diff --git "a/checkpoint-3000/trainer_state.json" "b/checkpoint-3000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-3000/trainer_state.json"
@@ -0,0 +1,3694 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.6694486173705,
+  "eval_steps": 100,
+  "global_step": 3000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.6993201158940794,
+      "epoch": 0.00890224225226729,
+      "grad_norm": 264.0,
+      "learning_rate": 1.8e-07,
+      "loss": 6.8158,
+      "mean_token_accuracy": 0.21653764923103153,
+      "num_tokens": 129942.0,
+      "step": 10
+    },
+    {
+      "entropy": 2.748416267335415,
+      "epoch": 0.01780448450453458,
+      "grad_norm": 342.0,
+      "learning_rate": 3.8e-07,
+      "loss": 6.842,
+      "mean_token_accuracy": 0.21704859719611705,
+      "num_tokens": 259048.0,
+      "step": 20
+    },
+    {
+      "entropy": 2.856283613294363,
+      "epoch": 0.026706726756801868,
+      "grad_norm": 260.0,
+      "learning_rate": 5.800000000000001e-07,
+      "loss": 6.5796,
+      "mean_token_accuracy": 0.21546755363233389,
+      "num_tokens": 397518.0,
+      "step": 30
+    },
+    {
+      "entropy": 3.1255638763308524,
+      "epoch": 0.03560896900906916,
+      "grad_norm": 235.0,
+      "learning_rate": 7.8e-07,
+      "loss": 6.0702,
+      "mean_token_accuracy": 0.22630739556625484,
+      "num_tokens": 532019.0,
+      "step": 40
+    },
+    {
+      "entropy": 3.437747097015381,
+      "epoch": 0.04451121126133645,
+      "grad_norm": 120.0,
+      "learning_rate": 9.800000000000001e-07,
+      "loss": 5.3567,
+      "mean_token_accuracy": 0.24060304677113892,
+      "num_tokens": 667753.0,
+      "step": 50
+    },
+    {
+      "entropy": 3.8268849313259126,
+      "epoch": 0.053413453513603736,
+      "grad_norm": 69.0,
+      "learning_rate": 1.1800000000000001e-06,
+      "loss": 4.7458,
+      "mean_token_accuracy": 0.25712613044306637,
+      "num_tokens": 807220.0,
+      "step": 60
+    },
+    {
+      "entropy": 4.098993194103241,
+      "epoch": 0.06231569576587103,
+      "grad_norm": 44.0,
+      "learning_rate": 1.3800000000000001e-06,
+      "loss": 4.1873,
+      "mean_token_accuracy": 0.2863635550253093,
+      "num_tokens": 938024.0,
+      "step": 70
+    },
+    {
+      "entropy": 3.7632771015167235,
+      "epoch": 0.07121793801813832,
+      "grad_norm": 39.75,
+      "learning_rate": 1.5800000000000001e-06,
+      "loss": 3.8191,
+      "mean_token_accuracy": 0.3185409564524889,
+      "num_tokens": 1068429.0,
+      "step": 80
+    },
+    {
+      "entropy": 3.4629461631178855,
+      "epoch": 0.08012018027040561,
+      "grad_norm": 32.0,
+      "learning_rate": 1.7800000000000001e-06,
+      "loss": 3.5033,
+      "mean_token_accuracy": 0.3425675223581493,
+      "num_tokens": 1209109.0,
+      "step": 90
+    },
+    {
+      "entropy": 3.3570878595113753,
+      "epoch": 0.0890224225226729,
+      "grad_norm": 22.375,
+      "learning_rate": 1.98e-06,
+      "loss": 3.3283,
+      "mean_token_accuracy": 0.36780647775158287,
+      "num_tokens": 1336698.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.0890224225226729,
+      "eval_coding_entropy": 2.025357067108154,
+      "eval_coding_loss": 2.231159210205078,
+      "eval_coding_mean_token_accuracy": 0.49899543941020963,
+      "eval_coding_num_tokens": 1336698.0,
+      "eval_coding_runtime": 81.3311,
+      "eval_coding_samples_per_second": 6.148,
+      "eval_coding_steps_per_second": 3.074,
+      "step": 100
+    },
+    {
+      "epoch": 0.0890224225226729,
+      "eval_chemistry_entropy": 3.242468201637268,
+      "eval_chemistry_loss": 3.3159916400909424,
+      "eval_chemistry_mean_token_accuracy": 0.3768811701536179,
+      "eval_chemistry_num_tokens": 1336698.0,
+      "eval_chemistry_runtime": 44.3638,
+      "eval_chemistry_samples_per_second": 11.27,
+      "eval_chemistry_steps_per_second": 5.635,
+      "step": 100
+    },
+    {
+      "entropy": 3.208672270178795,
+      "epoch": 0.09792466477494019,
+      "grad_norm": 26.625,
+      "learning_rate": 2.1800000000000003e-06,
+      "loss": 3.2262,
+      "mean_token_accuracy": 0.38037093859165905,
+      "num_tokens": 1470326.0,
+      "step": 110
+    },
+    {
+      "entropy": 3.073743937909603,
+      "epoch": 0.10682690702720747,
+      "grad_norm": 27.75,
+      "learning_rate": 2.38e-06,
+      "loss": 3.0418,
+      "mean_token_accuracy": 0.4051793536171317,
+      "num_tokens": 1608574.0,
+      "step": 120
+    },
+    {
+      "entropy": 3.0620595172047613,
+      "epoch": 0.11572914927947477,
+      "grad_norm": 23.0,
+      "learning_rate": 2.5800000000000003e-06,
+      "loss": 3.0619,
+      "mean_token_accuracy": 0.407612837664783,
+      "num_tokens": 1740382.0,
+      "step": 130
+    },
+    {
+      "entropy": 2.9336021929979323,
+      "epoch": 0.12463139153174206,
+      "grad_norm": 24.0,
+      "learning_rate": 2.7800000000000005e-06,
+      "loss": 2.8918,
+      "mean_token_accuracy": 0.4315342627465725,
+      "num_tokens": 1872747.0,
+      "step": 140
+    },
+    {
+      "entropy": 2.933651026338339,
+      "epoch": 0.13353363378400934,
+      "grad_norm": 21.125,
+      "learning_rate": 2.9800000000000003e-06,
+      "loss": 2.8977,
+      "mean_token_accuracy": 0.4332870215177536,
+      "num_tokens": 1999786.0,
+      "step": 150
+    },
+    {
+      "entropy": 2.78707619830966,
+      "epoch": 0.14243587603627664,
+      "grad_norm": 26.75,
+      "learning_rate": 3.1800000000000005e-06,
+      "loss": 2.713,
+      "mean_token_accuracy": 0.45073309261351824,
+      "num_tokens": 2133173.0,
+      "step": 160
+    },
+    {
+      "entropy": 2.686560308933258,
+      "epoch": 0.15133811828854393,
