "loss": 2.3906,
+      "theoretical_loss": 3.4275832576183145,
+      "tokens_seen": 2019753984
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039464646464646466,
+      "loss": 2.5083,
+      "theoretical_loss": 3.4275740027210917,
+      "tokens_seen": 2019819520
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003946262626262626,
+      "loss": 2.5144,
+      "theoretical_loss": 3.4275647482082308,
+      "tokens_seen": 2019885056
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003946060606060606,
+      "loss": 2.6141,
+      "theoretical_loss": 3.4275554940797033,
+      "tokens_seen": 2019950592
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003945858585858586,
+      "loss": 2.5868,
+      "theoretical_loss": 3.42754624033548,
+      "tokens_seen": 2020016128
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003945656565656566,
+      "loss": 2.3907,
+      "theoretical_loss": 3.427536986975533,
+      "tokens_seen": 2020081664
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039454545454545455,
+      "loss": 2.5512,
+      "theoretical_loss": 3.427527733999834,
+      "tokens_seen": 2020147200
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003945252525252525,
+      "loss": 2.4261,
+      "theoretical_loss": 3.427518481408354,
+      "tokens_seen": 2020212736
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003945050505050505,
+      "loss": 2.4513,
+      "theoretical_loss": 3.427509229201066,
+      "tokens_seen": 2020278272
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003944848484848485,
+      "loss": 2.4909,
+      "theoretical_loss": 3.4274999773779395,
+      "tokens_seen": 2020343808
+    },
+    {
+      "epoch": 0.22,
+      "objective/train/docs_used": 1112399,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.484231472015381,
+      "objective/train/theoretical_loss": 3.4274907259389478,
+      "objective/train/tokens_used": 379268576,
+      "theoretical_loss": 3.4274907259389478,
+      "tokens_seen": 2020409344
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003944646464646465,
+      "loss": 2.6462,
+      "theoretical_loss": 3.4274907259389478,
+      "tokens_seen": 2020409344
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039444444444444444,
+      "loss": 2.5147,
+      "theoretical_loss": 3.427481474884061,
+      "tokens_seen": 2020474880
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003944242424242424,
+      "loss": 2.5727,
+      "theoretical_loss": 3.4274722242132523,
+      "tokens_seen": 2020540416
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003944040404040404,
+      "loss": 2.646,
+      "theoretical_loss": 3.4274629739264926,
+      "tokens_seen": 2020605952
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003943838383838384,
+      "loss": 2.3592,
+      "theoretical_loss": 3.427453724023753,
+      "tokens_seen": 2020671488
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003943636363636364,
+      "loss": 2.5641,
+      "theoretical_loss": 3.427444474505006,
+      "tokens_seen": 2020737024
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039434343434343434,
+      "loss": 2.6167,
+      "theoretical_loss": 3.427435225370222,
+      "tokens_seen": 2020802560
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003943232323232323,
+      "loss": 2.5598,
+      "theoretical_loss": 3.427425976619374,
+      "tokens_seen": 2020868096
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003943030303030303,
+      "loss": 2.7719,
+      "theoretical_loss": 3.427416728252433,
+      "tokens_seen": 2020933632
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003942828282828283,
+      "loss": 2.5257,
+      "theoretical_loss": 3.4274074802693706,
+      "tokens_seen": 2020999168
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039426262626262627,
+      "loss": 2.7512,
+      "theoretical_loss": 3.427398232670158,
+      "tokens_seen": 2021064704
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039424242424242423,
+      "loss": 2.4711,
+      "theoretical_loss": 3.4273889854547672,
+      "tokens_seen": 2021130240
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003942222222222222,
+      "loss": 2.6363,
+      "theoretical_loss": 3.4273797386231704,
+      "tokens_seen": 2021195776
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039420202020202025,
+      "loss": 2.66,
+      "theoretical_loss": 3.4273704921753385,
+      "tokens_seen": 2021261312
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003941818181818182,
+      "loss": 2.5107,
+      "theoretical_loss": 3.427361246111243,
+      "tokens_seen": 2021326848
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039416161616161616,
+      "loss": 2.5211,
+      "theoretical_loss": 3.427352000430856,
+      "tokens_seen": 2021392384
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003941414141414141,
+      "loss": 2.6448,
+      "theoretical_loss": 3.4273427551341493,
+      "tokens_seen": 2021457920
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003941212121212121,
+      "loss": 2.4784,
+      "theoretical_loss": 3.427333510221094,
+      "tokens_seen": 2021523456
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039410101010101014,
+      "loss": 2.6683,
+      "theoretical_loss": 3.4273242656916616,
+      "tokens_seen": 2021588992
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003940808080808081,
+      "loss": 2.4676,
+      "theoretical_loss": 3.4273150215458243,
+      "tokens_seen": 2021654528
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039406060606060606,
+      "loss": 2.4458,
+      "theoretical_loss": 3.4273057777835536,
+      "tokens_seen": 2021720064
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000394040404040404,
+      "loss": 2.4926,
+      "theoretical_loss": 3.4272965344048214,
+      "tokens_seen": 2021785600
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000394020202020202,
+      "loss": 2.4265,
+      "theoretical_loss": 3.427287291409599,
+      "tokens_seen": 2021851136
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039400000000000004,
+      "loss": 2.54,
+      "theoretical_loss": 3.4272780487978576,
+      "tokens_seen": 2021916672
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000393979797979798,
+      "loss": 2.4683,
+      "theoretical_loss": 3.42726880656957,
+      "tokens_seen": 2021982208
+    },
+    {
+      "epoch": 0.22,
+      "objective/train/docs_used": 1113663,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.153092622756958,
+      "objective/train/theoretical_loss": 3.4272595647247073,
+      "objective/train/tokens_used": 380906976,
+      "theoretical_loss": 3.4272595647247073,
+      "tokens_seen": 2022047744
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039395959595959595,
+      "loss": 2.5295,
+      "theoretical_loss": 3.4272595647247073,
+      "tokens_seen": 2022047744
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003939393939393939,
+      "loss": 2.4037,
+      "theoretical_loss": 3.42725032326324,
+      "tokens_seen": 2022113280
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039391919191919197,
+      "loss": 2.6336,
+      "theoretical_loss": 3.4272410821851422,
+      "tokens_seen": 2022178816
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039389898989898993,
+      "loss": 2.4541,
+      "theoretical_loss": 3.427231841490384,
+      "tokens_seen": 2022244352
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003938787878787879,
+      "loss": 2.4697,
+      "theoretical_loss": 3.427222601178937,
+      "tokens_seen": 2022309888
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039385858585858584,
+      "loss": 2.6019,
+      "theoretical_loss": 3.4272133612507734,
+      "tokens_seen": 2022375424
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039383838383838385,
+      "loss": 2.6577,
+      "theoretical_loss": 3.427204121705864,
+      "tokens_seen": 2022440960
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039381818181818186,
+      "loss": 2.6264,
+      "theoretical_loss": 3.427194882544182,
+      "tokens_seen": 2022506496
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003937979797979798,
+      "loss": 2.5512,
+      "theoretical_loss": 3.4271856437656982,
+      "tokens_seen": 2022572032
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003937777777777778,
+      "loss": 2.5572,
+      "theoretical_loss": 3.4271764053703837,
+      "tokens_seen": 2022637568
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039375757575757574,
+      "loss": 2.7579,
+      "theoretical_loss": 3.4271671673582116,
+      "tokens_seen": 2022703104
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039373737373737375,
+      "loss": 2.4911,
+      "theoretical_loss": 3.4271579297291526,
+      "tokens_seen": 2022768640
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039371717171717176,
+      "loss": 2.6171,
+      "theoretical_loss": 3.4271486924831787,
+      "tokens_seen": 2022834176
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003936969696969697,
+      "loss": 2.635,
+      "theoretical_loss": 3.4271394556202615,
+      "tokens_seen": 2022899712
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039367676767676767,
+      "loss": 2.4454,
+      "theoretical_loss": 3.4271302191403725,
+      "tokens_seen": 2022965248
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003936565656565657,
+      "loss": 2.5407,
+      "theoretical_loss": 3.4271209830434843,
+      "tokens_seen": 2023030784
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039363636363636364,
+      "loss": 2.4783,
+      "theoretical_loss": 3.427111747329567,
+      "tokens_seen": 2023096320
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039361616161616165,
+      "loss": 2.815,
+      "theoretical_loss": 3.4271025119985943,
+      "tokens_seen": 2023161856
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003935959595959596,
+      "loss": 2.2448,
+      "theoretical_loss": 3.4270932770505365,
+      "tokens_seen": 2023227392
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039357575757575756,
+      "loss": 2.5646,
+      "theoretical_loss": 3.4270840424853657,
+      "tokens_seen": 2023292928
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003935555555555556,
+      "loss": 2.5724,
+      "theoretical_loss": 3.427074808303054,
+      "tokens_seen": 2023358464
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039353535353535353,
+      "loss": 2.5373,
+      "theoretical_loss": 3.427065574503572,
+      "tokens_seen": 2023424000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039351515151515154,
+      "loss": 2.6068,
+      "theoretical_loss": 3.427056341086893,
+      "tokens_seen": 2023489536
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003934949494949495,
+      "loss": 2.4802,
+      "theoretical_loss": 3.427047108052988,
+      "tokens_seen": 2023555072
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003934747474747475,
+      "loss": 2.5053,
+      "theoretical_loss": 3.427037875401828,
+      "tokens_seen": 2023620608
+    },
+    {
+      "epoch": 0.22,
+      "objective/train/docs_used": 1114958,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.473273754119873,
+      "objective/train/theoretical_loss": 3.4270286431333856,
+      "objective/train/tokens_used": 382545376,
+      "theoretical_loss": 3.4270286431333856,
+      "tokens_seen": 2023686144
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039345454545454547,
+      "loss": 2.3698,
+      "theoretical_loss": 3.4270286431333856,
+      "tokens_seen": 2023686144
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003934343434343434,
+      "loss": 2.5781,
+      "theoretical_loss": 3.4270194112476324,
+      "tokens_seen": 2023751680
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039341414141414144,
+      "loss": 2.6656,
+      "theoretical_loss": 3.4270101797445403,
+      "tokens_seen": 2023817216
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003933939393939394,
+      "loss": 2.383,
+      "theoretical_loss": 3.4270009486240807,
+      "tokens_seen": 2023882752
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003933737373737374,
+      "loss": 2.6131,
+      "theoretical_loss": 3.4269917178862253,
+      "tokens_seen": 2023948288
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039335353535353536,
+      "loss": 2.5708,
+      "theoretical_loss": 3.426982487530947,
+      "tokens_seen": 2024013824
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003933333333333333,
+      "loss": 2.3579,
+      "theoretical_loss": 3.426973257558216,
+      "tokens_seen": 2024079360
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039331313131313133,
+      "loss": 2.57,
+      "theoretical_loss": 3.4269640279680043,
+      "tokens_seen": 2024144896
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039329292929292934,
+      "loss": 2.662,
+      "theoretical_loss": 3.4269547987602844,
+      "tokens_seen": 2024210432
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003932727272727273,
+      "loss": 2.6302,
+      "theoretical_loss": 3.4269455699350275,
+      "tokens_seen": 2024275968
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039325252525252525,
+      "loss": 2.3562,
+      "theoretical_loss": 3.4269363414922056,
+      "tokens_seen": 2024341504
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003932323232323232,
+      "loss": 2.6728,
+      "theoretical_loss": 3.426927113431791,
+      "tokens_seen": 2024407040
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003932121212121212,
+      "loss": 2.5875,
+      "theoretical_loss": 3.4269178857537543,
+      "tokens_seen": 2024472576
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039319191919191923,
+      "loss": 2.8369,
+      "theoretical_loss": 3.426908658458068,
+      "tokens_seen": 2024538112
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003931717171717172,
+      "loss": 2.7961,
+      "theoretical_loss": 3.426899431544704,
+      "tokens_seen": 2024603648
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039315151515151515,
+      "loss": 2.401,
+      "theoretical_loss": 3.426890205013634,
+      "tokens_seen": 2024669184
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003931313131313131,
+      "loss": 2.5557,
+      "theoretical_loss": 3.4268809788648293,
+      "tokens_seen": 2024734720
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039311111111111117,
+      "loss": 2.4,
+      "theoretical_loss": 3.426871753098262,
+      "tokens_seen": 2024800256
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003930909090909091,
+      "loss": 2.7028,
+      "theoretical_loss": 3.4268625277139044,
+      "tokens_seen": 2024865792
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003930707070707071,
+      "loss": 2.5791,
+      "theoretical_loss": 3.426853302711727,
+      "tokens_seen": 2024931328
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039305050505050504,
+      "loss": 2.6614,
+      "theoretical_loss": 3.4268440780917038,
+      "tokens_seen": 2024996864
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000393030303030303,
+      "loss": 2.5811,
+      "theoretical_loss": 3.426834853853804,
+      "tokens_seen": 2025062400
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039301010101010106,
+      "loss": 2.6949,
+      "theoretical_loss": 3.4268256299980013,
+      "tokens_seen": 2025127936
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000392989898989899,
+      "loss": 2.6975,
+      "theoretical_loss": 3.4268164065242663,
+      "tokens_seen": 2025193472
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000392969696969697,
+      "loss": 2.5428,
+      "theoretical_loss": 3.4268071834325715,
+      "tokens_seen": 2025259008
+    },
+    {
+      "epoch": 0.22,
+      "objective/train/docs_used": 1115580,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.886110544204712,
+      "objective/train/theoretical_loss": 3.4267979607228893,
+      "objective/train/tokens_used": 384183776,
+      "theoretical_loss": 3.4267979607228893,
+      "tokens_seen": 2025324544
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039294949494949493,
+      "loss": 2.5216,
+      "theoretical_loss": 3.4267979607228893,
+      "tokens_seen": 2025324544
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003929292929292929,
+      "loss": 2.5705,
+      "theoretical_loss": 3.42678873839519,
+      "tokens_seen": 2025390080
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039290909090909095,
+      "loss": 2.735,
+      "theoretical_loss": 3.4267795164494466,
+      "tokens_seen": 2025455616
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003928888888888889,
+      "loss": 2.6675,
+      "theoretical_loss": 3.42677029488563,
+      "tokens_seen": 2025521152
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039286868686868687,
+      "loss": 2.453,
+      "theoretical_loss": 3.4267610737037133,
+      "tokens_seen": 2025586688
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003928484848484848,
+      "loss": 2.6699,
+      "theoretical_loss": 3.426751852903667,
+      "tokens_seen": 2025652224
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039282828282828283,
+      "loss": 2.5889,
+      "theoretical_loss": 3.426742632485464,
+      "tokens_seen": 2025717760
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039280808080808085,
+      "loss": 2.5707,
+      "theoretical_loss": 3.4267334124490754,
+      "tokens_seen": 2025783296
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003927878787878788,
+      "loss": 2.5891,
+      "theoretical_loss": 3.426724192794473,
+      "tokens_seen": 2025848832
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039276767676767676,
+      "loss": 2.5181,
+      "theoretical_loss": 3.426714973521629,
+      "tokens_seen": 2025914368
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003927474747474747,
+      "loss": 2.4398,
+      "theoretical_loss": 3.4267057546305155,
+      "tokens_seen": 2025979904
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039272727272727273,
+      "loss": 2.5029,
+      "theoretical_loss": 3.426696536121104,
+      "tokens_seen": 2026045440
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039270707070707074,
+      "loss": 2.6772,
+      "theoretical_loss": 3.426687317993366,
+      "tokens_seen": 2026110976
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003926868686868687,
+      "loss": 2.4527,
+      "theoretical_loss": 3.4266781002472744,
+      "tokens_seen": 2026176512
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039266666666666665,
+      "loss": 2.5978,
+      "theoretical_loss": 3.4266688828828,
+      "tokens_seen": 2026242048
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039264646464646466,
+      "loss": 2.4343,
+      "theoretical_loss": 3.4266596658999147,
+      "tokens_seen": 2026307584
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003926262626262626,
+      "loss": 2.6968,
+      "theoretical_loss": 3.426650449298591,
+      "tokens_seen": 2026373120
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039260606060606063,
+      "loss": 2.589,
+      "theoretical_loss": 3.4266412330788008,
+      "tokens_seen": 2026438656
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003925858585858586,
+      "loss": 2.604,
+      "theoretical_loss": 3.4266320172405154,
+      "tokens_seen": 2026504192
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039256565656565654,
+      "loss": 2.5602,
+      "theoretical_loss": 3.426622801783707,
+      "tokens_seen": 2026569728
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039254545454545456,
+      "loss": 2.5963,
+      "theoretical_loss": 3.426613586708347,
+      "tokens_seen": 2026635264
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003925252525252525,
+      "loss": 2.4481,
+      "theoretical_loss": 3.426604372014408,
+      "tokens_seen": 2026700800
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003925050505050505,
+      "loss": 2.5384,
+      "theoretical_loss": 3.426595157701861,
+      "tokens_seen": 2026766336
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003924848484848485,
+      "loss": 2.496,
+      "theoretical_loss": 3.426585943770679,
+      "tokens_seen": 2026831872
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003924646464646465,
+      "loss": 2.6017,
+      "theoretical_loss": 3.426576730220833,
+      "tokens_seen": 2026897408
+    },
+    {
+      "epoch": 0.22,
+      "objective/train/docs_used": 1116847,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7708523273468018,
+      "objective/train/theoretical_loss": 3.4265675170522956,
+      "objective/train/tokens_used": 385822176,
+      "theoretical_loss": 3.4265675170522956,
+      "tokens_seen": 2026962944
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039244444444444445,
+      "loss": 2.6355,
+      "theoretical_loss": 3.4265675170522956,
+      "tokens_seen": 2026962944
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003924242424242424,
+      "loss": 2.7949,
+      "theoretical_loss": 3.426558304265038,
+      "tokens_seen": 2027028480
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003924040404040404,
+      "loss": 2.6705,
+      "theoretical_loss": 3.426549091859032,
+      "tokens_seen": 2027094016
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003923838383838384,
+      "loss": 2.7159,
+      "theoretical_loss": 3.426539879834251,
+      "tokens_seen": 2027159552
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003923636363636364,
+      "loss": 2.689,
+      "theoretical_loss": 3.4265306681906647,
+      "tokens_seen": 2027225088
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039234343434343434,
+      "loss": 2.5202,
+      "theoretical_loss": 3.4265214569282465,
+      "tokens_seen": 2027290624
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003923232323232323,
+      "loss": 2.5135,
+      "theoretical_loss": 3.4265122460469675,
+      "tokens_seen": 2027356160
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003923030303030303,
+      "loss": 2.5385,
+      "theoretical_loss": 3.4265030355468005,
+      "tokens_seen": 2027421696
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003922828282828283,
+      "loss": 2.635,
+      "theoretical_loss": 3.4264938254277166,
+      "tokens_seen": 2027487232
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003922626262626263,
+      "loss": 2.5788,
+      "theoretical_loss": 3.426484615689688,
+      "tokens_seen": 2027552768
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039224242424242423,
+      "loss": 2.5681,
+      "theoretical_loss": 3.4264754063326865,
+      "tokens_seen": 2027618304
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039222222222222225,
+      "loss": 2.441,
+      "theoretical_loss": 3.426466197356685,
+      "tokens_seen": 2027683840
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003922020202020202,
+      "loss": 2.5216,
+      "theoretical_loss": 3.426456988761654,
+      "tokens_seen": 2027749376
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003921818181818182,
+      "loss": 2.4985,
+      "theoretical_loss": 3.4264477805475657,
+      "tokens_seen": 2027814912
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039216161616161617,
+      "loss": 2.5328,
+      "theoretical_loss": 3.426438572714393,
+      "tokens_seen": 2027880448
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003921414141414141,
+      "loss": 2.3552,
+      "theoretical_loss": 3.4264293652621065,
+      "tokens_seen": 2027945984
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039212121212121214,
+      "loss": 2.5907,
+      "theoretical_loss": 3.4264201581906795,
+      "tokens_seen": 2028011520
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039210101010101015,
+      "loss": 2.4451,
+      "theoretical_loss": 3.426410951500083,
+      "tokens_seen": 2028077056
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003920808080808081,
+      "loss": 2.6231,
+      "theoretical_loss": 3.4264017451902893,
+      "tokens_seen": 2028142592
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039206060606060606,
+      "loss": 2.6823,
+      "theoretical_loss": 3.4263925392612697,
+      "tokens_seen": 2028208128
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000392040404040404,
+      "loss": 2.5544,
+      "theoretical_loss": 3.426383333712997,
+      "tokens_seen": 2028273664
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039202020202020203,
+      "loss": 2.832,
+      "theoretical_loss": 3.4263741285454428,
+      "tokens_seen": 2028339200
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039200000000000004,
+      "loss": 2.3735,
+      "theoretical_loss": 3.4263649237585794,
+      "tokens_seen": 2028404736
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000391979797979798,
+      "loss": 2.6263,
+      "theoretical_loss": 3.426355719352378,
+      "tokens_seen": 2028470272
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039195959595959596,
+      "loss": 2.4902,
+      "theoretical_loss": 3.4263465153268116,
+      "tokens_seen": 2028535808
+    },
+    {
+      "epoch": 0.22,
+      "objective/train/docs_used": 1117380,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6172401905059814,
+      "objective/train/theoretical_loss": 3.426337311681851,
+      "objective/train/tokens_used": 387460576,
+      "theoretical_loss": 3.426337311681851,
+      "tokens_seen": 2028601344
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003919393939393939,
+      "loss": 2.4793,
+      "theoretical_loss": 3.426337311681851,
+      "tokens_seen": 2028601344
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000391919191919192,
+      "loss": 2.4009,
+      "theoretical_loss": 3.4263281084174695,
+      "tokens_seen": 2028666880
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039189898989898993,
+      "loss": 2.492,
+      "theoretical_loss": 3.4263189055336376,
+      "tokens_seen": 2028732416
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003918787878787879,
+      "loss": 2.4839,
+      "theoretical_loss": 3.426309703030328,
+      "tokens_seen": 2028797952
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039185858585858585,
+      "loss": 2.7173,
+      "theoretical_loss": 3.426300500907513,
+      "tokens_seen": 2028863488
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003918383838383838,
+      "loss": 2.5024,
+      "theoretical_loss": 3.4262912991651637,
+      "tokens_seen": 2028929024
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039181818181818187,
+      "loss": 2.6256,
+      "theoretical_loss": 3.426282097803253,
+      "tokens_seen": 2028994560
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003917979797979798,
+      "loss": 2.638,
+      "theoretical_loss": 3.426272896821753,
+      "tokens_seen": 2029060096
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003917777777777778,
+      "loss": 2.3982,
+      "theoretical_loss": 3.4262636962206345,
+      "tokens_seen": 2029125632
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039175757575757574,
+      "loss": 2.3516,
+      "theoretical_loss": 3.42625449599987,
+      "tokens_seen": 2029191168
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039173737373737375,
+      "loss": 2.4793,
+      "theoretical_loss": 3.4262452961594323,
+      "tokens_seen": 2029256704
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039171717171717176,
+      "loss": 2.5811,
+      "theoretical_loss": 3.4262360966992924,
+      "tokens_seen": 2029322240
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003916969696969697,
+      "loss": 2.6168,
+      "theoretical_loss": 3.426226897619423,
+      "tokens_seen": 2029387776
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003916767676767677,
+      "loss": 2.4779,
+      "theoretical_loss": 3.426217698919795,
+      "tokens_seen": 2029453312
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039165656565656563,
+      "loss": 2.4986,
+      "theoretical_loss": 3.426208500600382,
+      "tokens_seen": 2029518848
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039163636363636364,
+      "loss": 2.7759,
+      "theoretical_loss": 3.426199302661155,
+      "tokens_seen": 2029584384
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039161616161616166,
+      "loss": 2.4451,
+      "theoretical_loss": 3.4261901051020858,
+      "tokens_seen": 2029649920
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003915959595959596,
+      "loss": 2.4326,
+      "theoretical_loss": 3.426180907923147,
+      "tokens_seen": 2029715456
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039157575757575757,
+      "loss": 2.4658,
+      "theoretical_loss": 3.4261717111243106,
+      "tokens_seen": 2029780992
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003915555555555556,
+      "loss": 2.4223,
+      "theoretical_loss": 3.426162514705548,
+      "tokens_seen": 2029846528
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039153535353535354,
+      "loss": 2.7919,
+      "theoretical_loss": 3.426153318666832,
+      "tokens_seen": 2029912064
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039151515151515155,
+      "loss": 2.6321,
+      "theoretical_loss": 3.426144123008134,
+      "tokens_seen": 2029977600
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003914949494949495,
+      "loss": 2.5509,
+      "theoretical_loss": 3.426134927729427,
+      "tokens_seen": 2030043136
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00039147474747474746,
+      "loss": 2.5512,
+      "theoretical_loss": 3.426125732830682,
+      "tokens_seen": 2030108672
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003914545454545455,
+      "loss": 2.5388,
+      "theoretical_loss": 3.426116538311871,
+      "tokens_seen": 2030174208
+    },
+    {
+      "epoch": 0.22,
+      "objective/train/docs_used": 1118494,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7309558391571045,
+      "objective/train/theoretical_loss": 3.4261073441729666,
+      "objective/train/tokens_used": 389098976,
+      "theoretical_loss": 3.4261073441729666,
+      "tokens_seen": 2030239744
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039143434343434343,
+      "loss": 2.6567,
+      "theoretical_loss": 3.4261073441729666,
+      "tokens_seen": 2030239744
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039141414141414144,
+      "loss": 2.4831,
+      "theoretical_loss": 3.426098150413941,
+      "tokens_seen": 2030305280
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003913939393939394,
+      "loss": 2.6372,
+      "theoretical_loss": 3.4260889570347652,
+      "tokens_seen": 2030370816
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003913737373737374,
+      "loss": 2.5891,
+      "theoretical_loss": 3.426079764035413,
+      "tokens_seen": 2030436352
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039135353535353537,
+      "loss": 2.5801,
+      "theoretical_loss": 3.4260705714158544,
+      "tokens_seen": 2030501888
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003913333333333333,
+      "loss": 2.8033,
+      "theoretical_loss": 3.426061379176063,
+      "tokens_seen": 2030567424
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039131313131313133,
+      "loss": 2.4674,
+      "theoretical_loss": 3.42605218731601,
+      "tokens_seen": 2030632960
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003912929292929293,
+      "loss": 2.469,
+      "theoretical_loss": 3.426042995835668,
+      "tokens_seen": 2030698496
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003912727272727273,
+      "loss": 2.672,
+      "theoretical_loss": 3.4260338047350087,
+      "tokens_seen": 2030764032
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039125252525252526,
+      "loss": 2.5594,
+      "theoretical_loss": 3.426024614014004,
+      "tokens_seen": 2030829568
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003912323232323232,
+      "loss": 2.5454,
+      "theoretical_loss": 3.426015423672627,
+      "tokens_seen": 2030895104
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003912121212121212,
+      "loss": 2.48,
+      "theoretical_loss": 3.426006233710848,
+      "tokens_seen": 2030960640
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003911919191919192,
+      "loss": 2.5707,
+      "theoretical_loss": 3.425997044128641,
+      "tokens_seen": 2031026176
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003911717171717172,
+      "loss": 2.5987,
+      "theoretical_loss": 3.425987854925977,
+      "tokens_seen": 2031091712
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039115151515151515,
+      "loss": 2.5819,
+      "theoretical_loss": 3.4259786661028278,
+      "tokens_seen": 2031157248
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003911313131313131,
+      "loss": 2.6741,
+      "theoretical_loss": 3.4259694776591663,
+      "tokens_seen": 2031222784
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003911111111111111,
+      "loss": 2.6763,
+      "theoretical_loss": 3.425960289594964,
+      "tokens_seen": 2031288320
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039109090909090913,
+      "loss": 2.6218,
+      "theoretical_loss": 3.4259511019101936,
+      "tokens_seen": 2031353856
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003910707070707071,
+      "loss": 2.5182,
+      "theoretical_loss": 3.4259419146048264,
+      "tokens_seen": 2031419392
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039105050505050504,
+      "loss": 2.7424,
+      "theoretical_loss": 3.425932727678835,
+      "tokens_seen": 2031484928
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000391030303030303,
+      "loss": 2.5233,
+      "theoretical_loss": 3.425923541132191,
+      "tokens_seen": 2031550464
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000391010101010101,
+      "loss": 2.8571,
+      "theoretical_loss": 3.4259143549648674,
+      "tokens_seen": 2031616000
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000390989898989899,
+      "loss": 2.4963,
+      "theoretical_loss": 3.4259051691768354,
+      "tokens_seen": 2031681536
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000390969696969697,
+      "loss": 2.6932,
+      "theoretical_loss": 3.425895983768068,
+      "tokens_seen": 2031747072
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039094949494949494,
+      "loss": 2.5079,
+      "theoretical_loss": 3.425886798738537,
+      "tokens_seen": 2031812608
+    },
+    {
+      "epoch": 0.23,
+      "objective/train/docs_used": 1119152,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6228444576263428,
+      "objective/train/theoretical_loss": 3.425877614088213,
+      "objective/train/tokens_used": 390737376,
+      "theoretical_loss": 3.425877614088213,
+      "tokens_seen": 2031878144
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003909292929292929,
+      "loss": 2.5856,
+      "theoretical_loss": 3.425877614088213,
+      "tokens_seen": 2031878144
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039090909090909096,
+      "loss": 2.6037,
+      "theoretical_loss": 3.4258684298170703,
+      "tokens_seen": 2031943680
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003908888888888889,
+      "loss": 2.8355,
+      "theoretical_loss": 3.4258592459250803,
+      "tokens_seen": 2032009216
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039086868686868687,
+      "loss": 2.6795,
+      "theoretical_loss": 3.4258500624122146,
+      "tokens_seen": 2032074752
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039084848484848483,
+      "loss": 2.496,
+      "theoretical_loss": 3.4258408792784456,
+      "tokens_seen": 2032140288
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003908282828282828,
+      "loss": 2.5187,
+      "theoretical_loss": 3.425831696523746,
+      "tokens_seen": 2032205824
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039080808080808085,
+      "loss": 2.5352,
+      "theoretical_loss": 3.425822514148087,
+      "tokens_seen": 2032271360
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003907878787878788,
+      "loss": 2.571,
+      "theoretical_loss": 3.4258133321514412,
+      "tokens_seen": 2032336896
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039076767676767677,
+      "loss": 2.4906,
+      "theoretical_loss": 3.4258041505337813,
+      "tokens_seen": 2032402432
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003907474747474747,
+      "loss": 2.6286,
+      "theoretical_loss": 3.425794969295078,
+      "tokens_seen": 2032467968
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039072727272727273,
+      "loss": 2.4928,
+      "theoretical_loss": 3.425785788435305,
+      "tokens_seen": 2032533504
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039070707070707074,
+      "loss": 2.7237,
+      "theoretical_loss": 3.4257766079544334,
+      "tokens_seen": 2032599040
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003906868686868687,
+      "loss": 2.4682,
+      "theoretical_loss": 3.4257674278524353,
+      "tokens_seen": 2032664576
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039066666666666666,
+      "loss": 2.624,
+      "theoretical_loss": 3.4257582481292834,
+      "tokens_seen": 2032730112
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003906464646464646,
+      "loss": 2.5292,
+      "theoretical_loss": 3.42574906878495,
+      "tokens_seen": 2032795648
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003906262626262626,
+      "loss": 2.6357,
+      "theoretical_loss": 3.4257398898194067,
+      "tokens_seen": 2032861184
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039060606060606064,
+      "loss": 2.6523,
+      "theoretical_loss": 3.425730711232626,
+      "tokens_seen": 2032926720
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003905858585858586,
+      "loss": 2.5825,
+      "theoretical_loss": 3.4257215330245794,
+      "tokens_seen": 2032992256
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039056565656565655,
+      "loss": 2.5612,
+      "theoretical_loss": 3.42571235519524,
+      "tokens_seen": 2033057792
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039054545454545456,
+      "loss": 2.4326,
+      "theoretical_loss": 3.4257031777445803,
+      "tokens_seen": 2033123328
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003905252525252525,
+      "loss": 2.8784,
+      "theoretical_loss": 3.425694000672571,
+      "tokens_seen": 2033188864
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039050505050505053,
+      "loss": 2.4251,
+      "theoretical_loss": 3.425684823979185,
+      "tokens_seen": 2033254400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003904848484848485,
+      "loss": 2.722,
+      "theoretical_loss": 3.4256756476643946,
+      "tokens_seen": 2033319936
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039046464646464644,
+      "loss": 2.7486,
+      "theoretical_loss": 3.4256664717281717,
+      "tokens_seen": 2033385472
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039044444444444445,
+      "loss": 2.4272,
+      "theoretical_loss": 3.425657296170489,
+      "tokens_seen": 2033451008
+    },
+    {
+      "epoch": 0.23,
+      "objective/train/docs_used": 1120297,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.3779661655426025,
+      "objective/train/theoretical_loss": 3.425648120991318,
+      "objective/train/tokens_used": 392375776,
+      "theoretical_loss": 3.425648120991318,
+      "tokens_seen": 2033516544
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039042424242424247,
+      "loss": 2.5723,
+      "theoretical_loss": 3.425648120991318,
+      "tokens_seen": 2033516544
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003904040404040404,
+      "loss": 2.9231,
+      "theoretical_loss": 3.425638946190632,
+      "tokens_seen": 2033582080
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003903838383838384,
+      "loss": 2.6368,
+      "theoretical_loss": 3.4256297717684014,
+      "tokens_seen": 2033647616
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003903636363636364,
+      "loss": 2.6167,
+      "theoretical_loss": 3.4256205977246,
+      "tokens_seen": 2033713152
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039034343434343435,
+      "loss": 2.5206,
+      "theoretical_loss": 3.4256114240591993,
+      "tokens_seen": 2033778688
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039032323232323236,
+      "loss": 2.6265,
+      "theoretical_loss": 3.4256022507721715,
+      "tokens_seen": 2033844224
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003903030303030303,
+      "loss": 2.5845,
+      "theoretical_loss": 3.425593077863489,
+      "tokens_seen": 2033909760
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039028282828282827,
+      "loss": 2.7537,
+      "theoretical_loss": 3.4255839053331236,
+      "tokens_seen": 2033975296
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003902626262626263,
+      "loss": 2.8376,
+      "theoretical_loss": 3.4255747331810484,
+      "tokens_seen": 2034040832
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039024242424242424,
+      "loss": 2.4043,
+      "theoretical_loss": 3.4255655614072342,
+      "tokens_seen": 2034106368
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039022222222222225,
+      "loss": 2.5765,
+      "theoretical_loss": 3.4255563900116552,
+      "tokens_seen": 2034171904
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003902020202020202,
+      "loss": 2.7122,
+      "theoretical_loss": 3.4255472189942813,
+      "tokens_seen": 2034237440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003901818181818182,
+      "loss": 2.5745,
+      "theoretical_loss": 3.4255380483550866,
+      "tokens_seen": 2034302976
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003901616161616162,
+      "loss": 2.4317,
+      "theoretical_loss": 3.4255288780940427,
+      "tokens_seen": 2034368512
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039014141414141413,
+      "loss": 2.5789,
+      "theoretical_loss": 3.4255197082111213,
+      "tokens_seen": 2034434048
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039012121212121214,
+      "loss": 2.5604,
+      "theoretical_loss": 3.425510538706295,
+      "tokens_seen": 2034499584
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003901010101010101,
+      "loss": 2.6917,
+      "theoretical_loss": 3.4255013695795364,
+      "tokens_seen": 2034565120
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003900808080808081,
+      "loss": 2.6604,
+      "theoretical_loss": 3.4254922008308175,
+      "tokens_seen": 2034630656
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039006060606060607,
+      "loss": 2.5923,
+      "theoretical_loss": 3.4254830324601104,
+      "tokens_seen": 2034696192
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000390040404040404,
+      "loss": 2.6612,
+      "theoretical_loss": 3.425473864467387,
+      "tokens_seen": 2034761728
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039002020202020204,
+      "loss": 2.7573,
+      "theoretical_loss": 3.425464696852621,
+      "tokens_seen": 2034827264
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00039000000000000005,
+      "loss": 2.567,
+      "theoretical_loss": 3.4254555296157827,
+      "tokens_seen": 2034892800
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000389979797979798,
+      "loss": 2.8421,
+      "theoretical_loss": 3.425446362756845,
+      "tokens_seen": 2034958336
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038995959595959596,
+      "loss": 2.605,
+      "theoretical_loss": 3.425437196275781,
+      "tokens_seen": 2035023872
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003899393939393939,
+      "loss": 2.7614,
+      "theoretical_loss": 3.4254280301725624,
+      "tokens_seen": 2035089408
+    },
+    {
+      "epoch": 0.23,
+      "objective/train/docs_used": 1121395,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6363847255706787,
+      "objective/train/theoretical_loss": 3.425418864447161,
+      "objective/train/tokens_used": 394014176,
+      "theoretical_loss": 3.425418864447161,
+      "tokens_seen": 2035154944
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038991919191919193,
+      "loss": 2.5577,
+      "theoretical_loss": 3.425418864447161,
+      "tokens_seen": 2035154944
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038989898989898994,
+      "loss": 2.5876,
+      "theoretical_loss": 3.42540969909955,
+      "tokens_seen": 2035220480
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003898787878787879,
+      "loss": 2.4677,
+      "theoretical_loss": 3.4254005341297007,
+      "tokens_seen": 2035286016
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038985858585858585,
+      "loss": 2.5843,
+      "theoretical_loss": 3.4253913695375857,
+      "tokens_seen": 2035351552
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003898383838383838,
+      "loss": 2.536,
+      "theoretical_loss": 3.425382205323178,
+      "tokens_seen": 2035417088
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003898181818181819,
+      "loss": 2.59,
+      "theoretical_loss": 3.4253730414864485,
+      "tokens_seen": 2035482624
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038979797979797983,
+      "loss": 2.6873,
+      "theoretical_loss": 3.4253638780273707,
+      "tokens_seen": 2035548160
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003897777777777778,
+      "loss": 2.8808,
+      "theoretical_loss": 3.4253547149459163,
+      "tokens_seen": 2035613696
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038975757575757575,
+      "loss": 2.7569,
+      "theoretical_loss": 3.4253455522420575,
+      "tokens_seen": 2035679232
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003897373737373737,
+      "loss": 2.608,
+      "theoretical_loss": 3.425336389915767,
+      "tokens_seen": 2035744768
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038971717171717177,
+      "loss": 2.6626,
+      "theoretical_loss": 3.4253272279670166,
+      "tokens_seen": 2035810304
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003896969696969697,
+      "loss": 2.7393,
+      "theoretical_loss": 3.4253180663957794,
+      "tokens_seen": 2035875840
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003896767676767677,
+      "loss": 2.4374,
+      "theoretical_loss": 3.4253089052020265,
+      "tokens_seen": 2035941376
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038965656565656564,
+      "loss": 2.5749,
+      "theoretical_loss": 3.425299744385731,
+      "tokens_seen": 2036006912
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003896363636363636,
+      "loss": 2.7259,
+      "theoretical_loss": 3.4252905839468655,
+      "tokens_seen": 2036072448
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038961616161616166,
+      "loss": 2.6453,
+      "theoretical_loss": 3.425281423885401,
+      "tokens_seen": 2036137984
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003895959595959596,
+      "loss": 2.7484,
+      "theoretical_loss": 3.425272264201311,
+      "tokens_seen": 2036203520
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003895757575757576,
+      "loss": 2.701,
+      "theoretical_loss": 3.425263104894568,
+      "tokens_seen": 2036269056
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038955555555555553,
+      "loss": 2.6245,
+      "theoretical_loss": 3.425253945965143,
+      "tokens_seen": 2036334592
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038953535353535354,
+      "loss": 2.8589,
+      "theoretical_loss": 3.4252447874130096,
+      "tokens_seen": 2036400128
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038951515151515155,
+      "loss": 2.7356,
+      "theoretical_loss": 3.425235629238139,
+      "tokens_seen": 2036465664
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003894949494949495,
+      "loss": 2.572,
+      "theoretical_loss": 3.4252264714405047,
+      "tokens_seen": 2036531200
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038947474747474747,
+      "loss": 2.6346,
+      "theoretical_loss": 3.4252173140200783,
+      "tokens_seen": 2036596736
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003894545454545454,
+      "loss": 2.6631,
+      "theoretical_loss": 3.425208156976832,
+      "tokens_seen": 2036662272
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038943434343434344,
+      "loss": 2.7026,
+      "theoretical_loss": 3.4251990003107387,
+      "tokens_seen": 2036727808
+    },
+    {
+      "epoch": 0.23,
+      "objective/train/docs_used": 1121841,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7600228786468506,
+      "objective/train/theoretical_loss": 3.4251898440217703,
+      "objective/train/tokens_used": 395652576,
+      "theoretical_loss": 3.4251898440217703,
+      "tokens_seen": 2036793344
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038941414141414145,
+      "loss": 2.7747,
+      "theoretical_loss": 3.4251898440217703,
+      "tokens_seen": 2036793344
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003893939393939394,
+      "loss": 2.5716,
+      "theoretical_loss": 3.425180688109899,
+      "tokens_seen": 2036858880
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038937373737373736,
+      "loss": 2.3788,
+      "theoretical_loss": 3.4251715325750975,
+      "tokens_seen": 2036924416
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038935353535353537,
+      "loss": 2.5345,
+      "theoretical_loss": 3.4251623774173385,
+      "tokens_seen": 2036989952
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038933333333333333,
+      "loss": 2.5381,
+      "theoretical_loss": 3.4251532226365935,
+      "tokens_seen": 2037055488
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038931313131313134,
+      "loss": 2.7236,
+      "theoretical_loss": 3.425144068232835,
+      "tokens_seen": 2037121024
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003892929292929293,
+      "loss": 2.685,
+      "theoretical_loss": 3.425134914206036,
+      "tokens_seen": 2037186560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038927272727272725,
+      "loss": 2.528,
+      "theoretical_loss": 3.425125760556168,
+      "tokens_seen": 2037252096