+      "grad_norm": 21.0,
+      "learning_rate": 3.3800000000000007e-06,
+      "loss": 2.6608,
+      "mean_token_accuracy": 0.45698707569390534,
+      "num_tokens": 2268889.0,
+      "step": 170
+    },
+    {
+      "entropy": 2.688250133395195,
+      "epoch": 0.16024036054081123,
+      "grad_norm": 23.0,
+      "learning_rate": 3.58e-06,
+      "loss": 2.6085,
+      "mean_token_accuracy": 0.4693053809925914,
+      "num_tokens": 2406145.0,
+      "step": 180
+    },
+    {
+      "entropy": 2.606059934198856,
+      "epoch": 0.1691426027930785,
+      "grad_norm": 19.0,
+      "learning_rate": 3.7800000000000002e-06,
+      "loss": 2.5776,
+      "mean_token_accuracy": 0.4732652172446251,
+      "num_tokens": 2533703.0,
+      "step": 190
+    },
+    {
+      "entropy": 2.6039498060941697,
+      "epoch": 0.1780448450453458,
+      "grad_norm": 20.625,
+      "learning_rate": 3.980000000000001e-06,
+      "loss": 2.5669,
+      "mean_token_accuracy": 0.4830343261361122,
+      "num_tokens": 2663709.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.1780448450453458,
+      "eval_coding_entropy": 1.940526816368103,
+      "eval_coding_loss": 2.1769461631774902,
+      "eval_coding_mean_token_accuracy": 0.5133515602350235,
+      "eval_coding_num_tokens": 2663709.0,
+      "eval_coding_runtime": 80.7684,
+      "eval_coding_samples_per_second": 6.191,
+      "eval_coding_steps_per_second": 3.095,
+      "step": 200
+    },
+    {
+      "epoch": 0.1780448450453458,
+      "eval_chemistry_entropy": 2.574181836128235,
+      "eval_chemistry_loss": 2.5678696632385254,
+      "eval_chemistry_mean_token_accuracy": 0.4837975095510483,
+      "eval_chemistry_num_tokens": 2663709.0,
+      "eval_chemistry_runtime": 44.3338,
+      "eval_chemistry_samples_per_second": 11.278,
+      "eval_chemistry_steps_per_second": 5.639,
+      "step": 200
+    },
+    {
+      "entropy": 2.517921609431505,
+      "epoch": 0.18694708729761308,
+      "grad_norm": 21.875,
+      "learning_rate": 4.18e-06,
+      "loss": 2.4289,
+      "mean_token_accuracy": 0.4946174016222358,
+      "num_tokens": 2803638.0,
+      "step": 210
+    },
+    {
+      "entropy": 2.4712012231349947,
+      "epoch": 0.19584932954988038,
+      "grad_norm": 20.625,
+      "learning_rate": 4.38e-06,
+      "loss": 2.4073,
+      "mean_token_accuracy": 0.4968462478369474,
+      "num_tokens": 2934980.0,
+      "step": 220
+    },
+    {
+      "entropy": 2.402497890591621,
+      "epoch": 0.20475157180214767,
+      "grad_norm": 19.375,
+      "learning_rate": 4.58e-06,
+      "loss": 2.3435,
+      "mean_token_accuracy": 0.507471177726984,
+      "num_tokens": 3071339.0,
+      "step": 230
+    },
+    {
+      "entropy": 2.4326496072113515,
+      "epoch": 0.21365381405441494,
+      "grad_norm": 19.875,
+      "learning_rate": 4.78e-06,
+      "loss": 2.3438,
+      "mean_token_accuracy": 0.5086988668888808,
+      "num_tokens": 3208702.0,
+      "step": 240
+    },
+    {
+      "entropy": 2.321472028642893,
+      "epoch": 0.22255605630668224,
+      "grad_norm": 20.0,
+      "learning_rate": 4.980000000000001e-06,
+      "loss": 2.2871,
+      "mean_token_accuracy": 0.5164231477305293,
+      "num_tokens": 3347298.0,
+      "step": 250
+    },
+    {
+      "entropy": 2.342081367969513,
+      "epoch": 0.23145829855894953,
+      "grad_norm": 20.375,
+      "learning_rate": 5.18e-06,
+      "loss": 2.3225,
+      "mean_token_accuracy": 0.5174826141446829,
+      "num_tokens": 3480119.0,
+      "step": 260
+    },
+    {
+      "entropy": 2.30421741977334,
+      "epoch": 0.24036054081121683,
+      "grad_norm": 25.125,
+      "learning_rate": 5.380000000000001e-06,
+      "loss": 2.2163,
+      "mean_token_accuracy": 0.5236617928370834,
+      "num_tokens": 3615777.0,
+      "step": 270
+    },
+    {
+      "entropy": 2.3181479632854463,
+      "epoch": 0.24926278306348412,
+      "grad_norm": 19.625,
+      "learning_rate": 5.580000000000001e-06,
+      "loss": 2.2445,
+      "mean_token_accuracy": 0.5264638463035226,
+      "num_tokens": 3746392.0,
+      "step": 280
+    },
+    {
+      "entropy": 2.3754931248724462,
+      "epoch": 0.2581650253157514,
+      "grad_norm": 19.5,
+      "learning_rate": 5.78e-06,
+      "loss": 2.3122,
+      "mean_token_accuracy": 0.5208129966631532,
+      "num_tokens": 3880425.0,
+      "step": 290
+    },
+    {
+      "entropy": 2.2281612612307073,
+      "epoch": 0.2670672675680187,
+      "grad_norm": 21.5,
+      "learning_rate": 5.98e-06,
+      "loss": 2.1504,
+      "mean_token_accuracy": 0.5389395957812667,
+      "num_tokens": 4019487.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.2670672675680187,
+      "eval_coding_entropy": 1.8798885374069214,
+      "eval_coding_loss": 2.1364073753356934,
+      "eval_coding_mean_token_accuracy": 0.5203545212745666,
+      "eval_coding_num_tokens": 4019487.0,
+      "eval_coding_runtime": 80.6635,
+      "eval_coding_samples_per_second": 6.199,
+      "eval_coding_steps_per_second": 3.099,
+      "step": 300
+    },
+    {
+      "epoch": 0.2670672675680187,
+      "eval_chemistry_entropy": 2.227835373401642,
+      "eval_chemistry_loss": 2.2248311042785645,
+      "eval_chemistry_mean_token_accuracy": 0.5376261820793152,
+      "eval_chemistry_num_tokens": 4019487.0,
+      "eval_chemistry_runtime": 44.3841,
+      "eval_chemistry_samples_per_second": 11.265,
+      "eval_chemistry_steps_per_second": 5.633,
+      "step": 300
+    },
+    {
+      "entropy": 2.2015350952744486,
+      "epoch": 0.275969509820286,
+      "grad_norm": 21.0,
+      "learning_rate": 6.18e-06,
+      "loss": 2.1543,
+      "mean_token_accuracy": 0.5409426406025887,
+      "num_tokens": 4162935.0,
+      "step": 310
+    },
+    {
+      "entropy": 2.148462438583374,
+      "epoch": 0.2848717520725533,
+      "grad_norm": 23.5,
+      "learning_rate": 6.380000000000001e-06,
+      "loss": 2.0957,
+      "mean_token_accuracy": 0.5452342376112937,
+      "num_tokens": 4298100.0,
+      "step": 320
+    },
+    {
+      "entropy": 2.0984656289219856,
+      "epoch": 0.29377399432482054,
+      "grad_norm": 19.5,
+      "learning_rate": 6.5800000000000005e-06,
+      "loss": 2.0404,
+      "mean_token_accuracy": 0.5558066863566637,
+      "num_tokens": 4431990.0,
+      "step": 330
+    },
+    {
+      "entropy": 2.1295887492597103,
+      "epoch": 0.30267623657708786,
+      "grad_norm": 17.625,
+      "learning_rate": 6.780000000000001e-06,
+      "loss": 2.0432,
+      "mean_token_accuracy": 0.5554485151544213,
+      "num_tokens": 4568116.0,
+      "step": 340
+    },
+    {
+      "entropy": 2.0176764823496343,
+      "epoch": 0.31157847882935513,
+      "grad_norm": 15.6875,
+      "learning_rate": 6.98e-06,
+      "loss": 1.9609,
+      "mean_token_accuracy": 0.5690282396972179,
+      "num_tokens": 4696575.0,
+      "step": 350
+    },
+    {
+      "entropy": 2.138897743076086,
+      "epoch": 0.32048072108162245,
+      "grad_norm": 20.0,
+      "learning_rate": 7.180000000000001e-06,
+      "loss": 2.0699,
+      "mean_token_accuracy": 0.5525342913344502,
+      "num_tokens": 4831334.0,
+      "step": 360
+    },
+    {
+      "entropy": 1.9979716561734677,
+      "epoch": 0.3293829633338897,
+      "grad_norm": 19.875,
+      "learning_rate": 7.3800000000000005e-06,
+      "loss": 1.9417,
+      "mean_token_accuracy": 0.5726395657286047,
+      "num_tokens": 4968132.0,
+      "step": 370
+    },
+    {
+      "entropy": 2.0710657715797423,
+      "epoch": 0.338285205586157,
+      "grad_norm": 23.125,
+      "learning_rate": 7.58e-06,
+      "loss": 1.9983,
+      "mean_token_accuracy": 0.5640093663707375,
+      "num_tokens": 5103301.0,
+      "step": 380
+    },
+    {
+      "entropy": 2.0690263971686362,
+      "epoch": 0.3471874478384243,
+      "grad_norm": 19.125,
+      "learning_rate": 7.78e-06,
+      "loss": 2.0099,
+      "mean_token_accuracy": 0.5661305790767074,
+      "num_tokens": 5230700.0,
+      "step": 390
+    },
+    {
+      "entropy": 2.0010731495916843,
+      "epoch": 0.3560896900906916,
+      "grad_norm": 15.625,
+      "learning_rate": 7.980000000000002e-06,
+      "loss": 1.9302,
+      "mean_token_accuracy": 0.5734745427966118,
+      "num_tokens": 5362915.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.3560896900906916,
+      "eval_coding_entropy": 1.9175666098594666,
+      "eval_coding_loss": 2.1198952198028564,
+      "eval_coding_mean_token_accuracy": 0.5218169560432434,
+      "eval_coding_num_tokens": 5362915.0,
+      "eval_coding_runtime": 80.6411,
+      "eval_coding_samples_per_second": 6.2,
+      "eval_coding_steps_per_second": 3.1,
+      "step": 400
+    },
+    {
+      "epoch": 0.3560896900906916,
+      "eval_chemistry_entropy": 2.0164283666610716,
+      "eval_chemistry_loss": 1.9814609289169312,
+      "eval_chemistry_mean_token_accuracy": 0.5761862490177154,
+      "eval_chemistry_num_tokens": 5362915.0,
+      "eval_chemistry_runtime": 44.318,
+      "eval_chemistry_samples_per_second": 11.282,
+      "eval_chemistry_steps_per_second": 5.641,
+      "step": 400
+    },
+    {
+      "entropy": 1.9029628101736307,
+      "epoch": 0.3649919323429589,
+      "grad_norm": 20.625,
+      "learning_rate": 8.18e-06,
+      "loss": 1.861,
+      "mean_token_accuracy": 0.5885544015094638,
+      "num_tokens": 5497551.0,
+      "step": 410
+    },
+    {
+      "entropy": 1.8858691919595003,
+      "epoch": 0.37389417459522617,
+      "grad_norm": 18.75,
+      "learning_rate": 8.380000000000001e-06,
+      "loss": 1.8473,
+      "mean_token_accuracy": 0.5886353930458427,
+      "num_tokens": 5625455.0,
+      "step": 420
+    },
+    {
+      "entropy": 1.917143502831459,
+      "epoch": 0.38279641684749344,
+      "grad_norm": 18.0,
+      "learning_rate": 8.580000000000001e-06,
+      "loss": 1.8792,
+      "mean_token_accuracy": 0.5864967351779342,
+      "num_tokens": 5757241.0,
+      "step": 430
+    },
+    {
+      "entropy": 1.9342061571776867,
+      "epoch": 0.39169865909976076,
+      "grad_norm": 18.25,
+      "learning_rate": 8.78e-06,
+      "loss": 1.8538,
+      "mean_token_accuracy": 0.5882492190226912,
+      "num_tokens": 5890806.0,
+      "step": 440
+    },
+    {
+      "entropy": 1.8766419854015113,
+      "epoch": 0.400600901352028,
+      "grad_norm": 19.25,
+      "learning_rate": 8.98e-06,
+      "loss": 1.8162,
+      "mean_token_accuracy": 0.5928706232458353,
+      "num_tokens": 6023088.0,
+      "step": 450
+    },
+    {
+      "entropy": 1.8140752121806145,
+      "epoch": 0.40950314360429535,
+      "grad_norm": 17.375,
+      "learning_rate": 9.180000000000002e-06,
+      "loss": 1.7678,
+      "mean_token_accuracy": 0.6031433455646038,
+      "num_tokens": 6158751.0,
+      "step": 460
+    },
+    {
+      "entropy": 1.8589354373514653,
+      "epoch": 0.4184053858565626,
+      "grad_norm": 17.25,
+      "learning_rate": 9.38e-06,
+      "loss": 1.8104,
+      "mean_token_accuracy": 0.5982734605669975,
+      "num_tokens": 6289699.0,
+      "step": 470
+    },
+    {
+      "entropy": 1.8204911682754754,
+      "epoch": 0.4273076281088299,
+      "grad_norm": 18.125,
+      "learning_rate": 9.58e-06,
+      "loss": 1.7427,
+      "mean_token_accuracy": 0.6014625774696469,
+      "num_tokens": 6423747.0,
+      "step": 480
+    },
+    {
+      "entropy": 1.8097763508558273,
+      "epoch": 0.4362098703610972,
+      "grad_norm": 18.0,