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038925252525252526,
+      "loss": 2.6716,
+      "theoretical_loss": 3.425116607283204,
+      "tokens_seen": 2037317632
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003892323232323232,
+      "loss": 2.7653,
+      "theoretical_loss": 3.425107454387116,
+      "tokens_seen": 2037383168
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038921212121212123,
+      "loss": 2.6616,
+      "theoretical_loss": 3.4250983018678767,
+      "tokens_seen": 2037448704
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003891919191919192,
+      "loss": 2.7058,
+      "theoretical_loss": 3.425089149725458,
+      "tokens_seen": 2037514240
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003891717171717172,
+      "loss": 2.8981,
+      "theoretical_loss": 3.4250799979598328,
+      "tokens_seen": 2037579776
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038915151515151516,
+      "loss": 2.6551,
+      "theoretical_loss": 3.4250708465709727,
+      "tokens_seen": 2037645312
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003891313131313131,
+      "loss": 2.4252,
+      "theoretical_loss": 3.425061695558851,
+      "tokens_seen": 2037710848
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003891111111111111,
+      "loss": 2.7006,
+      "theoretical_loss": 3.4250525449234397,
+      "tokens_seen": 2037776384
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003890909090909091,
+      "loss": 2.4131,
+      "theoretical_loss": 3.4250433946647107,
+      "tokens_seen": 2037841920
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003890707070707071,
+      "loss": 2.5942,
+      "theoretical_loss": 3.425034244782637,
+      "tokens_seen": 2037907456
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038905050505050505,
+      "loss": 2.6065,
+      "theoretical_loss": 3.4250250952771912,
+      "tokens_seen": 2037972992
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000389030303030303,
+      "loss": 2.6553,
+      "theoretical_loss": 3.4250159461483447,
+      "tokens_seen": 2038038528
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000389010101010101,
+      "loss": 2.7689,
+      "theoretical_loss": 3.4250067973960707,
+      "tokens_seen": 2038104064
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038898989898989903,
+      "loss": 2.6705,
+      "theoretical_loss": 3.424997649020341,
+      "tokens_seen": 2038169600
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000388969696969697,
+      "loss": 2.5966,
+      "theoretical_loss": 3.4249885010211285,
+      "tokens_seen": 2038235136
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038894949494949494,
+      "loss": 2.5179,
+      "theoretical_loss": 3.4249793533984056,
+      "tokens_seen": 2038300672
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003889292929292929,
+      "loss": 2.5465,
+      "theoretical_loss": 3.4249702061521443,
+      "tokens_seen": 2038366208
+    },
+    {
+      "epoch": 0.23,
+      "objective/train/docs_used": 1123215,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.411255121231079,
+      "objective/train/theoretical_loss": 3.424961059282318,
+      "objective/train/tokens_used": 397290976,
+      "theoretical_loss": 3.424961059282318,
+      "tokens_seen": 2038431744
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003889090909090909,
+      "loss": 2.5545,
+      "theoretical_loss": 3.424961059282318,
+      "tokens_seen": 2038431744
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003888888888888889,
+      "loss": 2.5991,
+      "theoretical_loss": 3.4249519127888974,
+      "tokens_seen": 2038497280
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003888686868686869,
+      "loss": 2.6284,
+      "theoretical_loss": 3.4249427666718564,
+      "tokens_seen": 2038562816
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038884848484848483,
+      "loss": 2.661,
+      "theoretical_loss": 3.4249336209311667,
+      "tokens_seen": 2038628352
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003888282828282828,
+      "loss": 2.5987,
+      "theoretical_loss": 3.424924475566801,
+      "tokens_seen": 2038693888
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038880808080808086,
+      "loss": 2.6224,
+      "theoretical_loss": 3.4249153305787314,
+      "tokens_seen": 2038759424
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003887878787878788,
+      "loss": 2.4543,
+      "theoretical_loss": 3.4249061859669307,
+      "tokens_seen": 2038824960
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038876767676767677,
+      "loss": 2.6108,
+      "theoretical_loss": 3.424897041731371,
+      "tokens_seen": 2038890496
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038874747474747473,
+      "loss": 2.4736,
+      "theoretical_loss": 3.424887897872025,
+      "tokens_seen": 2038956032
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038872727272727274,
+      "loss": 2.8445,
+      "theoretical_loss": 3.4248787543888652,
+      "tokens_seen": 2039021568
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038870707070707075,
+      "loss": 2.6671,
+      "theoretical_loss": 3.4248696112818635,
+      "tokens_seen": 2039087104
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003886868686868687,
+      "loss": 2.6505,
+      "theoretical_loss": 3.424860468550993,
+      "tokens_seen": 2039152640
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038866666666666666,
+      "loss": 2.3294,
+      "theoretical_loss": 3.4248513261962255,
+      "tokens_seen": 2039218176
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003886464646464646,
+      "loss": 2.5899,
+      "theoretical_loss": 3.4248421842175336,
+      "tokens_seen": 2039283712
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003886262626262627,
+      "loss": 2.9221,
+      "theoretical_loss": 3.4248330426148907,
+      "tokens_seen": 2039349248
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038860606060606064,
+      "loss": 2.5795,
+      "theoretical_loss": 3.424823901388268,
+      "tokens_seen": 2039414784
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003885858585858586,
+      "loss": 2.593,
+      "theoretical_loss": 3.424814760537638,
+      "tokens_seen": 2039480320
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038856565656565656,
+      "loss": 2.592,
+      "theoretical_loss": 3.4248056200629744,
+      "tokens_seen": 2039545856
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003885454545454545,
+      "loss": 2.6182,
+      "theoretical_loss": 3.424796479964248,
+      "tokens_seen": 2039611392
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003885252525252526,
+      "loss": 2.8057,
+      "theoretical_loss": 3.4247873402414326,
+      "tokens_seen": 2039676928
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038850505050505054,
+      "loss": 2.5074,
+      "theoretical_loss": 3.4247782008945,
+      "tokens_seen": 2039742464
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003884848484848485,
+      "loss": 2.8014,
+      "theoretical_loss": 3.4247690619234223,
+      "tokens_seen": 2039808000
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038846464646464645,
+      "loss": 2.5553,
+      "theoretical_loss": 3.424759923328173,
+      "tokens_seen": 2039873536
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038844444444444446,
+      "loss": 2.6588,
+      "theoretical_loss": 3.424750785108724,
+      "tokens_seen": 2039939072
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038842424242424247,
+      "loss": 2.6811,
+      "theoretical_loss": 3.424741647265048,
+      "tokens_seen": 2040004608
+    },
+    {
+      "epoch": 0.23,
+      "objective/train/docs_used": 1124379,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.8337912559509277,
+      "objective/train/theoretical_loss": 3.424732509797117,
+      "objective/train/tokens_used": 398929376,
+      "theoretical_loss": 3.424732509797117,
+      "tokens_seen": 2040070144
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038840404040404043,
+      "loss": 2.8508,
+      "theoretical_loss": 3.424732509797117,
+      "tokens_seen": 2040070144
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003883838383838384,
+      "loss": 2.7414,
+      "theoretical_loss": 3.4247233727049036,
+      "tokens_seen": 2040135680
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038836363636363634,
+      "loss": 2.5021,
+      "theoretical_loss": 3.4247142359883807,
+      "tokens_seen": 2040201216
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038834343434343435,
+      "loss": 2.6071,
+      "theoretical_loss": 3.424705099647521,
+      "tokens_seen": 2040266752
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038832323232323236,
+      "loss": 2.665,
+      "theoretical_loss": 3.4246959636822956,
+      "tokens_seen": 2040332288
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003883030303030303,
+      "loss": 2.6956,
+      "theoretical_loss": 3.4246868280926783,
+      "tokens_seen": 2040397824
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003882828282828283,
+      "loss": 2.6893,
+      "theoretical_loss": 3.4246776928786415,
+      "tokens_seen": 2040463360
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003882626262626263,
+      "loss": 2.4344,
+      "theoretical_loss": 3.4246685580401577,
+      "tokens_seen": 2040528896
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038824242424242425,
+      "loss": 2.6557,
+      "theoretical_loss": 3.4246594235771983,
+      "tokens_seen": 2040594432
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038822222222222226,
+      "loss": 2.6863,
+      "theoretical_loss": 3.424650289489737,
+      "tokens_seen": 2040659968
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003882020202020202,
+      "loss": 2.6083,
+      "theoretical_loss": 3.4246411557777456,
+      "tokens_seen": 2040725504
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038818181818181817,
+      "loss": 2.397,
+      "theoretical_loss": 3.4246320224411972,
+      "tokens_seen": 2040791040
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003881616161616162,
+      "loss": 2.54,
+      "theoretical_loss": 3.424622889480064,
+      "tokens_seen": 2040856576
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038814141414141414,
+      "loss": 2.4706,
+      "theoretical_loss": 3.4246137568943187,
+      "tokens_seen": 2040922112
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038812121212121215,
+      "loss": 2.5738,
+      "theoretical_loss": 3.4246046246839335,
+      "tokens_seen": 2040987648
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003881010101010101,
+      "loss": 2.5798,
+      "theoretical_loss": 3.424595492848881,
+      "tokens_seen": 2041053184
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003880808080808081,
+      "loss": 2.5736,
+      "theoretical_loss": 3.4245863613891343,
+      "tokens_seen": 2041118720
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003880606060606061,
+      "loss": 2.5231,
+      "theoretical_loss": 3.424577230304665,
+      "tokens_seen": 2041184256
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038804040404040403,
+      "loss": 2.4575,
+      "theoretical_loss": 3.424568099595446,
+      "tokens_seen": 2041249792
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038802020202020204,
+      "loss": 2.5567,
+      "theoretical_loss": 3.42455896926145,
+      "tokens_seen": 2041315328
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000388,
+      "loss": 2.5012,
+      "theoretical_loss": 3.4245498393026494,
+      "tokens_seen": 2041380864
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000387979797979798,
+      "loss": 2.6823,
+      "theoretical_loss": 3.4245407097190173,
+      "tokens_seen": 2041446400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038795959595959597,
+      "loss": 2.5196,
+      "theoretical_loss": 3.424531580510525,
+      "tokens_seen": 2041511936
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003879393939393939,
+      "loss": 2.5629,
+      "theoretical_loss": 3.424522451677146,
+      "tokens_seen": 2041577472
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038791919191919193,
+      "loss": 2.5199,
+      "theoretical_loss": 3.4245133232188527,
+      "tokens_seen": 2041643008
+    },
+    {
+      "epoch": 0.23,
+      "objective/train/docs_used": 1125100,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.047553062438965,
+      "objective/train/theoretical_loss": 3.4245041951356177,
+      "objective/train/tokens_used": 400567776,
+      "theoretical_loss": 3.4245041951356177,
+      "tokens_seen": 2041708544
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003878989898989899,
+      "loss": 2.7422,
+      "theoretical_loss": 3.4245041951356177,
+      "tokens_seen": 2041708544
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003878787878787879,
+      "loss": 2.6362,
+      "theoretical_loss": 3.424495067427413,
+      "tokens_seen": 2041774080
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038785858585858586,
+      "loss": 2.6407,
+      "theoretical_loss": 3.4244859400942116,
+      "tokens_seen": 2041839616
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003878383838383838,
+      "loss": 2.5186,
+      "theoretical_loss": 3.424476813135986,
+      "tokens_seen": 2041905152
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003878181818181818,
+      "loss": 2.421,
+      "theoretical_loss": 3.424467686552709,
+      "tokens_seen": 2041970688
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038779797979797984,
+      "loss": 2.5184,
+      "theoretical_loss": 3.4244585603443527,
+      "tokens_seen": 2042036224
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003877777777777778,
+      "loss": 2.6667,
+      "theoretical_loss": 3.4244494345108905,
+      "tokens_seen": 2042101760
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038775757575757575,
+      "loss": 2.564,
+      "theoretical_loss": 3.4244403090522937,
+      "tokens_seen": 2042167296
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003877373737373737,
+      "loss": 2.5341,
+      "theoretical_loss": 3.4244311839685357,
+      "tokens_seen": 2042232832
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003877171717171717,
+      "loss": 2.4584,
+      "theoretical_loss": 3.424422059259589,
+      "tokens_seen": 2042298368
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038769696969696973,
+      "loss": 2.5476,
+      "theoretical_loss": 3.424412934925426,
+      "tokens_seen": 2042363904
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003876767676767677,
+      "loss": 2.4415,
+      "theoretical_loss": 3.4244038109660195,
+      "tokens_seen": 2042429440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038765656565656564,
+      "loss": 2.5881,
+      "theoretical_loss": 3.4243946873813425,
+      "tokens_seen": 2042494976
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003876363636363636,
+      "loss": 2.6939,
+      "theoretical_loss": 3.4243855641713665,
+      "tokens_seen": 2042560512
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038761616161616167,
+      "loss": 2.2827,
+      "theoretical_loss": 3.4243764413360642,
+      "tokens_seen": 2042626048
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003875959595959596,
+      "loss": 2.7188,
+      "theoretical_loss": 3.4243673188754093,
+      "tokens_seen": 2042691584
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003875757575757576,
+      "loss": 2.4194,
+      "theoretical_loss": 3.4243581967893735,
+      "tokens_seen": 2042757120
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038755555555555554,
+      "loss": 2.6265,
+      "theoretical_loss": 3.42434907507793,
+      "tokens_seen": 2042822656
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003875353535353535,
+      "loss": 2.5645,
+      "theoretical_loss": 3.4243399537410504,
+      "tokens_seen": 2042888192
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038751515151515156,
+      "loss": 2.5277,
+      "theoretical_loss": 3.424330832778708,
+      "tokens_seen": 2042953728
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003874949494949495,
+      "loss": 2.7011,
+      "theoretical_loss": 3.424321712190876,
+      "tokens_seen": 2043019264
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003874747474747475,
+      "loss": 2.5774,
+      "theoretical_loss": 3.4243125919775257,
+      "tokens_seen": 2043084800
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038745454545454543,
+      "loss": 2.6009,
+      "theoretical_loss": 3.4243034721386305,
+      "tokens_seen": 2043150336
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038743434343434344,
+      "loss": 2.6962,
+      "theoretical_loss": 3.424294352674163,
+      "tokens_seen": 2043215872
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038741414141414145,
+      "loss": 2.8634,
+      "theoretical_loss": 3.4242852335840954,
+      "tokens_seen": 2043281408
+    },
+    {
+      "epoch": 0.23,
+      "objective/train/docs_used": 1126113,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.695091962814331,
+      "objective/train/theoretical_loss": 3.424276114868401,
+      "objective/train/tokens_used": 402206176,
+      "theoretical_loss": 3.424276114868401,
+      "tokens_seen": 2043346944
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003873939393939394,
+      "loss": 2.7604,
+      "theoretical_loss": 3.424276114868401,
+      "tokens_seen": 2043346944
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038737373737373737,
+      "loss": 2.4762,
+      "theoretical_loss": 3.424266996527052,
+      "tokens_seen": 2043412480
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003873535353535353,
+      "loss": 2.6634,
+      "theoretical_loss": 3.424257878560021,
+      "tokens_seen": 2043478016
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038733333333333333,
+      "loss": 2.4992,
+      "theoretical_loss": 3.4242487609672807,
+      "tokens_seen": 2043543552
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038731313131313134,
+      "loss": 2.8986,
+      "theoretical_loss": 3.4242396437488036,
+      "tokens_seen": 2043609088
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003872929292929293,
+      "loss": 2.6341,
+      "theoretical_loss": 3.4242305269045623,
+      "tokens_seen": 2043674624
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038727272727272726,
+      "loss": 2.5154,
+      "theoretical_loss": 3.42422141043453,
+      "tokens_seen": 2043740160
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038725252525252527,
+      "loss": 2.5922,
+      "theoretical_loss": 3.4242122943386786,
+      "tokens_seen": 2043805696
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003872323232323232,
+      "loss": 2.5824,
+      "theoretical_loss": 3.4242031786169815,
+      "tokens_seen": 2043871232
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038721212121212124,
+      "loss": 2.6942,
+      "theoretical_loss": 3.424194063269411,
+      "tokens_seen": 2043936768
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003871919191919192,
+      "loss": 2.6585,
+      "theoretical_loss": 3.424184948295939,
+      "tokens_seen": 2044002304
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038717171717171715,
+      "loss": 2.4877,
+      "theoretical_loss": 3.4241758336965393,
+      "tokens_seen": 2044067840
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038715151515151516,
+      "loss": 2.5554,
+      "theoretical_loss": 3.4241667194711845,
+      "tokens_seen": 2044133376
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003871313131313131,
+      "loss": 2.6856,
+      "theoretical_loss": 3.424157605619846,
+      "tokens_seen": 2044198912
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038711111111111113,
+      "loss": 2.4869,
+      "theoretical_loss": 3.424148492142498,
+      "tokens_seen": 2044264448
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003870909090909091,
+      "loss": 2.5563,
+      "theoretical_loss": 3.4241393790391124,
+      "tokens_seen": 2044329984
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003870707070707071,
+      "loss": 2.5721,
+      "theoretical_loss": 3.4241302663096613,
+      "tokens_seen": 2044395520
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038705050505050505,
+      "loss": 2.4328,
+      "theoretical_loss": 3.4241211539541188,
+      "tokens_seen": 2044461056
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038703030303030307,
+      "loss": 2.4588,
+      "theoretical_loss": 3.4241120419724567,
+      "tokens_seen": 2044526592
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000387010101010101,
+      "loss": 2.4166,
+      "theoretical_loss": 3.424102930364647,
+      "tokens_seen": 2044592128
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000386989898989899,
+      "loss": 2.4133,
+      "theoretical_loss": 3.424093819130664,
+      "tokens_seen": 2044657664
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000386969696969697,
+      "loss": 2.6739,
+      "theoretical_loss": 3.4240847082704793,
+      "tokens_seen": 2044723200
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038694949494949495,
+      "loss": 2.4544,
+      "theoretical_loss": 3.4240755977840656,
+      "tokens_seen": 2044788736
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038692929292929296,
+      "loss": 2.4786,
+      "theoretical_loss": 3.4240664876713955,
+      "tokens_seen": 2044854272
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003869090909090909,
+      "loss": 2.6681,
+      "theoretical_loss": 3.424057377932443,
+      "tokens_seen": 2044919808
+    },
+    {
+      "epoch": 0.23,
+      "objective/train/docs_used": 1126722,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.8349125385284424,
+      "objective/train/theoretical_loss": 3.424048268567179,
+      "objective/train/tokens_used": 403844576,
+      "theoretical_loss": 3.424048268567179,
+      "tokens_seen": 2044985344
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003868888888888889,
+      "loss": 2.7018,
+      "theoretical_loss": 3.424048268567179,
+      "tokens_seen": 2044985344
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003868686868686869,
+      "loss": 2.6153,
+      "theoretical_loss": 3.4240391595755773,
+      "tokens_seen": 2045050880
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038684848484848484,
+      "loss": 2.6144,
+      "theoretical_loss": 3.42403005095761,
+      "tokens_seen": 2045116416
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038682828282828285,
+      "loss": 2.3565,
+      "theoretical_loss": 3.42402094271325,
+      "tokens_seen": 2045181952
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003868080808080808,
+      "loss": 2.513,
+      "theoretical_loss": 3.424011834842471,
+      "tokens_seen": 2045247488
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003867878787878788,
+      "loss": 2.8146,
+      "theoretical_loss": 3.424002727345244,
+      "tokens_seen": 2045313024
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003867676767676768,
+      "loss": 2.6684,
+      "theoretical_loss": 3.4239936202215424,
+      "tokens_seen": 2045378560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038674747474747473,
+      "loss": 2.3647,
+      "theoretical_loss": 3.423984513471339,
+      "tokens_seen": 2045444096
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038672727272727274,
+      "loss": 2.5581,
+      "theoretical_loss": 3.423975407094607,
+      "tokens_seen": 2045509632
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038670707070707076,
+      "loss": 2.6833,
+      "theoretical_loss": 3.423966301091318,
+      "tokens_seen": 2045575168
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003866868686868687,
+      "loss": 2.6042,
+      "theoretical_loss": 3.423957195461446,
+      "tokens_seen": 2045640704
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038666666666666667,
+      "loss": 2.7104,
+      "theoretical_loss": 3.423948090204963,
+      "tokens_seen": 2045706240
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003866464646464646,
+      "loss": 2.6974,
+      "theoretical_loss": 3.4239389853218416,
+      "tokens_seen": 2045771776
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038662626262626264,
+      "loss": 2.5431,
+      "theoretical_loss": 3.4239298808120546,
+      "tokens_seen": 2045837312
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038660606060606065,
+      "loss": 2.4468,
+      "theoretical_loss": 3.4239207766755753,
+      "tokens_seen": 2045902848
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003865858585858586,
+      "loss": 2.3854,
+      "theoretical_loss": 3.423911672912376,
+      "tokens_seen": 2045968384
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038656565656565656,
+      "loss": 2.5403,
+      "theoretical_loss": 3.423902569522429,
+      "tokens_seen": 2046033920
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003865454545454545,
+      "loss": 2.4529,
+      "theoretical_loss": 3.4238934665057084,
+      "tokens_seen": 2046099456
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003865252525252526,
+      "loss": 2.5285,
+      "theoretical_loss": 3.4238843638621854,
+      "tokens_seen": 2046164992
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038650505050505054,
+      "loss": 2.7578,
+      "theoretical_loss": 3.4238752615918333,
+      "tokens_seen": 2046230528
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003864848484848485,
+      "loss": 2.6517,
+      "theoretical_loss": 3.423866159694625,
+      "tokens_seen": 2046296064
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038646464646464645,
+      "loss": 2.595,
+      "theoretical_loss": 3.4238570581705337,
+      "tokens_seen": 2046361600
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003864444444444444,
+      "loss": 2.5574,
+      "theoretical_loss": 3.4238479570195315,
+      "tokens_seen": 2046427136
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003864242424242425,
+      "loss": 2.3707,
+      "theoretical_loss": 3.423838856241591,
+      "tokens_seen": 2046492672
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00038640404040404043,
+      "loss": 2.7492,
+      "theoretical_loss": 3.4238297558366853,
+      "tokens_seen": 2046558208
+    },
+    {
+      "epoch": 0.23,
+      "objective/train/docs_used": 1127892,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5316739082336426,
+      "objective/train/theoretical_loss": 3.4238206558047874,
+      "objective/train/tokens_used": 405482976,
+      "theoretical_loss": 3.4238206558047874,
+      "tokens_seen": 2046623744
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003863838383838384,
+      "loss": 2.4914,
+      "theoretical_loss": 3.4238206558047874,
+      "tokens_seen": 2046623744
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038636363636363635,
+      "loss": 2.6275,
+      "theoretical_loss": 3.4238115561458695,
+      "tokens_seen": 2046689280
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003863434343434343,
+      "loss": 2.3453,
+      "theoretical_loss": 3.423802456859905,
+      "tokens_seen": 2046754816
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038632323232323237,
+      "loss": 2.355,
+      "theoretical_loss": 3.423793357946866,
+      "tokens_seen": 2046820352
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003863030303030303,
+      "loss": 2.4346,
+      "theoretical_loss": 3.423784259406726,
+      "tokens_seen": 2046885888
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003862828282828283,
+      "loss": 2.4405,
+      "theoretical_loss": 3.423775161239457,
+      "tokens_seen": 2046951424
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038626262626262624,
+      "loss": 2.6577,
+      "theoretical_loss": 3.4237660634450324,
+      "tokens_seen": 2047016960
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038624242424242425,
+      "loss": 2.4727,
+      "theoretical_loss": 3.423756966023425,
+      "tokens_seen": 2047082496
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038622222222222226,
+      "loss": 2.6148,
+      "theoretical_loss": 3.4237478689746066,
+      "tokens_seen": 2047148032
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003862020202020202,
+      "loss": 2.5595,
+      "theoretical_loss": 3.4237387722985515,
+      "tokens_seen": 2047213568
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003861818181818182,
+      "loss": 2.6433,
+      "theoretical_loss": 3.4237296759952316,
+      "tokens_seen": 2047279104
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038616161616161613,
+      "loss": 2.4541,
+      "theoretical_loss": 3.4237205800646198,
+      "tokens_seen": 2047344640
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038614141414141414,
+      "loss": 2.6971,
+      "theoretical_loss": 3.423711484506689,
+      "tokens_seen": 2047410176
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038612121212121215,
+      "loss": 2.3779,
+      "theoretical_loss": 3.423702389321412,
+      "tokens_seen": 2047475712
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003861010101010101,
+      "loss": 2.7949,
+      "theoretical_loss": 3.4236932945087615,
+      "tokens_seen": 2047541248
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038608080808080807,
+      "loss": 2.5279,
+      "theoretical_loss": 3.42368420006871,
+      "tokens_seen": 2047606784
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003860606060606061,
+      "loss": 2.692,
+      "theoretical_loss": 3.423675106001231,
+      "tokens_seen": 2047672320
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038604040404040404,
+      "loss": 2.4909,
+      "theoretical_loss": 3.423666012306297,
+      "tokens_seen": 2047737856
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038602020202020205,
+      "loss": 2.4489,
+      "theoretical_loss": 3.423656918983881,
+      "tokens_seen": 2047803392
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000386,
+      "loss": 2.6628,
+      "theoretical_loss": 3.423647826033955,
+      "tokens_seen": 2047868928
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038597979797979796,
+      "loss": 2.4424,
+      "theoretical_loss": 3.4236387334564933,
+      "tokens_seen": 2047934464
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038595959595959597,
+      "loss": 2.7376,
+      "theoretical_loss": 3.423629641251467,
+      "tokens_seen": 2048000000
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038593939393939393,
+      "loss": 2.4449,
+      "theoretical_loss": 3.42362054941885,
+      "tokens_seen": 2048065536
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038591919191919194,
+      "loss": 2.4178,
+      "theoretical_loss": 3.423611457958615,
+      "tokens_seen": 2048131072
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003858989898989899,
+      "loss": 2.3096,
+      "theoretical_loss": 3.4236023668707354,
+      "tokens_seen": 2048196608
+    },
+    {
+      "epoch": 0.24,
+      "objective/train/docs_used": 1128475,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.619576930999756,
+      "objective/train/theoretical_loss": 3.4235932761551826,
+      "objective/train/tokens_used": 407121376,
+      "theoretical_loss": 3.4235932761551826,
+      "tokens_seen": 2048262144
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003858787878787879,
+      "loss": 2.5873,
+      "theoretical_loss": 3.4235932761551826,
+      "tokens_seen": 2048262144
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038585858585858586,
+      "loss": 2.8021,
+      "theoretical_loss": 3.4235841858119302,
+      "tokens_seen": 2048327680
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003858383838383838,
+      "loss": 2.5011,
+      "theoretical_loss": 3.4235750958409517,
+      "tokens_seen": 2048393216
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038581818181818183,
+      "loss": 2.7487,
+      "theoretical_loss": 3.4235660062422184,
+      "tokens_seen": 2048458752
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003857979797979798,
+      "loss": 2.5642,
+      "theoretical_loss": 3.4235569170157047,
+      "tokens_seen": 2048524288
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003857777777777778,
+      "loss": 2.6144,
+      "theoretical_loss": 3.423547828161383,
+      "tokens_seen": 2048589824
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038575757575757576,
+      "loss": 2.6756,
+      "theoretical_loss": 3.4235387396792256,
+      "tokens_seen": 2048655360
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003857373737373737,
+      "loss": 2.501,
+      "theoretical_loss": 3.423529651569206,
+      "tokens_seen": 2048720896
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003857171717171717,
+      "loss": 2.4741,
+      "theoretical_loss": 3.423520563831296,
+      "tokens_seen": 2048786432
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038569696969696974,
+      "loss": 2.6386,
+      "theoretical_loss": 3.42351147646547,
+      "tokens_seen": 2048851968
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003856767676767677,
+      "loss": 2.8198,
+      "theoretical_loss": 3.4235023894717003,
+      "tokens_seen": 2048917504
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038565656565656565,
+      "loss": 2.8276,
+      "theoretical_loss": 3.423493302849959,
+      "tokens_seen": 2048983040
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003856363636363636,
+      "loss": 2.6402,
+      "theoretical_loss": 3.4234842166002197,
+      "tokens_seen": 2049048576
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003856161616161616,
+      "loss": 2.4989,
+      "theoretical_loss": 3.4234751307224554,
+      "tokens_seen": 2049114112
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038559595959595963,
+      "loss": 2.5747,
+      "theoretical_loss": 3.4234660452166383,
+      "tokens_seen": 2049179648
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003855757575757576,
+      "loss": 2.3174,
+      "theoretical_loss": 3.423456960082742,
+      "tokens_seen": 2049245184
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038555555555555554,
+      "loss": 2.5597,
+      "theoretical_loss": 3.423447875320739,
+      "tokens_seen": 2049310720
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003855353535353535,
+      "loss": 2.396,
+      "theoretical_loss": 3.423438790930602,
+      "tokens_seen": 2049376256
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038551515151515156,
+      "loss": 2.3726,
+      "theoretical_loss": 3.4234297069123043,
+      "tokens_seen": 2049441792
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003854949494949495,
+      "loss": 2.4606,
+      "theoretical_loss": 3.4234206232658186,
+      "tokens_seen": 2049507328
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003854747474747475,
+      "loss": 2.6207,
+      "theoretical_loss": 3.423411539991118,
+      "tokens_seen": 2049572864
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038545454545454544,
+      "loss": 2.3535,
+      "theoretical_loss": 3.4234024570881747,
+      "tokens_seen": 2049638400
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003854343434343434,
+      "loss": 2.5528,
+      "theoretical_loss": 3.4233933745569627,
+      "tokens_seen": 2049703936
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038541414141414146,
+      "loss": 2.4746,
+      "theoretical_loss": 3.4233842923974542,
+      "tokens_seen": 2049769472
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003853939393939394,
+      "loss": 2.4524,
+      "theoretical_loss": 3.423375210609622,
+      "tokens_seen": 2049835008
+    },
+    {
+      "epoch": 0.24,
+      "objective/train/docs_used": 1129543,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6075730323791504,
+      "objective/train/theoretical_loss": 3.423366129193439,
+      "objective/train/tokens_used": 408759776,
+      "theoretical_loss": 3.423366129193439,
+      "tokens_seen": 2049900544
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038537373737373737,
+      "loss": 2.5068,
+      "theoretical_loss": 3.423366129193439,
+      "tokens_seen": 2049900544
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038535353535353533,
+      "loss": 2.8476,
+      "theoretical_loss": 3.423357048148879,
+      "tokens_seen": 2049966080
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003853333333333334,
+      "loss": 2.5966,
+      "theoretical_loss": 3.4233479674759137,
+      "tokens_seen": 2050031616
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038531313131313135,
+      "loss": 2.5537,
+      "theoretical_loss": 3.4233388871745167,
+      "tokens_seen": 2050097152
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003852929292929293,
+      "loss": 2.5581,
+      "theoretical_loss": 3.4233298072446607,
+      "tokens_seen": 2050162688
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038527272727272726,
+      "loss": 2.5214,
+      "theoretical_loss": 3.4233207276863187,
+      "tokens_seen": 2050228224
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003852525252525252,
+      "loss": 2.3995,
+      "theoretical_loss": 3.423311648499464,
+      "tokens_seen": 2050293760
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003852323232323233,
+      "loss": 2.682,
+      "theoretical_loss": 3.4233025696840684,
+      "tokens_seen": 2050359296
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038521212121212124,
+      "loss": 2.5779,
+      "theoretical_loss": 3.4232934912401065,
+      "tokens_seen": 2050424832
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003851919191919192,
+      "loss": 2.7086,
+      "theoretical_loss": 3.4232844131675497,
+      "tokens_seen": 2050490368
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038517171717171716,
+      "loss": 2.5114,
+      "theoretical_loss": 3.4232753354663714,
+      "tokens_seen": 2050555904
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038515151515151517,
+      "loss": 2.6217,
+      "theoretical_loss": 3.4232662581365454,
+      "tokens_seen": 2050621440
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003851313131313132,
+      "loss": 2.7702,
+      "theoretical_loss": 3.4232571811780437,
+      "tokens_seen": 2050686976
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038511111111111114,
+      "loss": 2.4582,
+      "theoretical_loss": 3.423248104590839,
+      "tokens_seen": 2050752512
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003850909090909091,
+      "loss": 2.4804,
+      "theoretical_loss": 3.423239028374905,
+      "tokens_seen": 2050818048
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038507070707070705,
+      "loss": 2.4453,
+      "theoretical_loss": 3.423229952530215,
+      "tokens_seen": 2050883584
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038505050505050506,
+      "loss": 2.5674,
+      "theoretical_loss": 3.4232208770567403,
+      "tokens_seen": 2050949120
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038503030303030307,
+      "loss": 2.5093,
+      "theoretical_loss": 3.423211801954455,
+      "tokens_seen": 2051014656
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038501010101010103,
+      "loss": 2.4003,
+      "theoretical_loss": 3.4232027272233325,
+      "tokens_seen": 2051080192
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000384989898989899,
+      "loss": 2.4936,
+      "theoretical_loss": 3.423193652863345,
+      "tokens_seen": 2051145728
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000384969696969697,
+      "loss": 2.6507,
+      "theoretical_loss": 3.4231845788744657,
+      "tokens_seen": 2051211264
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038494949494949495,
+      "loss": 2.6008,
+      "theoretical_loss": 3.4231755052566673,
+      "tokens_seen": 2051276800
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038492929292929296,
+      "loss": 2.6672,
+      "theoretical_loss": 3.423166432009923,
+      "tokens_seen": 2051342336
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003849090909090909,
+      "loss": 2.4185,
+      "theoretical_loss": 3.4231573591342057,
+      "tokens_seen": 2051407872
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003848888888888889,
+      "loss": 2.4452,
+      "theoretical_loss": 3.4231482866294884,
+      "tokens_seen": 2051473408
+    },
+    {
+      "epoch": 0.24,
+      "objective/train/docs_used": 1129971,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5400750637054443,
+      "objective/train/theoretical_loss": 3.4231392144957447,
+      "objective/train/tokens_used": 410398176,
+      "theoretical_loss": 3.4231392144957447,
+      "tokens_seen": 2051538944
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003848686868686869,
+      "loss": 2.544,
+      "theoretical_loss": 3.4231392144957447,
+      "tokens_seen": 2051538944
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038484848484848485,
+      "loss": 2.5103,
+      "theoretical_loss": 3.423130142732947,
+      "tokens_seen": 2051604480
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038482828282828286,
+      "loss": 2.6622,
+      "theoretical_loss": 3.423121071341068,
+      "tokens_seen": 2051670016
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003848080808080808,
+      "loss": 2.6093,
+      "theoretical_loss": 3.4231120003200806,
+      "tokens_seen": 2051735552
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038478787878787877,
+      "loss": 2.7409,
+      "theoretical_loss": 3.4231029296699584,
+      "tokens_seen": 2051801088
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003847676767676768,
+      "loss": 2.3119,
+      "theoretical_loss": 3.4230938593906743,
+      "tokens_seen": 2051866624
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038474747474747474,
+      "loss": 2.9422,
+      "theoretical_loss": 3.423084789482201,
+      "tokens_seen": 2051932160
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038472727272727275,
+      "loss": 2.5798,
+      "theoretical_loss": 3.423075719944512,
+      "tokens_seen": 2051997696
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003847070707070707,
+      "loss": 2.6104,
+      "theoretical_loss": 3.4230666507775798,
+      "tokens_seen": 2052063232
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003846868686868687,
+      "loss": 2.6882,
+      "theoretical_loss": 3.4230575819813778,
+      "tokens_seen": 2052128768
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003846666666666667,
+      "loss": 2.7611,
+      "theoretical_loss": 3.4230485135558784,
+      "tokens_seen": 2052194304
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038464646464646463,
+      "loss": 2.6087,
+      "theoretical_loss": 3.423039445501055,
+      "tokens_seen": 2052259840
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038462626262626264,
+      "loss": 2.5599,
+      "theoretical_loss": 3.4230303778168802,
+      "tokens_seen": 2052325376
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003846060606060606,
+      "loss": 2.5465,
+      "theoretical_loss": 3.423021310503328,
+      "tokens_seen": 2052390912
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003845858585858586,
+      "loss": 2.5225,
+      "theoretical_loss": 3.4230122435603705,
+      "tokens_seen": 2052456448
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038456565656565657,
+      "loss": 2.5076,
+      "theoretical_loss": 3.423003176987981,
+      "tokens_seen": 2052521984
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003845454545454545,
+      "loss": 2.4484,
+      "theoretical_loss": 3.422994110786133,
+      "tokens_seen": 2052587520
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038452525252525254,
+      "loss": 2.4406,
+      "theoretical_loss": 3.4229850449547987,
+      "tokens_seen": 2052653056
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038450505050505055,
+      "loss": 2.5912,
+      "theoretical_loss": 3.422975979493952,
+      "tokens_seen": 2052718592
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003844848484848485,
+      "loss": 2.6731,
+      "theoretical_loss": 3.422966914403565,
+      "tokens_seen": 2052784128
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038446464646464646,
+      "loss": 2.5045,
+      "theoretical_loss": 3.422957849683611,
+      "tokens_seen": 2052849664
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003844444444444444,
+      "loss": 2.5084,
+      "theoretical_loss": 3.4229487853340634,
+      "tokens_seen": 2052915200
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038442424242424243,
+      "loss": 2.4632,
+      "theoretical_loss": 3.4229397213548953,
+      "tokens_seen": 2052980736
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038440404040404044,
+      "loss": 2.79,
+      "theoretical_loss": 3.422930657746079,
+      "tokens_seen": 2053046272
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003843838383838384,
+      "loss": 2.653,
+      "theoretical_loss": 3.4229215945075886,
+      "tokens_seen": 2053111808
+    },
+    {
+      "epoch": 0.24,
+      "objective/train/docs_used": 1131091,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.304107666015625,
+      "objective/train/theoretical_loss": 3.4229125316393967,
+      "objective/train/tokens_used": 412036576,
+      "theoretical_loss": 3.4229125316393967,
+      "tokens_seen": 2053177344
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038436363636363635,
+      "loss": 2.3587,
+      "theoretical_loss": 3.4229125316393967,
+      "tokens_seen": 2053177344
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003843434343434343,
+      "loss": 2.6467,
+      "theoretical_loss": 3.4229034691414757,
+      "tokens_seen": 2053242880
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003843232323232324,
+      "loss": 2.6997,
+      "theoretical_loss": 3.4228944070137994,
+      "tokens_seen": 2053308416
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038430303030303033,
+      "loss": 2.6112,
+      "theoretical_loss": 3.422885345256341,
+      "tokens_seen": 2053373952
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003842828282828283,
+      "loss": 2.9658,
+      "theoretical_loss": 3.4228762838690727,
+      "tokens_seen": 2053439488
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038426262626262625,
+      "loss": 2.5993,
+      "theoretical_loss": 3.422867222851968,
+      "tokens_seen": 2053505024
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003842424242424242,
+      "loss": 2.3219,
+      "theoretical_loss": 3.4228581622050007,
+      "tokens_seen": 2053570560
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038422222222222227,
+      "loss": 2.2982,
+      "theoretical_loss": 3.4228491019281426,
+      "tokens_seen": 2053636096
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003842020202020202,
+      "loss": 2.5316,
+      "theoretical_loss": 3.422840042021368,
+      "tokens_seen": 2053701632
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003841818181818182,
+      "loss": 2.5537,
+      "theoretical_loss": 3.4228309824846495,
+      "tokens_seen": 2053767168
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038416161616161614,
+      "loss": 2.6209,
+      "theoretical_loss": 3.422821923317959,
+      "tokens_seen": 2053832704
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038414141414141415,
+      "loss": 2.3912,
+      "theoretical_loss": 3.4228128645212714,
+      "tokens_seen": 2053898240
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038412121212121216,
+      "loss": 2.6452,
+      "theoretical_loss": 3.422803806094559,
+      "tokens_seen": 2053963776
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003841010101010101,
+      "loss": 2.6819,
+      "theoretical_loss": 3.422794748037795,
+      "tokens_seen": 2054029312
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003840808080808081,
+      "loss": 2.8093,
+      "theoretical_loss": 3.4227856903509517,
+      "tokens_seen": 2054094848
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038406060606060603,
+      "loss": 2.5679,
+      "theoretical_loss": 3.4227766330340033,
+      "tokens_seen": 2054160384
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038404040404040404,
+      "loss": 2.4692,
+      "theoretical_loss": 3.4227675760869225,
+      "tokens_seen": 2054225920
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038402020202020205,
+      "loss": 2.4967,
+      "theoretical_loss": 3.4227585195096824,
+      "tokens_seen": 2054291456
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000384,
+      "loss": 2.5947,
+      "theoretical_loss": 3.4227494633022557,
+      "tokens_seen": 2054356992
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038397979797979797,
+      "loss": 2.4263,
+      "theoretical_loss": 3.422740407464616,
+      "tokens_seen": 2054422528
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000383959595959596,
+      "loss": 2.5193,
+      "theoretical_loss": 3.4227313519967364,
+      "tokens_seen": 2054488064
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038393939393939393,
+      "loss": 2.2884,
+      "theoretical_loss": 3.4227222968985904,
+      "tokens_seen": 2054553600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038391919191919195,
+      "loss": 2.5641,
+      "theoretical_loss": 3.4227132421701496,
+      "tokens_seen": 2054619136
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003838989898989899,
+      "loss": 2.3804,
+      "theoretical_loss": 3.422704187811389,
+      "tokens_seen": 2054684672
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038387878787878786,
+      "loss": 2.5944,
+      "theoretical_loss": 3.4226951338222804,
+      "tokens_seen": 2054750208
+    },
+    {
+      "epoch": 0.24,
+      "objective/train/docs_used": 1131564,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 1.9922235012054443,
+      "objective/train/theoretical_loss": 3.422686080202797,
+      "objective/train/tokens_used": 413674976,
+      "theoretical_loss": 3.422686080202797,
+      "tokens_seen": 2054815744
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038385858585858587,