+      "learning_rate": 9.780000000000001e-06,
+      "loss": 1.7556,
+      "mean_token_accuracy": 0.606043404340744,
+      "num_tokens": 6556651.0,
+      "step": 490
+    },
+    {
+      "entropy": 1.7938172847032547,
+      "epoch": 0.4451121126133645,
+      "grad_norm": 16.125,
+      "learning_rate": 9.980000000000001e-06,
+      "loss": 1.7349,
+      "mean_token_accuracy": 0.6063602799549699,
+      "num_tokens": 6684517.0,
+      "step": 500
+    },
+    {
+      "epoch": 0.4451121126133645,
+      "eval_coding_entropy": 1.9086761941909791,
+      "eval_coding_loss": 2.1499152183532715,
+      "eval_coding_mean_token_accuracy": 0.5191502428054809,
+      "eval_coding_num_tokens": 6684517.0,
+      "eval_coding_runtime": 80.761,
+      "eval_coding_samples_per_second": 6.191,
+      "eval_coding_steps_per_second": 3.096,
+      "step": 500
+    },
+    {
+      "epoch": 0.4451121126133645,
+      "eval_chemistry_entropy": 1.8382492520809173,
+      "eval_chemistry_loss": 1.8092410564422607,
+      "eval_chemistry_mean_token_accuracy": 0.6044963012933731,
+      "eval_chemistry_num_tokens": 6684517.0,
+      "eval_chemistry_runtime": 43.9817,
+      "eval_chemistry_samples_per_second": 11.368,
+      "eval_chemistry_steps_per_second": 5.684,
+      "step": 500
+    },
+    {
+      "entropy": 1.7506890654563905,
+      "epoch": 0.4540143548656318,
+      "grad_norm": 17.0,
+      "learning_rate": 1.018e-05,
+      "loss": 1.6717,
+      "mean_token_accuracy": 0.6144311860203743,
+      "num_tokens": 6814065.0,
+      "step": 510
+    },
+    {
+      "entropy": 1.752192370966077,
+      "epoch": 0.46291659711789906,
+      "grad_norm": 18.5,
+      "learning_rate": 1.038e-05,
+      "loss": 1.6926,
+      "mean_token_accuracy": 0.6115166075527668,
+      "num_tokens": 6945659.0,
+      "step": 520
+    },
+    {
+      "entropy": 1.7476748272776603,
+      "epoch": 0.4718188393701664,
+      "grad_norm": 18.875,
+      "learning_rate": 1.0580000000000002e-05,
+      "loss": 1.6791,
+      "mean_token_accuracy": 0.614359668083489,
+      "num_tokens": 7083465.0,
+      "step": 530
+    },
+    {
+      "entropy": 1.7340088631957769,
+      "epoch": 0.48072108162243365,
+      "grad_norm": 14.125,
+      "learning_rate": 1.0780000000000002e-05,
+      "loss": 1.7051,
+      "mean_token_accuracy": 0.6148024540394544,
+      "num_tokens": 7217797.0,
+      "step": 540
+    },
+    {
+      "entropy": 1.7364882446825505,
+      "epoch": 0.4896233238747009,
+      "grad_norm": 16.25,
+      "learning_rate": 1.0980000000000002e-05,
+      "loss": 1.6523,
+      "mean_token_accuracy": 0.6168902778998018,
+      "num_tokens": 7364316.0,
+      "step": 550
+    },
+    {
+      "entropy": 1.6127678520977498,
+      "epoch": 0.49852556612696824,
+      "grad_norm": 20.5,
+      "learning_rate": 1.1180000000000001e-05,
+      "loss": 1.5842,
+      "mean_token_accuracy": 0.6330170270055533,
+      "num_tokens": 7498182.0,
+      "step": 560
+    },
+    {
+      "entropy": 1.7071593157947063,
+      "epoch": 0.5074278083792355,
+      "grad_norm": 15.1875,
+      "learning_rate": 1.138e-05,
+      "loss": 1.6117,
+      "mean_token_accuracy": 0.6257967110723257,
+      "num_tokens": 7630581.0,
+      "step": 570
+    },
+    {
+      "entropy": 1.6894463405013085,
+      "epoch": 0.5163300506315028,
+      "grad_norm": 17.875,
+      "learning_rate": 1.1580000000000001e-05,
+      "loss": 1.6253,
+      "mean_token_accuracy": 0.6228752240538598,
+      "num_tokens": 7761686.0,
+      "step": 580
+    },
+    {
+      "entropy": 1.7127487953752278,
+      "epoch": 0.52523229288377,
+      "grad_norm": 14.8125,
+      "learning_rate": 1.178e-05,
+      "loss": 1.6567,
+      "mean_token_accuracy": 0.6184359768405556,
+      "num_tokens": 7894733.0,
+      "step": 590
+    },
+    {
+      "entropy": 1.7086713213473559,
+      "epoch": 0.5341345351360374,
+      "grad_norm": 15.75,
+      "learning_rate": 1.198e-05,
+      "loss": 1.6548,
+      "mean_token_accuracy": 0.6228462293744087,
+      "num_tokens": 8026923.0,
+      "step": 600
+    },
+    {
+      "epoch": 0.5341345351360374,
+      "eval_coding_entropy": 1.9360160393714905,
+      "eval_coding_loss": 2.150575876235962,
+      "eval_coding_mean_token_accuracy": 0.5199909027814865,
+      "eval_coding_num_tokens": 8026923.0,
+      "eval_coding_runtime": 80.6899,
+      "eval_coding_samples_per_second": 6.197,
+      "eval_coding_steps_per_second": 3.098,
+      "step": 600
+    },
+    {
+      "epoch": 0.5341345351360374,
+      "eval_chemistry_entropy": 1.7303061096668244,
+      "eval_chemistry_loss": 1.6744147539138794,
+      "eval_chemistry_mean_token_accuracy": 0.6269070360660552,
+      "eval_chemistry_num_tokens": 8026923.0,
+      "eval_chemistry_runtime": 44.5775,
+      "eval_chemistry_samples_per_second": 11.216,
+      "eval_chemistry_steps_per_second": 5.608,
+      "step": 600
+    },
+    {
+      "entropy": 1.692430131137371,
+      "epoch": 0.5430367773883047,
+      "grad_norm": 14.75,
+      "learning_rate": 1.218e-05,
+      "loss": 1.6309,
+      "mean_token_accuracy": 0.6261043848469854,
+      "num_tokens": 8162179.0,
+      "step": 610
+    },
+    {
+      "entropy": 1.6273927468806506,
+      "epoch": 0.551939019640572,
+      "grad_norm": 17.875,
+      "learning_rate": 1.2380000000000002e-05,
+      "loss": 1.6052,
+      "mean_token_accuracy": 0.6333680363371968,
+      "num_tokens": 8301506.0,
+      "step": 620
+    },
+    {
+      "entropy": 1.6359726019203662,
+      "epoch": 0.5608412618928392,
+      "grad_norm": 16.625,
+      "learning_rate": 1.2580000000000002e-05,
+      "loss": 1.574,
+      "mean_token_accuracy": 0.6349722359329462,
+      "num_tokens": 8425846.0,
+      "step": 630
+    },
+    {
+      "entropy": 1.6046722147613763,
+      "epoch": 0.5697435041451065,
+      "grad_norm": 16.125,
+      "learning_rate": 1.2780000000000001e-05,
+      "loss": 1.5789,
+      "mean_token_accuracy": 0.6387555615976452,
+      "num_tokens": 8559126.0,
+      "step": 640
+    },
+    {
+      "entropy": 1.641243514046073,
+      "epoch": 0.5786457463973739,
+      "grad_norm": 18.375,
+      "learning_rate": 1.2980000000000001e-05,
+      "loss": 1.5626,
+      "mean_token_accuracy": 0.6389757834374905,
+      "num_tokens": 8685256.0,
+      "step": 650
+    },
+    {
+      "entropy": 1.6162871330976487,
+      "epoch": 0.5875479886496411,
+      "grad_norm": 13.375,
+      "learning_rate": 1.3180000000000001e-05,
+      "loss": 1.585,
+      "mean_token_accuracy": 0.6322070205584168,
+      "num_tokens": 8821088.0,
+      "step": 660
+    },
+    {
+      "entropy": 1.5711755152791738,
+      "epoch": 0.5964502309019084,
+      "grad_norm": 16.125,
+      "learning_rate": 1.3380000000000002e-05,
+      "loss": 1.5036,
+      "mean_token_accuracy": 0.6462705478072166,
+      "num_tokens": 8950665.0,
+      "step": 670
+    },
+    {
+      "entropy": 1.5002975206822158,
+      "epoch": 0.6053524731541757,
+      "grad_norm": 11.875,
+      "learning_rate": 1.3580000000000002e-05,
+      "loss": 1.455,
+      "mean_token_accuracy": 0.6536407358944416,
+      "num_tokens": 9089866.0,
+      "step": 680
+    },
+    {
+      "entropy": 1.5762819994241,
+      "epoch": 0.614254715406443,
+      "grad_norm": 15.3125,
+      "learning_rate": 1.378e-05,
+      "loss": 1.5317,
+      "mean_token_accuracy": 0.6418447192758322,
+      "num_tokens": 9219941.0,
+      "step": 690
+    },
+    {
+      "entropy": 1.5155543368309736,
+      "epoch": 0.6231569576587103,
+      "grad_norm": 14.1875,
+      "learning_rate": 1.398e-05,
+      "loss": 1.4688,
+      "mean_token_accuracy": 0.6533356573432684,
+      "num_tokens": 9358866.0,
+      "step": 700
+    },
+    {
+      "epoch": 0.6231569576587103,
+      "eval_coding_entropy": 1.9042859573364257,
+      "eval_coding_loss": 2.178054094314575,
+      "eval_coding_mean_token_accuracy": 0.517359386563301,
+      "eval_coding_num_tokens": 9358866.0,
+      "eval_coding_runtime": 80.7138,
+      "eval_coding_samples_per_second": 6.195,
+      "eval_coding_steps_per_second": 3.097,
+      "step": 700
+    },
+    {
+      "epoch": 0.6231569576587103,
+      "eval_chemistry_entropy": 1.56386332988739,
+      "eval_chemistry_loss": 1.572709560394287,
+      "eval_chemistry_mean_token_accuracy": 0.6433606895208359,
+      "eval_chemistry_num_tokens": 9358866.0,
+      "eval_chemistry_runtime": 44.295,
+      "eval_chemistry_samples_per_second": 11.288,
+      "eval_chemistry_steps_per_second": 5.644,
+      "step": 700
+    },
+    {
+      "entropy": 1.5002719465643168,
+      "epoch": 0.6320591999109776,
+      "grad_norm": 15.75,
+      "learning_rate": 1.418e-05,
+      "loss": 1.4804,
+      "mean_token_accuracy": 0.6501881824806333,
+      "num_tokens": 9493065.0,
+      "step": 710
+    },
+    {
+      "entropy": 1.522989520803094,
+      "epoch": 0.6409614421632449,
+      "grad_norm": 16.75,
+      "learning_rate": 1.4380000000000001e-05,
+      "loss": 1.4427,
+      "mean_token_accuracy": 0.6548429084941745,
+      "num_tokens": 9629977.0,
+      "step": 720
+    },
+    {
+      "entropy": 1.4675043471157552,
+      "epoch": 0.6498636844155121,
+      "grad_norm": 11.9375,
+      "learning_rate": 1.4580000000000001e-05,
+      "loss": 1.4134,
+      "mean_token_accuracy": 0.6610828962177038,
+      "num_tokens": 9765477.0,
+      "step": 730
+    },
+    {
+      "entropy": 1.491576162353158,
+      "epoch": 0.6587659266677794,
+      "grad_norm": 14.625,
+      "learning_rate": 1.478e-05,
+      "loss": 1.4587,
+      "mean_token_accuracy": 0.6557513184845447,
+      "num_tokens": 9897757.0,
+      "step": 740
+    },
+    {
+      "entropy": 1.4832498595118522,
+      "epoch": 0.6676681689200468,
+      "grad_norm": 13.0625,
+      "learning_rate": 1.498e-05,
+      "loss": 1.4184,
+      "mean_token_accuracy": 0.6580261953175068,
+      "num_tokens": 10026698.0,
+      "step": 750
+    },
+    {
+      "entropy": 1.4681301727890967,
+      "epoch": 0.676570411172314,
+      "grad_norm": 12.6875,
+      "learning_rate": 1.5180000000000002e-05,
+      "loss": 1.4207,
+      "mean_token_accuracy": 0.6607962453737855,
+      "num_tokens": 10168367.0,
+      "step": 760
+    },
+    {
+      "entropy": 1.431799278780818,
+      "epoch": 0.6854726534245813,
+      "grad_norm": 12.375,
+      "learning_rate": 1.5380000000000002e-05,
+      "loss": 1.4078,
+      "mean_token_accuracy": 0.6646994914859533,
+      "num_tokens": 10307268.0,
+      "step": 770
+    },
+    {
+      "entropy": 1.4440745569765567,
+      "epoch": 0.6943748956768486,
+      "grad_norm": 11.4375,
+      "learning_rate": 1.5580000000000003e-05,
+      "loss": 1.3797,
+      "mean_token_accuracy": 0.666821115091443,
+      "num_tokens": 10439039.0,
+      "step": 780
+    },
+    {
+      "entropy": 1.4467374693602324,
+      "epoch": 0.703277137929116,
+      "grad_norm": 12.3125,
+      "learning_rate": 1.578e-05,
+      "loss": 1.4202,
+      "mean_token_accuracy": 0.6628865763545037,
+      "num_tokens": 10576127.0,
+      "step": 790
+    },
+    {
+      "entropy": 1.476112574338913,
+      "epoch": 0.7121793801813832,
+      "grad_norm": 12.375,
+      "learning_rate": 1.5980000000000003e-05,
+      "loss": 1.397,
+      "mean_token_accuracy": 0.663382687792182,
+      "num_tokens": 10709549.0,
+      "step": 800
+    },
+    {
+      "epoch": 0.7121793801813832,