+      "loss": 2.3964,
+      "theoretical_loss": 3.422686080202797,
+      "tokens_seen": 2054815744
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038383838383838383,
+      "loss": 2.4571,
+      "theoretical_loss": 3.422677026952913,
+      "tokens_seen": 2054881280
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038381818181818184,
+      "loss": 2.6484,
+      "theoretical_loss": 3.4226679740726,
+      "tokens_seen": 2054946816
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003837979797979798,
+      "loss": 2.5657,
+      "theoretical_loss": 3.4226589215618324,
+      "tokens_seen": 2055012352
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003837777777777778,
+      "loss": 2.4732,
+      "theoretical_loss": 3.4226498694205825,
+      "tokens_seen": 2055077888
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038375757575757576,
+      "loss": 2.5445,
+      "theoretical_loss": 3.4226408176488246,
+      "tokens_seen": 2055143424
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003837373737373737,
+      "loss": 2.3885,
+      "theoretical_loss": 3.42263176624653,
+      "tokens_seen": 2055208960
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038371717171717173,
+      "loss": 2.7094,
+      "theoretical_loss": 3.4226227152136737,
+      "tokens_seen": 2055274496
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003836969696969697,
+      "loss": 2.5966,
+      "theoretical_loss": 3.4226136645502274,
+      "tokens_seen": 2055340032
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003836767676767677,
+      "loss": 2.7365,
+      "theoretical_loss": 3.422604614256165,
+      "tokens_seen": 2055405568
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038365656565656566,
+      "loss": 2.5426,
+      "theoretical_loss": 3.422595564331459,
+      "tokens_seen": 2055471104
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003836363636363636,
+      "loss": 2.4673,
+      "theoretical_loss": 3.422586514776084,
+      "tokens_seen": 2055536640
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003836161616161616,
+      "loss": 2.4799,
+      "theoretical_loss": 3.4225774655900114,
+      "tokens_seen": 2055602176
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038359595959595963,
+      "loss": 2.804,
+      "theoretical_loss": 3.422568416773216,
+      "tokens_seen": 2055667712
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003835757575757576,
+      "loss": 2.7354,
+      "theoretical_loss": 3.422559368325669,
+      "tokens_seen": 2055733248
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038355555555555555,
+      "loss": 2.6043,
+      "theoretical_loss": 3.4225503202473453,
+      "tokens_seen": 2055798784
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038353535353535356,
+      "loss": 2.5097,
+      "theoretical_loss": 3.4225412725382176,
+      "tokens_seen": 2055864320
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003835151515151515,
+      "loss": 2.6064,
+      "theoretical_loss": 3.4225322251982586,
+      "tokens_seen": 2055929856
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038349494949494953,
+      "loss": 2.5131,
+      "theoretical_loss": 3.4225231782274417,
+      "tokens_seen": 2055995392
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003834747474747475,
+      "loss": 2.4516,
+      "theoretical_loss": 3.42251413162574,
+      "tokens_seen": 2056060928
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038345454545454544,
+      "loss": 2.7433,
+      "theoretical_loss": 3.4225050853931274,
+      "tokens_seen": 2056126464
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038343434343434345,
+      "loss": 2.7684,
+      "theoretical_loss": 3.4224960395295763,
+      "tokens_seen": 2056192000
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038341414141414146,
+      "loss": 2.5327,
+      "theoretical_loss": 3.4224869940350597,
+      "tokens_seen": 2056257536
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003833939393939394,
+      "loss": 2.4565,
+      "theoretical_loss": 3.4224779489095516,
+      "tokens_seen": 2056323072
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003833737373737374,
+      "loss": 2.6484,
+      "theoretical_loss": 3.4224689041530247,
+      "tokens_seen": 2056388608
+    },
+    {
+      "epoch": 0.24,
+      "objective/train/docs_used": 1132626,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.2589941024780273,
+      "objective/train/theoretical_loss": 3.422459859765452,
+      "objective/train/tokens_used": 415313376,
+      "theoretical_loss": 3.422459859765452,
+      "tokens_seen": 2056454144
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038335353535353533,
+      "loss": 2.3534,
+      "theoretical_loss": 3.422459859765452,
+      "tokens_seen": 2056454144
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038333333333333334,
+      "loss": 2.7033,
+      "theoretical_loss": 3.422450815746807,
+      "tokens_seen": 2056519680
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038331313131313136,
+      "loss": 2.8655,
+      "theoretical_loss": 3.422441772097063,
+      "tokens_seen": 2056585216
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003832929292929293,
+      "loss": 2.6004,
+      "theoretical_loss": 3.422432728816193,
+      "tokens_seen": 2056650752
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038327272727272727,
+      "loss": 2.8118,
+      "theoretical_loss": 3.42242368590417,
+      "tokens_seen": 2056716288
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003832525252525252,
+      "loss": 2.6264,
+      "theoretical_loss": 3.422414643360967,
+      "tokens_seen": 2056781824
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003832323232323233,
+      "loss": 2.329,
+      "theoretical_loss": 3.4224056011865587,
+      "tokens_seen": 2056847360
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038321212121212125,
+      "loss": 2.7078,
+      "theoretical_loss": 3.4223965593809167,
+      "tokens_seen": 2056912896
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003831919191919192,
+      "loss": 2.6651,
+      "theoretical_loss": 3.4223875179440144,
+      "tokens_seen": 2056978432
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038317171717171716,
+      "loss": 2.6018,
+      "theoretical_loss": 3.4223784768758256,
+      "tokens_seen": 2057043968
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003831515151515151,
+      "loss": 2.7493,
+      "theoretical_loss": 3.4223694361763233,
+      "tokens_seen": 2057109504
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003831313131313132,
+      "loss": 2.7182,
+      "theoretical_loss": 3.422360395845481,
+      "tokens_seen": 2057175040
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038311111111111114,
+      "loss": 2.656,
+      "theoretical_loss": 3.422351355883271,
+      "tokens_seen": 2057240576
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003830909090909091,
+      "loss": 2.6502,
+      "theoretical_loss": 3.422342316289668,
+      "tokens_seen": 2057306112
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038307070707070706,
+      "loss": 2.4645,
+      "theoretical_loss": 3.4223332770646437,
+      "tokens_seen": 2057371648
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000383050505050505,
+      "loss": 2.7262,
+      "theoretical_loss": 3.422324238208172,
+      "tokens_seen": 2057437184
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003830303030303031,
+      "loss": 2.5653,
+      "theoretical_loss": 3.422315199720227,
+      "tokens_seen": 2057502720
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038301010101010103,
+      "loss": 2.7167,
+      "theoretical_loss": 3.42230616160078,
+      "tokens_seen": 2057568256
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000382989898989899,
+      "loss": 2.6084,
+      "theoretical_loss": 3.422297123849806,
+      "tokens_seen": 2057633792
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038296969696969695,
+      "loss": 2.6191,
+      "theoretical_loss": 3.4222880864672773,
+      "tokens_seen": 2057699328
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038294949494949496,
+      "loss": 2.7294,
+      "theoretical_loss": 3.4222790494531674,
+      "tokens_seen": 2057764864
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038292929292929297,
+      "loss": 2.4276,
+      "theoretical_loss": 3.4222700128074495,
+      "tokens_seen": 2057830400
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003829090909090909,
+      "loss": 2.2369,
+      "theoretical_loss": 3.4222609765300973,
+      "tokens_seen": 2057895936
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003828888888888889,
+      "loss": 2.7929,
+      "theoretical_loss": 3.4222519406210834,
+      "tokens_seen": 2057961472
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038286868686868684,
+      "loss": 2.5678,
+      "theoretical_loss": 3.422242905080381,
+      "tokens_seen": 2058027008
+    },
+    {
+      "epoch": 0.24,
+      "objective/train/docs_used": 1133082,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.546114683151245,
+      "objective/train/theoretical_loss": 3.422233869907964,
+      "objective/train/tokens_used": 416951776,
+      "theoretical_loss": 3.422233869907964,
+      "tokens_seen": 2058092544
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038284848484848485,
+      "loss": 2.5294,
+      "theoretical_loss": 3.422233869907964,
+      "tokens_seen": 2058092544
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038282828282828286,
+      "loss": 2.6326,
+      "theoretical_loss": 3.4222248351038056,
+      "tokens_seen": 2058158080
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003828080808080808,
+      "loss": 2.7425,
+      "theoretical_loss": 3.4222158006678787,
+      "tokens_seen": 2058223616
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003827878787878788,
+      "loss": 2.7415,
+      "theoretical_loss": 3.4222067666001563,
+      "tokens_seen": 2058289152
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003827676767676768,
+      "loss": 2.4934,
+      "theoretical_loss": 3.4221977329006124,
+      "tokens_seen": 2058354688
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038274747474747474,
+      "loss": 2.7173,
+      "theoretical_loss": 3.4221886995692197,
+      "tokens_seen": 2058420224
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038272727272727276,
+      "loss": 2.5987,
+      "theoretical_loss": 3.4221796666059516,
+      "tokens_seen": 2058485760
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003827070707070707,
+      "loss": 2.8897,
+      "theoretical_loss": 3.422170634010782,
+      "tokens_seen": 2058551296
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038268686868686867,
+      "loss": 2.9005,
+      "theoretical_loss": 3.4221616017836833,
+      "tokens_seen": 2058616832
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003826666666666667,
+      "loss": 2.8517,
+      "theoretical_loss": 3.422152569924629,
+      "tokens_seen": 2058682368
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038264646464646464,
+      "loss": 2.697,
+      "theoretical_loss": 3.422143538433593,
+      "tokens_seen": 2058747904
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038262626262626265,
+      "loss": 2.5326,
+      "theoretical_loss": 3.422134507310548,
+      "tokens_seen": 2058813440
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003826060606060606,
+      "loss": 2.5776,
+      "theoretical_loss": 3.422125476555467,
+      "tokens_seen": 2058878976
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003825858585858586,
+      "loss": 2.4844,
+      "theoretical_loss": 3.4221164461683236,
+      "tokens_seen": 2058944512
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038256565656565657,
+      "loss": 2.6396,
+      "theoretical_loss": 3.4221074161490916,
+      "tokens_seen": 2059010048
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038254545454545453,
+      "loss": 2.9819,
+      "theoretical_loss": 3.422098386497744,
+      "tokens_seen": 2059075584
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038252525252525254,
+      "loss": 2.5535,
+      "theoretical_loss": 3.422089357214254,
+      "tokens_seen": 2059141120
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003825050505050505,
+      "loss": 2.7637,
+      "theoretical_loss": 3.4220803282985948,
+      "tokens_seen": 2059206656
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003824848484848485,
+      "loss": 2.4363,
+      "theoretical_loss": 3.4220712997507396,
+      "tokens_seen": 2059272192
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038246464646464647,
+      "loss": 2.6999,
+      "theoretical_loss": 3.4220622715706623,
+      "tokens_seen": 2059337728
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003824444444444444,
+      "loss": 2.7858,
+      "theoretical_loss": 3.4220532437583353,
+      "tokens_seen": 2059403264
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038242424242424243,
+      "loss": 2.5095,
+      "theoretical_loss": 3.4220442163137332,
+      "tokens_seen": 2059468800
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038240404040404044,
+      "loss": 2.5913,
+      "theoretical_loss": 3.422035189236828,
+      "tokens_seen": 2059534336
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003823838383838384,
+      "loss": 2.3923,
+      "theoretical_loss": 3.4220261625275943,
+      "tokens_seen": 2059599872
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038236363636363636,
+      "loss": 2.629,
+      "theoretical_loss": 3.422017136186004,
+      "tokens_seen": 2059665408
+    },
+    {
+      "epoch": 0.24,
+      "objective/train/docs_used": 1134328,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.9984428882598877,
+      "objective/train/theoretical_loss": 3.422008110212032,
+      "objective/train/tokens_used": 418590176,
+      "theoretical_loss": 3.422008110212032,
+      "tokens_seen": 2059730944
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003823434343434343,
+      "loss": 2.7817,
+      "theoretical_loss": 3.422008110212032,
+      "tokens_seen": 2059730944
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003823232323232323,
+      "loss": 2.604,
+      "theoretical_loss": 3.42199908460565,
+      "tokens_seen": 2059796480
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038230303030303034,
+      "loss": 2.6303,
+      "theoretical_loss": 3.4219900593668324,
+      "tokens_seen": 2059862016
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003822828282828283,
+      "loss": 2.594,
+      "theoretical_loss": 3.421981034495553,
+      "tokens_seen": 2059927552
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038226262626262625,
+      "loss": 2.6818,
+      "theoretical_loss": 3.421972009991783,
+      "tokens_seen": 2059993088
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003822424242424242,
+      "loss": 2.5402,
+      "theoretical_loss": 3.4219629858554983,
+      "tokens_seen": 2060058624
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003822222222222223,
+      "loss": 2.3345,
+      "theoretical_loss": 3.4219539620866706,
+      "tokens_seen": 2060124160
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038220202020202023,
+      "loss": 2.7212,
+      "theoretical_loss": 3.421944938685274,
+      "tokens_seen": 2060189696
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003821818181818182,
+      "loss": 2.6767,
+      "theoretical_loss": 3.4219359156512814,
+      "tokens_seen": 2060255232
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038216161616161614,
+      "loss": 2.4343,
+      "theoretical_loss": 3.4219268929846662,
+      "tokens_seen": 2060320768
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003821414141414141,
+      "loss": 2.4259,
+      "theoretical_loss": 3.4219178706854025,
+      "tokens_seen": 2060386304
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038212121212121217,
+      "loss": 2.6288,
+      "theoretical_loss": 3.4219088487534624,
+      "tokens_seen": 2060451840
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003821010101010101,
+      "loss": 2.6574,
+      "theoretical_loss": 3.4218998271888204,
+      "tokens_seen": 2060517376
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003820808080808081,
+      "loss": 2.5343,
+      "theoretical_loss": 3.4218908059914495,
+      "tokens_seen": 2060582912
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038206060606060604,
+      "loss": 2.6432,
+      "theoretical_loss": 3.4218817851613226,
+      "tokens_seen": 2060648448
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038204040404040405,
+      "loss": 2.6204,
+      "theoretical_loss": 3.4218727646984135,
+      "tokens_seen": 2060713984
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038202020202020206,
+      "loss": 2.6521,
+      "theoretical_loss": 3.4218637446026956,
+      "tokens_seen": 2060779520
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000382,
+      "loss": 2.7458,
+      "theoretical_loss": 3.4218547248741427,
+      "tokens_seen": 2060845056
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038197979797979797,
+      "loss": 2.8135,
+      "theoretical_loss": 3.4218457055127267,
+      "tokens_seen": 2060910592
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038195959595959593,
+      "loss": 2.5271,
+      "theoretical_loss": 3.4218366865184224,
+      "tokens_seen": 2060976128
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038193939393939394,
+      "loss": 2.7021,
+      "theoretical_loss": 3.421827667891203,
+      "tokens_seen": 2061041664
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038191919191919195,
+      "loss": 2.5729,
+      "theoretical_loss": 3.4218186496310414,
+      "tokens_seen": 2061107200
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003818989898989899,
+      "loss": 2.416,
+      "theoretical_loss": 3.4218096317379114,
+      "tokens_seen": 2061172736
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038187878787878786,
+      "loss": 2.6395,
+      "theoretical_loss": 3.4218006142117856,
+      "tokens_seen": 2061238272
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003818585858585859,
+      "loss": 2.6007,
+      "theoretical_loss": 3.4217915970526387,
+      "tokens_seen": 2061303808
+    },
+    {
+      "epoch": 0.24,
+      "objective/train/docs_used": 1135184,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6123404502868652,
+      "objective/train/theoretical_loss": 3.421782580260443,
+      "objective/train/tokens_used": 420228576,
+      "theoretical_loss": 3.421782580260443,
+      "tokens_seen": 2061369344
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003818383838383839,
+      "loss": 2.6084,
+      "theoretical_loss": 3.421782580260443,
+      "tokens_seen": 2061369344
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038181818181818184,
+      "loss": 2.6643,
+      "theoretical_loss": 3.4217735638351727,
+      "tokens_seen": 2061434880
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003817979797979798,
+      "loss": 2.6587,
+      "theoretical_loss": 3.4217645477768004,
+      "tokens_seen": 2061500416
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038177777777777776,
+      "loss": 2.6285,
+      "theoretical_loss": 3.4217555320853004,
+      "tokens_seen": 2061565952
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038175757575757577,
+      "loss": 2.5553,
+      "theoretical_loss": 3.4217465167606456,
+      "tokens_seen": 2061631488
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003817373737373738,
+      "loss": 2.6293,
+      "theoretical_loss": 3.4217375018028084,
+      "tokens_seen": 2061697024
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038171717171717174,
+      "loss": 2.6729,
+      "theoretical_loss": 3.421728487211764,
+      "tokens_seen": 2061762560
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003816969696969697,
+      "loss": 2.617,
+      "theoretical_loss": 3.4217194729874856,
+      "tokens_seen": 2061828096
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003816767676767677,
+      "loss": 2.7,
+      "theoretical_loss": 3.4217104591299456,
+      "tokens_seen": 2061893632
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038165656565656566,
+      "loss": 2.7145,
+      "theoretical_loss": 3.4217014456391177,
+      "tokens_seen": 2061959168
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038163636363636367,
+      "loss": 2.501,
+      "theoretical_loss": 3.421692432514976,
+      "tokens_seen": 2062024704
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038161616161616163,
+      "loss": 2.4279,
+      "theoretical_loss": 3.421683419757493,
+      "tokens_seen": 2062090240
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003815959595959596,
+      "loss": 2.4535,
+      "theoretical_loss": 3.421674407366643,
+      "tokens_seen": 2062155776
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003815757575757576,
+      "loss": 2.5424,
+      "theoretical_loss": 3.4216653953423988,
+      "tokens_seen": 2062221312
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038155555555555555,
+      "loss": 2.5376,
+      "theoretical_loss": 3.4216563836847342,
+      "tokens_seen": 2062286848
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038153535353535357,
+      "loss": 2.6659,
+      "theoretical_loss": 3.4216473723936223,
+      "tokens_seen": 2062352384
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003815151515151515,
+      "loss": 2.5682,
+      "theoretical_loss": 3.421638361469037,
+      "tokens_seen": 2062417920
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003814949494949495,
+      "loss": 2.52,
+      "theoretical_loss": 3.421629350910951,
+      "tokens_seen": 2062483456
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003814747474747475,
+      "loss": 2.4654,
+      "theoretical_loss": 3.421620340719339,
+      "tokens_seen": 2062548992
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038145454545454545,
+      "loss": 2.5279,
+      "theoretical_loss": 3.421611330894174,
+      "tokens_seen": 2062614528
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038143434343434346,
+      "loss": 2.3815,
+      "theoretical_loss": 3.421602321435428,
+      "tokens_seen": 2062680064
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003814141414141414,
+      "loss": 2.8111,
+      "theoretical_loss": 3.4215933123430764,
+      "tokens_seen": 2062745600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003813939393939394,
+      "loss": 2.4246,
+      "theoretical_loss": 3.4215843036170916,
+      "tokens_seen": 2062811136
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003813737373737374,
+      "loss": 2.5496,
+      "theoretical_loss": 3.421575295257447,
+      "tokens_seen": 2062876672
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038135353535353534,
+      "loss": 2.5605,
+      "theoretical_loss": 3.421566287264117,
+      "tokens_seen": 2062942208
+    },
+    {
+      "epoch": 0.24,
+      "objective/train/docs_used": 1135835,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.9515092372894287,
+      "objective/train/theoretical_loss": 3.4215572796370743,
+      "objective/train/tokens_used": 421866976,
+      "theoretical_loss": 3.4215572796370743,
+      "tokens_seen": 2063007744
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00038133333333333335,
+      "loss": 2.7702,
+      "theoretical_loss": 3.4215572796370743,
+      "tokens_seen": 2063007744
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003813131313131313,
+      "loss": 2.7661,
+      "theoretical_loss": 3.4215482723762927,
+      "tokens_seen": 2063073280
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003812929292929293,
+      "loss": 2.5597,
+      "theoretical_loss": 3.421539265481745,
+      "tokens_seen": 2063138816
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003812727272727273,
+      "loss": 2.8065,
+      "theoretical_loss": 3.421530258953406,
+      "tokens_seen": 2063204352
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038125252525252523,
+      "loss": 2.5136,
+      "theoretical_loss": 3.4215212527912477,
+      "tokens_seen": 2063269888
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038123232323232324,
+      "loss": 2.638,
+      "theoretical_loss": 3.4215122469952446,
+      "tokens_seen": 2063335424
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038121212121212125,
+      "loss": 2.6101,
+      "theoretical_loss": 3.4215032415653694,
+      "tokens_seen": 2063400960
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003811919191919192,
+      "loss": 2.5277,
+      "theoretical_loss": 3.421494236501596,
+      "tokens_seen": 2063466496
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038117171717171717,
+      "loss": 2.683,
+      "theoretical_loss": 3.4214852318038984,
+      "tokens_seen": 2063532032
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003811515151515151,
+      "loss": 2.6695,
+      "theoretical_loss": 3.421476227472249,
+      "tokens_seen": 2063597568
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038113131313131314,
+      "loss": 2.6012,
+      "theoretical_loss": 3.421467223506623,
+      "tokens_seen": 2063663104
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038111111111111115,
+      "loss": 2.6724,
+      "theoretical_loss": 3.421458219906992,
+      "tokens_seen": 2063728640
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003810909090909091,
+      "loss": 2.5402,
+      "theoretical_loss": 3.4214492166733304,
+      "tokens_seen": 2063794176
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038107070707070706,
+      "loss": 2.6403,
+      "theoretical_loss": 3.421440213805612,
+      "tokens_seen": 2063859712
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000381050505050505,
+      "loss": 2.5712,
+      "theoretical_loss": 3.4214312113038092,
+      "tokens_seen": 2063925248
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003810303030303031,
+      "loss": 2.7039,
+      "theoretical_loss": 3.4214222091678965,
+      "tokens_seen": 2063990784
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038101010101010104,
+      "loss": 2.5423,
+      "theoretical_loss": 3.421413207397847,
+      "tokens_seen": 2064056320
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000380989898989899,
+      "loss": 2.7038,
+      "theoretical_loss": 3.4214042059936345,
+      "tokens_seen": 2064121856
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038096969696969695,
+      "loss": 2.7188,
+      "theoretical_loss": 3.4213952049552327,
+      "tokens_seen": 2064187392
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003809494949494949,
+      "loss": 2.5841,
+      "theoretical_loss": 3.421386204282615,
+      "tokens_seen": 2064252928
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000380929292929293,
+      "loss": 2.9913,
+      "theoretical_loss": 3.4213772039757537,
+      "tokens_seen": 2064318464
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038090909090909093,
+      "loss": 2.5172,
+      "theoretical_loss": 3.421368204034624,
+      "tokens_seen": 2064384000
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003808888888888889,
+      "loss": 2.6884,
+      "theoretical_loss": 3.4213592044591987,
+      "tokens_seen": 2064449536
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038086868686868685,
+      "loss": 2.5919,
+      "theoretical_loss": 3.421350205249451,
+      "tokens_seen": 2064515072
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038084848484848486,
+      "loss": 2.545,
+      "theoretical_loss": 3.421341206405355,
+      "tokens_seen": 2064580608
+    },
+    {
+      "epoch": 0.25,
+      "objective/train/docs_used": 1136767,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7607131004333496,
+      "objective/train/theoretical_loss": 3.421332207926884,
+      "objective/train/tokens_used": 423505376,
+      "theoretical_loss": 3.421332207926884,
+      "tokens_seen": 2064646144
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038082828282828287,
+      "loss": 2.6132,
+      "theoretical_loss": 3.421332207926884,
+      "tokens_seen": 2064646144
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003808080808080808,
+      "loss": 2.5181,
+      "theoretical_loss": 3.421323209814012,
+      "tokens_seen": 2064711680
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003807878787878788,
+      "loss": 2.7202,
+      "theoretical_loss": 3.4213142120667124,
+      "tokens_seen": 2064777216
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038076767676767674,
+      "loss": 2.6562,
+      "theoretical_loss": 3.421305214684958,
+      "tokens_seen": 2064842752
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038074747474747475,
+      "loss": 2.6323,
+      "theoretical_loss": 3.421296217668723,
+      "tokens_seen": 2064908288
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038072727272727276,
+      "loss": 2.5255,
+      "theoretical_loss": 3.4212872210179803,
+      "tokens_seen": 2064973824
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003807070707070707,
+      "loss": 2.6225,
+      "theoretical_loss": 3.4212782247327045,
+      "tokens_seen": 2065039360
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003806868686868687,
+      "loss": 2.7296,
+      "theoretical_loss": 3.4212692288128683,
+      "tokens_seen": 2065104896
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003806666666666667,
+      "loss": 2.5503,
+      "theoretical_loss": 3.4212602332584456,
+      "tokens_seen": 2065170432
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038064646464646464,
+      "loss": 2.7646,
+      "theoretical_loss": 3.4212512380694102,
+      "tokens_seen": 2065235968
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038062626262626265,
+      "loss": 2.461,
+      "theoretical_loss": 3.421242243245735,
+      "tokens_seen": 2065301504
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003806060606060606,
+      "loss": 2.5117,
+      "theoretical_loss": 3.421233248787394,
+      "tokens_seen": 2065367040
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038058585858585857,
+      "loss": 2.7507,
+      "theoretical_loss": 3.421224254694361,
+      "tokens_seen": 2065432576
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003805656565656566,
+      "loss": 2.4589,
+      "theoretical_loss": 3.4212152609666093,
+      "tokens_seen": 2065498112
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038054545454545454,
+      "loss": 2.7945,
+      "theoretical_loss": 3.421206267604112,
+      "tokens_seen": 2065563648
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038052525252525255,
+      "loss": 2.7459,
+      "theoretical_loss": 3.4211972746068438,
+      "tokens_seen": 2065629184
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003805050505050505,
+      "loss": 2.5679,
+      "theoretical_loss": 3.421188281974777,
+      "tokens_seen": 2065694720
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003804848484848485,
+      "loss": 2.7287,
+      "theoretical_loss": 3.4211792897078865,
+      "tokens_seen": 2065760256
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038046464646464647,
+      "loss": 2.8253,
+      "theoretical_loss": 3.4211702978061447,
+      "tokens_seen": 2065825792
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038044444444444443,
+      "loss": 2.938,
+      "theoretical_loss": 3.421161306269526,
+      "tokens_seen": 2065891328
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038042424242424244,
+      "loss": 2.7181,
+      "theoretical_loss": 3.421152315098003,
+      "tokens_seen": 2065956864
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003804040404040404,
+      "loss": 2.5896,
+      "theoretical_loss": 3.4211433242915508,
+      "tokens_seen": 2066022400
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003803838383838384,
+      "loss": 2.5979,
+      "theoretical_loss": 3.4211343338501417,
+      "tokens_seen": 2066087936
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038036363636363636,
+      "loss": 2.7055,
+      "theoretical_loss": 3.42112534377375,
+      "tokens_seen": 2066153472
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003803434343434343,
+      "loss": 2.8343,
+      "theoretical_loss": 3.421116354062349,
+      "tokens_seen": 2066219008
+    },
+    {
+      "epoch": 0.25,
+      "objective/train/docs_used": 1137905,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.325758457183838,
+      "objective/train/theoretical_loss": 3.4211073647159123,
+      "objective/train/tokens_used": 425143776,
+      "theoretical_loss": 3.4211073647159123,
+      "tokens_seen": 2066284544
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038032323232323233,
+      "loss": 2.7325,
+      "theoretical_loss": 3.4211073647159123,
+      "tokens_seen": 2066284544
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038030303030303034,
+      "loss": 2.5732,
+      "theoretical_loss": 3.421098375734414,
+      "tokens_seen": 2066350080
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003802828282828283,
+      "loss": 2.7549,
+      "theoretical_loss": 3.4210893871178265,
+      "tokens_seen": 2066415616
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038026262626262626,
+      "loss": 2.768,
+      "theoretical_loss": 3.421080398866125,
+      "tokens_seen": 2066481152
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003802424242424242,
+      "loss": 2.6855,
+      "theoretical_loss": 3.4210714109792817,
+      "tokens_seen": 2066546688
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003802222222222222,
+      "loss": 2.7331,
+      "theoretical_loss": 3.421062423457271,
+      "tokens_seen": 2066612224
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038020202020202024,
+      "loss": 2.4095,
+      "theoretical_loss": 3.421053436300067,
+      "tokens_seen": 2066677760
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003801818181818182,
+      "loss": 2.695,
+      "theoretical_loss": 3.421044449507642,
+      "tokens_seen": 2066743296
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038016161616161615,
+      "loss": 2.7255,
+      "theoretical_loss": 3.4210354630799706,
+      "tokens_seen": 2066808832
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038014141414141416,
+      "loss": 2.5248,
+      "theoretical_loss": 3.4210264770170262,
+      "tokens_seen": 2066874368
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038012121212121217,
+      "loss": 2.6969,
+      "theoretical_loss": 3.421017491318782,
+      "tokens_seen": 2066939904
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038010101010101013,
+      "loss": 2.613,
+      "theoretical_loss": 3.4210085059852124,
+      "tokens_seen": 2067005440
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003800808080808081,
+      "loss": 2.6488,
+      "theoretical_loss": 3.4209995210162907,
+      "tokens_seen": 2067070976
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038006060606060604,
+      "loss": 2.6582,
+      "theoretical_loss": 3.42099053641199,
+      "tokens_seen": 2067136512
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038004040404040405,
+      "loss": 2.4411,
+      "theoretical_loss": 3.420981552172285,
+      "tokens_seen": 2067202048
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038002020202020206,
+      "loss": 2.6109,
+      "theoretical_loss": 3.420972568297149,
+      "tokens_seen": 2067267584
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00038,
+      "loss": 2.5581,
+      "theoretical_loss": 3.4209635847865547,
+      "tokens_seen": 2067333120
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000379979797979798,
+      "loss": 2.7596,
+      "theoretical_loss": 3.420954601640477,
+      "tokens_seen": 2067398656
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037995959595959593,
+      "loss": 2.564,
+      "theoretical_loss": 3.420945618858889,
+      "tokens_seen": 2067464192
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000379939393939394,
+      "loss": 2.7477,
+      "theoretical_loss": 3.4209366364417644,
+      "tokens_seen": 2067529728
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037991919191919196,
+      "loss": 2.4975,
+      "theoretical_loss": 3.420927654389077,
+      "tokens_seen": 2067595264
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003798989898989899,
+      "loss": 2.7735,
+      "theoretical_loss": 3.4209186727008,
+      "tokens_seen": 2067660800
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037987878787878787,
+      "loss": 2.5042,
+      "theoretical_loss": 3.4209096913769073,
+      "tokens_seen": 2067726336
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037985858585858583,
+      "loss": 2.7599,
+      "theoretical_loss": 3.420900710417373,
+      "tokens_seen": 2067791872
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003798383838383839,
+      "loss": 2.7375,
+      "theoretical_loss": 3.4208917298221704,
+      "tokens_seen": 2067857408
+    },
+    {
+      "epoch": 0.25,
+      "objective/train/docs_used": 1138616,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.164562940597534,
+      "objective/train/theoretical_loss": 3.4208827495912733,
+      "objective/train/tokens_used": 426782176,
+      "theoretical_loss": 3.4208827495912733,
+      "tokens_seen": 2067922944
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037981818181818185,
+      "loss": 2.6866,
+      "theoretical_loss": 3.4208827495912733,
+      "tokens_seen": 2067922944
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003797979797979798,
+      "loss": 2.5767,
+      "theoretical_loss": 3.420873769724655,
+      "tokens_seen": 2067988480
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037977777777777776,
+      "loss": 2.4466,
+      "theoretical_loss": 3.4208647902222893,
+      "tokens_seen": 2068054016
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003797575757575757,
+      "loss": 2.6539,
+      "theoretical_loss": 3.4208558110841505,
+      "tokens_seen": 2068119552
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003797373737373738,
+      "loss": 2.6817,
+      "theoretical_loss": 3.420846832310212,
+      "tokens_seen": 2068185088
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037971717171717174,
+      "loss": 2.6445,
+      "theoretical_loss": 3.420837853900447,
+      "tokens_seen": 2068250624
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003796969696969697,
+      "loss": 2.6812,
+      "theoretical_loss": 3.4208288758548293,
+      "tokens_seen": 2068316160
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037967676767676766,
+      "loss": 2.5428,
+      "theoretical_loss": 3.420819898173333,
+      "tokens_seen": 2068381696
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037965656565656567,
+      "loss": 2.6958,
+      "theoretical_loss": 3.420810920855932,
+      "tokens_seen": 2068447232
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003796363636363637,
+      "loss": 2.5625,
+      "theoretical_loss": 3.4208019439025987,
+      "tokens_seen": 2068512768
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037961616161616163,
+      "loss": 2.7598,
+      "theoretical_loss": 3.4207929673133086,
+      "tokens_seen": 2068578304
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003795959595959596,
+      "loss": 2.6718,
+      "theoretical_loss": 3.420783991088034,
+      "tokens_seen": 2068643840
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037957575757575755,
+      "loss": 2.8103,
+      "theoretical_loss": 3.4207750152267495,
+      "tokens_seen": 2068709376
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037955555555555556,
+      "loss": 2.5705,
+      "theoretical_loss": 3.420766039729428,
+      "tokens_seen": 2068774912
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037953535353535357,
+      "loss": 2.6372,
+      "theoretical_loss": 3.420757064596044,
+      "tokens_seen": 2068840448
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037951515151515153,
+      "loss": 2.8598,
+      "theoretical_loss": 3.4207480898265707,
+      "tokens_seen": 2068905984
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003794949494949495,
+      "loss": 2.3985,
+      "theoretical_loss": 3.4207391154209823,
+      "tokens_seen": 2068971520
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003794747474747475,
+      "loss": 2.8237,
+      "theoretical_loss": 3.4207301413792512,
+      "tokens_seen": 2069037056
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037945454545454545,
+      "loss": 2.5604,
+      "theoretical_loss": 3.420721167701353,
+      "tokens_seen": 2069102592
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037943434343434346,
+      "loss": 2.4821,
+      "theoretical_loss": 3.4207121943872605,
+      "tokens_seen": 2069168128
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003794141414141414,
+      "loss": 2.6009,
+      "theoretical_loss": 3.4207032214369475,
+      "tokens_seen": 2069233664
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003793939393939394,
+      "loss": 2.7229,
+      "theoretical_loss": 3.4206942488503875,
+      "tokens_seen": 2069299200
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003793737373737374,
+      "loss": 2.7748,
+      "theoretical_loss": 3.4206852766275544,
+      "tokens_seen": 2069364736
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037935353535353535,
+      "loss": 2.6456,
+      "theoretical_loss": 3.420676304768422,
+      "tokens_seen": 2069430272
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037933333333333336,
+      "loss": 2.8453,
+      "theoretical_loss": 3.420667333272964,
+      "tokens_seen": 2069495808
+    },
+    {
+      "epoch": 0.25,
+      "objective/train/docs_used": 1139859,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.9761197566986084,
+      "objective/train/theoretical_loss": 3.420658362141154,
+      "objective/train/tokens_used": 428420576,
+      "theoretical_loss": 3.420658362141154,
+      "tokens_seen": 2069561344
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003793131313131313,
+      "loss": 2.645,
+      "theoretical_loss": 3.420658362141154,
+      "tokens_seen": 2069561344
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003792929292929293,
+      "loss": 2.4449,
+      "theoretical_loss": 3.420649391372966,
+      "tokens_seen": 2069626880
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003792727272727273,
+      "loss": 2.6226,
+      "theoretical_loss": 3.4206404209683736,
+      "tokens_seen": 2069692416
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037925252525252524,
+      "loss": 2.5629,
+      "theoretical_loss": 3.4206314509273508,
+      "tokens_seen": 2069757952
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037923232323232325,
+      "loss": 2.7043,
+      "theoretical_loss": 3.4206224812498713,
+      "tokens_seen": 2069823488
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003792121212121212,
+      "loss": 2.7883,
+      "theoretical_loss": 3.4206135119359082,
+      "tokens_seen": 2069889024
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003791919191919192,
+      "loss": 2.4786,
+      "theoretical_loss": 3.420604542985436,
+      "tokens_seen": 2069954560
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003791717171717172,
+      "loss": 2.811,
+      "theoretical_loss": 3.420595574398429,
+      "tokens_seen": 2070020096
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037915151515151513,
+      "loss": 2.5502,
+      "theoretical_loss": 3.420586606174859,
+      "tokens_seen": 2070085632
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037913131313131314,
+      "loss": 2.7804,
+      "theoretical_loss": 3.4205776383147013,
+      "tokens_seen": 2070151168
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037911111111111115,
+      "loss": 2.5411,
+      "theoretical_loss": 3.4205686708179295,
+      "tokens_seen": 2070216704
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003790909090909091,
+      "loss": 2.6516,
+      "theoretical_loss": 3.4205597036845177,
+      "tokens_seen": 2070282240
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037907070707070707,
+      "loss": 2.5921,
+      "theoretical_loss": 3.4205507369144383,
+      "tokens_seen": 2070347776
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000379050505050505,
+      "loss": 2.7129,
+      "theoretical_loss": 3.4205417705076666,
+      "tokens_seen": 2070413312
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037903030303030303,
+      "loss": 2.4977,
+      "theoretical_loss": 3.420532804464175,
+      "tokens_seen": 2070478848
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037901010101010105,
+      "loss": 2.6412,
+      "theoretical_loss": 3.4205238387839385,
+      "tokens_seen": 2070544384
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000378989898989899,
+      "loss": 2.4797,
+      "theoretical_loss": 3.4205148734669306,
+      "tokens_seen": 2070609920
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037896969696969696,
+      "loss": 2.4803,
+      "theoretical_loss": 3.4205059085131246,
+      "tokens_seen": 2070675456
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003789494949494949,
+      "loss": 2.5416,
+      "theoretical_loss": 3.420496943922495,
+      "tokens_seen": 2070740992
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000378929292929293,
+      "loss": 2.7728,
+      "theoretical_loss": 3.4204879796950145,
+      "tokens_seen": 2070806528
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037890909090909094,
+      "loss": 2.6949,
+      "theoretical_loss": 3.420479015830658,
+      "tokens_seen": 2070872064
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003788888888888889,
+      "loss": 2.6882,
+      "theoretical_loss": 3.4204700523293985,
+      "tokens_seen": 2070937600
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037886868686868685,
+      "loss": 2.7507,
+      "theoretical_loss": 3.4204610891912104,
+      "tokens_seen": 2071003136
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003788484848484848,
+      "loss": 2.6074,
+      "theoretical_loss": 3.4204521264160674,
+      "tokens_seen": 2071068672
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003788282828282829,
+      "loss": 2.6161,
+      "theoretical_loss": 3.4204431640039434,
+      "tokens_seen": 2071134208
+    },
+    {
+      "epoch": 0.25,
+      "objective/train/docs_used": 1140861,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.942692518234253,
+      "objective/train/theoretical_loss": 3.4204342019548113,
+      "objective/train/tokens_used": 430058976,
+      "theoretical_loss": 3.4204342019548113,
+      "tokens_seen": 2071199744
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037880808080808083,
+      "loss": 2.9214,
+      "theoretical_loss": 3.4204342019548113,
+      "tokens_seen": 2071199744
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003787878787878788,
+      "loss": 2.4544,
+      "theoretical_loss": 3.4204252402686466,
+      "tokens_seen": 2071265280
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037876767676767674,
+      "loss": 2.8409,
+      "theoretical_loss": 3.4204162789454213,
+      "tokens_seen": 2071330816
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037874747474747476,
+      "loss": 2.4468,
+      "theoretical_loss": 3.42040731798511,
+      "tokens_seen": 2071396352
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037872727272727277,
+      "loss": 2.4653,
+      "theoretical_loss": 3.420398357387687,
+      "tokens_seen": 2071461888
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003787070707070707,
+      "loss": 2.7628,
+      "theoretical_loss": 3.4203893971531256,
+      "tokens_seen": 2071527424
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003786868686868687,
+      "loss": 2.4264,
+      "theoretical_loss": 3.4203804372814,
+      "tokens_seen": 2071592960
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037866666666666664,
+      "loss": 2.6893,
+      "theoretical_loss": 3.420371477772483,
+      "tokens_seen": 2071658496
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037864646464646465,
+      "loss": 2.8688,
+      "theoretical_loss": 3.4203625186263498,
+      "tokens_seen": 2071724032
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037862626262626266,
+      "loss": 2.7391,
+      "theoretical_loss": 3.4203535598429737,
+      "tokens_seen": 2071789568
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003786060606060606,
+      "loss": 2.783,
+      "theoretical_loss": 3.4203446014223284,
+      "tokens_seen": 2071855104
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003785858585858586,
+      "loss": 2.6383,
+      "theoretical_loss": 3.4203356433643877,
+      "tokens_seen": 2071920640
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003785656565656566,
+      "loss": 2.61,
+      "theoretical_loss": 3.4203266856691252,
+      "tokens_seen": 2071986176
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037854545454545454,
+      "loss": 2.6615,
+      "theoretical_loss": 3.4203177283365154,
+      "tokens_seen": 2072051712
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037852525252525255,
+      "loss": 2.6566,
+      "theoretical_loss": 3.420308771366532,
+      "tokens_seen": 2072117248
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003785050505050505,
+      "loss": 2.6085,
+      "theoretical_loss": 3.420299814759148,
+      "tokens_seen": 2072182784
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037848484848484847,
+      "loss": 2.6634,
+      "theoretical_loss": 3.420290858514339,
+      "tokens_seen": 2072248320
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003784646464646465,
+      "loss": 2.6108,
+      "theoretical_loss": 3.4202819026320768,
+      "tokens_seen": 2072313856
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003784444444444445,
+      "loss": 2.7928,
+      "theoretical_loss": 3.4202729471123368,
+      "tokens_seen": 2072379392
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037842424242424244,
+      "loss": 2.8207,
+      "theoretical_loss": 3.4202639919550926,
+      "tokens_seen": 2072444928
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003784040404040404,
+      "loss": 2.9374,
+      "theoretical_loss": 3.420255037160317,
+      "tokens_seen": 2072510464
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003783838383838384,
+      "loss": 2.6872,
+      "theoretical_loss": 3.420246082727985,
+      "tokens_seen": 2072576000
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037836363636363637,
+      "loss": 2.6963,
+      "theoretical_loss": 3.42023712865807,
+      "tokens_seen": 2072641536
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003783434343434344,
+      "loss": 2.674,
+      "theoretical_loss": 3.4202281749505463,
+      "tokens_seen": 2072707072
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037832323232323234,
+      "loss": 2.6523,
+      "theoretical_loss": 3.4202192216053873,
+      "tokens_seen": 2072772608
+    },
+    {
+      "epoch": 0.25,
+      "objective/train/docs_used": 1141535,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.150759696960449,
+      "objective/train/theoretical_loss": 3.420210268622567,
+      "objective/train/tokens_used": 431697376,
+      "theoretical_loss": 3.420210268622567,
+      "tokens_seen": 2072838144
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003783030303030303,
+      "loss": 2.7664,
+      "theoretical_loss": 3.420210268622567,
+      "tokens_seen": 2072838144
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003782828282828283,
+      "loss": 2.7415,
+      "theoretical_loss": 3.4202013160020597,
+      "tokens_seen": 2072903680
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037826262626262626,
+      "loss": 2.9062,
+      "theoretical_loss": 3.4201923637438383,
+      "tokens_seen": 2072969216
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003782424242424243,
+      "loss": 2.8399,
+      "theoretical_loss": 3.4201834118478773,
+      "tokens_seen": 2073034752
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037822222222222223,
+      "loss": 2.7105,
+      "theoretical_loss": 3.4201744603141515,
+      "tokens_seen": 2073100288
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003782020202020202,
+      "loss": 2.4727,
+      "theoretical_loss": 3.420165509142633,
+      "tokens_seen": 2073165824
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003781818181818182,
+      "loss": 2.3444,
+      "theoretical_loss": 3.420156558333297,
+      "tokens_seen": 2073231360
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037816161616161615,
+      "loss": 2.5233,
+      "theoretical_loss": 3.4201476078861166,
+      "tokens_seen": 2073296896
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037814141414141417,
+      "loss": 2.4147,
+      "theoretical_loss": 3.420138657801066,
+      "tokens_seen": 2073362432
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003781212121212121,
+      "loss": 2.7456,
+      "theoretical_loss": 3.42012970807812,
+      "tokens_seen": 2073427968
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037810101010101013,
+      "loss": 2.935,
+      "theoretical_loss": 3.4201207587172506,
+      "tokens_seen": 2073493504
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003780808080808081,
+      "loss": 2.6897,
+      "theoretical_loss": 3.4201118097184335,
+      "tokens_seen": 2073559040
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037806060606060605,
+      "loss": 2.645,
+      "theoretical_loss": 3.4201028610816415,
+      "tokens_seen": 2073624576
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037804040404040406,
+      "loss": 2.6722,
+      "theoretical_loss": 3.4200939128068493,
+      "tokens_seen": 2073690112
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000378020202020202,
+      "loss": 2.544,
+      "theoretical_loss": 3.4200849648940306,
+      "tokens_seen": 2073755648
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000378,
+      "loss": 2.7328,
+      "theoretical_loss": 3.420076017343159,
+      "tokens_seen": 2073821184
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000377979797979798,
+      "loss": 3.0095,
+      "theoretical_loss": 3.420067070154208,
+      "tokens_seen": 2073886720
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037795959595959594,
+      "loss": 2.5682,
+      "theoretical_loss": 3.4200581233271525,
+      "tokens_seen": 2073952256
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037793939393939395,
+      "loss": 2.576,
+      "theoretical_loss": 3.420049176861966,
+      "tokens_seen": 2074017792
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037791919191919196,
+      "loss": 2.7552,
+      "theoretical_loss": 3.420040230758622,
+      "tokens_seen": 2074083328
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003778989898989899,
+      "loss": 2.6943,
+      "theoretical_loss": 3.420031285017096,
+      "tokens_seen": 2074148864
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003778787878787879,
+      "loss": 2.6718,
+      "theoretical_loss": 3.42002233963736,
+      "tokens_seen": 2074214400
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037785858585858583,
+      "loss": 2.6333,
+      "theoretical_loss": 3.4200133946193887,
+      "tokens_seen": 2074279936
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037783838383838384,
+      "loss": 2.4596,
+      "theoretical_loss": 3.420004449963156,
+      "tokens_seen": 2074345472
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037781818181818186,
+      "loss": 2.4134,
+      "theoretical_loss": 3.419995505668637,
+      "tokens_seen": 2074411008
+    },
+    {
+      "epoch": 0.25,
+      "objective/train/docs_used": 1142675,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.8896119594573975,
+      "objective/train/theoretical_loss": 3.4199865617358034,
+      "objective/train/tokens_used": 433335776,
+      "theoretical_loss": 3.4199865617358034,
+      "tokens_seen": 2074476544
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003777979797979798,
+      "loss": 2.7336,
+      "theoretical_loss": 3.4199865617358034,
+      "tokens_seen": 2074476544
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037777777777777777,
+      "loss": 2.6664,
+      "theoretical_loss": 3.419977618164631,
+      "tokens_seen": 2074542080
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003777575757575757,
+      "loss": 2.5695,
+      "theoretical_loss": 3.4199686749550926,
+      "tokens_seen": 2074607616
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003777373737373738,
+      "loss": 2.7597,
+      "theoretical_loss": 3.419959732107163,
+      "tokens_seen": 2074673152
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037771717171717175,
+      "loss": 2.3906,
+      "theoretical_loss": 3.419950789620816,
+      "tokens_seen": 2074738688
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003776969696969697,
+      "loss": 2.5256,
+      "theoretical_loss": 3.419941847496025,
+      "tokens_seen": 2074804224
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037767676767676766,
+      "loss": 2.6481,
+      "theoretical_loss": 3.419932905732764,
+      "tokens_seen": 2074869760
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003776565656565656,
+      "loss": 2.6393,
+      "theoretical_loss": 3.4199239643310078,
+      "tokens_seen": 2074935296
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003776363636363637,
+      "loss": 2.4312,
+      "theoretical_loss": 3.41991502329073,
+      "tokens_seen": 2075000832
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037761616161616164,
+      "loss": 2.6489,
+      "theoretical_loss": 3.4199060826119037,
+      "tokens_seen": 2075066368
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003775959595959596,
+      "loss": 2.7394,
+      "theoretical_loss": 3.419897142294504,
+      "tokens_seen": 2075131904
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037757575757575755,
+      "loss": 2.649,
+      "theoretical_loss": 3.4198882023385044,
+      "tokens_seen": 2075197440
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037755555555555557,
+      "loss": 2.7409,
+      "theoretical_loss": 3.419879262743879,
+      "tokens_seen": 2075262976
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003775353535353536,
+      "loss": 2.7992,
+      "theoretical_loss": 3.419870323510602,
+      "tokens_seen": 2075328512
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037751515151515153,
+      "loss": 2.7074,
+      "theoretical_loss": 3.4198613846386468,
+      "tokens_seen": 2075394048
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003774949494949495,
+      "loss": 2.4963,
+      "theoretical_loss": 3.4198524461279876,
+      "tokens_seen": 2075459584
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037747474747474745,
+      "loss": 2.8299,
+      "theoretical_loss": 3.419843507978599,
+      "tokens_seen": 2075525120
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037745454545454546,
+      "loss": 2.6015,
+      "theoretical_loss": 3.419834570190454,
+      "tokens_seen": 2075590656
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037743434343434347,
+      "loss": 2.6814,
+      "theoretical_loss": 3.4198256327635272,
+      "tokens_seen": 2075656192
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003774141414141414,
+      "loss": 2.8352,
+      "theoretical_loss": 3.4198166956977927,
+      "tokens_seen": 2075721728
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003773939393939394,
+      "loss": 2.7546,
+      "theoretical_loss": 3.4198077589932234,
+      "tokens_seen": 2075787264
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003773737373737374,
+      "loss": 2.6698,
+      "theoretical_loss": 3.4197988226497955,
+      "tokens_seen": 2075852800
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037735353535353535,
+      "loss": 2.7075,
+      "theoretical_loss": 3.4197898866674805,
+      "tokens_seen": 2075918336
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037733333333333336,
+      "loss": 2.5608,
+      "theoretical_loss": 3.4197809510462545,
+      "tokens_seen": 2075983872
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003773131313131313,
+      "loss": 2.3638,
+      "theoretical_loss": 3.41977201578609,
+      "tokens_seen": 2076049408
+    },
+    {
+      "epoch": 0.25,
+      "objective/train/docs_used": 1143849,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.351544141769409,
+      "objective/train/theoretical_loss": 3.4197630808869617,
+      "objective/train/tokens_used": 434974176,
+      "theoretical_loss": 3.4197630808869617,
+      "tokens_seen": 2076114944
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003772929292929293,
+      "loss": 2.6191,
+      "theoretical_loss": 3.4197630808869617,
+      "tokens_seen": 2076114944
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003772727272727273,
+      "loss": 2.8404,
+      "theoretical_loss": 3.4197541463488434,
+      "tokens_seen": 2076180480
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037725252525252524,
+      "loss": 2.7983,
+      "theoretical_loss": 3.41974521217171,
+      "tokens_seen": 2076246016
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037723232323232325,
+      "loss": 2.6373,
+      "theoretical_loss": 3.419736278355534,
+      "tokens_seen": 2076311552
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003772121212121212,
+      "loss": 2.6044,
+      "theoretical_loss": 3.4197273449002905,
+      "tokens_seen": 2076377088
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003771919191919192,
+      "loss": 2.3621,
+      "theoretical_loss": 3.4197184118059534,
+      "tokens_seen": 2076442624
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003771717171717172,
+      "loss": 2.8032,
+      "theoretical_loss": 3.419709479072496,
+      "tokens_seen": 2076508160
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037715151515151514,
+      "loss": 2.7246,
+      "theoretical_loss": 3.419700546699893,
+      "tokens_seen": 2076573696
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037713131313131315,
+      "loss": 2.773,
+      "theoretical_loss": 3.4196916146881184,
+      "tokens_seen": 2076639232
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003771111111111111,
+      "loss": 2.87,
+      "theoretical_loss": 3.4196826830371467,
+      "tokens_seen": 2076704768
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003770909090909091,
+      "loss": 2.6496,
+      "theoretical_loss": 3.419673751746951,
+      "tokens_seen": 2076770304
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037707070707070707,
+      "loss": 2.5852,
+      "theoretical_loss": 3.4196648208175056,
+      "tokens_seen": 2076835840
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037705050505050503,
+      "loss": 2.7811,
+      "theoretical_loss": 3.4196558902487846,
+      "tokens_seen": 2076901376
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037703030303030304,
+      "loss": 2.5733,
+      "theoretical_loss": 3.4196469600407626,
+      "tokens_seen": 2076966912
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037701010101010105,
+      "loss": 2.5622,
+      "theoretical_loss": 3.419638030193413,
+      "tokens_seen": 2077032448
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000376989898989899,
+      "loss": 2.6142,
+      "theoretical_loss": 3.4196291007067097,
+      "tokens_seen": 2077097984
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037696969696969696,
+      "loss": 2.6837,
+      "theoretical_loss": 3.419620171580627,
+      "tokens_seen": 2077163520
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003769494949494949,
+      "loss": 2.5987,
+      "theoretical_loss": 3.4196112428151397,
+      "tokens_seen": 2077229056
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037692929292929293,
+      "loss": 2.7628,
+      "theoretical_loss": 3.419602314410221,
+      "tokens_seen": 2077294592
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037690909090909094,
+      "loss": 2.5048,
+      "theoretical_loss": 3.419593386365845,
+      "tokens_seen": 2077360128
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003768888888888889,
+      "loss": 2.5455,
+      "theoretical_loss": 3.4195844586819857,
+      "tokens_seen": 2077425664
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037686868686868686,
+      "loss": 2.6995,
+      "theoretical_loss": 3.4195755313586176,
+      "tokens_seen": 2077491200
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003768484848484848,
+      "loss": 2.901,
+      "theoretical_loss": 3.419566604395715,
+      "tokens_seen": 2077556736
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003768282828282829,
+      "loss": 2.6293,
+      "theoretical_loss": 3.419557677793251,
+      "tokens_seen": 2077622272
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037680808080808084,
+      "loss": 2.4327,
+      "theoretical_loss": 3.4195487515512006,
+      "tokens_seen": 2077687808
+    },
+    {
+      "epoch": 0.25,
+      "objective/train/docs_used": 1144478,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.4185636043548584,
+      "objective/train/theoretical_loss": 3.4195398256695375,
+      "objective/train/tokens_used": 436612576,
+      "theoretical_loss": 3.4195398256695375,
+      "tokens_seen": 2077753344
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003767878787878788,
+      "loss": 2.756,
+      "theoretical_loss": 3.4195398256695375,
+      "tokens_seen": 2077753344
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037676767676767675,
+      "loss": 2.6662,
+      "theoretical_loss": 3.4195309001482355,
+      "tokens_seen": 2077818880
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003767474747474747,
+      "loss": 2.5164,
+      "theoretical_loss": 3.419521974987269,
+      "tokens_seen": 2077884416
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037672727272727277,
+      "loss": 2.7099,
+      "theoretical_loss": 3.4195130501866124,
+      "tokens_seen": 2077949952
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037670707070707073,
+      "loss": 2.5342,
+      "theoretical_loss": 3.4195041257462395,
+      "tokens_seen": 2078015488
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003766868686868687,
+      "loss": 2.5396,
+      "theoretical_loss": 3.419495201666124,
+      "tokens_seen": 2078081024
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037666666666666664,
+      "loss": 2.5388,
+      "theoretical_loss": 3.4194862779462407,
+      "tokens_seen": 2078146560
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003766464646464647,
+      "loss": 2.6906,
+      "theoretical_loss": 3.4194773545865633,
+      "tokens_seen": 2078212096
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037662626262626266,
+      "loss": 2.7033,
+      "theoretical_loss": 3.419468431587066,
+      "tokens_seen": 2078277632
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003766060606060606,
+      "loss": 2.6427,
+      "theoretical_loss": 3.419459508947723,
+      "tokens_seen": 2078343168
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003765858585858586,
+      "loss": 2.6233,
+      "theoretical_loss": 3.419450586668508,
+      "tokens_seen": 2078408704
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037656565656565654,
+      "loss": 2.671,
+      "theoretical_loss": 3.419441664749395,
+      "tokens_seen": 2078474240
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003765454545454546,
+      "loss": 2.5917,
+      "theoretical_loss": 3.419432743190359,
+      "tokens_seen": 2078539776
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037652525252525256,
+      "loss": 2.7356,
+      "theoretical_loss": 3.4194238219913737,
+      "tokens_seen": 2078605312
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003765050505050505,
+      "loss": 2.5018,
+      "theoretical_loss": 3.419414901152413,
+      "tokens_seen": 2078670848
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037648484848484847,
+      "loss": 2.6064,
+      "theoretical_loss": 3.4194059806734507,
+      "tokens_seen": 2078736384
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037646464646464643,
+      "loss": 2.8587,
+      "theoretical_loss": 3.419397060554462,
+      "tokens_seen": 2078801920
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003764444444444445,
+      "loss": 2.5207,
+      "theoretical_loss": 3.41938814079542,
+      "tokens_seen": 2078867456
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037642424242424245,
+      "loss": 2.644,
+      "theoretical_loss": 3.4193792213962992,
+      "tokens_seen": 2078932992
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003764040404040404,
+      "loss": 2.4817,
+      "theoretical_loss": 3.4193703023570743,
+      "tokens_seen": 2078998528
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037638383838383836,
+      "loss": 2.4036,
+      "theoretical_loss": 3.4193613836777184,
+      "tokens_seen": 2079064064
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003763636363636364,
+      "loss": 2.5825,
+      "theoretical_loss": 3.419352465358206,
+      "tokens_seen": 2079129600
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003763434343434344,
+      "loss": 2.6791,
+      "theoretical_loss": 3.419343547398512,
+      "tokens_seen": 2079195136
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00037632323232323234,
+      "loss": 2.6596,
+      "theoretical_loss": 3.4193346297986094,
+      "tokens_seen": 2079260672
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003763030303030303,
+      "loss": 2.6833,
+      "theoretical_loss": 3.419325712558473,
+      "tokens_seen": 2079326208
+    },
+    {
+      "epoch": 0.25,
+      "objective/train/docs_used": 1145756,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6733858585357666,
+      "objective/train/theoretical_loss": 3.4193167956780766,
+      "objective/train/tokens_used": 438250976,
+      "theoretical_loss": 3.4193167956780766,
+      "tokens_seen": 2079391744
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037628282828282826,
+      "loss": 2.4661,
+      "theoretical_loss": 3.4193167956780766,
+      "tokens_seen": 2079391744
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037626262626262627,
+      "loss": 2.744,
+      "theoretical_loss": 3.4193078791573948,
+      "tokens_seen": 2079457280
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003762424242424243,
+      "loss": 2.611,
+      "theoretical_loss": 3.4192989629964012,
+      "tokens_seen": 2079522816
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037622222222222224,
+      "loss": 2.7055,
+      "theoretical_loss": 3.4192900471950707,
+      "tokens_seen": 2079588352
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003762020202020202,
+      "loss": 2.6647,
+      "theoretical_loss": 3.419281131753377,
+      "tokens_seen": 2079653888
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003761818181818182,
+      "loss": 2.732,
+      "theoretical_loss": 3.419272216671294,
+      "tokens_seen": 2079719424
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037616161616161616,
+      "loss": 2.6474,
+      "theoretical_loss": 3.4192633019487966,
+      "tokens_seen": 2079784960
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037614141414141417,
+      "loss": 2.6054,
+      "theoretical_loss": 3.4192543875858576,
+      "tokens_seen": 2079850496
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037612121212121213,
+      "loss": 2.714,
+      "theoretical_loss": 3.419245473582453,
+      "tokens_seen": 2079916032
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003761010101010101,
+      "loss": 2.6577,
+      "theoretical_loss": 3.4192365599385557,
+      "tokens_seen": 2079981568
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003760808080808081,
+      "loss": 2.7123,
+      "theoretical_loss": 3.4192276466541403,
+      "tokens_seen": 2080047104
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037606060606060605,
+      "loss": 2.9202,
+      "theoretical_loss": 3.4192187337291804,
+      "tokens_seen": 2080112640
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037604040404040406,
+      "loss": 2.7607,
+      "theoretical_loss": 3.4192098211636512,
+      "tokens_seen": 2080178176
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000376020202020202,
+      "loss": 2.6153,
+      "theoretical_loss": 3.419200908957526,
+      "tokens_seen": 2080243712
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037600000000000003,
+      "loss": 2.6651,
+      "theoretical_loss": 3.41919199711078,
+      "tokens_seen": 2080309248
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000375979797979798,
+      "loss": 2.7883,
+      "theoretical_loss": 3.419183085623386,
+      "tokens_seen": 2080374784
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037595959595959595,
+      "loss": 2.5957,
+      "theoretical_loss": 3.4191741744953195,
+      "tokens_seen": 2080440320
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037593939393939396,
+      "loss": 2.4795,
+      "theoretical_loss": 3.419165263726554,
+      "tokens_seen": 2080505856
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003759191919191919,
+      "loss": 2.4591,
+      "theoretical_loss": 3.4191563533170637,
+      "tokens_seen": 2080571392
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003758989898989899,
+      "loss": 2.5381,
+      "theoretical_loss": 3.419147443266823,
+      "tokens_seen": 2080636928
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003758787878787879,
+      "loss": 2.76,
+      "theoretical_loss": 3.4191385335758055,
+      "tokens_seen": 2080702464
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037585858585858584,
+      "loss": 2.8718,
+      "theoretical_loss": 3.419129624243986,
+      "tokens_seen": 2080768000
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037583838383838385,
+      "loss": 2.6215,
+      "theoretical_loss": 3.4191207152713394,
+      "tokens_seen": 2080833536
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037581818181818186,
+      "loss": 2.7257,
+      "theoretical_loss": 3.4191118066578383,
+      "tokens_seen": 2080899072
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003757979797979798,
+      "loss": 2.7604,
+      "theoretical_loss": 3.419102898403459,
+      "tokens_seen": 2080964608
+    },
+    {
+      "epoch": 0.26,
+      "objective/train/docs_used": 1146429,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.32552433013916,
+      "objective/train/theoretical_loss": 3.419093990508173,
+      "objective/train/tokens_used": 439889376,
+      "theoretical_loss": 3.419093990508173,
+      "tokens_seen": 2081030144
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003757777777777778,
+      "loss": 2.6839,
+      "theoretical_loss": 3.419093990508173,
+      "tokens_seen": 2081030144
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037575757575757573,
+      "loss": 2.6432,
+      "theoretical_loss": 3.4190850829719563,
+      "tokens_seen": 2081095680
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037573737373737374,
+      "loss": 2.557,
+      "theoretical_loss": 3.4190761757947836,
+      "tokens_seen": 2081161216
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037571717171717175,
+      "loss": 2.5485,
+      "theoretical_loss": 3.4190672689766277,
+      "tokens_seen": 2081226752
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003756969696969697,
+      "loss": 2.7005,
+      "theoretical_loss": 3.4190583625174633,
+      "tokens_seen": 2081292288
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037567676767676767,
+      "loss": 2.6643,
+      "theoretical_loss": 3.419049456417265,
+      "tokens_seen": 2081357824
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003756565656565656,
+      "loss": 2.5289,
+      "theoretical_loss": 3.419040550676007,
+      "tokens_seen": 2081423360
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003756363636363637,
+      "loss": 2.8114,
+      "theoretical_loss": 3.419031645293663,
+      "tokens_seen": 2081488896
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037561616161616165,
+      "loss": 2.7906,
+      "theoretical_loss": 3.4190227402702074,
+      "tokens_seen": 2081554432
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003755959595959596,
+      "loss": 2.9096,
+      "theoretical_loss": 3.419013835605615,
+      "tokens_seen": 2081619968
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037557575757575756,
+      "loss": 2.5331,
+      "theoretical_loss": 3.4190049312998596,
+      "tokens_seen": 2081685504
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003755555555555555,
+      "loss": 2.7058,
+      "theoretical_loss": 3.4189960273529154,
+      "tokens_seen": 2081751040
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003755353535353536,
+      "loss": 2.726,
+      "theoretical_loss": 3.4189871237647567,
+      "tokens_seen": 2081816576
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037551515151515154,
+      "loss": 2.562,
+      "theoretical_loss": 3.4189782205353576,
+      "tokens_seen": 2081882112
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003754949494949495,
+      "loss": 2.5914,
+      "theoretical_loss": 3.4189693176646925,
+      "tokens_seen": 2081947648
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037547474747474745,
+      "loss": 2.7903,
+      "theoretical_loss": 3.4189604151527364,
+      "tokens_seen": 2082013184
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037545454545454546,
+      "loss": 2.6528,
+      "theoretical_loss": 3.4189515129994623,
+      "tokens_seen": 2082078720
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003754343434343435,
+      "loss": 2.6828,
+      "theoretical_loss": 3.418942611204845,
+      "tokens_seen": 2082144256
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037541414141414143,
+      "loss": 2.5347,
+      "theoretical_loss": 3.418933709768859,
+      "tokens_seen": 2082209792
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003753939393939394,
+      "loss": 2.6722,
+      "theoretical_loss": 3.418924808691478,
+      "tokens_seen": 2082275328
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037537373737373735,
+      "loss": 2.4734,
+      "theoretical_loss": 3.4189159079726767,
+      "tokens_seen": 2082340864
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037535353535353536,
+      "loss": 2.6157,
+      "theoretical_loss": 3.4189070076124293,
+      "tokens_seen": 2082406400
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037533333333333337,
+      "loss": 2.7297,
+      "theoretical_loss": 3.41889810761071,
+      "tokens_seen": 2082471936
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003753131313131313,
+      "loss": 2.6283,
+      "theoretical_loss": 3.418889207967493,
+      "tokens_seen": 2082537472
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003752929292929293,
+      "loss": 2.3652,
+      "theoretical_loss": 3.418880308682753,
+      "tokens_seen": 2082603008
+    },
+    {
+      "epoch": 0.26,
+      "objective/train/docs_used": 1147399,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.0194616317749023,
+      "objective/train/theoretical_loss": 3.4188714097564636,
+      "objective/train/tokens_used": 441527776,
+      "theoretical_loss": 3.4188714097564636,
+      "tokens_seen": 2082668544
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003752727272727273,
+      "loss": 2.8635,
+      "theoretical_loss": 3.4188714097564636,
+      "tokens_seen": 2082668544
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037525252525252525,
+      "loss": 2.7142,
+      "theoretical_loss": 3.4188625111886,
+      "tokens_seen": 2082734080
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037523232323232326,
+      "loss": 2.5603,
+      "theoretical_loss": 3.4188536129791354,
+      "tokens_seen": 2082799616
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003752121212121212,
+      "loss": 2.668,
+      "theoretical_loss": 3.418844715128045,
+      "tokens_seen": 2082865152
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003751919191919192,
+      "loss": 2.9046,
+      "theoretical_loss": 3.418835817635303,
+      "tokens_seen": 2082930688
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003751717171717172,
+      "loss": 2.5116,
+      "theoretical_loss": 3.4188269205008828,
+      "tokens_seen": 2082996224
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037515151515151514,
+      "loss": 2.5802,
+      "theoretical_loss": 3.41881802372476,
+      "tokens_seen": 2083061760
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037513131313131315,
+      "loss": 2.6405,
+      "theoretical_loss": 3.4188091273069072,
+      "tokens_seen": 2083127296
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003751111111111111,
+      "loss": 2.613,
+      "theoretical_loss": 3.4188002312473005,
+      "tokens_seen": 2083192832
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003750909090909091,
+      "loss": 2.6961,
+      "theoretical_loss": 3.4187913355459134,
+      "tokens_seen": 2083258368
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003750707070707071,
+      "loss": 2.5561,
+      "theoretical_loss": 3.41878244020272,
+      "tokens_seen": 2083323904
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037505050505050503,
+      "loss": 2.6632,
+      "theoretical_loss": 3.418773545217695,
+      "tokens_seen": 2083389440
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037503030303030305,
+      "loss": 3.0064,
+      "theoretical_loss": 3.418764650590812,
+      "tokens_seen": 2083454976
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000375010101010101,
+      "loss": 2.7303,
+      "theoretical_loss": 3.4187557563220463,
+      "tokens_seen": 2083520512
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000374989898989899,
+      "loss": 2.7951,
+      "theoretical_loss": 3.4187468624113717,
+      "tokens_seen": 2083586048
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037496969696969697,
+      "loss": 2.4562,
+      "theoretical_loss": 3.4187379688587627,
+      "tokens_seen": 2083651584
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000374949494949495,
+      "loss": 2.6969,
+      "theoretical_loss": 3.4187290756641935,
+      "tokens_seen": 2083717120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037492929292929294,
+      "loss": 2.3717,
+      "theoretical_loss": 3.4187201828276383,
+      "tokens_seen": 2083782656
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003749090909090909,
+      "loss": 2.6182,
+      "theoretical_loss": 3.418711290349072,
+      "tokens_seen": 2083848192
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003748888888888889,
+      "loss": 2.7406,
+      "theoretical_loss": 3.418702398228468,
+      "tokens_seen": 2083913728
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037486868686868686,
+      "loss": 2.6609,
+      "theoretical_loss": 3.418693506465801,
+      "tokens_seen": 2083979264
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003748484848484849,
+      "loss": 2.6927,
+      "theoretical_loss": 3.4186846150610464,
+      "tokens_seen": 2084044800
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037482828282828283,
+      "loss": 2.9106,
+      "theoretical_loss": 3.418675724014177,
+      "tokens_seen": 2084110336
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037480808080808084,
+      "loss": 2.7649,
+      "theoretical_loss": 3.418666833325168,
+      "tokens_seen": 2084175872
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003747878787878788,
+      "loss": 2.6435,
+      "theoretical_loss": 3.4186579429939927,
+      "tokens_seen": 2084241408
+    },
+    {
+      "epoch": 0.26,
+      "objective/train/docs_used": 1148024,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.3189051151275635,
+      "objective/train/theoretical_loss": 3.4186490530206273,
+      "objective/train/tokens_used": 443166176,
+      "theoretical_loss": 3.4186490530206273,
+      "tokens_seen": 2084306944
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037476767676767676,
+      "loss": 2.4753,
+      "theoretical_loss": 3.4186490530206273,
+      "tokens_seen": 2084306944
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037474747474747477,
+      "loss": 2.7101,
+      "theoretical_loss": 3.4186401634050445,
+      "tokens_seen": 2084372480
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003747272727272727,
+      "loss": 2.6549,
+      "theoretical_loss": 3.4186312741472196,
+      "tokens_seen": 2084438016
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037470707070707073,
+      "loss": 2.6999,
+      "theoretical_loss": 3.4186223852471262,
+      "tokens_seen": 2084503552
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003746868686868687,
+      "loss": 2.6416,
+      "theoretical_loss": 3.4186134967047392,
+      "tokens_seen": 2084569088
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037466666666666665,
+      "loss": 2.6206,
+      "theoretical_loss": 3.4186046085200332,
+      "tokens_seen": 2084634624
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037464646464646466,
+      "loss": 2.6309,
+      "theoretical_loss": 3.418595720692982,
+      "tokens_seen": 2084700160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037462626262626267,
+      "loss": 2.7098,
+      "theoretical_loss": 3.41858683322356,
+      "tokens_seen": 2084765696
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037460606060606063,
+      "loss": 2.3725,
+      "theoretical_loss": 3.418577946111742,
+      "tokens_seen": 2084831232
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003745858585858586,
+      "loss": 2.7087,
+      "theoretical_loss": 3.4185690593575018,
+      "tokens_seen": 2084896768
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037456565656565654,
+      "loss": 2.7147,
+      "theoretical_loss": 3.418560172960814,
+      "tokens_seen": 2084962304
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037454545454545455,
+      "loss": 2.5222,
+      "theoretical_loss": 3.418551286921653,
+      "tokens_seen": 2085027840
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037452525252525256,
+      "loss": 2.4876,
+      "theoretical_loss": 3.4185424012399936,
+      "tokens_seen": 2085093376
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003745050505050505,
+      "loss": 2.6135,
+      "theoretical_loss": 3.4185335159158097,
+      "tokens_seen": 2085158912
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003744848484848485,
+      "loss": 2.5551,
+      "theoretical_loss": 3.4185246309490758,
+      "tokens_seen": 2085224448
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037446464646464643,
+      "loss": 2.5469,
+      "theoretical_loss": 3.418515746339766,
+      "tokens_seen": 2085289984
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003744444444444445,
+      "loss": 2.536,
+      "theoretical_loss": 3.418506862087855,
+      "tokens_seen": 2085355520
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037442424242424246,
+      "loss": 2.4808,
+      "theoretical_loss": 3.4184979781933174,
+      "tokens_seen": 2085421056
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003744040404040404,
+      "loss": 2.5116,
+      "theoretical_loss": 3.418489094656127,
+      "tokens_seen": 2085486592
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037438383838383837,
+      "loss": 2.597,
+      "theoretical_loss": 3.418480211476259,
+      "tokens_seen": 2085552128
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003743636363636363,
+      "loss": 2.7988,
+      "theoretical_loss": 3.4184713286536867,
+      "tokens_seen": 2085617664
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003743434343434344,
+      "loss": 2.6326,
+      "theoretical_loss": 3.4184624461883857,
+      "tokens_seen": 2085683200
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037432323232323235,
+      "loss": 2.5852,
+      "theoretical_loss": 3.41845356408033,
+      "tokens_seen": 2085748736
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003743030303030303,
+      "loss": 2.6487,
+      "theoretical_loss": 3.4184446823294934,
+      "tokens_seen": 2085814272
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037428282828282826,
+      "loss": 2.6216,
+      "theoretical_loss": 3.418435800935851,
+      "tokens_seen": 2085879808
+    },
+    {
+      "epoch": 0.26,
+      "objective/train/docs_used": 1149004,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6764867305755615,
+      "objective/train/theoretical_loss": 3.4184269198993773,
+      "objective/train/tokens_used": 444804576,
+      "theoretical_loss": 3.4184269198993773,
+      "tokens_seen": 2085945344
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003742626262626263,
+      "loss": 2.7363,
+      "theoretical_loss": 3.4184269198993773,
+      "tokens_seen": 2085945344
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003742424242424243,
+      "loss": 2.4787,
+      "theoretical_loss": 3.418418039220046,
+      "tokens_seen": 2086010880
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037422222222222224,
+      "loss": 2.5255,
+      "theoretical_loss": 3.4184091588978314,
+      "tokens_seen": 2086076416
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003742020202020202,
+      "loss": 2.8832,
+      "theoretical_loss": 3.418400278932709,
+      "tokens_seen": 2086141952
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037418181818181815,
+      "loss": 2.6207,
+      "theoretical_loss": 3.418391399324653,
+      "tokens_seen": 2086207488
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037416161616161617,
+      "loss": 2.5225,
+      "theoretical_loss": 3.418382520073637,
+      "tokens_seen": 2086273024
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003741414141414142,
+      "loss": 2.5866,
+      "theoretical_loss": 3.4183736411796364,
+      "tokens_seen": 2086338560
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037412121212121213,
+      "loss": 2.7921,
+      "theoretical_loss": 3.4183647626426246,
+      "tokens_seen": 2086404096
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003741010101010101,
+      "loss": 2.6267,
+      "theoretical_loss": 3.4183558844625765,
+      "tokens_seen": 2086469632
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003740808080808081,
+      "loss": 2.736,
+      "theoretical_loss": 3.418347006639467,
+      "tokens_seen": 2086535168
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037406060606060606,
+      "loss": 2.5893,
+      "theoretical_loss": 3.41833812917327,
+      "tokens_seen": 2086600704
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037404040404040407,
+      "loss": 2.4624,
+      "theoretical_loss": 3.4183292520639603,
+      "tokens_seen": 2086666240
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000374020202020202,
+      "loss": 2.7353,
+      "theoretical_loss": 3.418320375311512,
+      "tokens_seen": 2086731776
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000374,
+      "loss": 2.7195,
+      "theoretical_loss": 3.4183114989159,
+      "tokens_seen": 2086797312
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000373979797979798,
+      "loss": 2.7483,
+      "theoretical_loss": 3.4183026228770976,
+      "tokens_seen": 2086862848
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037395959595959595,
+      "loss": 2.7238,
+      "theoretical_loss": 3.4182937471950807,
+      "tokens_seen": 2086928384
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037393939393939396,
+      "loss": 2.6748,
+      "theoretical_loss": 3.4182848718698233,
+      "tokens_seen": 2086993920
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003739191919191919,
+      "loss": 2.4875,
+      "theoretical_loss": 3.4182759969012992,
+      "tokens_seen": 2087059456
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037389898989898993,
+      "loss": 2.6688,
+      "theoretical_loss": 3.4182671222894836,
+      "tokens_seen": 2087124992
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003738787878787879,
+      "loss": 2.831,
+      "theoretical_loss": 3.418258248034351,
+      "tokens_seen": 2087190528
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037385858585858584,
+      "loss": 2.4438,
+      "theoretical_loss": 3.4182493741358755,
+      "tokens_seen": 2087256064
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037383838383838386,
+      "loss": 2.6071,
+      "theoretical_loss": 3.418240500594031,
+      "tokens_seen": 2087321600
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003738181818181818,
+      "loss": 2.7333,
+      "theoretical_loss": 3.4182316274087934,
+      "tokens_seen": 2087387136
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003737979797979798,
+      "loss": 2.7107,
+      "theoretical_loss": 3.4182227545801362,
+      "tokens_seen": 2087452672
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003737777777777778,
+      "loss": 2.491,
+      "theoretical_loss": 3.418213882108034,
+      "tokens_seen": 2087518208
+    },
+    {
+      "epoch": 0.26,
+      "objective/train/docs_used": 1149506,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7207868099212646,
+      "objective/train/theoretical_loss": 3.4182050099924615,
+      "objective/train/tokens_used": 446442976,
+      "theoretical_loss": 3.4182050099924615,
+      "tokens_seen": 2087583744
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037375757575757574,
+      "loss": 2.615,
+      "theoretical_loss": 3.4182050099924615,
+      "tokens_seen": 2087583744
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037373737373737375,
+      "loss": 2.6447,
+      "theoretical_loss": 3.4181961382333927,
+      "tokens_seen": 2087649280
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037371717171717176,
+      "loss": 2.7587,
+      "theoretical_loss": 3.4181872668308024,
+      "tokens_seen": 2087714816
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003736969696969697,
+      "loss": 2.8354,
+      "theoretical_loss": 3.4181783957846656,
+      "tokens_seen": 2087780352
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037367676767676767,
+      "loss": 2.4997,
+      "theoretical_loss": 3.418169525094956,
+      "tokens_seen": 2087845888
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037365656565656563,
+      "loss": 2.6017,
+      "theoretical_loss": 3.4181606547616488,
+      "tokens_seen": 2087911424
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037363636363636364,
+      "loss": 2.7573,
+      "theoretical_loss": 3.4181517847847176,
+      "tokens_seen": 2087976960
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037361616161616165,
+      "loss": 2.5815,
+      "theoretical_loss": 3.418142915164138,
+      "tokens_seen": 2088042496
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003735959595959596,
+      "loss": 2.8231,
+      "theoretical_loss": 3.4181340458998832,
+      "tokens_seen": 2088108032
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037357575757575757,
+      "loss": 2.6708,
+      "theoretical_loss": 3.4181251769919285,
+      "tokens_seen": 2088173568
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003735555555555555,
+      "loss": 2.6358,
+      "theoretical_loss": 3.418116308440248,
+      "tokens_seen": 2088239104
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003735353535353536,
+      "loss": 2.6792,
+      "theoretical_loss": 3.4181074402448175,
+      "tokens_seen": 2088304640
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037351515151515154,
+      "loss": 2.8498,
+      "theoretical_loss": 3.4180985724056097,
+      "tokens_seen": 2088370176
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003734949494949495,
+      "loss": 2.8895,
+      "theoretical_loss": 3.4180897049226004,
+      "tokens_seen": 2088435712
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037347474747474746,
+      "loss": 2.4677,
+      "theoretical_loss": 3.418080837795763,
+      "tokens_seen": 2088501248
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003734545454545454,
+      "loss": 2.5268,
+      "theoretical_loss": 3.418071971025073,
+      "tokens_seen": 2088566784
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003734343434343435,
+      "loss": 2.797,
+      "theoretical_loss": 3.4180631046105048,
+      "tokens_seen": 2088632320
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037341414141414144,
+      "loss": 2.8185,
+      "theoretical_loss": 3.4180542385520325,
+      "tokens_seen": 2088697856
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003733939393939394,
+      "loss": 2.8718,
+      "theoretical_loss": 3.4180453728496305,
+      "tokens_seen": 2088763392
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037337373737373735,
+      "loss": 2.769,
+      "theoretical_loss": 3.4180365075032744,
+      "tokens_seen": 2088828928
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003733535353535353,
+      "loss": 2.4134,
+      "theoretical_loss": 3.4180276425129374,
+      "tokens_seen": 2088894464
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003733333333333334,
+      "loss": 2.7009,
+      "theoretical_loss": 3.4180187778785953,
+      "tokens_seen": 2088960000
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037331313131313133,
+      "loss": 2.6487,
+      "theoretical_loss": 3.418009913600221,
+      "tokens_seen": 2089025536
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003732929292929293,
+      "loss": 2.809,
+      "theoretical_loss": 3.4180010496777906,
+      "tokens_seen": 2089091072
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037327272727272724,
+      "loss": 2.6093,
+      "theoretical_loss": 3.417992186111278,
+      "tokens_seen": 2089156608
+    },
+    {
+      "epoch": 0.26,
+      "objective/train/docs_used": 1150478,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.582503080368042,
+      "objective/train/theoretical_loss": 3.4179833229006578,
+      "objective/train/tokens_used": 448081376,
+      "theoretical_loss": 3.4179833229006578,
+      "tokens_seen": 2089222144
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003732525252525253,
+      "loss": 2.5808,
+      "theoretical_loss": 3.4179833229006578,
+      "tokens_seen": 2089222144
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037323232323232327,
+      "loss": 2.5832,
+      "theoretical_loss": 3.417974460045904,
+      "tokens_seen": 2089287680
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003732121212121212,
+      "loss": 2.8097,
+      "theoretical_loss": 3.417965597546992,
+      "tokens_seen": 2089353216
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003731919191919192,
+      "loss": 2.8212,
+      "theoretical_loss": 3.417956735403896,
+      "tokens_seen": 2089418752
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037317171717171714,
+      "loss": 2.7588,
+      "theoretical_loss": 3.4179478736165914,
+      "tokens_seen": 2089484288
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003731515151515152,
+      "loss": 2.6283,
+      "theoretical_loss": 3.417939012185051,
+      "tokens_seen": 2089549824
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037313131313131316,
+      "loss": 2.7115,
+      "theoretical_loss": 3.4179301511092506,
+      "tokens_seen": 2089615360
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003731111111111111,
+      "loss": 2.4287,
+      "theoretical_loss": 3.4179212903891645,
+      "tokens_seen": 2089680896
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037309090909090907,
+      "loss": 2.5453,
+      "theoretical_loss": 3.417912430024767,
+      "tokens_seen": 2089746432
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003730707070707071,
+      "loss": 2.5528,
+      "theoretical_loss": 3.4179035700160334,
+      "tokens_seen": 2089811968
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003730505050505051,
+      "loss": 2.6557,
+      "theoretical_loss": 3.417894710362937,
+      "tokens_seen": 2089877504
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037303030303030305,
+      "loss": 2.5877,
+      "theoretical_loss": 3.4178858510654537,
+      "tokens_seen": 2089943040
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000373010101010101,