+      "eval_coding_entropy": 1.8322501955032349,
+      "eval_coding_loss": 2.2156379222869873,
+      "eval_coding_mean_token_accuracy": 0.5140866179466248,
+      "eval_coding_num_tokens": 10709549.0,
+      "eval_coding_runtime": 80.6982,
+      "eval_coding_samples_per_second": 6.196,
+      "eval_coding_steps_per_second": 3.098,
+      "step": 800
+    },
+    {
+      "epoch": 0.7121793801813832,
+      "eval_chemistry_entropy": 1.456686591386795,
+      "eval_chemistry_loss": 1.4780359268188477,
+      "eval_chemistry_mean_token_accuracy": 0.6602535841464996,
+      "eval_chemistry_num_tokens": 10709549.0,
+      "eval_chemistry_runtime": 44.3155,
+      "eval_chemistry_samples_per_second": 11.283,
+      "eval_chemistry_steps_per_second": 5.641,
+      "step": 800
+    },
+    {
+      "entropy": 1.3352404598146677,
+      "epoch": 0.7210816224336505,
+      "grad_norm": 13.75,
+      "learning_rate": 1.618e-05,
+      "loss": 1.3165,
+      "mean_token_accuracy": 0.6821591448038816,
+      "num_tokens": 10841474.0,
+      "step": 810
+    },
+    {
+      "entropy": 1.4511364437639713,
+      "epoch": 0.7299838646859178,
+      "grad_norm": 11.75,
+      "learning_rate": 1.638e-05,
+      "loss": 1.4194,
+      "mean_token_accuracy": 0.6614615445956588,
+      "num_tokens": 10976011.0,
+      "step": 820
+    },
+    {
+      "entropy": 1.4821706857532262,
+      "epoch": 0.738886106938185,
+      "grad_norm": 10.875,
+      "learning_rate": 1.658e-05,
+      "loss": 1.4022,
+      "mean_token_accuracy": 0.662100600078702,
+      "num_tokens": 11109618.0,
+      "step": 830
+    },
+    {
+      "entropy": 1.3720035433769227,
+      "epoch": 0.7477883491904523,
+      "grad_norm": 13.0625,
+      "learning_rate": 1.6780000000000002e-05,
+      "loss": 1.3532,
+      "mean_token_accuracy": 0.6731239786371589,
+      "num_tokens": 11245251.0,
+      "step": 840
+    },
+    {
+      "entropy": 1.428163754194975,
+      "epoch": 0.7566905914427197,
+      "grad_norm": 12.625,
+      "learning_rate": 1.698e-05,
+      "loss": 1.384,
+      "mean_token_accuracy": 0.6671924598515033,
+      "num_tokens": 11379950.0,
+      "step": 850
+    },
+    {
+      "entropy": 1.425218838453293,
+      "epoch": 0.7655928336949869,
+      "grad_norm": 12.4375,
+      "learning_rate": 1.718e-05,
+      "loss": 1.3805,
+      "mean_token_accuracy": 0.6696572992950678,
+      "num_tokens": 11518514.0,
+      "step": 860
+    },
+    {
+      "entropy": 1.3892541889101266,
+      "epoch": 0.7744950759472542,
+      "grad_norm": 10.0,
+      "learning_rate": 1.7380000000000003e-05,
+      "loss": 1.3583,
+      "mean_token_accuracy": 0.6718392688781023,
+      "num_tokens": 11650235.0,
+      "step": 870
+    },
+    {
+      "entropy": 1.409494410455227,
+      "epoch": 0.7833973181995215,
+      "grad_norm": 9.6875,
+      "learning_rate": 1.758e-05,
+      "loss": 1.3757,
+      "mean_token_accuracy": 0.6687737111002207,
+      "num_tokens": 11787117.0,
+      "step": 880
+    },
+    {
+      "entropy": 1.3409042585641147,
+      "epoch": 0.7922995604517888,
+      "grad_norm": 13.625,
+      "learning_rate": 1.7780000000000003e-05,
+      "loss": 1.3062,
+      "mean_token_accuracy": 0.6828067794442176,
+      "num_tokens": 11924960.0,
+      "step": 890
+    },
+    {
+      "entropy": 1.461272732913494,
+      "epoch": 0.801201802704056,
+      "grad_norm": 12.8125,
+      "learning_rate": 1.798e-05,
+      "loss": 1.4331,
+      "mean_token_accuracy": 0.6593699859455228,
+      "num_tokens": 12050993.0,
+      "step": 900
+    },
+    {
+      "epoch": 0.801201802704056,
+      "eval_coding_entropy": 1.9756887073516847,
+      "eval_coding_loss": 2.2261524200439453,
+      "eval_coding_mean_token_accuracy": 0.5107356268167496,
+      "eval_coding_num_tokens": 12050993.0,
+      "eval_coding_runtime": 80.8151,
+      "eval_coding_samples_per_second": 6.187,
+      "eval_coding_steps_per_second": 3.093,
+      "step": 900
+    },
+    {
+      "epoch": 0.801201802704056,
+      "eval_chemistry_entropy": 1.4363152146339417,
+      "eval_chemistry_loss": 1.4172585010528564,
+      "eval_chemistry_mean_token_accuracy": 0.6695912653207778,
+      "eval_chemistry_num_tokens": 12050993.0,
+      "eval_chemistry_runtime": 44.4444,
+      "eval_chemistry_samples_per_second": 11.25,
+      "eval_chemistry_steps_per_second": 5.625,
+      "step": 900
+    },
+    {
+      "entropy": 1.409971272200346,
+      "epoch": 0.8101040449563234,
+      "grad_norm": 9.75,
+      "learning_rate": 1.8180000000000002e-05,
+      "loss": 1.3519,
+      "mean_token_accuracy": 0.6748842839151621,
+      "num_tokens": 12184261.0,
+      "step": 910
+    },
+    {
+      "entropy": 1.3415631584823131,
+      "epoch": 0.8190062872085907,
+      "grad_norm": 11.1875,
+      "learning_rate": 1.8380000000000004e-05,
+      "loss": 1.3134,
+      "mean_token_accuracy": 0.6776627466082573,
+      "num_tokens": 12311688.0,
+      "step": 920
+    },
+    {
+      "entropy": 1.3450417269021273,
+      "epoch": 0.8279085294608579,
+      "grad_norm": 9.6875,
+      "learning_rate": 1.858e-05,
+      "loss": 1.2681,
+      "mean_token_accuracy": 0.6853212669491768,
+      "num_tokens": 12453515.0,
+      "step": 930
+    },
+    {
+      "entropy": 1.3162209324538707,
+      "epoch": 0.8368107717131252,
+      "grad_norm": 9.3125,
+      "learning_rate": 1.878e-05,
+      "loss": 1.2935,
+      "mean_token_accuracy": 0.6805886892601848,
+      "num_tokens": 12598408.0,
+      "step": 940
+    },
+    {
+      "entropy": 1.3450543098151684,
+      "epoch": 0.8457130139653926,
+      "grad_norm": 11.125,
+      "learning_rate": 1.898e-05,
+      "loss": 1.3186,