+      "loss": 2.4979,
+      "theoretical_loss": 3.4178769921235572,
+      "tokens_seen": 2090008576
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037298989898989896,
+      "loss": 2.6663,
+      "theoretical_loss": 3.4178681335372225,
+      "tokens_seen": 2090074112
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000372969696969697,
+      "loss": 2.735,
+      "theoretical_loss": 3.4178592753064243,
+      "tokens_seen": 2090139648
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000372949494949495,
+      "loss": 2.4537,
+      "theoretical_loss": 3.417850417431137,
+      "tokens_seen": 2090205184
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037292929292929294,
+      "loss": 2.5839,
+      "theoretical_loss": 3.417841559911335,
+      "tokens_seen": 2090270720
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003729090909090909,
+      "loss": 2.6842,
+      "theoretical_loss": 3.417832702746993,
+      "tokens_seen": 2090336256
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003728888888888889,
+      "loss": 2.6375,
+      "theoretical_loss": 3.4178238459380856,
+      "tokens_seen": 2090401792
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037286868686868687,
+      "loss": 2.6688,
+      "theoretical_loss": 3.4178149894845875,
+      "tokens_seen": 2090467328
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003728484848484849,
+      "loss": 2.6437,
+      "theoretical_loss": 3.417806133386473,
+      "tokens_seen": 2090532864
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037282828282828284,
+      "loss": 2.6284,
+      "theoretical_loss": 3.417797277643717,
+      "tokens_seen": 2090598400
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003728080808080808,
+      "loss": 2.5581,
+      "theoretical_loss": 3.417788422256294,
+      "tokens_seen": 2090663936
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003727878787878788,
+      "loss": 2.6807,
+      "theoretical_loss": 3.417779567224179,
+      "tokens_seen": 2090729472
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037276767676767676,
+      "loss": 2.5903,
+      "theoretical_loss": 3.417770712547346,
+      "tokens_seen": 2090795008
+    },
+    {
+      "epoch": 0.26,
+      "objective/train/docs_used": 1151517,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.077552556991577,
+      "objective/train/theoretical_loss": 3.41776185822577,
+      "objective/train/tokens_used": 449719776,
+      "theoretical_loss": 3.41776185822577,
+      "tokens_seen": 2090860544
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037274747474747477,
+      "loss": 2.8183,
+      "theoretical_loss": 3.41776185822577,
+      "tokens_seen": 2090860544
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037272727272727273,
+      "loss": 2.757,
+      "theoretical_loss": 3.417753004259425,
+      "tokens_seen": 2090926080
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037270707070707074,
+      "loss": 2.7178,
+      "theoretical_loss": 3.4177441506482866,
+      "tokens_seen": 2090991616
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003726868686868687,
+      "loss": 2.8016,
+      "theoretical_loss": 3.417735297392328,
+      "tokens_seen": 2091057152
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037266666666666665,
+      "loss": 2.603,
+      "theoretical_loss": 3.417726444491526,
+      "tokens_seen": 2091122688
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037264646464646466,
+      "loss": 2.6375,
+      "theoretical_loss": 3.417717591945853,
+      "tokens_seen": 2091188224
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003726262626262626,
+      "loss": 2.7086,
+      "theoretical_loss": 3.4177087397552848,
+      "tokens_seen": 2091253760
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037260606060606063,
+      "loss": 2.4693,
+      "theoretical_loss": 3.4176998879197957,
+      "tokens_seen": 2091319296
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003725858585858586,
+      "loss": 2.766,
+      "theoretical_loss": 3.4176910364393605,
+      "tokens_seen": 2091384832
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037256565656565655,
+      "loss": 2.6232,
+      "theoretical_loss": 3.4176821853139536,
+      "tokens_seen": 2091450368
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037254545454545456,
+      "loss": 2.5957,
+      "theoretical_loss": 3.41767333454355,
+      "tokens_seen": 2091515904
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037252525252525257,
+      "loss": 2.6096,
+      "theoretical_loss": 3.417664484128124,
+      "tokens_seen": 2091581440
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003725050505050505,
+      "loss": 2.8325,
+      "theoretical_loss": 3.4176556340676507,
+      "tokens_seen": 2091646976
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003724848484848485,
+      "loss": 2.8292,
+      "theoretical_loss": 3.4176467843621037,
+      "tokens_seen": 2091712512
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037246464646464644,
+      "loss": 2.7331,
+      "theoretical_loss": 3.417637935011459,
+      "tokens_seen": 2091778048
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037244444444444445,
+      "loss": 2.6483,
+      "theoretical_loss": 3.4176290860156904,
+      "tokens_seen": 2091843584
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037242424242424246,
+      "loss": 2.7649,
+      "theoretical_loss": 3.4176202373747726,
+      "tokens_seen": 2091909120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003724040404040404,
+      "loss": 2.4776,
+      "theoretical_loss": 3.4176113890886803,
+      "tokens_seen": 2091974656
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003723838383838384,
+      "loss": 2.5944,
+      "theoretical_loss": 3.4176025411573887,
+      "tokens_seen": 2092040192
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037236363636363633,
+      "loss": 2.5043,
+      "theoretical_loss": 3.417593693580872,
+      "tokens_seen": 2092105728
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003723434343434344,
+      "loss": 2.4376,
+      "theoretical_loss": 3.4175848463591043,
+      "tokens_seen": 2092171264
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037232323232323235,
+      "loss": 2.9385,
+      "theoretical_loss": 3.4175759994920614,
+      "tokens_seen": 2092236800
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003723030303030303,
+      "loss": 2.5103,
+      "theoretical_loss": 3.417567152979717,
+      "tokens_seen": 2092302336
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037228282828282827,
+      "loss": 2.6269,
+      "theoretical_loss": 3.4175583068220465,
+      "tokens_seen": 2092367872
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003722626262626262,
+      "loss": 2.7398,
+      "theoretical_loss": 3.417549461019024,
+      "tokens_seen": 2092433408
+    },
+    {
+      "epoch": 0.26,
+      "objective/train/docs_used": 1152178,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.0097758769989014,
+      "objective/train/theoretical_loss": 3.4175406155706245,
+      "objective/train/tokens_used": 451358176,
+      "theoretical_loss": 3.4175406155706245,
+      "tokens_seen": 2092498944
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003722424242424243,
+      "loss": 2.4834,
+      "theoretical_loss": 3.4175406155706245,
+      "tokens_seen": 2092498944
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037222222222222225,
+      "loss": 2.4869,
+      "theoretical_loss": 3.417531770476823,
+      "tokens_seen": 2092564480
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003722020202020202,
+      "loss": 2.7087,
+      "theoretical_loss": 3.4175229257375936,
+      "tokens_seen": 2092630016
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037218181818181816,
+      "loss": 2.6086,
+      "theoretical_loss": 3.417514081352911,
+      "tokens_seen": 2092695552
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037216161616161617,
+      "loss": 2.666,
+      "theoretical_loss": 3.41750523732275,
+      "tokens_seen": 2092761088
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003721414141414142,
+      "loss": 2.7373,
+      "theoretical_loss": 3.4174963936470855,
+      "tokens_seen": 2092826624
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037212121212121214,
+      "loss": 2.7108,
+      "theoretical_loss": 3.417487550325892,
+      "tokens_seen": 2092892160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003721010101010101,
+      "loss": 2.6518,
+      "theoretical_loss": 3.417478707359144,
+      "tokens_seen": 2092957696
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037208080808080805,
+      "loss": 2.5699,
+      "theoretical_loss": 3.417469864746817,
+      "tokens_seen": 2093023232
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037206060606060606,
+      "loss": 2.6461,
+      "theoretical_loss": 3.417461022488885,
+      "tokens_seen": 2093088768
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003720404040404041,
+      "loss": 2.4359,
+      "theoretical_loss": 3.417452180585322,
+      "tokens_seen": 2093154304
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037202020202020203,
+      "loss": 2.5586,
+      "theoretical_loss": 3.4174433390361045,
+      "tokens_seen": 2093219840
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000372,
+      "loss": 2.4325,
+      "theoretical_loss": 3.4174344978412057,
+      "tokens_seen": 2093285376
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000371979797979798,
+      "loss": 2.569,
+      "theoretical_loss": 3.417425657000601,
+      "tokens_seen": 2093350912
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037195959595959596,
+      "loss": 2.6863,
+      "theoretical_loss": 3.417416816514265,
+      "tokens_seen": 2093416448
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037193939393939397,
+      "loss": 2.3295,
+      "theoretical_loss": 3.4174079763821723,
+      "tokens_seen": 2093481984
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003719191919191919,
+      "loss": 2.5572,
+      "theoretical_loss": 3.4173991366042973,
+      "tokens_seen": 2093547520
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003718989898989899,
+      "loss": 2.4566,
+      "theoretical_loss": 3.417390297180616,
+      "tokens_seen": 2093613056
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003718787878787879,
+      "loss": 2.4297,
+      "theoretical_loss": 3.4173814581111017,
+      "tokens_seen": 2093678592
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037185858585858585,
+      "loss": 2.8283,
+      "theoretical_loss": 3.4173726193957297,
+      "tokens_seen": 2093744128
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037183838383838386,
+      "loss": 2.6182,
+      "theoretical_loss": 3.4173637810344744,
+      "tokens_seen": 2093809664
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003718181818181818,
+      "loss": 2.4856,
+      "theoretical_loss": 3.417354943027311,
+      "tokens_seen": 2093875200
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037179797979797983,
+      "loss": 2.6159,
+      "theoretical_loss": 3.417346105374214,
+      "tokens_seen": 2093940736
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003717777777777778,
+      "loss": 2.5668,
+      "theoretical_loss": 3.4173372680751584,
+      "tokens_seen": 2094006272
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037175757575757574,
+      "loss": 2.6343,
+      "theoretical_loss": 3.4173284311301186,
+      "tokens_seen": 2094071808
+    },
+    {
+      "epoch": 0.26,
+      "objective/train/docs_used": 1153480,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.500739574432373,
+      "objective/train/theoretical_loss": 3.417319594539069,
+      "objective/train/tokens_used": 452996576,
+      "theoretical_loss": 3.417319594539069,
+      "tokens_seen": 2094137344
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037173737373737375,
+      "loss": 2.6112,
+      "theoretical_loss": 3.417319594539069,
+      "tokens_seen": 2094137344
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003717171717171717,
+      "loss": 2.6866,
+      "theoretical_loss": 3.4173107583019853,
+      "tokens_seen": 2094202880
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003716969696969697,
+      "loss": 2.6229,
+      "theoretical_loss": 3.417301922418842,
+      "tokens_seen": 2094268416
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003716767676767677,
+      "loss": 2.606,
+      "theoretical_loss": 3.4172930868896128,
+      "tokens_seen": 2094333952
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037165656565656564,
+      "loss": 2.5647,
+      "theoretical_loss": 3.417284251714274,
+      "tokens_seen": 2094399488
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037163636363636365,
+      "loss": 2.6784,
+      "theoretical_loss": 3.4172754168927986,
+      "tokens_seen": 2094465024
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003716161616161616,
+      "loss": 2.5724,
+      "theoretical_loss": 3.417266582425163,
+      "tokens_seen": 2094530560
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003715959595959596,
+      "loss": 2.3217,
+      "theoretical_loss": 3.417257748311341,
+      "tokens_seen": 2094596096
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037157575757575757,
+      "loss": 2.4304,
+      "theoretical_loss": 3.4172489145513083,
+      "tokens_seen": 2094661632
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037155555555555553,
+      "loss": 2.2137,
+      "theoretical_loss": 3.4172400811450387,
+      "tokens_seen": 2094727168
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037153535353535354,
+      "loss": 2.6197,
+      "theoretical_loss": 3.4172312480925067,
+      "tokens_seen": 2094792704
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037151515151515155,
+      "loss": 2.603,
+      "theoretical_loss": 3.417222415393688,
+      "tokens_seen": 2094858240
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003714949494949495,
+      "loss": 2.5942,
+      "theoretical_loss": 3.4172135830485573,
+      "tokens_seen": 2094923776
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037147474747474746,
+      "loss": 2.5866,
+      "theoretical_loss": 3.4172047510570893,
+      "tokens_seen": 2094989312
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003714545454545455,
+      "loss": 2.4817,
+      "theoretical_loss": 3.417195919419258,
+      "tokens_seen": 2095054848
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037143434343434343,
+      "loss": 2.6287,
+      "theoretical_loss": 3.417187088135039,
+      "tokens_seen": 2095120384
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037141414141414144,
+      "loss": 2.3498,
+      "theoretical_loss": 3.417178257204407,
+      "tokens_seen": 2095185920
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003713939393939394,
+      "loss": 2.454,
+      "theoretical_loss": 3.4171694266273365,
+      "tokens_seen": 2095251456
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037137373737373736,
+      "loss": 2.5625,
+      "theoretical_loss": 3.4171605964038023,
+      "tokens_seen": 2095316992
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037135353535353537,
+      "loss": 2.6528,
+      "theoretical_loss": 3.417151766533779,
+      "tokens_seen": 2095382528
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003713333333333334,
+      "loss": 2.5552,
+      "theoretical_loss": 3.4171429370172426,
+      "tokens_seen": 2095448064
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037131313131313134,
+      "loss": 2.4329,
+      "theoretical_loss": 3.4171341078541664,
+      "tokens_seen": 2095513600
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003712929292929293,
+      "loss": 2.6025,
+      "theoretical_loss": 3.4171252790445257,
+      "tokens_seen": 2095579136
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037127272727272725,
+      "loss": 2.6012,
+      "theoretical_loss": 3.4171164505882956,
+      "tokens_seen": 2095644672
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00037125252525252526,
+      "loss": 2.6446,
+      "theoretical_loss": 3.4171076224854504,
+      "tokens_seen": 2095710208
+    },
+    {
+      "epoch": 0.26,
+      "objective/train/docs_used": 1154065,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.8171300888061523,
+      "objective/train/theoretical_loss": 3.4170987947359652,
+      "objective/train/tokens_used": 454634976,
+      "theoretical_loss": 3.4170987947359652,
+      "tokens_seen": 2095775744
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037123232323232327,
+      "loss": 2.7618,
+      "theoretical_loss": 3.4170987947359652,
+      "tokens_seen": 2095775744
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037121212121212123,
+      "loss": 2.3248,
+      "theoretical_loss": 3.4170899673398147,
+      "tokens_seen": 2095841280
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003711919191919192,
+      "loss": 2.4875,
+      "theoretical_loss": 3.4170811402969745,
+      "tokens_seen": 2095906816
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037117171717171714,
+      "loss": 2.416,
+      "theoretical_loss": 3.417072313607418,
+      "tokens_seen": 2095972352
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003711515151515152,
+      "loss": 2.5119,
+      "theoretical_loss": 3.417063487271121,
+      "tokens_seen": 2096037888
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037113131313131316,
+      "loss": 2.7506,
+      "theoretical_loss": 3.4170546612880575,
+      "tokens_seen": 2096103424
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003711111111111111,
+      "loss": 2.6149,
+      "theoretical_loss": 3.417045835658204,
+      "tokens_seen": 2096168960
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003710909090909091,
+      "loss": 2.7236,
+      "theoretical_loss": 3.4170370103815335,
+      "tokens_seen": 2096234496
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037107070707070703,
+      "loss": 2.5956,
+      "theoretical_loss": 3.417028185458021,
+      "tokens_seen": 2096300032
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003710505050505051,
+      "loss": 2.506,
+      "theoretical_loss": 3.4170193608876427,
+      "tokens_seen": 2096365568
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037103030303030306,
+      "loss": 2.578,
+      "theoretical_loss": 3.4170105366703725,
+      "tokens_seen": 2096431104
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000371010101010101,
+      "loss": 2.5963,
+      "theoretical_loss": 3.4170017128061847,
+      "tokens_seen": 2096496640
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037098989898989897,
+      "loss": 2.7297,
+      "theoretical_loss": 3.416992889295055,
+      "tokens_seen": 2096562176
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000370969696969697,
+      "loss": 2.8348,
+      "theoretical_loss": 3.4169840661369584,
+      "tokens_seen": 2096627712
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000370949494949495,
+      "loss": 2.5569,
+      "theoretical_loss": 3.4169752433318688,
+      "tokens_seen": 2096693248
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037092929292929295,
+      "loss": 2.6674,
+      "theoretical_loss": 3.4169664208797617,
+      "tokens_seen": 2096758784
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003709090909090909,
+      "loss": 2.9471,
+      "theoretical_loss": 3.416957598780612,
+      "tokens_seen": 2096824320
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037088888888888886,
+      "loss": 2.4803,
+      "theoretical_loss": 3.416948777034394,
+      "tokens_seen": 2096889856
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003708686868686869,
+      "loss": 2.9403,
+      "theoretical_loss": 3.416939955641083,
+      "tokens_seen": 2096955392
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003708484848484849,
+      "loss": 2.6036,
+      "theoretical_loss": 3.416931134600654,
+      "tokens_seen": 2097020928
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037082828282828284,
+      "loss": 2.6213,
+      "theoretical_loss": 3.4169223139130813,
+      "tokens_seen": 2097086464
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003708080808080808,
+      "loss": 2.5021,
+      "theoretical_loss": 3.41691349357834,
+      "tokens_seen": 2097152000
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003707878787878788,
+      "loss": 2.5866,
+      "theoretical_loss": 3.416904673596406,
+      "tokens_seen": 2097217536
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037076767676767677,
+      "loss": 2.7012,
+      "theoretical_loss": 3.416895853967252,
+      "tokens_seen": 2097283072
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003707474747474748,
+      "loss": 2.6695,
+      "theoretical_loss": 3.4168870346908546,
+      "tokens_seen": 2097348608
+    },
+    {
+      "epoch": 0.27,
+      "objective/train/docs_used": 1155055,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.3634607791900635,
+      "objective/train/theoretical_loss": 3.416878215767188,
+      "objective/train/tokens_used": 456273376,
+      "theoretical_loss": 3.416878215767188,
+      "tokens_seen": 2097414144
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037072727272727273,
+      "loss": 2.6518,
+      "theoretical_loss": 3.416878215767188,
+      "tokens_seen": 2097414144
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003707070707070707,
+      "loss": 2.7283,
+      "theoretical_loss": 3.416869397196227,
+      "tokens_seen": 2097479680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003706868686868687,
+      "loss": 2.6649,
+      "theoretical_loss": 3.416860578977947,
+      "tokens_seen": 2097545216
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037066666666666666,
+      "loss": 2.6906,
+      "theoretical_loss": 3.416851761112323,
+      "tokens_seen": 2097610752
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037064646464646467,
+      "loss": 2.6143,
+      "theoretical_loss": 3.416842943599329,
+      "tokens_seen": 2097676288
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037062626262626263,
+      "loss": 2.3629,
+      "theoretical_loss": 3.41683412643894,
+      "tokens_seen": 2097741824
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037060606060606064,
+      "loss": 2.8845,
+      "theoretical_loss": 3.4168253096311316,
+      "tokens_seen": 2097807360
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003705858585858586,
+      "loss": 2.7976,
+      "theoretical_loss": 3.4168164931758778,
+      "tokens_seen": 2097872896
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037056565656565655,
+      "loss": 2.7928,
+      "theoretical_loss": 3.4168076770731544,
+      "tokens_seen": 2097938432
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037054545454545456,
+      "loss": 2.6009,
+      "theoretical_loss": 3.416798861322936,
+      "tokens_seen": 2098003968
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003705252525252525,
+      "loss": 2.7349,
+      "theoretical_loss": 3.416790045925197,
+      "tokens_seen": 2098069504
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037050505050505053,
+      "loss": 2.6537,
+      "theoretical_loss": 3.416781230879913,
+      "tokens_seen": 2098135040
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003704848484848485,
+      "loss": 2.5315,
+      "theoretical_loss": 3.416772416187058,
+      "tokens_seen": 2098200576
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037046464646464644,
+      "loss": 2.3986,
+      "theoretical_loss": 3.416763601846608,
+      "tokens_seen": 2098266112
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037044444444444446,
+      "loss": 2.4969,
+      "theoretical_loss": 3.4167547878585376,
+      "tokens_seen": 2098331648
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037042424242424247,
+      "loss": 2.6538,
+      "theoretical_loss": 3.416745974222821,
+      "tokens_seen": 2098397184
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003704040404040404,
+      "loss": 2.5273,
+      "theoretical_loss": 3.4167371609394337,
+      "tokens_seen": 2098462720
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003703838383838384,
+      "loss": 2.7185,
+      "theoretical_loss": 3.41672834800835,
+      "tokens_seen": 2098528256
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037036363636363634,
+      "loss": 2.7387,
+      "theoretical_loss": 3.416719535429546,
+      "tokens_seen": 2098593792
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037034343434343435,
+      "loss": 2.6902,
+      "theoretical_loss": 3.416710723202996,
+      "tokens_seen": 2098659328
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037032323232323236,
+      "loss": 2.6166,
+      "theoretical_loss": 3.4167019113286745,
+      "tokens_seen": 2098724864
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003703030303030303,
+      "loss": 2.2139,
+      "theoretical_loss": 3.4166930998065568,
+      "tokens_seen": 2098790400
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003702828282828283,
+      "loss": 2.6439,
+      "theoretical_loss": 3.4166842886366178,
+      "tokens_seen": 2098855936
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037026262626262623,
+      "loss": 2.8034,
+      "theoretical_loss": 3.416675477818832,
+      "tokens_seen": 2098921472
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003702424242424243,
+      "loss": 2.5019,
+      "theoretical_loss": 3.4166666673531756,
+      "tokens_seen": 2098987008
+    },
+    {
+      "epoch": 0.27,
+      "objective/train/docs_used": 1156556,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7254951000213623,
+      "objective/train/theoretical_loss": 3.4166578572396222,
+      "objective/train/tokens_used": 457911776,
+      "theoretical_loss": 3.4166578572396222,
+      "tokens_seen": 2099052544
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037022222222222225,
+      "loss": 2.4085,
+      "theoretical_loss": 3.4166578572396222,
+      "tokens_seen": 2099052544
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003702020202020202,
+      "loss": 2.739,
+      "theoretical_loss": 3.4166490474781477,
+      "tokens_seen": 2099118080
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037018181818181817,
+      "loss": 2.5515,
+      "theoretical_loss": 3.416640238068726,
+      "tokens_seen": 2099183616
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003701616161616161,
+      "loss": 2.6694,
+      "theoretical_loss": 3.416631429011333,
+      "tokens_seen": 2099249152
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003701414141414142,
+      "loss": 2.6967,
+      "theoretical_loss": 3.416622620305943,
+      "tokens_seen": 2099314688
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037012121212121215,
+      "loss": 2.5425,
+      "theoretical_loss": 3.4166138119525313,
+      "tokens_seen": 2099380224
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003701010101010101,
+      "loss": 2.8095,
+      "theoretical_loss": 3.416605003951073,
+      "tokens_seen": 2099445760
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037008080808080806,
+      "loss": 2.5167,
+      "theoretical_loss": 3.4165961963015423,
+      "tokens_seen": 2099511296
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000370060606060606,
+      "loss": 2.6135,
+      "theoretical_loss": 3.416587389003915,
+      "tokens_seen": 2099576832
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003700404040404041,
+      "loss": 2.635,
+      "theoretical_loss": 3.4165785820581656,
+      "tokens_seen": 2099642368
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037002020202020204,
+      "loss": 2.6785,
+      "theoretical_loss": 3.416569775464269,
+      "tokens_seen": 2099707904
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00037,
+      "loss": 2.7011,
+      "theoretical_loss": 3.4165609692222008,
+      "tokens_seen": 2099773440
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036997979797979795,
+      "loss": 2.7845,
+      "theoretical_loss": 3.416552163331935,
+      "tokens_seen": 2099838976
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036995959595959596,
+      "loss": 2.5003,
+      "theoretical_loss": 3.416543357793447,
+      "tokens_seen": 2099904512
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000369939393939394,
+      "loss": 2.5218,
+      "theoretical_loss": 3.4165345526067123,
+      "tokens_seen": 2099970048
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036991919191919193,
+      "loss": 2.5563,
+      "theoretical_loss": 3.416525747771705,
+      "tokens_seen": 2100035584
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003698989898989899,
+      "loss": 2.7174,
+      "theoretical_loss": 3.416516943288401,
+      "tokens_seen": 2100101120
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036987878787878784,
+      "loss": 2.7711,
+      "theoretical_loss": 3.4165081391567744,
+      "tokens_seen": 2100166656
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036985858585858586,
+      "loss": 2.6654,
+      "theoretical_loss": 3.4164993353768005,
+      "tokens_seen": 2100232192
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036983838383838387,
+      "loss": 2.6669,
+      "theoretical_loss": 3.4164905319484546,
+      "tokens_seen": 2100297728
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003698181818181818,
+      "loss": 2.6784,
+      "theoretical_loss": 3.4164817288717106,
+      "tokens_seen": 2100363264
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003697979797979798,
+      "loss": 2.4752,
+      "theoretical_loss": 3.416472926146545,
+      "tokens_seen": 2100428800
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003697777777777778,
+      "loss": 2.4664,
+      "theoretical_loss": 3.416464123772932,
+      "tokens_seen": 2100494336
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003697575757575758,
+      "loss": 2.6676,
+      "theoretical_loss": 3.4164553217508464,
+      "tokens_seen": 2100559872
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036973737373737376,
+      "loss": 2.5018,
+      "theoretical_loss": 3.416446520080264,
+      "tokens_seen": 2100625408
+    },
+    {
+      "epoch": 0.27,
+      "objective/train/docs_used": 1157195,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.8589794635772705,
+      "objective/train/theoretical_loss": 3.4164377187611588,
+      "objective/train/tokens_used": 459550176,
+      "theoretical_loss": 3.4164377187611588,
+      "tokens_seen": 2100690944
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003697171717171717,
+      "loss": 2.7354,
+      "theoretical_loss": 3.4164377187611588,
+      "tokens_seen": 2100690944
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036969696969696967,
+      "loss": 2.545,
+      "theoretical_loss": 3.4164289177935063,
+      "tokens_seen": 2100756480
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003696767676767677,
+      "loss": 2.6809,
+      "theoretical_loss": 3.4164201171772817,
+      "tokens_seen": 2100822016
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003696565656565657,
+      "loss": 2.7996,
+      "theoretical_loss": 3.4164113169124595,
+      "tokens_seen": 2100887552
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036963636363636365,
+      "loss": 2.4825,
+      "theoretical_loss": 3.416402516999015,
+      "tokens_seen": 2100953088
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003696161616161616,
+      "loss": 2.5685,
+      "theoretical_loss": 3.4163937174369234,
+      "tokens_seen": 2101018624
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003695959595959596,
+      "loss": 2.7909,
+      "theoretical_loss": 3.416384918226159,
+      "tokens_seen": 2101084160
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003695757575757576,
+      "loss": 2.41,
+      "theoretical_loss": 3.4163761193666975,
+      "tokens_seen": 2101149696
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003695555555555556,
+      "loss": 2.5918,
+      "theoretical_loss": 3.416367320858514,
+      "tokens_seen": 2101215232
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036953535353535354,
+      "loss": 2.5184,
+      "theoretical_loss": 3.416358522701583,
+      "tokens_seen": 2101280768
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003695151515151515,
+      "loss": 2.5719,
+      "theoretical_loss": 3.4163497248958805,
+      "tokens_seen": 2101346304
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003694949494949495,
+      "loss": 2.7546,
+      "theoretical_loss": 3.4163409274413796,
+      "tokens_seen": 2101411840
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036947474747474747,
+      "loss": 2.5471,
+      "theoretical_loss": 3.4163321303380574,
+      "tokens_seen": 2101477376
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003694545454545455,
+      "loss": 2.6367,
+      "theoretical_loss": 3.416323333585888,
+      "tokens_seen": 2101542912
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036943434343434344,
+      "loss": 2.7109,
+      "theoretical_loss": 3.416314537184846,
+      "tokens_seen": 2101608448
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036941414141414145,
+      "loss": 2.8343,
+      "theoretical_loss": 3.416305741134907,
+      "tokens_seen": 2101673984
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003693939393939394,
+      "loss": 2.5901,
+      "theoretical_loss": 3.4162969454360463,
+      "tokens_seen": 2101739520
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036937373737373736,
+      "loss": 2.6116,
+      "theoretical_loss": 3.4162881500882385,
+      "tokens_seen": 2101805056
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003693535353535354,
+      "loss": 2.5429,
+      "theoretical_loss": 3.4162793550914587,
+      "tokens_seen": 2101870592
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036933333333333333,
+      "loss": 2.7821,
+      "theoretical_loss": 3.416270560445682,
+      "tokens_seen": 2101936128
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036931313131313134,
+      "loss": 2.7114,
+      "theoretical_loss": 3.4162617661508836,
+      "tokens_seen": 2102001664
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003692929292929293,
+      "loss": 2.6573,
+      "theoretical_loss": 3.4162529722070385,
+      "tokens_seen": 2102067200
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036927272727272725,
+      "loss": 2.5542,
+      "theoretical_loss": 3.416244178614121,
+      "tokens_seen": 2102132736
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036925252525252527,
+      "loss": 2.4011,
+      "theoretical_loss": 3.416235385372107,
+      "tokens_seen": 2102198272
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003692323232323233,
+      "loss": 2.4574,
+      "theoretical_loss": 3.416226592480972,
+      "tokens_seen": 2102263808
+    },
+    {
+      "epoch": 0.27,
+      "objective/train/docs_used": 1158634,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.4343600273132324,
+      "objective/train/theoretical_loss": 3.41621779994069,
+      "objective/train/tokens_used": 461188576,
+      "theoretical_loss": 3.41621779994069,
+      "tokens_seen": 2102329344
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036921212121212123,
+      "loss": 2.7046,
+      "theoretical_loss": 3.41621779994069,
+      "tokens_seen": 2102329344
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003691919191919192,
+      "loss": 2.8016,
+      "theoretical_loss": 3.4162090077512364,
+      "tokens_seen": 2102394880
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036917171717171715,
+      "loss": 2.6305,
+      "theoretical_loss": 3.4162002159125864,
+      "tokens_seen": 2102460416
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036915151515151516,
+      "loss": 2.6215,
+      "theoretical_loss": 3.416191424424715,
+      "tokens_seen": 2102525952
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036913131313131317,
+      "loss": 2.7357,
+      "theoretical_loss": 3.4161826332875975,
+      "tokens_seen": 2102591488
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003691111111111111,
+      "loss": 2.4573,
+      "theoretical_loss": 3.4161738425012085,
+      "tokens_seen": 2102657024
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003690909090909091,
+      "loss": 2.7079,
+      "theoretical_loss": 3.4161650520655233,
+      "tokens_seen": 2102722560
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036907070707070704,
+      "loss": 2.522,
+      "theoretical_loss": 3.4161562619805173,
+      "tokens_seen": 2102788096
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003690505050505051,
+      "loss": 2.8353,
+      "theoretical_loss": 3.4161474722461653,
+      "tokens_seen": 2102853632
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036903030303030306,
+      "loss": 2.5757,
+      "theoretical_loss": 3.4161386828624423,
+      "tokens_seen": 2102919168
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000369010101010101,
+      "loss": 2.6205,
+      "theoretical_loss": 3.416129893829323,
+      "tokens_seen": 2102984704
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000368989898989899,
+      "loss": 2.631,
+      "theoretical_loss": 3.4161211051467837,
+      "tokens_seen": 2103050240
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036896969696969693,
+      "loss": 2.4927,
+      "theoretical_loss": 3.4161123168147984,
+      "tokens_seen": 2103115776
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000368949494949495,
+      "loss": 2.7321,
+      "theoretical_loss": 3.4161035288333426,
+      "tokens_seen": 2103181312
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036892929292929295,
+      "loss": 2.5872,
+      "theoretical_loss": 3.416094741202391,
+      "tokens_seen": 2103246848
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003689090909090909,
+      "loss": 2.9086,
+      "theoretical_loss": 3.41608595392192,
+      "tokens_seen": 2103312384
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036888888888888887,
+      "loss": 2.5979,
+      "theoretical_loss": 3.416077166991903,
+      "tokens_seen": 2103377920
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003688686868686869,
+      "loss": 2.522,
+      "theoretical_loss": 3.416068380412316,
+      "tokens_seen": 2103443456
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003688484848484849,
+      "loss": 2.62,
+      "theoretical_loss": 3.416059594183134,
+      "tokens_seen": 2103508992
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036882828282828285,
+      "loss": 2.7018,
+      "theoretical_loss": 3.416050808304332,
+      "tokens_seen": 2103574528
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003688080808080808,
+      "loss": 2.4874,
+      "theoretical_loss": 3.416042022775885,
+      "tokens_seen": 2103640064
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036878787878787876,
+      "loss": 2.7253,
+      "theoretical_loss": 3.4160332375977687,
+      "tokens_seen": 2103705600
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036876767676767677,
+      "loss": 2.6079,
+      "theoretical_loss": 3.4160244527699577,
+      "tokens_seen": 2103771136
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003687474747474748,
+      "loss": 2.808,
+      "theoretical_loss": 3.4160156682924274,
+      "tokens_seen": 2103836672
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036872727272727274,
+      "loss": 2.5926,
+      "theoretical_loss": 3.416006884165152,
+      "tokens_seen": 2103902208
+    },
+    {
+      "epoch": 0.27,
+      "objective/train/docs_used": 1159301,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5771374702453613,
+      "objective/train/theoretical_loss": 3.4159981003881086,
+      "objective/train/tokens_used": 462826976,
+      "theoretical_loss": 3.4159981003881086,
+      "tokens_seen": 2103967744
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003687070707070707,
+      "loss": 2.724,
+      "theoretical_loss": 3.4159981003881086,
+      "tokens_seen": 2103967744
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003686868686868687,
+      "loss": 2.6223,
+      "theoretical_loss": 3.4159893169612703,
+      "tokens_seen": 2104033280
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036866666666666667,
+      "loss": 2.5754,
+      "theoretical_loss": 3.415980533884613,
+      "tokens_seen": 2104098816
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003686464646464647,
+      "loss": 2.6732,
+      "theoretical_loss": 3.4159717511581116,
+      "tokens_seen": 2104164352
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036862626262626263,
+      "loss": 2.7595,
+      "theoretical_loss": 3.4159629687817423,
+      "tokens_seen": 2104229888
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003686060606060606,
+      "loss": 2.4892,
+      "theoretical_loss": 3.415954186755479,
+      "tokens_seen": 2104295424
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003685858585858586,
+      "loss": 2.6573,
+      "theoretical_loss": 3.415945405079297,
+      "tokens_seen": 2104360960
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036856565656565656,
+      "loss": 2.7806,
+      "theoretical_loss": 3.4159366237531725,
+      "tokens_seen": 2104426496
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036854545454545457,
+      "loss": 2.5188,
+      "theoretical_loss": 3.415927842777079,
+      "tokens_seen": 2104492032
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003685252525252525,
+      "loss": 2.4961,
+      "theoretical_loss": 3.4159190621509934,
+      "tokens_seen": 2104557568
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036850505050505054,
+      "loss": 2.5838,
+      "theoretical_loss": 3.4159102818748894,
+      "tokens_seen": 2104623104
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003684848484848485,
+      "loss": 2.5448,
+      "theoretical_loss": 3.415901501948743,
+      "tokens_seen": 2104688640
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036846464646464645,
+      "loss": 2.6647,
+      "theoretical_loss": 3.415892722372529,
+      "tokens_seen": 2104754176
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036844444444444446,
+      "loss": 2.6084,
+      "theoretical_loss": 3.4158839431462225,
+      "tokens_seen": 2104819712
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003684242424242424,
+      "loss": 2.6192,
+      "theoretical_loss": 3.415875164269799,
+      "tokens_seen": 2104885248
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036840404040404043,
+      "loss": 2.6547,
+      "theoretical_loss": 3.415866385743233,
+      "tokens_seen": 2104950784
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003683838383838384,
+      "loss": 2.4461,
+      "theoretical_loss": 3.4158576075665006,
+      "tokens_seen": 2105016320
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036836363636363634,
+      "loss": 2.4437,
+      "theoretical_loss": 3.415848829739576,
+      "tokens_seen": 2105081856
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036834343434343435,
+      "loss": 2.6591,
+      "theoretical_loss": 3.4158400522624355,
+      "tokens_seen": 2105147392
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003683232323232323,
+      "loss": 2.7838,
+      "theoretical_loss": 3.4158312751350532,
+      "tokens_seen": 2105212928
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003683030303030303,
+      "loss": 2.541,
+      "theoretical_loss": 3.415822498357405,
+      "tokens_seen": 2105278464
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003682828282828283,
+      "loss": 2.6542,
+      "theoretical_loss": 3.4158137219294655,
+      "tokens_seen": 2105344000
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036826262626262624,
+      "loss": 2.4815,
+      "theoretical_loss": 3.4158049458512103,
+      "tokens_seen": 2105409536
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036824242424242425,
+      "loss": 2.7231,
+      "theoretical_loss": 3.415796170122615,
+      "tokens_seen": 2105475072
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036822222222222226,
+      "loss": 2.5703,
+      "theoretical_loss": 3.4157873947436537,
+      "tokens_seen": 2105540608
+    },
+    {
+      "epoch": 0.27,
+      "objective/train/docs_used": 1160284,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.0382537841796875,
+      "objective/train/theoretical_loss": 3.415778619714302,
+      "objective/train/tokens_used": 464465376,
+      "theoretical_loss": 3.415778619714302,
+      "tokens_seen": 2105606144
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003682020202020202,
+      "loss": 2.6972,
+      "theoretical_loss": 3.415778619714302,
+      "tokens_seen": 2105606144
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036818181818181817,
+      "loss": 2.4719,
+      "theoretical_loss": 3.4157698450345353,
+      "tokens_seen": 2105671680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036816161616161613,
+      "loss": 2.6513,
+      "theoretical_loss": 3.4157610707043284,
+      "tokens_seen": 2105737216
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036814141414141414,
+      "loss": 2.6362,
+      "theoretical_loss": 3.4157522967236575,
+      "tokens_seen": 2105802752
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036812121212121215,
+      "loss": 2.4002,
+      "theoretical_loss": 3.415743523092497,
+      "tokens_seen": 2105868288
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003681010101010101,
+      "loss": 2.7257,
+      "theoretical_loss": 3.415734749810822,
+      "tokens_seen": 2105933824
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036808080808080806,
+      "loss": 2.4798,
+      "theoretical_loss": 3.4157259768786075,
+      "tokens_seen": 2105999360
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003680606060606061,
+      "loss": 2.3792,
+      "theoretical_loss": 3.41571720429583,
+      "tokens_seen": 2106064896
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003680404040404041,
+      "loss": 2.7184,
+      "theoretical_loss": 3.415708432062463,
+      "tokens_seen": 2106130432
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036802020202020204,
+      "loss": 2.6301,
+      "theoretical_loss": 3.415699660178483,
+      "tokens_seen": 2106195968
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000368,
+      "loss": 2.629,
+      "theoretical_loss": 3.415690888643865,
+      "tokens_seen": 2106261504
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036797979797979796,
+      "loss": 2.8607,
+      "theoretical_loss": 3.415682117458583,
+      "tokens_seen": 2106327040
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036795959595959597,
+      "loss": 2.4124,
+      "theoretical_loss": 3.415673346622614,
+      "tokens_seen": 2106392576
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000367939393939394,
+      "loss": 2.5606,
+      "theoretical_loss": 3.4156645761359323,
+      "tokens_seen": 2106458112
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036791919191919194,
+      "loss": 2.471,
+      "theoretical_loss": 3.415655805998513,
+      "tokens_seen": 2106523648
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003678989898989899,
+      "loss": 2.4754,
+      "theoretical_loss": 3.4156470362103315,
+      "tokens_seen": 2106589184
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036787878787878785,
+      "loss": 2.6735,
+      "theoretical_loss": 3.415638266771363,
+      "tokens_seen": 2106654720
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003678585858585859,
+      "loss": 2.4972,
+      "theoretical_loss": 3.415629497681583,
+      "tokens_seen": 2106720256
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036783838383838387,
+      "loss": 2.6408,
+      "theoretical_loss": 3.415620728940967,
+      "tokens_seen": 2106785792
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036781818181818183,
+      "loss": 2.4854,
+      "theoretical_loss": 3.4156119605494895,
+      "tokens_seen": 2106851328
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003677979797979798,
+      "loss": 2.6254,
+      "theoretical_loss": 3.4156031925071257,
+      "tokens_seen": 2106916864
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036777777777777774,
+      "loss": 2.3775,
+      "theoretical_loss": 3.4155944248138512,
+      "tokens_seen": 2106982400
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003677575757575758,
+      "loss": 2.688,
+      "theoretical_loss": 3.4155856574696415,
+      "tokens_seen": 2107047936
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036773737373737376,
+      "loss": 2.6889,
+      "theoretical_loss": 3.4155768904744717,
+      "tokens_seen": 2107113472
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003677171717171717,
+      "loss": 2.2961,
+      "theoretical_loss": 3.415568123828317,
+      "tokens_seen": 2107179008
+    },
+    {
+      "epoch": 0.27,
+      "objective/train/docs_used": 1161731,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.1398634910583496,
+      "objective/train/theoretical_loss": 3.415559357531152,
+      "objective/train/tokens_used": 466103776,
+      "theoretical_loss": 3.415559357531152,
+      "tokens_seen": 2107244544
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003676969696969697,
+      "loss": 2.433,
+      "theoretical_loss": 3.415559357531152,
+      "tokens_seen": 2107244544
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003676767676767677,
+      "loss": 2.4995,
+      "theoretical_loss": 3.4155505915829525,
+      "tokens_seen": 2107310080
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003676565656565657,
+      "loss": 2.6416,
+      "theoretical_loss": 3.415541825983694,
+      "tokens_seen": 2107375616
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036763636363636366,
+      "loss": 2.4995,
+      "theoretical_loss": 3.4155330607333516,
+      "tokens_seen": 2107441152
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003676161616161616,
+      "loss": 2.7959,
+      "theoretical_loss": 3.4155242958319003,
+      "tokens_seen": 2107506688