+      "mean_token_accuracy": 0.6790889304131269,
+      "num_tokens": 12732830.0,
+      "step": 950
+    },
+    {
+      "entropy": 1.3630160838365555,
+      "epoch": 0.8546152562176598,
+      "grad_norm": 11.0,
+      "learning_rate": 1.918e-05,
+      "loss": 1.3251,
+      "mean_token_accuracy": 0.6795368468388915,
+      "num_tokens": 12869642.0,
+      "step": 960
+    },
+    {
+      "entropy": 1.2341018915176392,
+      "epoch": 0.8635174984699271,
+      "grad_norm": 9.4375,
+      "learning_rate": 1.938e-05,
+      "loss": 1.2047,
+      "mean_token_accuracy": 0.697525155171752,
+      "num_tokens": 13002795.0,
+      "step": 970
+    },
+    {
+      "entropy": 1.3432362783700227,
+      "epoch": 0.8724197407221944,
+      "grad_norm": 11.9375,
+      "learning_rate": 1.9580000000000002e-05,
+      "loss": 1.3242,
+      "mean_token_accuracy": 0.682424726523459,
+      "num_tokens": 13139555.0,
+      "step": 980
+    },
+    {
+      "entropy": 1.3661971539258957,
+      "epoch": 0.8813219829744617,
+      "grad_norm": 11.0625,
+      "learning_rate": 1.978e-05,
+      "loss": 1.3181,
+      "mean_token_accuracy": 0.678005202487111,
+      "num_tokens": 13261742.0,
+      "step": 990
+    },
+    {
+      "entropy": 1.33810890391469,
+      "epoch": 0.890224225226729,
+      "grad_norm": 9.3125,
+      "learning_rate": 1.9980000000000002e-05,
+      "loss": 1.2809,
+      "mean_token_accuracy": 0.6845312876626849,
+      "num_tokens": 13396545.0,
+      "step": 1000
+    },
+    {
+      "epoch": 0.890224225226729,
+      "eval_coding_entropy": 1.8324014544487,
+      "eval_coding_loss": 2.2719883918762207,
+      "eval_coding_mean_token_accuracy": 0.506569258928299,
+      "eval_coding_num_tokens": 13396545.0,
+      "eval_coding_runtime": 80.808,
+      "eval_coding_samples_per_second": 6.188,
+      "eval_coding_steps_per_second": 3.094,
+      "step": 1000
+    },
+    {
+      "epoch": 0.890224225226729,
+      "eval_chemistry_entropy": 1.2783789792060851,
+      "eval_chemistry_loss": 1.3424817323684692,
+      "eval_chemistry_mean_token_accuracy": 0.6828865656852722,
+      "eval_chemistry_num_tokens": 13396545.0,
+      "eval_chemistry_runtime": 43.9749,
+      "eval_chemistry_samples_per_second": 11.37,
+      "eval_chemistry_steps_per_second": 5.685,
+      "step": 1000
+    },
+    {
+      "entropy": 1.3035245168954135,
+      "epoch": 0.8991264674789963,
+      "grad_norm": 9.375,
+      "learning_rate": 1.9980000000000002e-05,
+      "loss": 1.2693,
+      "mean_token_accuracy": 0.688488537631929,
+      "num_tokens": 13530331.0,
+      "step": 1010
+    },
+    {
+      "entropy": 1.3113087877631187,
+      "epoch": 0.9080287097312636,
+      "grad_norm": 10.1875,
+      "learning_rate": 1.995777777777778e-05,
+      "loss": 1.2332,
+      "mean_token_accuracy": 0.685172756575048,
+      "num_tokens": 13667857.0,
+      "step": 1020
+    },
+    {
+      "entropy": 1.2651601791381837,
+      "epoch": 0.9169309519835308,
+      "grad_norm": 8.0625,
+      "learning_rate": 1.9935555555555557e-05,
+      "loss": 1.2144,
+      "mean_token_accuracy": 0.694831382855773,
+      "num_tokens": 13804327.0,
+      "step": 1030
+    },
+    {
+      "entropy": 1.375035434216261,
+      "epoch": 0.9258331942357981,
+      "grad_norm": 7.875,
+      "learning_rate": 1.9913333333333335e-05,
+      "loss": 1.3072,
+      "mean_token_accuracy": 0.6771694764494895,
+      "num_tokens": 13932771.0,
+      "step": 1040
+    },
+    {
+      "entropy": 1.2766119629144668,
+      "epoch": 0.9347354364880655,
+      "grad_norm": 17.875,
+      "learning_rate": 1.9891111111111112e-05,
+      "loss": 1.2383,
+      "mean_token_accuracy": 0.6927796453237534,
+      "num_tokens": 14054030.0,
+      "step": 1050
+    },
+    {
+      "entropy": 1.2739322971552611,
+      "epoch": 0.9436376787403328,
+      "grad_norm": 8.8125,
+      "learning_rate": 1.986888888888889e-05,
+      "loss": 1.2235,
+      "mean_token_accuracy": 0.697489945217967,
+      "num_tokens": 14194632.0,
+      "step": 1060
+    },
+    {
+      "entropy": 1.2921774670481683,
+      "epoch": 0.9525399209926,
+      "grad_norm": 10.5625,
+      "learning_rate": 1.9846666666666668e-05,
+      "loss": 1.2519,
+      "mean_token_accuracy": 0.6894336406141519,
+      "num_tokens": 14324321.0,
+      "step": 1070
+    },
+    {
+      "entropy": 1.3097396202385425,
+      "epoch": 0.9614421632448673,
+      "grad_norm": 8.1875,
+      "learning_rate": 1.9824444444444445e-05,
+      "loss": 1.2746,
+      "mean_token_accuracy": 0.6891787204891443,
+      "num_tokens": 14457650.0,
+      "step": 1080
+    },
+    {
+      "entropy": 1.2524930398911238,
+      "epoch": 0.9703444054971346,
+      "grad_norm": 9.5,
+      "learning_rate": 1.9802222222222226e-05,
+      "loss": 1.2239,
+      "mean_token_accuracy": 0.693185362778604,
+      "num_tokens": 14597103.0,
+      "step": 1090
+    },
+    {
+      "entropy": 1.2353097528219223,
+      "epoch": 0.9792466477494018,
+      "grad_norm": 9.0625,
+      "learning_rate": 1.978e-05,
+      "loss": 1.188,
+      "mean_token_accuracy": 0.7037162348628044,
+      "num_tokens": 14724967.0,
+      "step": 1100
+    },
+    {
+      "epoch": 0.9792466477494018,
+      "eval_coding_entropy": 1.9174934754371642,
+      "eval_coding_loss": 2.2787740230560303,
+      "eval_coding_mean_token_accuracy": 0.5043189586400986,
+      "eval_coding_num_tokens": 14724967.0,
+      "eval_coding_runtime": 80.7608,
+      "eval_coding_samples_per_second": 6.191,
+      "eval_coding_steps_per_second": 3.096,
+      "step": 1100
+    },
+    {
+      "epoch": 0.9792466477494018,
+      "eval_chemistry_entropy": 1.2794872944355011,
+      "eval_chemistry_loss": 1.2872111797332764,