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036759595959595957,
+      "loss": 2.4846,
+      "theoretical_loss": 3.4155155312793157,
+      "tokens_seen": 2107572224
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003675757575757576,
+      "loss": 2.7947,
+      "theoretical_loss": 3.415506767075573,
+      "tokens_seen": 2107637760
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003675555555555556,
+      "loss": 2.6484,
+      "theoretical_loss": 3.4154980032206472,
+      "tokens_seen": 2107703296
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036753535353535355,
+      "loss": 2.5693,
+      "theoretical_loss": 3.415489239714514,
+      "tokens_seen": 2107768832
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003675151515151515,
+      "loss": 2.754,
+      "theoretical_loss": 3.415480476557148,
+      "tokens_seen": 2107834368
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003674949494949495,
+      "loss": 2.6904,
+      "theoretical_loss": 3.4154717137485253,
+      "tokens_seen": 2107899904
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003674747474747475,
+      "loss": 2.6066,
+      "theoretical_loss": 3.4154629512886205,
+      "tokens_seen": 2107965440
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003674545454545455,
+      "loss": 2.6974,
+      "theoretical_loss": 3.4154541891774093,
+      "tokens_seen": 2108030976
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036743434343434344,
+      "loss": 2.5926,
+      "theoretical_loss": 3.415445427414867,
+      "tokens_seen": 2108096512
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003674141414141414,
+      "loss": 2.6391,
+      "theoretical_loss": 3.4154366660009687,
+      "tokens_seen": 2108162048
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003673939393939394,
+      "loss": 2.5497,
+      "theoretical_loss": 3.4154279049356897,
+      "tokens_seen": 2108227584
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036737373737373737,
+      "loss": 2.6752,
+      "theoretical_loss": 3.4154191442190056,
+      "tokens_seen": 2108293120
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003673535353535354,
+      "loss": 2.644,
+      "theoretical_loss": 3.4154103838508907,
+      "tokens_seen": 2108358656
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036733333333333334,
+      "loss": 2.6448,
+      "theoretical_loss": 3.415401623831322,
+      "tokens_seen": 2108424192
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036731313131313135,
+      "loss": 2.5945,
+      "theoretical_loss": 3.415392864160273,
+      "tokens_seen": 2108489728
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003672929292929293,
+      "loss": 2.6826,
+      "theoretical_loss": 3.41538410483772,
+      "tokens_seen": 2108555264
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036727272727272726,
+      "loss": 2.9419,
+      "theoretical_loss": 3.4153753458636382,
+      "tokens_seen": 2108620800
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036725252525252527,
+      "loss": 2.4951,
+      "theoretical_loss": 3.415366587238003,
+      "tokens_seen": 2108686336
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036723232323232323,
+      "loss": 2.6084,
+      "theoretical_loss": 3.415357828960789,
+      "tokens_seen": 2108751872
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036721212121212124,
+      "loss": 2.5701,
+      "theoretical_loss": 3.4153490710319723,
+      "tokens_seen": 2108817408
+    },
+    {
+      "epoch": 0.27,
+      "objective/train/docs_used": 1162303,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.828242063522339,
+      "objective/train/theoretical_loss": 3.4153403134515283,
+      "objective/train/tokens_used": 467742176,
+      "theoretical_loss": 3.4153403134515283,
+      "tokens_seen": 2108882944
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003671919191919192,
+      "loss": 2.588,
+      "theoretical_loss": 3.4153403134515283,
+      "tokens_seen": 2108882944
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036717171717171715,
+      "loss": 2.5231,
+      "theoretical_loss": 3.4153315562194315,
+      "tokens_seen": 2108948480
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036715151515151516,
+      "loss": 2.8862,
+      "theoretical_loss": 3.415322799335658,
+      "tokens_seen": 2109014016
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003671313131313132,
+      "loss": 2.7052,
+      "theoretical_loss": 3.4153140428001825,
+      "tokens_seen": 2109079552
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036711111111111113,
+      "loss": 2.4027,
+      "theoretical_loss": 3.4153052866129814,
+      "tokens_seen": 2109145088
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003670909090909091,
+      "loss": 2.4914,
+      "theoretical_loss": 3.4152965307740284,
+      "tokens_seen": 2109210624
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036707070707070705,
+      "loss": 2.8417,
+      "theoretical_loss": 3.4152877752833,
+      "tokens_seen": 2109276160
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036705050505050506,
+      "loss": 2.6103,
+      "theoretical_loss": 3.4152790201407717,
+      "tokens_seen": 2109341696
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036703030303030307,
+      "loss": 2.529,
+      "theoretical_loss": 3.4152702653464178,
+      "tokens_seen": 2109407232
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000367010101010101,
+      "loss": 2.6543,
+      "theoretical_loss": 3.4152615109002142,
+      "tokens_seen": 2109472768
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000366989898989899,
+      "loss": 2.958,
+      "theoretical_loss": 3.4152527568021362,
+      "tokens_seen": 2109538304
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036696969696969694,
+      "loss": 2.6509,
+      "theoretical_loss": 3.41524400305216,
+      "tokens_seen": 2109603840
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000366949494949495,
+      "loss": 2.5489,
+      "theoretical_loss": 3.4152352496502596,
+      "tokens_seen": 2109669376
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036692929292929296,
+      "loss": 2.6989,
+      "theoretical_loss": 3.4152264965964108,
+      "tokens_seen": 2109734912
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003669090909090909,
+      "loss": 2.483,
+      "theoretical_loss": 3.4152177438905893,
+      "tokens_seen": 2109800448
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003668888888888889,
+      "loss": 2.5866,
+      "theoretical_loss": 3.4152089915327695,
+      "tokens_seen": 2109865984
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036686868686868683,
+      "loss": 2.5572,
+      "theoretical_loss": 3.415200239522928,
+      "tokens_seen": 2109931520
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003668484848484849,
+      "loss": 2.5575,
+      "theoretical_loss": 3.4151914878610397,
+      "tokens_seen": 2109997056
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036682828282828285,
+      "loss": 2.4725,
+      "theoretical_loss": 3.4151827365470795,
+      "tokens_seen": 2110062592
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003668080808080808,
+      "loss": 2.7051,
+      "theoretical_loss": 3.4151739855810233,
+      "tokens_seen": 2110128128
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036678787878787877,
+      "loss": 2.5453,
+      "theoretical_loss": 3.4151652349628465,
+      "tokens_seen": 2110193664
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003667676767676767,
+      "loss": 2.5881,
+      "theoretical_loss": 3.415156484692524,
+      "tokens_seen": 2110259200
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003667474747474748,
+      "loss": 2.6439,
+      "theoretical_loss": 3.415147734770031,
+      "tokens_seen": 2110324736
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036672727272727275,
+      "loss": 2.6104,
+      "theoretical_loss": 3.4151389851953438,
+      "tokens_seen": 2110390272
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003667070707070707,
+      "loss": 2.6512,
+      "theoretical_loss": 3.4151302359684372,
+      "tokens_seen": 2110455808
+    },
+    {
+      "epoch": 0.27,
+      "objective/train/docs_used": 1163684,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.875441312789917,
+      "objective/train/theoretical_loss": 3.4151214870892868,
+      "objective/train/tokens_used": 469380576,
+      "theoretical_loss": 3.4151214870892868,
+      "tokens_seen": 2110521344
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036668686868686866,
+      "loss": 2.6088,
+      "theoretical_loss": 3.4151214870892868,
+      "tokens_seen": 2110521344
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036666666666666667,
+      "loss": 2.6473,
+      "theoretical_loss": 3.4151127385578675,
+      "tokens_seen": 2110586880
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003666464646464647,
+      "loss": 2.6037,
+      "theoretical_loss": 3.4151039903741554,
+      "tokens_seen": 2110652416
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036662626262626264,
+      "loss": 2.585,
+      "theoretical_loss": 3.415095242538125,
+      "tokens_seen": 2110717952
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003666060606060606,
+      "loss": 2.4103,
+      "theoretical_loss": 3.4150864950497524,
+      "tokens_seen": 2110783488
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036658585858585855,
+      "loss": 2.4533,
+      "theoretical_loss": 3.415077747909013,
+      "tokens_seen": 2110849024
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036656565656565656,
+      "loss": 2.5017,
+      "theoretical_loss": 3.4150690011158815,
+      "tokens_seen": 2110914560
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003665454545454546,
+      "loss": 2.5898,
+      "theoretical_loss": 3.415060254670334,
+      "tokens_seen": 2110980096
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036652525252525253,
+      "loss": 2.7279,
+      "theoretical_loss": 3.415051508572346,
+      "tokens_seen": 2111045632
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003665050505050505,
+      "loss": 2.6679,
+      "theoretical_loss": 3.415042762821892,
+      "tokens_seen": 2111111168
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003664848484848485,
+      "loss": 2.4494,
+      "theoretical_loss": 3.4150340174189484,
+      "tokens_seen": 2111176704
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036646464646464646,
+      "loss": 2.3438,
+      "theoretical_loss": 3.41502527236349,
+      "tokens_seen": 2111242240
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036644444444444447,
+      "loss": 2.6134,
+      "theoretical_loss": 3.4150165276554922,
+      "tokens_seen": 2111307776
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003664242424242424,
+      "loss": 2.7989,
+      "theoretical_loss": 3.4150077832949304,
+      "tokens_seen": 2111373312
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003664040404040404,
+      "loss": 2.7628,
+      "theoretical_loss": 3.4149990392817804,
+      "tokens_seen": 2111438848
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003663838383838384,
+      "loss": 2.7186,
+      "theoretical_loss": 3.4149902956160174,
+      "tokens_seen": 2111504384
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003663636363636364,
+      "loss": 2.6454,
+      "theoretical_loss": 3.414981552297617,
+      "tokens_seen": 2111569920
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036634343434343436,
+      "loss": 2.7228,
+      "theoretical_loss": 3.4149728093265543,
+      "tokens_seen": 2111635456
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003663232323232323,
+      "loss": 2.5719,
+      "theoretical_loss": 3.414964066702805,
+      "tokens_seen": 2111700992
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036630303030303033,
+      "loss": 2.6716,
+      "theoretical_loss": 3.4149553244263444,
+      "tokens_seen": 2111766528
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003662828282828283,
+      "loss": 2.6835,
+      "theoretical_loss": 3.414946582497148,
+      "tokens_seen": 2111832064
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003662626262626263,
+      "loss": 2.7341,
+      "theoretical_loss": 3.4149378409151905,
+      "tokens_seen": 2111897600
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036624242424242425,
+      "loss": 2.7706,
+      "theoretical_loss": 3.4149290996804487,
+      "tokens_seen": 2111963136
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003662222222222222,
+      "loss": 2.8759,
+      "theoretical_loss": 3.414920358792897,
+      "tokens_seen": 2112028672
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003662020202020202,
+      "loss": 2.6096,
+      "theoretical_loss": 3.414911618252511,
+      "tokens_seen": 2112094208
+    },
+    {
+      "epoch": 0.27,
+      "objective/train/docs_used": 1164108,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.35992169380188,
+      "objective/train/theoretical_loss": 3.4149028780592667,
+      "objective/train/tokens_used": 471018976,
+      "theoretical_loss": 3.4149028780592667,
+      "tokens_seen": 2112159744
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003661818181818182,
+      "loss": 2.5137,
+      "theoretical_loss": 3.4149028780592667,
+      "tokens_seen": 2112159744
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003661616161616162,
+      "loss": 2.7378,
+      "theoretical_loss": 3.4148941382131386,
+      "tokens_seen": 2112225280
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036614141414141415,
+      "loss": 2.695,
+      "theoretical_loss": 3.414885398714103,
+      "tokens_seen": 2112290816
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036612121212121216,
+      "loss": 2.3199,
+      "theoretical_loss": 3.414876659562135,
+      "tokens_seen": 2112356352
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003661010101010101,
+      "loss": 2.6186,
+      "theoretical_loss": 3.41486792075721,
+      "tokens_seen": 2112421888
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036608080808080807,
+      "loss": 2.558,
+      "theoretical_loss": 3.4148591822993035,
+      "tokens_seen": 2112487424
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003660606060606061,
+      "loss": 2.4394,
+      "theoretical_loss": 3.4148504441883913,
+      "tokens_seen": 2112552960
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036604040404040404,
+      "loss": 2.7204,
+      "theoretical_loss": 3.4148417064244487,
+      "tokens_seen": 2112618496
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036602020202020205,
+      "loss": 2.5691,
+      "theoretical_loss": 3.414832969007451,
+      "tokens_seen": 2112684032
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000366,
+      "loss": 2.4832,
+      "theoretical_loss": 3.4148242319373727,
+      "tokens_seen": 2112749568
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036597979797979796,
+      "loss": 2.6831,
+      "theoretical_loss": 3.414815495214191,
+      "tokens_seen": 2112815104
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000365959595959596,
+      "loss": 2.7641,
+      "theoretical_loss": 3.41480675883788,
+      "tokens_seen": 2112880640
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000365939393939394,
+      "loss": 2.7619,
+      "theoretical_loss": 3.4147980228084167,
+      "tokens_seen": 2112946176
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036591919191919194,
+      "loss": 2.6204,
+      "theoretical_loss": 3.4147892871257746,
+      "tokens_seen": 2113011712
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003658989898989899,
+      "loss": 2.7348,
+      "theoretical_loss": 3.414780551789931,
+      "tokens_seen": 2113077248
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036587878787878786,
+      "loss": 2.4798,
+      "theoretical_loss": 3.4147718168008607,
+      "tokens_seen": 2113142784
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036585858585858587,
+      "loss": 2.4059,
+      "theoretical_loss": 3.4147630821585384,
+      "tokens_seen": 2113208320
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003658383838383839,
+      "loss": 2.6539,
+      "theoretical_loss": 3.4147543478629405,
+      "tokens_seen": 2113273856
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036581818181818183,
+      "loss": 2.4668,
+      "theoretical_loss": 3.414745613914042,
+      "tokens_seen": 2113339392
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003657979797979798,
+      "loss": 2.6112,
+      "theoretical_loss": 3.4147368803118194,
+      "tokens_seen": 2113404928
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036577777777777775,
+      "loss": 2.7575,
+      "theoretical_loss": 3.4147281470562465,
+      "tokens_seen": 2113470464
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003657575757575758,
+      "loss": 2.5115,
+      "theoretical_loss": 3.4147194141473003,
+      "tokens_seen": 2113536000
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036573737373737377,
+      "loss": 2.756,
+      "theoretical_loss": 3.414710681584956,
+      "tokens_seen": 2113601536
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036571717171717173,
+      "loss": 2.88,
+      "theoretical_loss": 3.4147019493691877,
+      "tokens_seen": 2113667072
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003656969696969697,
+      "loss": 2.5343,
+      "theoretical_loss": 3.414693217499973,
+      "tokens_seen": 2113732608
+    },
+    {
+      "epoch": 0.28,
+      "objective/train/docs_used": 1165368,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.723964214324951,
+      "objective/train/theoretical_loss": 3.414684485977286,
+      "objective/train/tokens_used": 472657376,
+      "theoretical_loss": 3.414684485977286,
+      "tokens_seen": 2113798144
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036567676767676764,
+      "loss": 2.5892,
+      "theoretical_loss": 3.414684485977286,
+      "tokens_seen": 2113798144
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003656565656565657,
+      "loss": 2.8461,
+      "theoretical_loss": 3.4146757548011024,
+      "tokens_seen": 2113863680
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036563636363636366,
+      "loss": 2.599,
+      "theoretical_loss": 3.4146670239713983,
+      "tokens_seen": 2113929216
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003656161616161616,
+      "loss": 2.7025,
+      "theoretical_loss": 3.4146582934881486,
+      "tokens_seen": 2113994752
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003655959595959596,
+      "loss": 2.6218,
+      "theoretical_loss": 3.414649563351329,
+      "tokens_seen": 2114060288
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003655757575757576,
+      "loss": 2.7486,
+      "theoretical_loss": 3.4146408335609157,
+      "tokens_seen": 2114125824
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003655555555555556,
+      "loss": 2.6214,
+      "theoretical_loss": 3.414632104116883,
+      "tokens_seen": 2114191360
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036553535353535356,
+      "loss": 2.576,
+      "theoretical_loss": 3.4146233750192065,
+      "tokens_seen": 2114256896
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003655151515151515,
+      "loss": 2.5162,
+      "theoretical_loss": 3.4146146462678626,
+      "tokens_seen": 2114322432
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036549494949494947,
+      "loss": 2.4917,
+      "theoretical_loss": 3.4146059178628265,
+      "tokens_seen": 2114387968
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003654747474747475,
+      "loss": 2.6346,
+      "theoretical_loss": 3.414597189804074,
+      "tokens_seen": 2114453504
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003654545454545455,
+      "loss": 2.5658,
+      "theoretical_loss": 3.4145884620915803,
+      "tokens_seen": 2114519040
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036543434343434345,
+      "loss": 2.7962,
+      "theoretical_loss": 3.4145797347253204,
+      "tokens_seen": 2114584576
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003654141414141414,
+      "loss": 2.6246,
+      "theoretical_loss": 3.4145710077052707,
+      "tokens_seen": 2114650112
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003653939393939394,
+      "loss": 2.6092,
+      "theoretical_loss": 3.4145622810314062,
+      "tokens_seen": 2114715648
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003653737373737374,
+      "loss": 2.584,
+      "theoretical_loss": 3.4145535547037023,
+      "tokens_seen": 2114781184
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003653535353535354,
+      "loss": 2.6047,
+      "theoretical_loss": 3.4145448287221356,
+      "tokens_seen": 2114846720
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036533333333333334,
+      "loss": 2.5933,
+      "theoretical_loss": 3.4145361030866805,
+      "tokens_seen": 2114912256
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003653131313131313,
+      "loss": 2.4416,
+      "theoretical_loss": 3.414527377797313,
+      "tokens_seen": 2114977792
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003652929292929293,
+      "loss": 2.6182,
+      "theoretical_loss": 3.414518652854009,
+      "tokens_seen": 2115043328
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036527272727272727,
+      "loss": 2.7091,
+      "theoretical_loss": 3.4145099282567433,
+      "tokens_seen": 2115108864
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003652525252525253,
+      "loss": 2.5143,
+      "theoretical_loss": 3.414501204005492,
+      "tokens_seen": 2115174400
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036523232323232323,
+      "loss": 2.54,
+      "theoretical_loss": 3.41449248010023,
+      "tokens_seen": 2115239936
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036521212121212124,
+      "loss": 2.72,
+      "theoretical_loss": 3.4144837565409336,
+      "tokens_seen": 2115305472
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003651919191919192,
+      "loss": 2.6686,
+      "theoretical_loss": 3.4144750333275784,
+      "tokens_seen": 2115371008
+    },
+    {
+      "epoch": 0.28,
+      "objective/train/docs_used": 1165914,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.618112087249756,
+      "objective/train/theoretical_loss": 3.4144663104601394,
+      "objective/train/tokens_used": 474295776,
+      "theoretical_loss": 3.4144663104601394,
+      "tokens_seen": 2115436544
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036517171717171716,
+      "loss": 2.641,
+      "theoretical_loss": 3.4144663104601394,
+      "tokens_seen": 2115436544
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036515151515151517,
+      "loss": 2.5092,
+      "theoretical_loss": 3.4144575879385926,
+      "tokens_seen": 2115502080
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003651313131313131,
+      "loss": 2.537,
+      "theoretical_loss": 3.414448865762913,
+      "tokens_seen": 2115567616
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036511111111111114,
+      "loss": 2.8961,
+      "theoretical_loss": 3.414440143933077,
+      "tokens_seen": 2115633152
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003650909090909091,
+      "loss": 2.5186,
+      "theoretical_loss": 3.4144314224490597,
+      "tokens_seen": 2115698688
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036507070707070705,
+      "loss": 2.5034,
+      "theoretical_loss": 3.414422701310837,
+      "tokens_seen": 2115764224
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036505050505050506,
+      "loss": 2.5035,
+      "theoretical_loss": 3.4144139805183835,
+      "tokens_seen": 2115829760
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000365030303030303,
+      "loss": 2.7573,
+      "theoretical_loss": 3.4144052600716757,
+      "tokens_seen": 2115895296
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036501010101010103,
+      "loss": 2.6071,
+      "theoretical_loss": 3.4143965399706895,
+      "tokens_seen": 2115960832
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000364989898989899,
+      "loss": 2.6751,
+      "theoretical_loss": 3.414387820215399,
+      "tokens_seen": 2116026368
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036496969696969694,
+      "loss": 2.771,
+      "theoretical_loss": 3.4143791008057818,
+      "tokens_seen": 2116091904
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036494949494949496,
+      "loss": 2.4936,
+      "theoretical_loss": 3.4143703817418123,
+      "tokens_seen": 2116157440
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036492929292929297,
+      "loss": 2.551,
+      "theoretical_loss": 3.4143616630234654,
+      "tokens_seen": 2116222976
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003649090909090909,
+      "loss": 2.6326,
+      "theoretical_loss": 3.4143529446507186,
+      "tokens_seen": 2116288512
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003648888888888889,
+      "loss": 2.6491,
+      "theoretical_loss": 3.4143442266235455,
+      "tokens_seen": 2116354048
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036486868686868684,
+      "loss": 2.6087,
+      "theoretical_loss": 3.4143355089419227,
+      "tokens_seen": 2116419584
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036484848484848485,
+      "loss": 2.692,
+      "theoretical_loss": 3.414326791605826,
+      "tokens_seen": 2116485120
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036482828282828286,
+      "loss": 2.2571,
+      "theoretical_loss": 3.414318074615231,
+      "tokens_seen": 2116550656
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003648080808080808,
+      "loss": 2.6154,
+      "theoretical_loss": 3.414309357970113,
+      "tokens_seen": 2116616192
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036478787878787877,
+      "loss": 2.5167,
+      "theoretical_loss": 3.4143006416704473,
+      "tokens_seen": 2116681728
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036476767676767673,
+      "loss": 2.6605,
+      "theoretical_loss": 3.41429192571621,
+      "tokens_seen": 2116747264
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003647474747474748,
+      "loss": 2.6692,
+      "theoretical_loss": 3.414283210107377,
+      "tokens_seen": 2116812800
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036472727272727275,
+      "loss": 2.573,
+      "theoretical_loss": 3.414274494843923,
+      "tokens_seen": 2116878336
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003647070707070707,
+      "loss": 2.8211,
+      "theoretical_loss": 3.4142657799258243,
+      "tokens_seen": 2116943872
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036468686868686867,
+      "loss": 2.7455,
+      "theoretical_loss": 3.4142570653530564,
+      "tokens_seen": 2117009408
+    },
+    {
+      "epoch": 0.28,
+      "objective/train/docs_used": 1167104,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.4354066848754883,
+      "objective/train/theoretical_loss": 3.414248351125595,
+      "objective/train/tokens_used": 475934176,
+      "theoretical_loss": 3.414248351125595,
+      "tokens_seen": 2117074944
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003646666666666666,
+      "loss": 2.5663,
+      "theoretical_loss": 3.414248351125595,
+      "tokens_seen": 2117074944
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003646464646464647,
+      "loss": 2.4757,
+      "theoretical_loss": 3.414239637243415,
+      "tokens_seen": 2117140480
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036462626262626264,
+      "loss": 2.6941,
+      "theoretical_loss": 3.4142309237064934,
+      "tokens_seen": 2117206016
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003646060606060606,
+      "loss": 2.6617,
+      "theoretical_loss": 3.414222210514805,
+      "tokens_seen": 2117271552
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036458585858585856,
+      "loss": 2.5227,
+      "theoretical_loss": 3.4142134976683254,
+      "tokens_seen": 2117337088
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003645656565656566,
+      "loss": 2.5767,
+      "theoretical_loss": 3.41420478516703,
+      "tokens_seen": 2117402624
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003645454545454546,
+      "loss": 2.6952,
+      "theoretical_loss": 3.414196073010895,
+      "tokens_seen": 2117468160
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036452525252525254,
+      "loss": 2.3163,
+      "theoretical_loss": 3.414187361199896,
+      "tokens_seen": 2117533696
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003645050505050505,
+      "loss": 2.5076,
+      "theoretical_loss": 3.414178649734008,
+      "tokens_seen": 2117599232
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036448484848484845,
+      "loss": 2.5644,
+      "theoretical_loss": 3.414169938613208,
+      "tokens_seen": 2117664768
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003644646464646465,
+      "loss": 2.6019,
+      "theoretical_loss": 3.41416122783747,
+      "tokens_seen": 2117730304
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036444444444444447,
+      "loss": 2.6005,
+      "theoretical_loss": 3.4141525174067704,
+      "tokens_seen": 2117795840
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036442424242424243,
+      "loss": 2.6914,
+      "theoretical_loss": 3.4141438073210857,
+      "tokens_seen": 2117861376
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003644040404040404,
+      "loss": 2.4907,
+      "theoretical_loss": 3.41413509758039,
+      "tokens_seen": 2117926912
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003643838383838384,
+      "loss": 2.7162,
+      "theoretical_loss": 3.41412638818466,
+      "tokens_seen": 2117992448
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003643636363636364,
+      "loss": 2.4166,
+      "theoretical_loss": 3.414117679133871,
+      "tokens_seen": 2118057984
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036434343434343437,
+      "loss": 2.7053,
+      "theoretical_loss": 3.414108970427999,
+      "tokens_seen": 2118123520
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003643232323232323,
+      "loss": 2.5607,
+      "theoretical_loss": 3.414100262067019,
+      "tokens_seen": 2118189056
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003643030303030303,
+      "loss": 2.5588,
+      "theoretical_loss": 3.4140915540509074,
+      "tokens_seen": 2118254592
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003642828282828283,
+      "loss": 2.6071,
+      "theoretical_loss": 3.4140828463796398,
+      "tokens_seen": 2118320128
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003642626262626263,
+      "loss": 2.4847,
+      "theoretical_loss": 3.4140741390531915,
+      "tokens_seen": 2118385664
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036424242424242426,
+      "loss": 2.636,
+      "theoretical_loss": 3.4140654320715376,
+      "tokens_seen": 2118451200
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003642222222222222,
+      "loss": 2.6212,
+      "theoretical_loss": 3.414056725434655,
+      "tokens_seen": 2118516736
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003642020202020202,
+      "loss": 2.7221,
+      "theoretical_loss": 3.414048019142519,
+      "tokens_seen": 2118582272
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003641818181818182,
+      "loss": 2.551,
+      "theoretical_loss": 3.4140393131951052,
+      "tokens_seen": 2118647808
+    },
+    {
+      "epoch": 0.28,
+      "objective/train/docs_used": 1168338,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.4486634731292725,
+      "objective/train/theoretical_loss": 3.414030607592389,
+      "objective/train/tokens_used": 477572576,
+      "theoretical_loss": 3.414030607592389,
+      "tokens_seen": 2118713344
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003641616161616162,
+      "loss": 2.5584,
+      "theoretical_loss": 3.414030607592389,
+      "tokens_seen": 2118713344
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036414141414141415,
+      "loss": 2.6926,
+      "theoretical_loss": 3.4140219023343468,
+      "tokens_seen": 2118778880
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003641212121212121,
+      "loss": 2.5152,
+      "theoretical_loss": 3.414013197420953,
+      "tokens_seen": 2118844416
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003641010101010101,
+      "loss": 2.7168,
+      "theoretical_loss": 3.414004492852185,
+      "tokens_seen": 2118909952
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003640808080808081,
+      "loss": 2.3342,
+      "theoretical_loss": 3.413995788628017,
+      "tokens_seen": 2118975488
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003640606060606061,
+      "loss": 2.6007,
+      "theoretical_loss": 3.4139870847484257,
+      "tokens_seen": 2119041024
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036404040404040404,
+      "loss": 2.5816,
+      "theoretical_loss": 3.4139783812133864,
+      "tokens_seen": 2119106560
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036402020202020205,
+      "loss": 2.4268,
+      "theoretical_loss": 3.4139696780228745,
+      "tokens_seen": 2119172096
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000364,
+      "loss": 2.7488,
+      "theoretical_loss": 3.4139609751768663,
+      "tokens_seen": 2119237632
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036397979797979797,
+      "loss": 2.7172,
+      "theoretical_loss": 3.413952272675337,
+      "tokens_seen": 2119303168
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000363959595959596,
+      "loss": 2.4971,
+      "theoretical_loss": 3.413943570518263,
+      "tokens_seen": 2119368704
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036393939393939394,
+      "loss": 2.6373,
+      "theoretical_loss": 3.4139348687056192,
+      "tokens_seen": 2119434240
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036391919191919195,
+      "loss": 2.6311,
+      "theoretical_loss": 3.413926167237382,
+      "tokens_seen": 2119499776
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003638989898989899,
+      "loss": 2.621,
+      "theoretical_loss": 3.4139174661135265,
+      "tokens_seen": 2119565312
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036387878787878786,
+      "loss": 2.439,
+      "theoretical_loss": 3.413908765334029,
+      "tokens_seen": 2119630848
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036385858585858587,
+      "loss": 2.5993,
+      "theoretical_loss": 3.413900064898865,
+      "tokens_seen": 2119696384
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003638383838383839,
+      "loss": 2.5193,
+      "theoretical_loss": 3.41389136480801,
+      "tokens_seen": 2119761920
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036381818181818184,
+      "loss": 2.5325,
+      "theoretical_loss": 3.4138826650614402,
+      "tokens_seen": 2119827456
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003637979797979798,
+      "loss": 2.5291,
+      "theoretical_loss": 3.413873965659131,
+      "tokens_seen": 2119892992
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036377777777777775,
+      "loss": 2.6428,
+      "theoretical_loss": 3.413865266601058,
+      "tokens_seen": 2119958528
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036375757575757576,
+      "loss": 2.7483,
+      "theoretical_loss": 3.4138565678871973,
+      "tokens_seen": 2120024064
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003637373737373738,
+      "loss": 2.5676,
+      "theoretical_loss": 3.4138478695175243,
+      "tokens_seen": 2120089600
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036371717171717173,
+      "loss": 2.5161,
+      "theoretical_loss": 3.4138391714920147,
+      "tokens_seen": 2120155136
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003636969696969697,
+      "loss": 2.5857,
+      "theoretical_loss": 3.413830473810645,
+      "tokens_seen": 2120220672
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036367676767676765,
+      "loss": 2.5532,
+      "theoretical_loss": 3.41382177647339,
+      "tokens_seen": 2120286208
+    },
+    {
+      "epoch": 0.28,
+      "objective/train/docs_used": 1168872,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.698695182800293,
+      "objective/train/theoretical_loss": 3.413813079480226,
+      "objective/train/tokens_used": 479210976,
+      "theoretical_loss": 3.413813079480226,
+      "tokens_seen": 2120351744
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003636565656565657,
+      "loss": 2.5234,
+      "theoretical_loss": 3.413813079480226,
+      "tokens_seen": 2120351744
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036363636363636367,
+      "loss": 2.4915,
+      "theoretical_loss": 3.413804382831129,
+      "tokens_seen": 2120417280
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003636161616161616,
+      "loss": 2.7313,
+      "theoretical_loss": 3.4137956865260737,
+      "tokens_seen": 2120482816
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003635959595959596,
+      "loss": 2.7314,
+      "theoretical_loss": 3.413786990565037,
+      "tokens_seen": 2120548352
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036357575757575754,
+      "loss": 2.8578,
+      "theoretical_loss": 3.413778294947994,
+      "tokens_seen": 2120613888
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003635555555555556,
+      "loss": 2.6198,
+      "theoretical_loss": 3.4137695996749207,
+      "tokens_seen": 2120679424
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036353535353535356,
+      "loss": 2.5495,
+      "theoretical_loss": 3.413760904745793,
+      "tokens_seen": 2120744960
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003635151515151515,
+      "loss": 2.5978,
+      "theoretical_loss": 3.413752210160586,
+      "tokens_seen": 2120810496
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003634949494949495,
+      "loss": 2.6271,
+      "theoretical_loss": 3.413743515919276,
+      "tokens_seen": 2120876032
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036347474747474743,
+      "loss": 2.5787,
+      "theoretical_loss": 3.413734822021839,
+      "tokens_seen": 2120941568
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003634545454545455,
+      "loss": 2.6924,
+      "theoretical_loss": 3.4137261284682503,
+      "tokens_seen": 2121007104
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036343434343434345,
+      "loss": 2.723,
+      "theoretical_loss": 3.413717435258486,
+      "tokens_seen": 2121072640
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003634141414141414,
+      "loss": 2.3627,
+      "theoretical_loss": 3.4137087423925214,
+      "tokens_seen": 2121138176
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036339393939393937,
+      "loss": 2.3717,
+      "theoretical_loss": 3.413700049870333,
+      "tokens_seen": 2121203712
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003633737373737374,
+      "loss": 2.6768,
+      "theoretical_loss": 3.413691357691896,
+      "tokens_seen": 2121269248
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003633535353535354,
+      "loss": 2.5494,
+      "theoretical_loss": 3.413682665857187,
+      "tokens_seen": 2121334784
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036333333333333335,
+      "loss": 2.5297,
+      "theoretical_loss": 3.4136739743661804,
+      "tokens_seen": 2121400320
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003633131313131313,
+      "loss": 2.4224,
+      "theoretical_loss": 3.4136652832188528,
+      "tokens_seen": 2121465856
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036329292929292926,
+      "loss": 2.5031,
+      "theoretical_loss": 3.4136565924151805,
+      "tokens_seen": 2121531392
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036327272727272727,
+      "loss": 2.5477,
+      "theoretical_loss": 3.4136479019551382,
+      "tokens_seen": 2121596928
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003632525252525253,
+      "loss": 2.661,
+      "theoretical_loss": 3.413639211838703,
+      "tokens_seen": 2121662464
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036323232323232324,
+      "loss": 2.8096,
+      "theoretical_loss": 3.4136305220658496,
+      "tokens_seen": 2121728000
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003632121212121212,
+      "loss": 2.763,
+      "theoretical_loss": 3.413621832636554,
+      "tokens_seen": 2121793536
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003631919191919192,
+      "loss": 2.5383,
+      "theoretical_loss": 3.4136131435507924,
+      "tokens_seen": 2121859072
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036317171717171716,
+      "loss": 2.386,
+      "theoretical_loss": 3.4136044548085405,
+      "tokens_seen": 2121924608
+    },
+    {
+      "epoch": 0.28,
+      "objective/train/docs_used": 1169826,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.2446186542510986,
+      "objective/train/theoretical_loss": 3.413595766409774,
+      "objective/train/tokens_used": 480849376,
+      "theoretical_loss": 3.413595766409774,
+      "tokens_seen": 2121990144
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003631515151515152,
+      "loss": 2.4437,
+      "theoretical_loss": 3.413595766409774,
+      "tokens_seen": 2121990144
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036313131313131313,
+      "loss": 2.7176,
+      "theoretical_loss": 3.413587078354468,
+      "tokens_seen": 2122055680
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003631111111111111,
+      "loss": 2.379,
+      "theoretical_loss": 3.4135783906425994,
+      "tokens_seen": 2122121216
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003630909090909091,
+      "loss": 2.5955,
+      "theoretical_loss": 3.4135697032741437,
+      "tokens_seen": 2122186752
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036307070707070706,
+      "loss": 2.6066,
+      "theoretical_loss": 3.4135610162490773,
+      "tokens_seen": 2122252288
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036305050505050507,
+      "loss": 2.5977,
+      "theoretical_loss": 3.4135523295673744,
+      "tokens_seen": 2122317824
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000363030303030303,
+      "loss": 2.5323,
+      "theoretical_loss": 3.4135436432290125,
+      "tokens_seen": 2122383360
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036301010101010104,
+      "loss": 2.557,
+      "theoretical_loss": 3.413534957233966,
+      "tokens_seen": 2122448896
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000362989898989899,
+      "loss": 2.6608,
+      "theoretical_loss": 3.4135262715822123,
+      "tokens_seen": 2122514432
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036296969696969695,
+      "loss": 2.428,
+      "theoretical_loss": 3.413517586273726,
+      "tokens_seen": 2122579968
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036294949494949496,
+      "loss": 2.7387,
+      "theoretical_loss": 3.4135089013084836,
+      "tokens_seen": 2122645504
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003629292929292929,
+      "loss": 2.6331,
+      "theoretical_loss": 3.4135002166864608,
+      "tokens_seen": 2122711040
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036290909090909093,
+      "loss": 2.6378,
+      "theoretical_loss": 3.413491532407633,
+      "tokens_seen": 2122776576
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003628888888888889,
+      "loss": 2.6081,
+      "theoretical_loss": 3.4134828484719764,
+      "tokens_seen": 2122842112
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003628686868686869,
+      "loss": 2.6682,
+      "theoretical_loss": 3.4134741648794664,
+      "tokens_seen": 2122907648
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036284848484848485,
+      "loss": 2.5269,
+      "theoretical_loss": 3.4134654816300802,
+      "tokens_seen": 2122973184
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036282828282828286,
+      "loss": 2.6664,
+      "theoretical_loss": 3.413456798723792,
+      "tokens_seen": 2123038720
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003628080808080808,
+      "loss": 2.7007,
+      "theoretical_loss": 3.4134481161605787,
+      "tokens_seen": 2123104256
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003627878787878788,
+      "loss": 2.5572,
+      "theoretical_loss": 3.4134394339404155,
+      "tokens_seen": 2123169792
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003627676767676768,
+      "loss": 2.7016,
+      "theoretical_loss": 3.4134307520632787,
+      "tokens_seen": 2123235328
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036274747474747475,
+      "loss": 2.5064,
+      "theoretical_loss": 3.413422070529144,
+      "tokens_seen": 2123300864
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036272727272727276,
+      "loss": 2.6765,
+      "theoretical_loss": 3.413413389337988,
+      "tokens_seen": 2123366400
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003627070707070707,
+      "loss": 2.5349,
+      "theoretical_loss": 3.413404708489785,
+      "tokens_seen": 2123431936
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036268686868686867,
+      "loss": 2.7782,
+      "theoretical_loss": 3.4133960279845117,
+      "tokens_seen": 2123497472
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003626666666666667,
+      "loss": 2.6642,
+      "theoretical_loss": 3.4133873478221446,
+      "tokens_seen": 2123563008
+    },
+    {
+      "epoch": 0.28,
+      "objective/train/docs_used": 1170397,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.436032295227051,
+      "objective/train/theoretical_loss": 3.4133786680026583,
+      "objective/train/tokens_used": 482487776,
+      "theoretical_loss": 3.4133786680026583,
+      "tokens_seen": 2123628544
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003626464646464647,
+      "loss": 2.5855,
+      "theoretical_loss": 3.4133786680026583,
+      "tokens_seen": 2123628544
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036262626262626265,
+      "loss": 2.5702,
+      "theoretical_loss": 3.41336998852603,
+      "tokens_seen": 2123694080
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003626060606060606,
+      "loss": 2.4347,
+      "theoretical_loss": 3.4133613093922346,
+      "tokens_seen": 2123759616
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036258585858585856,
+      "loss": 2.8192,
+      "theoretical_loss": 3.4133526306012483,
+      "tokens_seen": 2123825152
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003625656565656566,
+      "loss": 2.6402,
+      "theoretical_loss": 3.413343952153047,
+      "tokens_seen": 2123890688
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003625454545454546,
+      "loss": 2.6956,
+      "theoretical_loss": 3.4133352740476064,
+      "tokens_seen": 2123956224
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036252525252525254,
+      "loss": 2.2476,
+      "theoretical_loss": 3.413326596284903,
+      "tokens_seen": 2124021760
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003625050505050505,
+      "loss": 2.7057,
+      "theoretical_loss": 3.413317918864912,
+      "tokens_seen": 2124087296
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036248484848484846,
+      "loss": 2.4141,
+      "theoretical_loss": 3.4133092417876094,
+      "tokens_seen": 2124152832
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003624646464646465,
+      "loss": 2.5951,
+      "theoretical_loss": 3.4133005650529715,
+      "tokens_seen": 2124218368
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003624444444444445,
+      "loss": 2.6807,
+      "theoretical_loss": 3.4132918886609733,
+      "tokens_seen": 2124283904
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036242424242424244,
+      "loss": 2.5597,
+      "theoretical_loss": 3.413283212611592,
+      "tokens_seen": 2124349440
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003624040404040404,
+      "loss": 2.4826,
+      "theoretical_loss": 3.4132745369048028,
+      "tokens_seen": 2124414976
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036238383838383835,
+      "loss": 2.698,
+      "theoretical_loss": 3.413265861540582,
+      "tokens_seen": 2124480512
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003623636363636364,
+      "loss": 2.5928,
+      "theoretical_loss": 3.413257186518904,
+      "tokens_seen": 2124546048
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036234343434343437,
+      "loss": 2.7756,
+      "theoretical_loss": 3.413248511839747,
+      "tokens_seen": 2124611584
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036232323232323233,
+      "loss": 2.6898,
+      "theoretical_loss": 3.413239837503085,
+      "tokens_seen": 2124677120
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003623030303030303,
+      "loss": 2.4804,
+      "theoretical_loss": 3.413231163508895,
+      "tokens_seen": 2124742656
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003622828282828283,
+      "loss": 2.5586,
+      "theoretical_loss": 3.4132224898571524,
+      "tokens_seen": 2124808192
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003622626262626263,
+      "loss": 2.4899,
+      "theoretical_loss": 3.4132138165478336,
+      "tokens_seen": 2124873728
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036224242424242426,
+      "loss": 2.7261,
+      "theoretical_loss": 3.413205143580914,
+      "tokens_seen": 2124939264
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003622222222222222,
+      "loss": 2.483,
+      "theoretical_loss": 3.41319647095637,
+      "tokens_seen": 2125004800
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003622020202020202,
+      "loss": 2.6043,
+      "theoretical_loss": 3.413187798674177,
+      "tokens_seen": 2125070336
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003621818181818182,
+      "loss": 2.6508,
+      "theoretical_loss": 3.4131791267343115,
+      "tokens_seen": 2125135872
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003621616161616162,
+      "loss": 2.4916,
+      "theoretical_loss": 3.4131704551367488,
+      "tokens_seen": 2125201408
+    },
+    {
+      "epoch": 0.28,
+      "objective/train/docs_used": 1171543,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5023303031921387,
+      "objective/train/theoretical_loss": 3.413161783881465,
+      "objective/train/tokens_used": 484126176,
+      "theoretical_loss": 3.413161783881465,
+      "tokens_seen": 2125266944
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036214141414141416,
+      "loss": 2.5373,
+      "theoretical_loss": 3.413161783881465,
+      "tokens_seen": 2125266944
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003621212121212121,
+      "loss": 2.7232,
+      "theoretical_loss": 3.4131531129684367,
+      "tokens_seen": 2125332480
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003621010101010101,
+      "loss": 2.3724,
+      "theoretical_loss": 3.4131444423976394,
+      "tokens_seen": 2125398016
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003620808080808081,
+      "loss": 2.594,
+      "theoretical_loss": 3.4131357721690487,
+      "tokens_seen": 2125463552
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003620606060606061,
+      "loss": 2.4723,
+      "theoretical_loss": 3.4131271022826413,
+      "tokens_seen": 2125529088
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036204040404040405,
+      "loss": 2.6411,
+      "theoretical_loss": 3.4131184327383917,
+      "tokens_seen": 2125594624
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000362020202020202,
+      "loss": 2.4801,
+      "theoretical_loss": 3.4131097635362777,
+      "tokens_seen": 2125660160
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000362,
+      "loss": 2.819,
+      "theoretical_loss": 3.413101094676274,
+      "tokens_seen": 2125725696
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000361979797979798,
+      "loss": 2.5704,
+      "theoretical_loss": 3.4130924261583573,
+      "tokens_seen": 2125791232
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000361959595959596,
+      "loss": 2.5092,
+      "theoretical_loss": 3.413083757982503,
+      "tokens_seen": 2125856768
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036193939393939394,
+      "loss": 2.5521,
+      "theoretical_loss": 3.413075090148687,
+      "tokens_seen": 2125922304
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003619191919191919,
+      "loss": 2.7827,
+      "theoretical_loss": 3.413066422656885,
+      "tokens_seen": 2125987840
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003618989898989899,
+      "loss": 2.3831,
+      "theoretical_loss": 3.413057755507075,
+      "tokens_seen": 2126053376
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036187878787878787,
+      "loss": 2.6201,
+      "theoretical_loss": 3.41304908869923,
+      "tokens_seen": 2126118912
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003618585858585859,
+      "loss": 2.576,
+      "theoretical_loss": 3.413040422233328,
+      "tokens_seen": 2126184448
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036183838383838383,
+      "loss": 2.7195,
+      "theoretical_loss": 3.413031756109344,
+      "tokens_seen": 2126249984
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036181818181818185,
+      "loss": 2.4798,
+      "theoretical_loss": 3.413023090327255,
+      "tokens_seen": 2126315520
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003617979797979798,
+      "loss": 2.6043,
+      "theoretical_loss": 3.413014424887036,
+      "tokens_seen": 2126381056
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036177777777777776,
+      "loss": 2.5492,
+      "theoretical_loss": 3.413005759788663,
+      "tokens_seen": 2126446592
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036175757575757577,
+      "loss": 2.4745,
+      "theoretical_loss": 3.4129970950321127,
+      "tokens_seen": 2126512128
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036173737373737373,
+      "loss": 2.5508,
+      "theoretical_loss": 3.412988430617361,
+      "tokens_seen": 2126577664
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036171717171717174,
+      "loss": 2.4879,
+      "theoretical_loss": 3.4129797665443826,
+      "tokens_seen": 2126643200
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003616969696969697,
+      "loss": 2.5086,
+      "theoretical_loss": 3.4129711028131546,
+      "tokens_seen": 2126708736
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036167676767676765,
+      "loss": 2.6245,
+      "theoretical_loss": 3.4129624394236533,
+      "tokens_seen": 2126774272
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036165656565656566,
+      "loss": 2.6815,
+      "theoretical_loss": 3.4129537763758537,
+      "tokens_seen": 2126839808
+    },
+    {
+      "epoch": 0.28,
+      "objective/train/docs_used": 1172068,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6468193531036377,
+      "objective/train/theoretical_loss": 3.412945113669733,
+      "objective/train/tokens_used": 485764576,
+      "theoretical_loss": 3.412945113669733,
+      "tokens_seen": 2126905344
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003616363636363637,
+      "loss": 2.75,
+      "theoretical_loss": 3.412945113669733,
+      "tokens_seen": 2126905344
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036161616161616163,
+      "loss": 2.6632,
+      "theoretical_loss": 3.4129364513052654,
+      "tokens_seen": 2126970880
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003615959595959596,
+      "loss": 2.6412,
+      "theoretical_loss": 3.4129277892824286,
+      "tokens_seen": 2127036416
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036157575757575754,
+      "loss": 2.6567,
+      "theoretical_loss": 3.4129191276011985,
+      "tokens_seen": 2127101952
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036155555555555556,
+      "loss": 2.5087,
+      "theoretical_loss": 3.41291046626155,
+      "tokens_seen": 2127167488
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036153535353535357,
+      "loss": 2.4819,
+      "theoretical_loss": 3.4129018052634597,
+      "tokens_seen": 2127233024
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003615151515151515,
+      "loss": 2.6294,
+      "theoretical_loss": 3.4128931446069037,
+      "tokens_seen": 2127298560
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003614949494949495,
+      "loss": 2.5302,
+      "theoretical_loss": 3.412884484291858,
+      "tokens_seen": 2127364096
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036147474747474744,
+      "loss": 2.6261,
+      "theoretical_loss": 3.4128758243182986,
+      "tokens_seen": 2127429632
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003614545454545455,
+      "loss": 2.4306,
+      "theoretical_loss": 3.412867164686201,
+      "tokens_seen": 2127495168
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036143434343434346,
+      "loss": 2.6665,
+      "theoretical_loss": 3.4128585053955423,
+      "tokens_seen": 2127560704
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003614141414141414,
+      "loss": 2.6082,
+      "theoretical_loss": 3.4128498464462975,
+      "tokens_seen": 2127626240
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003613939393939394,
+      "loss": 2.5125,
+      "theoretical_loss": 3.412841187838443,
+      "tokens_seen": 2127691776
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036137373737373733,
+      "loss": 2.8392,
+      "theoretical_loss": 3.412832529571955,
+      "tokens_seen": 2127757312
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003613535353535354,
+      "loss": 2.4665,
+      "theoretical_loss": 3.4128238716468093,
+      "tokens_seen": 2127822848
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036133333333333335,
+      "loss": 2.4348,
+      "theoretical_loss": 3.4128152140629817,
+      "tokens_seen": 2127888384
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003613131313131313,
+      "loss": 2.5413,
+      "theoretical_loss": 3.412806556820449,
+      "tokens_seen": 2127953920
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036129292929292927,
+      "loss": 2.6776,
+      "theoretical_loss": 3.4127978999191866,
+      "tokens_seen": 2128019456
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003612727272727273,
+      "loss": 2.6288,
+      "theoretical_loss": 3.412789243359171,
+      "tokens_seen": 2128084992
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003612525252525253,
+      "loss": 2.6286,
+      "theoretical_loss": 3.412780587140378,
+      "tokens_seen": 2128150528
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036123232323232324,
+      "loss": 2.4641,
+      "theoretical_loss": 3.4127719312627827,
+      "tokens_seen": 2128216064
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003612121212121212,
+      "loss": 2.5957,
+      "theoretical_loss": 3.4127632757263626,
+      "tokens_seen": 2128281600
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036119191919191916,
+      "loss": 2.7097,
+      "theoretical_loss": 3.4127546205310932,
+      "tokens_seen": 2128347136
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003611717171717172,
+      "loss": 2.5988,
+      "theoretical_loss": 3.4127459656769505,
+      "tokens_seen": 2128412672
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003611515151515152,
+      "loss": 2.4821,
+      "theoretical_loss": 3.4127373111639105,
+      "tokens_seen": 2128478208
+    },
+    {
+      "epoch": 0.28,
+      "objective/train/docs_used": 1173260,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5667147636413574,
+      "objective/train/theoretical_loss": 3.4127286569919493,
+      "objective/train/tokens_used": 487402976,
+      "theoretical_loss": 3.4127286569919493,
+      "tokens_seen": 2128543744
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036113131313131314,
+      "loss": 2.6403,
+      "theoretical_loss": 3.4127286569919493,
+      "tokens_seen": 2128543744
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003611111111111111,
+      "loss": 2.5117,
+      "theoretical_loss": 3.412720003161043,
+      "tokens_seen": 2128609280
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003610909090909091,
+      "loss": 2.5641,
+      "theoretical_loss": 3.412711349671168,
+      "tokens_seen": 2128674816
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003610707070707071,
+      "loss": 2.5267,
+      "theoretical_loss": 3.4127026965222997,
+      "tokens_seen": 2128740352
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003610505050505051,
+      "loss": 2.7131,
+      "theoretical_loss": 3.4126940437144144,
+      "tokens_seen": 2128805888
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036103030303030303,
+      "loss": 2.558,
+      "theoretical_loss": 3.4126853912474884,
+      "tokens_seen": 2128871424
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000361010101010101,
+      "loss": 2.5396,
+      "theoretical_loss": 3.4126767391214976,
+      "tokens_seen": 2128936960
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000360989898989899,
+      "loss": 2.6273,
+      "theoretical_loss": 3.412668087336418,
+      "tokens_seen": 2129002496
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000360969696969697,
+      "loss": 2.7662,
+      "theoretical_loss": 3.412659435892226,
+      "tokens_seen": 2129068032
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036094949494949497,
+      "loss": 2.6133,
+      "theoretical_loss": 3.412650784788897,
+      "tokens_seen": 2129133568
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003609292929292929,
+      "loss": 2.8016,
+      "theoretical_loss": 3.412642134026408,
+      "tokens_seen": 2129199104
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036090909090909093,
+      "loss": 2.5226,
+      "theoretical_loss": 3.4126334836047345,
+      "tokens_seen": 2129264640
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003608888888888889,
+      "loss": 2.4984,
+      "theoretical_loss": 3.4126248335238527,
+      "tokens_seen": 2129330176
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003608686868686869,
+      "loss": 2.6015,
+      "theoretical_loss": 3.4126161837837383,
+      "tokens_seen": 2129395712
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036084848484848486,
+      "loss": 2.6194,
+      "theoretical_loss": 3.412607534384368,
+      "tokens_seen": 2129461248
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003608282828282828,
+      "loss": 2.5166,
+      "theoretical_loss": 3.4125988853257176,
+      "tokens_seen": 2129526784
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003608080808080808,
+      "loss": 2.4291,
+      "theoretical_loss": 3.4125902366077634,
+      "tokens_seen": 2129592320
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003607878787878788,
+      "loss": 2.5511,
+      "theoretical_loss": 3.4125815882304806,
+      "tokens_seen": 2129657856
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003607676767676768,
+      "loss": 2.5663,
+      "theoretical_loss": 3.412572940193847,
+      "tokens_seen": 2129723392
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036074747474747475,
+      "loss": 2.7557,
+      "theoretical_loss": 3.412564292497837,
+      "tokens_seen": 2129788928
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036072727272727276,
+      "loss": 2.5695,
+      "theoretical_loss": 3.412555645142428,
+      "tokens_seen": 2129854464
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003607070707070707,
+      "loss": 2.5838,
+      "theoretical_loss": 3.4125469981275955,
+      "tokens_seen": 2129920000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003606868686868687,
+      "loss": 2.6765,
+      "theoretical_loss": 3.4125383514533154,
+      "tokens_seen": 2129985536
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003606666666666667,
+      "loss": 2.641,
+      "theoretical_loss": 3.412529705119564,
+      "tokens_seen": 2130051072
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036064646464646464,
+      "loss": 2.5517,
+      "theoretical_loss": 3.4125210591263175,
+      "tokens_seen": 2130116608
+    },
+    {
+      "epoch": 0.29,
+      "objective/train/docs_used": 1174166,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.371098041534424,
+      "objective/train/theoretical_loss": 3.4125124134735523,
+      "objective/train/tokens_used": 489041376,
+      "theoretical_loss": 3.4125124134735523,
+      "tokens_seen": 2130182144
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036062626262626266,
+      "loss": 2.5198,
+      "theoretical_loss": 3.4125124134735523,
+      "tokens_seen": 2130182144
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003606060606060606,
+      "loss": 2.5662,
+      "theoretical_loss": 3.412503768161244,
+      "tokens_seen": 2130247680
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036058585858585857,
+      "loss": 2.6296,
+      "theoretical_loss": 3.4124951231893688,
+      "tokens_seen": 2130313216
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003605656565656566,
+      "loss": 2.5507,
+      "theoretical_loss": 3.412486478557903,
+      "tokens_seen": 2130378752
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003605454545454546,
+      "loss": 2.799,
+      "theoretical_loss": 3.4124778342668227,
+      "tokens_seen": 2130444288
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036052525252525255,
+      "loss": 2.7239,
+      "theoretical_loss": 3.412469190316104,
+      "tokens_seen": 2130509824
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003605050505050505,
+      "loss": 2.7608,
+      "theoretical_loss": 3.412460546705723,
+      "tokens_seen": 2130575360
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036048484848484846,
+      "loss": 2.6064,
+      "theoretical_loss": 3.412451903435656,
+      "tokens_seen": 2130640896
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036046464646464647,
+      "loss": 2.5552,
+      "theoretical_loss": 3.4124432605058788,
+      "tokens_seen": 2130706432
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003604444444444445,
+      "loss": 2.9222,
+      "theoretical_loss": 3.412434617916368,
+      "tokens_seen": 2130771968
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036042424242424244,
+      "loss": 2.6447,
+      "theoretical_loss": 3.412425975667099,
+      "tokens_seen": 2130837504
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003604040404040404,
+      "loss": 2.6339,
+      "theoretical_loss": 3.4124173337580492,
+      "tokens_seen": 2130903040
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036038383838383835,
+      "loss": 2.6131,
+      "theoretical_loss": 3.412408692189193,
+      "tokens_seen": 2130968576
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003603636363636364,
+      "loss": 2.716,
+      "theoretical_loss": 3.412400050960508,
+      "tokens_seen": 2131034112
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003603434343434344,
+      "loss": 2.7042,
+      "theoretical_loss": 3.4123914100719697,
+      "tokens_seen": 2131099648
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036032323232323233,
+      "loss": 2.5032,
+      "theoretical_loss": 3.4123827695235542,
+      "tokens_seen": 2131165184
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003603030303030303,
+      "loss": 2.6595,
+      "theoretical_loss": 3.4123741293152383,
+      "tokens_seen": 2131230720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036028282828282825,
+      "loss": 2.715,
+      "theoretical_loss": 3.412365489446998,
+      "tokens_seen": 2131296256
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003602626262626263,
+      "loss": 2.547,
+      "theoretical_loss": 3.4123568499188086,
+      "tokens_seen": 2131361792
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036024242424242427,
+      "loss": 2.7751,
+      "theoretical_loss": 3.412348210730647,
+      "tokens_seen": 2131427328
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003602222222222222,
+      "loss": 2.4854,
+      "theoretical_loss": 3.412339571882489,
+      "tokens_seen": 2131492864
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003602020202020202,
+      "loss": 2.5917,
+      "theoretical_loss": 3.412330933374311,
+      "tokens_seen": 2131558400
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036018181818181814,
+      "loss": 2.6232,
+      "theoretical_loss": 3.4123222952060894,
+      "tokens_seen": 2131623936
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003601616161616162,
+      "loss": 2.6344,
+      "theoretical_loss": 3.4123136573777995,
+      "tokens_seen": 2131689472
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036014141414141416,
+      "loss": 2.6682,
+      "theoretical_loss": 3.412305019889419,
+      "tokens_seen": 2131755008
+    },
+    {
+      "epoch": 0.29,
+      "objective/train/docs_used": 1174844,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.656928300857544,
+      "objective/train/theoretical_loss": 3.412296382740922,
+      "objective/train/tokens_used": 490679776,
+      "theoretical_loss": 3.412296382740922,
+      "tokens_seen": 2131820544
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003601212121212121,
+      "loss": 2.5572,
+      "theoretical_loss": 3.412296382740922,
+      "tokens_seen": 2131820544
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003601010101010101,
+      "loss": 2.5109,
+      "theoretical_loss": 3.4122877459322867,
+      "tokens_seen": 2131886080
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003600808080808081,
+      "loss": 2.5999,
+      "theoretical_loss": 3.412279109463488,
+      "tokens_seen": 2131951616
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003600606060606061,
+      "loss": 2.4027,
+      "theoretical_loss": 3.4122704733345026,
+      "tokens_seen": 2132017152
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036004040404040405,
+      "loss": 2.482,
+      "theoretical_loss": 3.412261837545307,
+      "tokens_seen": 2132082688
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000360020202020202,
+      "loss": 2.4131,
+      "theoretical_loss": 3.4122532020958762,
+      "tokens_seen": 2132148224
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035999999999999997,
+      "loss": 2.6003,
+      "theoretical_loss": 3.412244566986187,
+      "tokens_seen": 2132213760
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000359979797979798,
+      "loss": 2.3046,
+      "theoretical_loss": 3.4122359322162166,
+      "tokens_seen": 2132279296
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000359959595959596,
+      "loss": 2.5533,
+      "theoretical_loss": 3.4122272977859396,
+      "tokens_seen": 2132344832
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035993939393939395,
+      "loss": 2.5854,
+      "theoretical_loss": 3.4122186636953336,
+      "tokens_seen": 2132410368
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003599191919191919,
+      "loss": 2.8106,
+      "theoretical_loss": 3.412210029944374,
+      "tokens_seen": 2132475904
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003598989898989899,
+      "loss": 2.5694,
+      "theoretical_loss": 3.4122013965330367,
+      "tokens_seen": 2132541440
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035987878787878787,
+      "loss": 2.6185,
+      "theoretical_loss": 3.4121927634612987,
+      "tokens_seen": 2132606976
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003598585858585859,
+      "loss": 2.4607,
+      "theoretical_loss": 3.4121841307291354,
+      "tokens_seen": 2132672512
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035983838383838384,
+      "loss": 2.7415,
+      "theoretical_loss": 3.4121754983365236,
+      "tokens_seen": 2132738048
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003598181818181818,
+      "loss": 2.5386,
+      "theoretical_loss": 3.4121668662834397,
+      "tokens_seen": 2132803584
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003597979797979798,
+      "loss": 2.5452,
+      "theoretical_loss": 3.4121582345698593,
+      "tokens_seen": 2132869120
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035977777777777776,
+      "loss": 2.4101,
+      "theoretical_loss": 3.4121496031957586,
+      "tokens_seen": 2132934656
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003597575757575758,
+      "loss": 2.7979,
+      "theoretical_loss": 3.4121409721611147,
+      "tokens_seen": 2133000192
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035973737373737373,
+      "loss": 2.6733,
+      "theoretical_loss": 3.412132341465903,
+      "tokens_seen": 2133065728
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035971717171717174,
+      "loss": 2.5141,
+      "theoretical_loss": 3.4121237111101,
+      "tokens_seen": 2133131264
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003596969696969697,
+      "loss": 2.503,
+      "theoretical_loss": 3.4121150810936816,
+      "tokens_seen": 2133196800
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035967676767676766,
+      "loss": 2.5298,
+      "theoretical_loss": 3.4121064514166246,
+      "tokens_seen": 2133262336
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035965656565656567,
+      "loss": 2.6807,
+      "theoretical_loss": 3.4120978220789047,
+      "tokens_seen": 2133327872
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003596363636363636,
+      "loss": 2.5674,
+      "theoretical_loss": 3.412089193080498,
+      "tokens_seen": 2133393408
+    },
+    {
+      "epoch": 0.29,
+      "objective/train/docs_used": 1175925,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5075652599334717,
+      "objective/train/theoretical_loss": 3.412080564421382,
+      "objective/train/tokens_used": 492318176,
+      "theoretical_loss": 3.412080564421382,
+      "tokens_seen": 2133458944
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035961616161616164,
+      "loss": 2.6865,
+      "theoretical_loss": 3.412080564421382,
+      "tokens_seen": 2133458944
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003595959595959596,
+      "loss": 2.822,
+      "theoretical_loss": 3.4120719361015315,
+      "tokens_seen": 2133524480
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035957575757575755,
+      "loss": 2.541,
+      "theoretical_loss": 3.4120633081209233,
+      "tokens_seen": 2133590016
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035955555555555556,
+      "loss": 2.6752,
+      "theoretical_loss": 3.4120546804795335,
+      "tokens_seen": 2133655552
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035953535353535357,
+      "loss": 2.6471,
+      "theoretical_loss": 3.412046053177339,
+      "tokens_seen": 2133721088
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035951515151515153,
+      "loss": 2.6086,
+      "theoretical_loss": 3.412037426214315,
+      "tokens_seen": 2133786624
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003594949494949495,
+      "loss": 2.5331,
+      "theoretical_loss": 3.4120287995904386,
+      "tokens_seen": 2133852160
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003594747474747475,
+      "loss": 2.739,
+      "theoretical_loss": 3.4120201733056854,
+      "tokens_seen": 2133917696
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035945454545454545,
+      "loss": 2.6104,
+      "theoretical_loss": 3.412011547360032,
+      "tokens_seen": 2133983232
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035943434343434347,
+      "loss": 2.6294,
+      "theoretical_loss": 3.412002921753455,
+      "tokens_seen": 2134048768
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003594141414141414,
+      "loss": 2.8417,
+      "theoretical_loss": 3.41199429648593,
+      "tokens_seen": 2134114304
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003593939393939394,
+      "loss": 2.5628,
+      "theoretical_loss": 3.4119856715574333,
+      "tokens_seen": 2134179840
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003593737373737374,
+      "loss": 2.6903,
+      "theoretical_loss": 3.411977046967942,
+      "tokens_seen": 2134245376
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003593535353535354,
+      "loss": 2.7785,
+      "theoretical_loss": 3.411968422717431,
+      "tokens_seen": 2134310912
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035933333333333336,
+      "loss": 2.5786,
+      "theoretical_loss": 3.411959798805878,
+      "tokens_seen": 2134376448
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003593131313131313,
+      "loss": 2.5277,
+      "theoretical_loss": 3.4119511752332583,
+      "tokens_seen": 2134441984
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035929292929292927,
+      "loss": 2.7237,
+      "theoretical_loss": 3.4119425519995485,
+      "tokens_seen": 2134507520
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003592727272727273,
+      "loss": 2.6926,
+      "theoretical_loss": 3.411933929104725,
+      "tokens_seen": 2134573056
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003592525252525253,
+      "loss": 2.6479,
+      "theoretical_loss": 3.411925306548764,
+      "tokens_seen": 2134638592
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035923232323232325,
+      "loss": 2.6249,
+      "theoretical_loss": 3.4119166843316413,
+      "tokens_seen": 2134704128
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003592121212121212,
+      "loss": 2.6359,
+      "theoretical_loss": 3.4119080624533336,
+      "tokens_seen": 2134769664
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035919191919191916,
+      "loss": 2.7272,
+      "theoretical_loss": 3.411899440913818,
+      "tokens_seen": 2134835200
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035917171717171723,
+      "loss": 2.7252,
+      "theoretical_loss": 3.4118908197130695,
+      "tokens_seen": 2134900736
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003591515151515152,
+      "loss": 2.7684,
+      "theoretical_loss": 3.411882198851065,
+      "tokens_seen": 2134966272
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035913131313131314,
+      "loss": 2.5349,
+      "theoretical_loss": 3.4118735783277803,
+      "tokens_seen": 2135031808
+    },
+    {
+      "epoch": 0.29,
+      "objective/train/docs_used": 1176552,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6072494983673096,
+      "objective/train/theoretical_loss": 3.411864958143192,
+      "objective/train/tokens_used": 493956576,
+      "theoretical_loss": 3.411864958143192,
+      "tokens_seen": 2135097344
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003591111111111111,
+      "loss": 2.4361,
+      "theoretical_loss": 3.411864958143192,
+      "tokens_seen": 2135097344
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035909090909090906,
+      "loss": 2.639,
+      "theoretical_loss": 3.4118563382972766,
+      "tokens_seen": 2135162880
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003590707070707071,
+      "loss": 2.608,
+      "theoretical_loss": 3.4118477187900105,
+      "tokens_seen": 2135228416
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003590505050505051,
+      "loss": 2.6096,
+      "theoretical_loss": 3.4118390996213694,
+      "tokens_seen": 2135293952
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035903030303030304,
+      "loss": 2.6053,
+      "theoretical_loss": 3.41183048079133,
+      "tokens_seen": 2135359488
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000359010101010101,
+      "loss": 2.5846,
+      "theoretical_loss": 3.4118218622998686,
+      "tokens_seen": 2135425024
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000358989898989899,
+      "loss": 2.4498,
+      "theoretical_loss": 3.4118132441469617,
+      "tokens_seen": 2135490560
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000358969696969697,
+      "loss": 2.7093,
+      "theoretical_loss": 3.411804626332585,
+      "tokens_seen": 2135556096
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035894949494949497,
+      "loss": 2.5592,
+      "theoretical_loss": 3.4117960088567156,
+      "tokens_seen": 2135621632
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035892929292929293,
+      "loss": 2.6863,
+      "theoretical_loss": 3.4117873917193293,
+      "tokens_seen": 2135687168
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003589090909090909,
+      "loss": 2.6119,
+      "theoretical_loss": 3.4117787749204025,
+      "tokens_seen": 2135752704
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003588888888888889,
+      "loss": 2.8139,
+      "theoretical_loss": 3.4117701584599116,
+      "tokens_seen": 2135818240
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003588686868686869,
+      "loss": 2.6701,
+      "theoretical_loss": 3.4117615423378327,
+      "tokens_seen": 2135883776
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035884848484848486,
+      "loss": 2.6039,
+      "theoretical_loss": 3.4117529265541426,
+      "tokens_seen": 2135949312
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003588282828282828,
+      "loss": 2.5645,
+      "theoretical_loss": 3.4117443111088175,
+      "tokens_seen": 2136014848
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035880808080808083,
+      "loss": 2.7743,
+      "theoretical_loss": 3.411735696001833,
+      "tokens_seen": 2136080384
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003587878787878788,
+      "loss": 2.654,
+      "theoretical_loss": 3.411727081233166,
+      "tokens_seen": 2136145920
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003587676767676768,
+      "loss": 2.6771,
+      "theoretical_loss": 3.4117184668027933,
+      "tokens_seen": 2136211456
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035874747474747476,
+      "loss": 2.7149,
+      "theoretical_loss": 3.41170985271069,
+      "tokens_seen": 2136276992
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003587272727272727,
+      "loss": 2.7189,
+      "theoretical_loss": 3.4117012389568337,
+      "tokens_seen": 2136342528
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003587070707070707,
+      "loss": 2.5876,
+      "theoretical_loss": 3.4116926255412006,
+      "tokens_seen": 2136408064
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003586868686868687,
+      "loss": 2.6411,
+      "theoretical_loss": 3.411684012463766,
+      "tokens_seen": 2136473600
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003586666666666667,
+      "loss": 2.5305,
+      "theoretical_loss": 3.4116753997245075,
+      "tokens_seen": 2136539136
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035864646464646465,
+      "loss": 2.5586,
+      "theoretical_loss": 3.4116667873234006,
+      "tokens_seen": 2136604672
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003586262626262626,
+      "loss": 2.6443,
+      "theoretical_loss": 3.4116581752604223,
+      "tokens_seen": 2136670208
+    },
+    {
+      "epoch": 0.29,
+      "objective/train/docs_used": 1177415,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.7517342567443848,
+      "objective/train/theoretical_loss": 3.4116495635355486,
+      "objective/train/tokens_used": 495594976,
+      "theoretical_loss": 3.4116495635355486,
+      "tokens_seen": 2136735744
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003586060606060606,
+      "loss": 2.6526,
+      "theoretical_loss": 3.4116495635355486,
+      "tokens_seen": 2136735744
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003585858585858586,
+      "loss": 2.8409,
+      "theoretical_loss": 3.411640952148755,
+      "tokens_seen": 2136801280
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003585656565656566,
+      "loss": 2.6958,
+      "theoretical_loss": 3.4116323411000193,
+      "tokens_seen": 2136866816
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035854545454545454,
+      "loss": 2.6911,
+      "theoretical_loss": 3.4116237303893175,
+      "tokens_seen": 2136932352
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035852525252525255,
+      "loss": 2.5204,
+      "theoretical_loss": 3.4116151200166254,
+      "tokens_seen": 2136997888
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003585050505050505,
+      "loss": 2.4892,
+      "theoretical_loss": 3.4116065099819197,
+      "tokens_seen": 2137063424
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035848484848484847,
+      "loss": 2.7625,
+      "theoretical_loss": 3.4115979002851775,
+      "tokens_seen": 2137128960
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003584646464646465,
+      "loss": 2.748,
+      "theoretical_loss": 3.411589290926374,
+      "tokens_seen": 2137194496
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035844444444444444,
+      "loss": 2.8012,
+      "theoretical_loss": 3.4115806819054857,
+      "tokens_seen": 2137260032
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035842424242424245,
+      "loss": 2.4286,
+      "theoretical_loss": 3.411572073222489,
+      "tokens_seen": 2137325568
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003584040404040404,
+      "loss": 2.7613,
+      "theoretical_loss": 3.411563464877361,
+      "tokens_seen": 2137391104
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035838383838383836,
+      "loss": 2.521,
+      "theoretical_loss": 3.4115548568700778,
+      "tokens_seen": 2137456640
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035836363636363637,
+      "loss": 2.5064,
+      "theoretical_loss": 3.4115462492006157,
+      "tokens_seen": 2137522176
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003583434343434344,
+      "loss": 2.687,
+      "theoretical_loss": 3.411537641868951,
+      "tokens_seen": 2137587712
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035832323232323234,
+      "loss": 2.6783,
+      "theoretical_loss": 3.41152903487506,
+      "tokens_seen": 2137653248
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003583030303030303,
+      "loss": 2.6258,
+      "theoretical_loss": 3.411520428218919,
+      "tokens_seen": 2137718784
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035828282828282825,
+      "loss": 2.4896,
+      "theoretical_loss": 3.411511821900505,
+      "tokens_seen": 2137784320
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035826262626262626,
+      "loss": 2.8056,
+      "theoretical_loss": 3.4115032159197938,
+      "tokens_seen": 2137849856
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003582424242424243,
+      "loss": 2.712,
+      "theoretical_loss": 3.411494610276762,
+      "tokens_seen": 2137915392
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035822222222222223,
+      "loss": 2.7067,
+      "theoretical_loss": 3.4114860049713855,
+      "tokens_seen": 2137980928
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003582020202020202,
+      "loss": 2.8186,
+      "theoretical_loss": 3.4114774000036414,
+      "tokens_seen": 2138046464
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035818181818181815,
+      "loss": 2.7648,
+      "theoretical_loss": 3.4114687953735063,
+      "tokens_seen": 2138112000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003581616161616162,
+      "loss": 2.7007,
+      "theoretical_loss": 3.411460191080956,
+      "tokens_seen": 2138177536
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035814141414141417,
+      "loss": 2.6543,
+      "theoretical_loss": 3.411451587125967,
+      "tokens_seen": 2138243072
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003581212121212121,
+      "loss": 2.6273,
+      "theoretical_loss": 3.4114429835085156,
+      "tokens_seen": 2138308608
+    },
+    {
+      "epoch": 0.29,
+      "objective/train/docs_used": 1177948,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.514267921447754,
+      "objective/train/theoretical_loss": 3.411434380228579,
+      "objective/train/tokens_used": 497233376,
+      "theoretical_loss": 3.411434380228579,
+      "tokens_seen": 2138374144
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003581010101010101,
+      "loss": 2.5898,
+      "theoretical_loss": 3.411434380228579,
+      "tokens_seen": 2138374144
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035808080808080804,
+      "loss": 2.7825,
+      "theoretical_loss": 3.4114257772861327,
+      "tokens_seen": 2138439680
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003580606060606061,
+      "loss": 2.7285,
+      "theoretical_loss": 3.4114171746811537,
+      "tokens_seen": 2138505216
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035804040404040406,
+      "loss": 2.5128,
+      "theoretical_loss": 3.411408572413618,
+      "tokens_seen": 2138570752
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000358020202020202,
+      "loss": 2.5706,
+      "theoretical_loss": 3.4113999704835023,
+      "tokens_seen": 2138636288
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000358,
+      "loss": 2.8487,
+      "theoretical_loss": 3.411391368890783,
+      "tokens_seen": 2138701824
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000357979797979798,
+      "loss": 2.7602,
+      "theoretical_loss": 3.4113827676354362,
+      "tokens_seen": 2138767360
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000357959595959596,
+      "loss": 2.5372,
+      "theoretical_loss": 3.4113741667174384,
+      "tokens_seen": 2138832896
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035793939393939395,
+      "loss": 2.7173,
+      "theoretical_loss": 3.411365566136767,
+      "tokens_seen": 2138898432
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003579191919191919,
+      "loss": 2.5033,
+      "theoretical_loss": 3.411356965893397,
+      "tokens_seen": 2138963968
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035789898989898987,
+      "loss": 2.7777,
+      "theoretical_loss": 3.4113483659873056,
+      "tokens_seen": 2139029504
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003578787878787879,
+      "loss": 2.8195,
+      "theoretical_loss": 3.4113397664184695,
+      "tokens_seen": 2139095040
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003578585858585859,
+      "loss": 2.4594,
+      "theoretical_loss": 3.4113311671868645,
+      "tokens_seen": 2139160576
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035783838383838385,
+      "loss": 2.7956,
+      "theoretical_loss": 3.4113225682924675,
+      "tokens_seen": 2139226112
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003578181818181818,
+      "loss": 2.8989,
+      "theoretical_loss": 3.4113139697352546,
+      "tokens_seen": 2139291648
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003577979797979798,
+      "loss": 2.5859,
+      "theoretical_loss": 3.411305371515202,
+      "tokens_seen": 2139357184
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035777777777777777,
+      "loss": 2.6077,
+      "theoretical_loss": 3.4112967736322872,
+      "tokens_seen": 2139422720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003577575757575758,
+      "loss": 2.5769,
+      "theoretical_loss": 3.411288176086486,
+      "tokens_seen": 2139488256
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035773737373737374,
+      "loss": 2.6175,
+      "theoretical_loss": 3.4112795788777746,
+      "tokens_seen": 2139553792
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003577171717171717,
+      "loss": 2.4836,
+      "theoretical_loss": 3.41127098200613,
+      "tokens_seen": 2139619328
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003576969696969697,
+      "loss": 2.6395,
+      "theoretical_loss": 3.4112623854715283,
+      "tokens_seen": 2139684864
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003576767676767677,
+      "loss": 2.7262,
+      "theoretical_loss": 3.411253789273946,
+      "tokens_seen": 2139750400
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003576565656565657,
+      "loss": 2.6886,
+      "theoretical_loss": 3.41124519341336,
+      "tokens_seen": 2139815936
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035763636363636363,
+      "loss": 2.6317,
+      "theoretical_loss": 3.4112365978897463,
+      "tokens_seen": 2139881472
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035761616161616164,
+      "loss": 2.5842,
+      "theoretical_loss": 3.411228002703081,
+      "tokens_seen": 2139947008
+    },
+    {
+      "epoch": 0.29,
+      "objective/train/docs_used": 1178547,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 3.098909378051758,
+      "objective/train/theoretical_loss": 3.4112194078533413,
+      "objective/train/tokens_used": 498871776,
+      "theoretical_loss": 3.4112194078533413,
+      "tokens_seen": 2140012544
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003575959595959596,
+      "loss": 2.8142,
+      "theoretical_loss": 3.4112194078533413,
+      "tokens_seen": 2140012544
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003575757575757576,
+      "loss": 2.6792,
+      "theoretical_loss": 3.4112108133405035,
+      "tokens_seen": 2140078080
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035755555555555557,
+      "loss": 2.9019,
+      "theoretical_loss": 3.4112022191645437,
+      "tokens_seen": 2140143616
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003575353535353535,
+      "loss": 2.6421,
+      "theoretical_loss": 3.4111936253254393,
+      "tokens_seen": 2140209152
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035751515151515153,
+      "loss": 2.5492,
+      "theoretical_loss": 3.4111850318231656,
+      "tokens_seen": 2140274688
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003574949494949495,
+      "loss": 2.6782,
+      "theoretical_loss": 3.4111764386576997,
+      "tokens_seen": 2140340224
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003574747474747475,
+      "loss": 2.6779,
+      "theoretical_loss": 3.411167845829018,
+      "tokens_seen": 2140405760
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035745454545454546,
+      "loss": 2.6727,
+      "theoretical_loss": 3.4111592533370976,
+      "tokens_seen": 2140471296
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035743434343434347,
+      "loss": 2.4204,
+      "theoretical_loss": 3.411150661181914,
+      "tokens_seen": 2140536832
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035741414141414143,
+      "loss": 2.8483,
+      "theoretical_loss": 3.4111420693634438,
+      "tokens_seen": 2140602368
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003573939393939394,
+      "loss": 2.5644,
+      "theoretical_loss": 3.4111334778816644,
+      "tokens_seen": 2140667904
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003573737373737374,
+      "loss": 2.618,
+      "theoretical_loss": 3.411124886736551,
+      "tokens_seen": 2140733440
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035735353535353535,
+      "loss": 2.5871,
+      "theoretical_loss": 3.4111162959280814,
+      "tokens_seen": 2140798976
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035733333333333336,
+      "loss": 2.865,
+      "theoretical_loss": 3.4111077054562315,
+      "tokens_seen": 2140864512
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003573131313131313,
+      "loss": 2.8388,
+      "theoretical_loss": 3.4110991153209778,
+      "tokens_seen": 2140930048
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003572929292929293,
+      "loss": 2.8237,
+      "theoretical_loss": 3.4110905255222965,
+      "tokens_seen": 2140995584
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003572727272727273,
+      "loss": 2.5144,
+      "theoretical_loss": 3.4110819360601647,
+      "tokens_seen": 2141061120
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003572525252525253,
+      "loss": 2.5991,
+      "theoretical_loss": 3.4110733469345584,
+      "tokens_seen": 2141126656
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035723232323232326,
+      "loss": 2.4743,
+      "theoretical_loss": 3.4110647581454545,
+      "tokens_seen": 2141192192
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003572121212121212,
+      "loss": 2.6059,
+      "theoretical_loss": 3.4110561696928294,
+      "tokens_seen": 2141257728
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035719191919191917,
+      "loss": 2.6451,
+      "theoretical_loss": 3.4110475815766597,
+      "tokens_seen": 2141323264
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003571717171717172,
+      "loss": 2.6165,
+      "theoretical_loss": 3.4110389937969217,
+      "tokens_seen": 2141388800
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003571515151515152,
+      "loss": 2.5672,
+      "theoretical_loss": 3.411030406353592,
+      "tokens_seen": 2141454336
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035713131313131315,
+      "loss": 2.6528,
+      "theoretical_loss": 3.4110218192466473,
+      "tokens_seen": 2141519872
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003571111111111111,
+      "loss": 2.8488,
+      "theoretical_loss": 3.411013232476064,
+      "tokens_seen": 2141585408
+    },
+    {
+      "epoch": 0.29,
+      "objective/train/docs_used": 1179742,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.639477014541626,
+      "objective/train/theoretical_loss": 3.4110046460418184,
+      "objective/train/tokens_used": 500510176,
+      "theoretical_loss": 3.4110046460418184,
+      "tokens_seen": 2141650944
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035709090909090906,
+      "loss": 2.6017,
+      "theoretical_loss": 3.4110046460418184,
+      "tokens_seen": 2141650944
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035707070707070713,
+      "loss": 2.7004,
+      "theoretical_loss": 3.410996059943887,
+      "tokens_seen": 2141716480
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003570505050505051,
+      "loss": 2.6781,
+      "theoretical_loss": 3.4109874741822477,
+      "tokens_seen": 2141782016
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035703030303030304,
+      "loss": 2.8444,
+      "theoretical_loss": 3.4109788887568753,
+      "tokens_seen": 2141847552
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000357010101010101,
+      "loss": 2.8455,
+      "theoretical_loss": 3.410970303667747,
+      "tokens_seen": 2141913088
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035698989898989896,
+      "loss": 2.8306,
+      "theoretical_loss": 3.410961718914839,
+      "tokens_seen": 2141978624
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000356969696969697,
+      "loss": 2.7306,
+      "theoretical_loss": 3.4109531344981283,
+      "tokens_seen": 2142044160
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000356949494949495,
+      "loss": 2.6625,
+      "theoretical_loss": 3.4109445504175917,
+      "tokens_seen": 2142109696
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035692929292929293,
+      "loss": 2.5678,
+      "theoretical_loss": 3.4109359666732053,
+      "tokens_seen": 2142175232
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003569090909090909,
+      "loss": 2.6553,
+      "theoretical_loss": 3.4109273832649456,
+      "tokens_seen": 2142240768
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035688888888888885,
+      "loss": 2.6125,
+      "theoretical_loss": 3.410918800192789,
+      "tokens_seen": 2142306304
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003568686868686869,
+      "loss": 2.6823,
+      "theoretical_loss": 3.4109102174567125,
+      "tokens_seen": 2142371840
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035684848484848487,
+      "loss": 2.6649,
+      "theoretical_loss": 3.4109016350566934,
+      "tokens_seen": 2142437376
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003568282828282828,
+      "loss": 2.7469,
+      "theoretical_loss": 3.410893052992706,
+      "tokens_seen": 2142502912
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003568080808080808,
+      "loss": 2.5226,
+      "theoretical_loss": 3.410884471264729,
+      "tokens_seen": 2142568448
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003567878787878788,
+      "loss": 2.3986,
+      "theoretical_loss": 3.410875889872738,
+      "tokens_seen": 2142633984
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003567676767676768,
+      "loss": 2.6273,
+      "theoretical_loss": 3.41086730881671,
+      "tokens_seen": 2142699520
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035674747474747476,
+      "loss": 2.5577,
+      "theoretical_loss": 3.4108587280966214,
+      "tokens_seen": 2142765056
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003567272727272727,
+      "loss": 2.5092,
+      "theoretical_loss": 3.4108501477124484,
+      "tokens_seen": 2142830592
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003567070707070707,
+      "loss": 2.4898,
+      "theoretical_loss": 3.4108415676641677,
+      "tokens_seen": 2142896128
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003566868686868687,
+      "loss": 2.7255,
+      "theoretical_loss": 3.4108329879517565,
+      "tokens_seen": 2142961664
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003566666666666667,
+      "loss": 2.6781,
+      "theoretical_loss": 3.4108244085751904,
+      "tokens_seen": 2143027200
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035664646464646466,
+      "loss": 2.6827,
+      "theoretical_loss": 3.4108158295344473,
+      "tokens_seen": 2143092736
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003566262626262626,
+      "loss": 2.7293,
+      "theoretical_loss": 3.4108072508295026,
+      "tokens_seen": 2143158272
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003566060606060606,
+      "loss": 2.4671,
+      "theoretical_loss": 3.4107986724603334,
+      "tokens_seen": 2143223808
+    },
+    {
+      "epoch": 0.29,
+      "objective/train/docs_used": 1180436,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.763584613800049,
+      "objective/train/theoretical_loss": 3.410790094426916,
+      "objective/train/tokens_used": 502148576,
+      "theoretical_loss": 3.410790094426916,
+      "tokens_seen": 2143289344
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003565858585858586,
+      "loss": 2.6251,
+      "theoretical_loss": 3.410790094426916,
+      "tokens_seen": 2143289344
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003565656565656566,
+      "loss": 2.7291,
+      "theoretical_loss": 3.4107815167292275,
+      "tokens_seen": 2143354880
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035654545454545455,
+      "loss": 2.4953,
+      "theoretical_loss": 3.410772939367244,
+      "tokens_seen": 2143420416
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003565252525252525,
+      "loss": 2.654,
+      "theoretical_loss": 3.4107643623409425,
+      "tokens_seen": 2143485952
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003565050505050505,
+      "loss": 2.6565,
+      "theoretical_loss": 3.410755785650299,
+      "tokens_seen": 2143551488
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003564848484848485,
+      "loss": 2.6607,
+      "theoretical_loss": 3.410747209295291,
+      "tokens_seen": 2143617024
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003564646464646465,
+      "loss": 2.7669,
+      "theoretical_loss": 3.410738633275894,
+      "tokens_seen": 2143682560
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035644444444444444,
+      "loss": 2.7056,
+      "theoretical_loss": 3.4107300575920854,
+      "tokens_seen": 2143748096
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035642424242424245,
+      "loss": 2.8577,
+      "theoretical_loss": 3.410721482243842,
+      "tokens_seen": 2143813632
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003564040404040404,
+      "loss": 2.6011,
+      "theoretical_loss": 3.4107129072311393,
+      "tokens_seen": 2143879168
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035638383838383837,
+      "loss": 2.4695,
+      "theoretical_loss": 3.410704332553955,
+      "tokens_seen": 2143944704
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003563636363636364,
+      "loss": 2.62,
+      "theoretical_loss": 3.410695758212266,
+      "tokens_seen": 2144010240
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035634343434343433,
+      "loss": 2.5418,
+      "theoretical_loss": 3.410687184206047,
+      "tokens_seen": 2144075776
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035632323232323234,
+      "loss": 2.6616,
+      "theoretical_loss": 3.4106786105352773,
+      "tokens_seen": 2144141312
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003563030303030303,
+      "loss": 2.6573,
+      "theoretical_loss": 3.410670037199931,
+      "tokens_seen": 2144206848
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035628282828282826,
+      "loss": 2.7227,
+      "theoretical_loss": 3.410661464199986,
+      "tokens_seen": 2144272384
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035626262626262627,
+      "loss": 2.9286,
+      "theoretical_loss": 3.4106528915354195,
+      "tokens_seen": 2144337920
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003562424242424243,
+      "loss": 2.7204,
+      "theoretical_loss": 3.410644319206207,
+      "tokens_seen": 2144403456
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035622222222222224,
+      "loss": 2.7795,
+      "theoretical_loss": 3.410635747212325,
+      "tokens_seen": 2144468992
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003562020202020202,
+      "loss": 2.6493,
+      "theoretical_loss": 3.4106271755537514,
+      "tokens_seen": 2144534528
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035618181818181815,
+      "loss": 2.7353,
+      "theoretical_loss": 3.4106186042304616,
+      "tokens_seen": 2144600064
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035616161616161616,
+      "loss": 2.6767,
+      "theoretical_loss": 3.410610033242433,
+      "tokens_seen": 2144665600
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003561414141414142,
+      "loss": 2.6527,
+      "theoretical_loss": 3.410601462589642,
+      "tokens_seen": 2144731136
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035612121212121213,
+      "loss": 2.5647,
+      "theoretical_loss": 3.4105928922720654,
+      "tokens_seen": 2144796672
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003561010101010101,
+      "loss": 2.6571,
+      "theoretical_loss": 3.4105843222896794,
+      "tokens_seen": 2144862208
+    },
+    {
+      "epoch": 0.29,
+      "objective/train/docs_used": 1181645,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.3639755249023438,
+      "objective/train/theoretical_loss": 3.410575752642461,
+      "objective/train/tokens_used": 503786976,
+      "theoretical_loss": 3.410575752642461,
+      "tokens_seen": 2144927744
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00035608080808080804,
+      "loss": 2.7949,
+      "theoretical_loss": 3.410575752642461,
+      "tokens_seen": 2144927744
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003560606060606061,
+      "loss": 2.5403,
+      "theoretical_loss": 3.4105671833303868,
+      "tokens_seen": 2144993280
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035604040404040407,
+      "loss": 2.71,
+      "theoretical_loss": 3.4105586143534334,
+      "tokens_seen": 2145058816
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000356020202020202,
+      "loss": 2.5801,
+      "theoretical_loss": 3.4105500457115774,
+      "tokens_seen": 2145124352
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000356,
+      "loss": 2.5069,
+      "theoretical_loss": 3.410541477404796,
+      "tokens_seen": 2145189888
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000355979797979798,
+      "loss": 2.815,
+      "theoretical_loss": 3.4105329094330648,
+      "tokens_seen": 2145255424
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000355959595959596,
+      "loss": 2.5818,
+      "theoretical_loss": 3.4105243417963615,
+      "tokens_seen": 2145320960
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035593939393939396,
+      "loss": 2.4357,
+      "theoretical_loss": 3.4105157744946624,
+      "tokens_seen": 2145386496
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003559191919191919,
+      "loss": 2.6195,
+      "theoretical_loss": 3.4105072075279432,
+      "tokens_seen": 2145452032
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035589898989898987,
+      "loss": 2.7188,
+      "theoretical_loss": 3.410498640896183,
+      "tokens_seen": 2145517568
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035587878787878794,
+      "loss": 2.6192,
+      "theoretical_loss": 3.410490074599356,
+      "tokens_seen": 2145583104
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003558585858585859,
+      "loss": 2.4677,
+      "theoretical_loss": 3.4104815086374396,
+      "tokens_seen": 2145648640
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035583838383838385,
+      "loss": 2.7138,
+      "theoretical_loss": 3.4104729430104115,
+      "tokens_seen": 2145714176
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003558181818181818,
+      "loss": 2.8676,
+      "theoretical_loss": 3.410464377718247,
+      "tokens_seen": 2145779712
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035579797979797977,
+      "loss": 2.7443,
+      "theoretical_loss": 3.4104558127609232,
+      "tokens_seen": 2145845248
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035577777777777783,
+      "loss": 2.9617,
+      "theoretical_loss": 3.4104472481384174,
+      "tokens_seen": 2145910784
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003557575757575758,
+      "loss": 2.5756,
+      "theoretical_loss": 3.410438683850706,
+      "tokens_seen": 2145976320
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035573737373737374,
+      "loss": 2.8851,
+      "theoretical_loss": 3.4104301198977653,
+      "tokens_seen": 2146041856
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003557171717171717,
+      "loss": 2.8399,
+      "theoretical_loss": 3.410421556279572,
+      "tokens_seen": 2146107392
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003556969696969697,
+      "loss": 2.5007,
+      "theoretical_loss": 3.410412992996103,
+      "tokens_seen": 2146172928
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003556767676767677,
+      "loss": 2.5867,
+      "theoretical_loss": 3.4104044300473353,
+      "tokens_seen": 2146238464
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003556565656565657,
+      "loss": 2.6169,
+      "theoretical_loss": 3.410395867433245,
+      "tokens_seen": 2146304000
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035563636363636364,
+      "loss": 2.5714,
+      "theoretical_loss": 3.4103873051538094,
+      "tokens_seen": 2146369536
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003556161616161616,
+      "loss": 2.6238,
+      "theoretical_loss": 3.410378743209005,
+      "tokens_seen": 2146435072
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003555959595959596,
+      "loss": 2.685,
+      "theoretical_loss": 3.4103701815988083,
+      "tokens_seen": 2146500608
+    },
+    {
+      "epoch": 0.3,
+      "objective/train/docs_used": 1182191,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.627549886703491,
+      "objective/train/theoretical_loss": 3.4103616203231955,
+      "objective/train/tokens_used": 505425376,
+      "theoretical_loss": 3.4103616203231955,
+      "tokens_seen": 2146566144
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003555757575757576,
+      "loss": 2.7703,
+      "theoretical_loss": 3.4103616203231955,
+      "tokens_seen": 2146566144
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035555555555555557,
+      "loss": 2.5253,
+      "theoretical_loss": 3.4103530593821447,
+      "tokens_seen": 2146631680
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035553535353535353,
+      "loss": 2.9456,
+      "theoretical_loss": 3.4103444987756317,
+      "tokens_seen": 2146697216
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035551515151515154,
+      "loss": 2.5637,
+      "theoretical_loss": 3.4103359385036334,
+      "tokens_seen": 2146762752
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003554949494949495,
+      "loss": 2.5552,
+      "theoretical_loss": 3.4103273785661266,
+      "tokens_seen": 2146828288
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003554747474747475,
+      "loss": 2.7312,
+      "theoretical_loss": 3.4103188189630878,
+      "tokens_seen": 2146893824
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035545454545454547,
+      "loss": 2.7148,
+      "theoretical_loss": 3.4103102596944934,
+      "tokens_seen": 2146959360
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003554343434343434,
+      "loss": 2.9333,
+      "theoretical_loss": 3.4103017007603205,
+      "tokens_seen": 2147024896
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035541414141414143,
+      "loss": 2.5406,
+      "theoretical_loss": 3.4102931421605467,
+      "tokens_seen": 2147090432
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003553939393939394,
+      "loss": 2.8978,
+      "theoretical_loss": 3.410284583895147,
+      "tokens_seen": 2147155968
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003553737373737374,
+      "loss": 2.7002,
+      "theoretical_loss": 3.4102760259641,
+      "tokens_seen": 2147221504
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035535353535353536,
+      "loss": 2.7219,
+      "theoretical_loss": 3.4102674683673806,
+      "tokens_seen": 2147287040
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003553333333333333,
+      "loss": 2.5009,
+      "theoretical_loss": 3.4102589111049664,
+      "tokens_seen": 2147352576
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003553131313131313,
+      "loss": 2.491,
+      "theoretical_loss": 3.4102503541768345,
+      "tokens_seen": 2147418112
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003552929292929293,
+      "loss": 2.7664,
+      "theoretical_loss": 3.4102417975829615,
+      "tokens_seen": 2147483648
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003552727272727273,
+      "loss": 2.6139,
+      "theoretical_loss": 3.4102332413233234,
+      "tokens_seen": 2147549184
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035525252525252525,
+      "loss": 2.7421,
+      "theoretical_loss": 3.410224685397898,
+      "tokens_seen": 2147614720
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035523232323232326,
+      "loss": 2.6755,
+      "theoretical_loss": 3.4102161298066607,
+      "tokens_seen": 2147680256
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003552121212121212,
+      "loss": 2.634,
+      "theoretical_loss": 3.41020757454959,
+      "tokens_seen": 2147745792
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003551919191919192,
+      "loss": 2.6799,
+      "theoretical_loss": 3.410199019626661,
+      "tokens_seen": 2147811328
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003551717171717172,
+      "loss": 2.7408,
+      "theoretical_loss": 3.410190465037852,
+      "tokens_seen": 2147876864
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035515151515151514,
+      "loss": 2.5752,
+      "theoretical_loss": 3.410181910783138,
+      "tokens_seen": 2147942400
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035513131313131315,
+      "loss": 2.582,
+      "theoretical_loss": 3.4101733568624972,
+      "tokens_seen": 2148007936
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003551111111111111,
+      "loss": 2.6301,
+      "theoretical_loss": 3.4101648032759053,
+      "tokens_seen": 2148073472
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035509090909090907,
+      "loss": 2.8678,
+      "theoretical_loss": 3.4101562500233404,
+      "tokens_seen": 2148139008
+    },
+    {
+      "epoch": 0.3,
+      "objective/train/docs_used": 1183255,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.5413014888763428,
+      "objective/train/theoretical_loss": 3.4101476971047777,
+      "objective/train/tokens_used": 507063776,
+      "theoretical_loss": 3.4101476971047777,
+      "tokens_seen": 2148204544
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003550707070707071,
+      "loss": 2.5406,
+      "theoretical_loss": 3.4101476971047777,
+      "tokens_seen": 2148204544
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003550505050505051,
+      "loss": 2.511,
+      "theoretical_loss": 3.4101391445201954,
+      "tokens_seen": 2148270080
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035503030303030305,
+      "loss": 2.6562,
+      "theoretical_loss": 3.410130592269569,
+      "tokens_seen": 2148335616
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000355010101010101,
+      "loss": 2.7914,
+      "theoretical_loss": 3.410122040352877,
+      "tokens_seen": 2148401152
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035498989898989896,
+      "loss": 2.6844,
+      "theoretical_loss": 3.4101134887700937,
+      "tokens_seen": 2148466688
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035496969696969697,
+      "loss": 2.5422,
+      "theoretical_loss": 3.410104937521198,
+      "tokens_seen": 2148532224
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000354949494949495,
+      "loss": 2.7876,
+      "theoretical_loss": 3.410096386606166,
+      "tokens_seen": 2148597760
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035492929292929294,
+      "loss": 2.7034,
+      "theoretical_loss": 3.4100878360249736,
+      "tokens_seen": 2148663296
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003549090909090909,
+      "loss": 2.4995,
+      "theoretical_loss": 3.4100792857775994,
+      "tokens_seen": 2148728832
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035488888888888885,
+      "loss": 2.6483,
+      "theoretical_loss": 3.4100707358640188,
+      "tokens_seen": 2148794368
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003548686868686869,
+      "loss": 2.775,
+      "theoretical_loss": 3.4100621862842084,
+      "tokens_seen": 2148859904
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003548484848484849,
+      "loss": 2.5338,
+      "theoretical_loss": 3.4100536370381462,
+      "tokens_seen": 2148925440
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035482828282828283,
+      "loss": 2.9031,
+      "theoretical_loss": 3.410045088125808,
+      "tokens_seen": 2148990976
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003548080808080808,
+      "loss": 2.5603,
+      "theoretical_loss": 3.4100365395471712,
+      "tokens_seen": 2149056512
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035478787878787875,
+      "loss": 2.5336,
+      "theoretical_loss": 3.4100279913022122,
+      "tokens_seen": 2149122048
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003547676767676768,
+      "loss": 2.7277,
+      "theoretical_loss": 3.410019443390908,
+      "tokens_seen": 2149187584
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035474747474747477,
+      "loss": 2.7138,
+      "theoretical_loss": 3.410010895813236,
+      "tokens_seen": 2149253120
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003547272727272727,
+      "loss": 2.5253,
+      "theoretical_loss": 3.4100023485691713,
+      "tokens_seen": 2149318656
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003547070707070707,
+      "loss": 2.7073,
+      "theoretical_loss": 3.4099938016586924,
+      "tokens_seen": 2149384192
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003546868686868687,
+      "loss": 2.5695,
+      "theoretical_loss": 3.409985255081775,
+      "tokens_seen": 2149449728
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003546666666666667,
+      "loss": 2.5718,
+      "theoretical_loss": 3.4099767088383968,
+      "tokens_seen": 2149515264
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035464646464646466,
+      "loss": 2.6058,
+      "theoretical_loss": 3.409968162928534,
+      "tokens_seen": 2149580800
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003546262626262626,
+      "loss": 2.6631,
+      "theoretical_loss": 3.409959617352164,
+      "tokens_seen": 2149646336
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003546060606060606,
+      "loss": 2.7089,
+      "theoretical_loss": 3.4099510721092625,
+      "tokens_seen": 2149711872
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003545858585858586,
+      "loss": 2.356,
+      "theoretical_loss": 3.4099425271998074,
+      "tokens_seen": 2149777408
+    },
+    {
+      "epoch": 0.3,
+      "objective/train/docs_used": 1183881,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.526425361633301,
+      "objective/train/theoretical_loss": 3.4099339826237753,
+      "objective/train/tokens_used": 508702176,
+      "theoretical_loss": 3.4099339826237753,
+      "tokens_seen": 2149842944
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003545656565656566,
+      "loss": 2.6612,
+      "theoretical_loss": 3.4099339826237753,
+      "tokens_seen": 2149842944
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035454545454545455,
+      "loss": 2.6441,
+      "theoretical_loss": 3.409925438381143,
+      "tokens_seen": 2149908480
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003545252525252525,
+      "loss": 2.4833,
+      "theoretical_loss": 3.4099168944718867,
+      "tokens_seen": 2149974016
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003545050505050505,
+      "loss": 2.6153,
+      "theoretical_loss": 3.409908350895984,
+      "tokens_seen": 2150039552
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003544848484848485,
+      "loss": 2.5933,
+      "theoretical_loss": 3.409899807653412,
+      "tokens_seen": 2150105088
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003544646464646465,
+      "loss": 2.4394,
+      "theoretical_loss": 3.4098912647441466,
+      "tokens_seen": 2150170624
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035444444444444445,
+      "loss": 2.4475,
+      "theoretical_loss": 3.409882722168165,
+      "tokens_seen": 2150236160
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003544242424242424,
+      "loss": 2.7372,
+      "theoretical_loss": 3.4098741799254446,
+      "tokens_seen": 2150301696
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003544040404040404,
+      "loss": 2.642,
+      "theoretical_loss": 3.409865638015961,
+      "tokens_seen": 2150367232
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035438383838383837,
+      "loss": 2.6953,
+      "theoretical_loss": 3.4098570964396924,
+      "tokens_seen": 2150432768
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003543636363636364,
+      "loss": 2.8193,
+      "theoretical_loss": 3.4098485551966147,
+      "tokens_seen": 2150498304
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035434343434343434,
+      "loss": 2.6868,
+      "theoretical_loss": 3.409840014286705,
+      "tokens_seen": 2150563840
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035432323232323235,
+      "loss": 2.6162,
+      "theoretical_loss": 3.409831473709941,
+      "tokens_seen": 2150629376
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003543030303030303,
+      "loss": 2.8,
+      "theoretical_loss": 3.409822933466298,
+      "tokens_seen": 2150694912
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003542828282828283,
+      "loss": 2.6104,
+      "theoretical_loss": 3.409814393555754,
+      "tokens_seen": 2150760448
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003542626262626263,
+      "loss": 2.7018,
+      "theoretical_loss": 3.4098058539782854,
+      "tokens_seen": 2150825984
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035424242424242423,
+      "loss": 2.5429,
+      "theoretical_loss": 3.4097973147338694,
+      "tokens_seen": 2150891520
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035422222222222224,
+      "loss": 2.4372,
+      "theoretical_loss": 3.4097887758224825,
+      "tokens_seen": 2150957056
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003542020202020202,
+      "loss": 2.6098,
+      "theoretical_loss": 3.4097802372441013,
+      "tokens_seen": 2151022592
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003541818181818182,
+      "loss": 2.6754,
+      "theoretical_loss": 3.4097716989987035,
+      "tokens_seen": 2151088128
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035416161616161617,
+      "loss": 2.4958,
+      "theoretical_loss": 3.409763161086266,
+      "tokens_seen": 2151153664
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003541414141414142,
+      "loss": 2.6487,
+      "theoretical_loss": 3.4097546235067644,
+      "tokens_seen": 2151219200
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035412121212121214,
+      "loss": 2.8716,
+      "theoretical_loss": 3.409746086260177,
+      "tokens_seen": 2151284736
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003541010101010101,
+      "loss": 2.6091,
+      "theoretical_loss": 3.409737549346479,
+      "tokens_seen": 2151350272
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003540808080808081,
+      "loss": 2.71,
+      "theoretical_loss": 3.4097290127656494,
+      "tokens_seen": 2151415808
+    },
+    {
+      "epoch": 0.3,
+      "objective/train/docs_used": 1184815,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.6544196605682373,
+      "objective/train/theoretical_loss": 3.409720476517664,
+      "objective/train/tokens_used": 510340576,
+      "theoretical_loss": 3.409720476517664,
+      "tokens_seen": 2151481344
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035406060606060606,
+      "loss": 2.6166,
+      "theoretical_loss": 3.409720476517664,
+      "tokens_seen": 2151481344
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035404040404040407,
+      "loss": 2.7207,
+      "theoretical_loss": 3.4097119406024996,
+      "tokens_seen": 2151546880
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035402020202020203,
+      "loss": 2.5904,
+      "theoretical_loss": 3.4097034050201334,
+      "tokens_seen": 2151612416
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000354,
+      "loss": 2.6907,
+      "theoretical_loss": 3.4096948697705414,
+      "tokens_seen": 2151677952
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000353979797979798,
+      "loss": 2.6082,
+      "theoretical_loss": 3.4096863348537014,
+      "tokens_seen": 2151743488
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000353959595959596,
+      "loss": 2.7023,
+      "theoretical_loss": 3.4096778002695904,
+      "tokens_seen": 2151809024
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035393939393939396,
+      "loss": 2.6989,
+      "theoretical_loss": 3.409669266018185,
+      "tokens_seen": 2151874560
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003539191919191919,
+      "loss": 2.5781,
+      "theoretical_loss": 3.4096607320994616,
+      "tokens_seen": 2151940096
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003538989898989899,
+      "loss": 2.4732,
+      "theoretical_loss": 3.409652198513398,
+      "tokens_seen": 2152005632
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003538787878787879,
+      "loss": 2.632,
+      "theoretical_loss": 3.4096436652599706,
+      "tokens_seen": 2152071168
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003538585858585859,
+      "loss": 2.7633,
+      "theoretical_loss": 3.4096351323391563,
+      "tokens_seen": 2152136704
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035383838383838386,
+      "loss": 2.5427,
+      "theoretical_loss": 3.409626599750932,
+      "tokens_seen": 2152202240
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003538181818181818,
+      "loss": 2.627,
+      "theoretical_loss": 3.4096180674952747,
+      "tokens_seen": 2152267776
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035379797979797977,
+      "loss": 2.4213,
+      "theoretical_loss": 3.4096095355721614,
+      "tokens_seen": 2152333312
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035377777777777784,
+      "loss": 2.4559,
+      "theoretical_loss": 3.409601003981569,
+      "tokens_seen": 2152398848
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003537575757575758,
+      "loss": 2.4325,
+      "theoretical_loss": 3.409592472723474,
+      "tokens_seen": 2152464384
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035373737373737375,
+      "loss": 2.6978,
+      "theoretical_loss": 3.4095839417978544,
+      "tokens_seen": 2152529920
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003537171717171717,
+      "loss": 2.6345,
+      "theoretical_loss": 3.4095754112046857,
+      "tokens_seen": 2152595456
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035369696969696966,
+      "loss": 2.704,
+      "theoretical_loss": 3.409566880943945,
+      "tokens_seen": 2152660992
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035367676767676773,
+      "loss": 2.4888,
+      "theoretical_loss": 3.409558351015611,
+      "tokens_seen": 2152726528
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003536565656565657,
+      "loss": 2.4636,
+      "theoretical_loss": 3.4095498214196587,
+      "tokens_seen": 2152792064
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035363636363636364,
+      "loss": 2.5985,
+      "theoretical_loss": 3.4095412921560655,
+      "tokens_seen": 2152857600
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003536161616161616,
+      "loss": 2.547,
+      "theoretical_loss": 3.409532763224809,
+      "tokens_seen": 2152923136
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035359595959595956,
+      "loss": 2.6535,
+      "theoretical_loss": 3.4095242346258656,
+      "tokens_seen": 2152988672
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003535757575757576,
+      "loss": 2.6403,
+      "theoretical_loss": 3.409515706359212,
+      "tokens_seen": 2153054208
+    },
+    {
+      "debugging/Self-BLEU-5": 0.39580322091234643,
+      "debugging/distinct-1-grams": 0.797421315948946,
+      "debugging/distinct-2-grams": 0.9612328752077172,
+      "debugging/entropy-1-grams": 5.484895349701254,
+      "debugging/entropy-2-grams": 6.169724221975992,
+      "debugging/length": 483.25,
+      "debugging/num_segments": 8,
+      "epoch": 0.3,
+      "objective/train/docs_used": 1185854,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.8254241943359375,
+      "objective/train/theoretical_loss": 3.4095071784248256,
+      "objective/train/tokens_used": 511978976,
+      "theoretical_loss": 3.4095071784248256,
+      "tokens_seen": 2153119744
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003535555555555556,
+      "loss": 2.5479,
+      "theoretical_loss": 3.4095071784248256,
+      "tokens_seen": 2153119744
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035353535353535354,
+      "loss": 2.6056,
+      "theoretical_loss": 3.409498650822683,
+      "tokens_seen": 2153185280
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003535151515151515,
+      "loss": 2.6974,
+      "theoretical_loss": 3.4094901235527613,
+      "tokens_seen": 2153250816
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003534949494949495,
+      "loss": 2.3753,
+      "theoretical_loss": 3.409481596615038,
+      "tokens_seen": 2153316352
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003534747474747475,
+      "loss": 2.5776,
+      "theoretical_loss": 3.4094730700094886,
+      "tokens_seen": 2153381888
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035345454545454547,
+      "loss": 2.5597,
+      "theoretical_loss": 3.4094645437360915,
+      "tokens_seen": 2153447424
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035343434343434343,
+      "loss": 2.773,
+      "theoretical_loss": 3.409456017794823,
+      "tokens_seen": 2153512960
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003534141414141414,
+      "loss": 2.6986,
+      "theoretical_loss": 3.4094474921856603,
+      "tokens_seen": 2153578496
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003533939393939394,
+      "loss": 2.6894,
+      "theoretical_loss": 3.40943896690858,
+      "tokens_seen": 2153644032
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003533737373737374,
+      "loss": 2.916,
+      "theoretical_loss": 3.4094304419635595,
+      "tokens_seen": 2153709568
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035335353535353536,
+      "loss": 2.6452,
+      "theoretical_loss": 3.409421917350575,
+      "tokens_seen": 2153775104
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003533333333333333,
+      "loss": 2.6145,
+      "theoretical_loss": 3.409413393069605,
+      "tokens_seen": 2153840640
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035331313131313133,
+      "loss": 2.8686,
+      "theoretical_loss": 3.4094048691206247,
+      "tokens_seen": 2153906176
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003532929292929293,
+      "loss": 2.8178,
+      "theoretical_loss": 3.4093963455036125,
+      "tokens_seen": 2153971712
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003532727272727273,
+      "loss": 2.6879,
+      "theoretical_loss": 3.4093878222185445,
+      "tokens_seen": 2154037248
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035325252525252526,
+      "loss": 2.623,
+      "theoretical_loss": 3.4093792992653977,
+      "tokens_seen": 2154102784
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003532323232323232,
+      "loss": 2.7944,
+      "theoretical_loss": 3.409370776644149,
+      "tokens_seen": 2154168320
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003532121212121212,
+      "loss": 2.7522,
+      "theoretical_loss": 3.409362254354776,
+      "tokens_seen": 2154233856
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003531919191919192,
+      "loss": 2.5473,
+      "theoretical_loss": 3.409353732397255,
+      "tokens_seen": 2154299392
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003531717171717172,
+      "loss": 2.6862,
+      "theoretical_loss": 3.409345210771564,
+      "tokens_seen": 2154364928
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035315151515151515,
+      "loss": 2.6817,
+      "theoretical_loss": 3.409336689477679,
+      "tokens_seen": 2154430464
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035313131313131316,
+      "loss": 2.5026,
+      "theoretical_loss": 3.409328168515577,
+      "tokens_seen": 2154496000
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003531111111111111,
+      "loss": 2.7158,
+      "theoretical_loss": 3.4093196478852352,
+      "tokens_seen": 2154561536
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003530909090909091,
+      "loss": 2.5761,
+      "theoretical_loss": 3.4093111275866312,
+      "tokens_seen": 2154627072
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003530707070707071,
+      "loss": 2.7904,
+      "theoretical_loss": 3.4093026076197415,
+      "tokens_seen": 2154692608
+    },
+    {
+      "epoch": 0.3,
+      "objective/train/docs_used": 1186317,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.778775691986084,
+      "objective/train/theoretical_loss": 3.4092940879845424,
+      "objective/train/tokens_used": 513617376,
+      "theoretical_loss": 3.4092940879845424,
+      "tokens_seen": 2154758144
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035305050505050504,
+      "loss": 2.5932,
+      "theoretical_loss": 3.4092940879845424,
+      "tokens_seen": 2154758144
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035303030303030305,
+      "loss": 2.6707,
+      "theoretical_loss": 3.409285568681012,
+      "tokens_seen": 2154823680
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000353010101010101,
+      "loss": 2.5527,
+      "theoretical_loss": 3.409277049709127,
+      "tokens_seen": 2154889216
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035298989898989897,
+      "loss": 2.5178,
+      "theoretical_loss": 3.409268531068864,
+      "tokens_seen": 2154954752
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000352969696969697,
+      "loss": 2.6367,
+      "theoretical_loss": 3.4092600127602006,
+      "tokens_seen": 2155020288
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000352949494949495,
+      "loss": 2.5147,
+      "theoretical_loss": 3.409251494783113,
+      "tokens_seen": 2155085824
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035292929292929295,
+      "loss": 2.6731,
+      "theoretical_loss": 3.4092429771375787,
+      "tokens_seen": 2155151360
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003529090909090909,
+      "loss": 2.5244,
+      "theoretical_loss": 3.4092344598235753,
+      "tokens_seen": 2155216896
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035288888888888886,
+      "loss": 2.6141,
+      "theoretical_loss": 3.4092259428410783,
+      "tokens_seen": 2155282432
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035286868686868687,
+      "loss": 2.6846,
+      "theoretical_loss": 3.4092174261900663,
+      "tokens_seen": 2155347968
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003528484848484849,
+      "loss": 2.5751,
+      "theoretical_loss": 3.409208909870516,
+      "tokens_seen": 2155413504
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035282828282828284,
+      "loss": 2.625,
+      "theoretical_loss": 3.4092003938824034,
+      "tokens_seen": 2155479040
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003528080808080808,
+      "loss": 2.5676,
+      "theoretical_loss": 3.4091918782257062,
+      "tokens_seen": 2155544576
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035278787878787875,
+      "loss": 2.5603,
+      "theoretical_loss": 3.409183362900402,
+      "tokens_seen": 2155610112
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003527676767676768,
+      "loss": 2.6779,
+      "theoretical_loss": 3.409174847906467,
+      "tokens_seen": 2155675648
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003527474747474748,
+      "loss": 2.597,
+      "theoretical_loss": 3.409166333243878,
+      "tokens_seen": 2155741184
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035272727272727273,
+      "loss": 2.6806,
+      "theoretical_loss": 3.409157818912613,
+      "tokens_seen": 2155806720
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003527070707070707,
+      "loss": 2.6874,
+      "theoretical_loss": 3.4091493049126487,
+      "tokens_seen": 2155872256
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035268686868686864,
+      "loss": 2.6295,
+      "theoretical_loss": 3.4091407912439617,
+      "tokens_seen": 2155937792
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003526666666666667,
+      "loss": 2.5543,
+      "theoretical_loss": 3.4091322779065294,
+      "tokens_seen": 2156003328
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035264646464646467,
+      "loss": 2.5731,
+      "theoretical_loss": 3.409123764900329,
+      "tokens_seen": 2156068864
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003526262626262626,
+      "loss": 2.6069,
+      "theoretical_loss": 3.4091152522253374,
+      "tokens_seen": 2156134400
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003526060606060606,
+      "loss": 2.6663,
+      "theoretical_loss": 3.409106739881531,
+      "tokens_seen": 2156199936
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035258585858585865,
+      "loss": 2.4908,
+      "theoretical_loss": 3.4090982278688875,
+      "tokens_seen": 2156265472
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003525656565656566,
+      "loss": 2.5468,
+      "theoretical_loss": 3.4090897161873848,
+      "tokens_seen": 2156331008
+    },
+    {
+      "epoch": 0.3,
+      "objective/train/docs_used": 1187470,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.505795955657959,
+      "objective/train/theoretical_loss": 3.409081204836998,
+      "objective/train/tokens_used": 515255776,
+      "theoretical_loss": 3.409081204836998,
+      "tokens_seen": 2156396544
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035254545454545456,
+      "loss": 2.6541,
+      "theoretical_loss": 3.409081204836998,
+      "tokens_seen": 2156396544
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003525252525252525,
+      "loss": 2.4961,
+      "theoretical_loss": 3.4090726938177056,
+      "tokens_seen": 2156462080
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003525050505050505,
+      "loss": 2.595,
+      "theoretical_loss": 3.409064183129484,
+      "tokens_seen": 2156527616
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035248484848484854,
+      "loss": 2.6051,
+      "theoretical_loss": 3.409055672772311,
+      "tokens_seen": 2156593152
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003524646464646465,
+      "loss": 2.6744,
+      "theoretical_loss": 3.4090471627461625,
+      "tokens_seen": 2156658688
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035244444444444445,
+      "loss": 2.4898,
+      "theoretical_loss": 3.409038653051017,
+      "tokens_seen": 2156724224
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003524242424242424,
+      "loss": 2.6977,
+      "theoretical_loss": 3.4090301436868504,
+      "tokens_seen": 2156789760
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003524040404040404,
+      "loss": 2.5083,
+      "theoretical_loss": 3.40902163465364,
+      "tokens_seen": 2156855296
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035238383838383843,
+      "loss": 2.6723,
+      "theoretical_loss": 3.4090131259513634,
+      "tokens_seen": 2156920832
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003523636363636364,
+      "loss": 2.6959,
+      "theoretical_loss": 3.4090046175799973,
+      "tokens_seen": 2156986368
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035234343434343434,
+      "loss": 2.4903,
+      "theoretical_loss": 3.408996109539519,
+      "tokens_seen": 2157051904
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003523232323232323,
+      "loss": 2.4713,
+      "theoretical_loss": 3.408987601829905,
+      "tokens_seen": 2157117440
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003523030303030303,
+      "loss": 2.7848,
+      "theoretical_loss": 3.4089790944511327,
+      "tokens_seen": 2157182976
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003522828282828283,
+      "loss": 2.8257,
+      "theoretical_loss": 3.4089705874031795,
+      "tokens_seen": 2157248512
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003522626262626263,
+      "loss": 2.4483,
+      "theoretical_loss": 3.408962080686022,
+      "tokens_seen": 2157314048
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035224242424242424,
+      "loss": 2.8532,
+      "theoretical_loss": 3.4089535742996375,
+      "tokens_seen": 2157379584
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035222222222222225,
+      "loss": 2.693,
+      "theoretical_loss": 3.4089450682440035,
+      "tokens_seen": 2157445120
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003522020202020202,
+      "loss": 2.6485,
+      "theoretical_loss": 3.408936562519096,
+      "tokens_seen": 2157510656
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003521818181818182,
+      "loss": 2.6114,
+      "theoretical_loss": 3.408928057124893,
+      "tokens_seen": 2157576192
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003521616161616162,
+      "loss": 2.567,
+      "theoretical_loss": 3.4089195520613718,
+      "tokens_seen": 2157641728
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035214141414141413,
+      "loss": 2.6924,
+      "theoretical_loss": 3.408911047328509,
+      "tokens_seen": 2157707264
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035212121212121214,
+      "loss": 2.4765,
+      "theoretical_loss": 3.4089025429262816,
+      "tokens_seen": 2157772800
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003521010101010101,
+      "loss": 2.6886,
+      "theoretical_loss": 3.408894038854667,
+      "tokens_seen": 2157838336
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003520808080808081,
+      "loss": 2.5828,
+      "theoretical_loss": 3.4088855351136425,
+      "tokens_seen": 2157903872
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035206060606060607,
+      "loss": 2.507,
+      "theoretical_loss": 3.4088770317031845,
+      "tokens_seen": 2157969408
+    },
+    {
+      "epoch": 0.3,
+      "objective/train/docs_used": 1188127,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.4245564937591553,
+      "objective/train/theoretical_loss": 3.4088685286232705,
+      "objective/train/tokens_used": 516894176,
+      "theoretical_loss": 3.4088685286232705,
+      "tokens_seen": 2158034944
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000352040404040404,
+      "loss": 2.5853,
+      "theoretical_loss": 3.4088685286232705,
+      "tokens_seen": 2158034944
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035202020202020203,
+      "loss": 2.6205,
+      "theoretical_loss": 3.408860025873878,
+      "tokens_seen": 2158100480
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000352,
+      "loss": 2.6775,
+      "theoretical_loss": 3.4088515234549837,
+      "tokens_seen": 2158166016
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000351979797979798,
+      "loss": 2.6753,
+      "theoretical_loss": 3.4088430213665646,
+      "tokens_seen": 2158231552
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035195959595959596,
+      "loss": 2.7198,
+      "theoretical_loss": 3.408834519608598,
+      "tokens_seen": 2158297088
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035193939393939397,
+      "loss": 2.6947,
+      "theoretical_loss": 3.408826018181061,
+      "tokens_seen": 2158362624
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003519191919191919,
+      "loss": 2.816,
+      "theoretical_loss": 3.408817517083931,
+      "tokens_seen": 2158428160
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003518989898989899,
+      "loss": 2.4772,
+      "theoretical_loss": 3.4088090163171847,
+      "tokens_seen": 2158493696
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003518787878787879,
+      "loss": 2.6107,
+      "theoretical_loss": 3.4088005158807992,
+      "tokens_seen": 2158559232
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035185858585858585,
+      "loss": 2.6709,
+      "theoretical_loss": 3.408792015774752,
+      "tokens_seen": 2158624768
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035183838383838386,
+      "loss": 2.3833,
+      "theoretical_loss": 3.40878351599902,
+      "tokens_seen": 2158690304
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003518181818181818,
+      "loss": 2.6363,
+      "theoretical_loss": 3.408775016553581,
+      "tokens_seen": 2158755840
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003517979797979798,
+      "loss": 2.389,
+      "theoretical_loss": 3.408766517438411,
+      "tokens_seen": 2158821376
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003517777777777778,
+      "loss": 2.5093,
+      "theoretical_loss": 3.408758018653488,
+      "tokens_seen": 2158886912
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003517575757575758,
+      "loss": 2.6983,
+      "theoretical_loss": 3.4087495201987883,
+      "tokens_seen": 2158952448
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035173737373737376,
+      "loss": 2.3098,
+      "theoretical_loss": 3.40874102207429,
+      "tokens_seen": 2159017984
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003517171717171717,
+      "loss": 2.6428,
+      "theoretical_loss": 3.4087325242799693,
+      "tokens_seen": 2159083520
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035169696969696967,
+      "loss": 2.5138,
+      "theoretical_loss": 3.4087240268158046,
+      "tokens_seen": 2159149056
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003516767676767677,
+      "loss": 2.6145,
+      "theoretical_loss": 3.4087155296817717,
+      "tokens_seen": 2159214592
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003516565656565657,
+      "loss": 2.631,
+      "theoretical_loss": 3.4087070328778486,
+      "tokens_seen": 2159280128
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035163636363636365,
+      "loss": 2.6058,
+      "theoretical_loss": 3.4086985364040125,
+      "tokens_seen": 2159345664
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003516161616161616,
+      "loss": 2.3295,
+      "theoretical_loss": 3.40869004026024,
+      "tokens_seen": 2159411200
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035159595959595956,
+      "loss": 2.6824,
+      "theoretical_loss": 3.408681544446509,
+      "tokens_seen": 2159476736
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003515757575757576,
+      "loss": 2.4669,
+      "theoretical_loss": 3.4086730489627954,
+      "tokens_seen": 2159542272
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003515555555555556,
+      "loss": 2.8928,
+      "theoretical_loss": 3.4086645538090776,
+      "tokens_seen": 2159607808
+    },
+    {
+      "epoch": 0.3,
+      "objective/train/docs_used": 1189561,
+      "objective/train/instantaneous_batch_size": 16,
+      "objective/train/instantaneous_microbatch_size": 16384,
+      "objective/train/original_loss": 2.801823139190674,
+      "objective/train/theoretical_loss": 3.4086560589853327,
+      "objective/train/tokens_used": 518532576,
+      "theoretical_loss": 3.4086560589853327,
+      "tokens_seen": 2159673344
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035153535353535354,
+      "loss": 2.7104,
+      "theoretical_loss": 3.4086560589853327,
+      "tokens_seen": 2159673344
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003515151515151515,
+      "loss": 2.534,
+      "theoretical_loss": 3.4086475644915373,
+      "tokens_seen": 2159738880
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035149494949494945,
+      "loss": 2.6664,
+      "theoretical_loss": 3.4086390703276686,
+      "tokens_seen": 2159804416
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003514747474747475,
+      "loss": 2.5441,
+      "theoretical_loss": 3.408630576493704,
+      "tokens_seen": 2159869952
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003514545454545455,
+      "loss": 2.495,
+      "theoretical_loss": 3.4086220829896208,
+      "tokens_seen": 2159935488
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00035143434343434343,
+      "loss": 2.7479,
+      "theoretical_loss": 3.408613589815396,
+      "tokens_seen": 2160001024
+    }
+  ],
+  "max_steps": 25000,
+  "num_train_epochs": 9223372036854775807,
+  "total_flos": 2.5435211563008e+17,
+  "trial_name": null,
+  "trial_params": null
+}