+      "eval_chemistry_mean_token_accuracy": 0.6942087370157242,
+      "eval_chemistry_num_tokens": 14724967.0,
+      "eval_chemistry_runtime": 44.348,
+      "eval_chemistry_samples_per_second": 11.274,
+      "eval_chemistry_steps_per_second": 5.637,
+      "step": 1100
+    },
+    {
+      "entropy": 1.2467465713620185,
+      "epoch": 0.9881488900016692,
+      "grad_norm": 8.0625,
+      "learning_rate": 1.975777777777778e-05,
+      "loss": 1.2251,
+      "mean_token_accuracy": 0.6985302936285734,
+      "num_tokens": 14866601.0,
+      "step": 1110
+    },
+    {
+      "entropy": 1.2711088940501214,
+      "epoch": 0.9970511322539365,
+      "grad_norm": 7.90625,
+      "learning_rate": 1.9735555555555556e-05,
+      "loss": 1.2636,
+      "mean_token_accuracy": 0.6943498553708196,
+      "num_tokens": 14993272.0,
+      "step": 1120
+    },
+    {
+      "entropy": 1.16374516967159,
+      "epoch": 1.0053413453513604,
+      "grad_norm": 8.625,
+      "learning_rate": 1.9713333333333337e-05,
+      "loss": 1.1153,
+      "mean_token_accuracy": 0.7124827076124665,
+      "num_tokens": 15117600.0,
+      "step": 1130
+    },
+    {
+      "entropy": 1.0887193260714412,
+      "epoch": 1.0142435876036278,
+      "grad_norm": 7.28125,
+      "learning_rate": 1.969111111111111e-05,
+      "loss": 1.0437,
+      "mean_token_accuracy": 0.7269797824323178,
+      "num_tokens": 15254936.0,
+      "step": 1140
+    },
+    {
+      "entropy": 1.182083160057664,
+      "epoch": 1.0231458298558949,
+      "grad_norm": 8.375,
+      "learning_rate": 1.9668888888888892e-05,
+      "loss": 1.1231,
+      "mean_token_accuracy": 0.7116198563948274,
+      "num_tokens": 15384044.0,
+      "step": 1150
+    },
+    {
+      "entropy": 1.149610160663724,
+      "epoch": 1.0320480721081622,
+      "grad_norm": 7.75,
+      "learning_rate": 1.9646666666666666e-05,
+      "loss": 1.121,
+      "mean_token_accuracy": 0.7131425060331822,
+      "num_tokens": 15515007.0,
+      "step": 1160
+    },
+    {
+      "entropy": 1.097679178789258,
+      "epoch": 1.0409503143604295,
+      "grad_norm": 7.90625,
+      "learning_rate": 1.9624444444444447e-05,
+      "loss": 1.0517,
+      "mean_token_accuracy": 0.7278640700504184,
+      "num_tokens": 15661927.0,
+      "step": 1170
+    },
+    {
+      "entropy": 1.1251065036281944,
+      "epoch": 1.0498525566126968,
+      "grad_norm": 7.90625,
+      "learning_rate": 1.9602222222222225e-05,
+      "loss": 1.1007,
+      "mean_token_accuracy": 0.7212187899276614,
+      "num_tokens": 15794203.0,
+      "step": 1180
+    },
+    {
+      "entropy": 1.139916059933603,
+      "epoch": 1.0587547988649642,
+      "grad_norm": 9.1875,
+      "learning_rate": 1.9580000000000002e-05,
+      "loss": 1.1087,
+      "mean_token_accuracy": 0.7153899630531668,
+      "num_tokens": 15929958.0,
+      "step": 1190
+    },
+    {
+      "entropy": 1.122027451172471,
+      "epoch": 1.0676570411172315,
+      "grad_norm": 10.5,
+      "learning_rate": 1.955777777777778e-05,
+      "loss": 1.1006,
+      "mean_token_accuracy": 0.722288279607892,
+      "num_tokens": 16063664.0,
+      "step": 1200
+    },
+    {
+      "epoch": 1.0676570411172315,
+      "eval_coding_entropy": 1.7862111530303955,
+      "eval_coding_loss": 2.3154520988464355,
+      "eval_coding_mean_token_accuracy": 0.503941803097725,
+      "eval_coding_num_tokens": 16063664.0,
+      "eval_coding_runtime": 80.6939,
+      "eval_coding_samples_per_second": 6.196,
+      "eval_coding_steps_per_second": 3.098,
+      "step": 1200
+    },
+    {
+      "epoch": 1.0676570411172315,
+      "eval_chemistry_entropy": 1.1847249876260757,
+      "eval_chemistry_loss": 1.2270034551620483,
+      "eval_chemistry_mean_token_accuracy": 0.7046360708475113,
+      "eval_chemistry_num_tokens": 16063664.0,
+      "eval_chemistry_runtime": 44.2926,
+      "eval_chemistry_samples_per_second": 11.289,
+      "eval_chemistry_steps_per_second": 5.644,
+      "step": 1200
+    },
+    {
+      "entropy": 1.0779426898807287,
+      "epoch": 1.0765592833694986,
+      "grad_norm": 7.78125,
+      "learning_rate": 1.9535555555555557e-05,
+      "loss": 1.0534,
+      "mean_token_accuracy": 0.730063559487462,
+      "num_tokens": 16203903.0,
+      "step": 1210
+    },
+    {
+      "entropy": 1.106826901063323,
+      "epoch": 1.085461525621766,
+      "grad_norm": 9.3125,
+      "learning_rate": 1.9513333333333335e-05,
+      "loss": 1.0825,
+      "mean_token_accuracy": 0.7251957911998034,
+      "num_tokens": 16338303.0,
+      "step": 1220
+    },
+    {
+      "entropy": 1.136346418224275,
+      "epoch": 1.0943637678740332,
+      "grad_norm": 8.375,
+      "learning_rate": 1.9491111111111113e-05,
+      "loss": 1.0946,
+      "mean_token_accuracy": 0.718695884756744,
+      "num_tokens": 16470818.0,
+      "step": 1230
+    },
+    {
+      "entropy": 1.136488999426365,
+      "epoch": 1.1032660101263005,
+      "grad_norm": 9.0,
+      "learning_rate": 1.946888888888889e-05,
+      "loss": 1.0941,
+      "mean_token_accuracy": 0.7204574767500163,
+      "num_tokens": 16606302.0,
+      "step": 1240
+    },
+    {
+      "entropy": 1.0911134846508503,
+      "epoch": 1.1121682523785679,
+      "grad_norm": 6.625,
+      "learning_rate": 1.9446666666666668e-05,
+      "loss": 1.0438,
+      "mean_token_accuracy": 0.7281744480133057,
+      "num_tokens": 16737209.0,
+      "step": 1250
+    },
+    {
+      "entropy": 1.1247851882129907,
+      "epoch": 1.1210704946308352,
+      "grad_norm": 7.375,
+      "learning_rate": 1.9424444444444446e-05,
+      "loss": 1.0708,
+      "mean_token_accuracy": 0.7217418739572168,
+      "num_tokens": 16877071.0,
+      "step": 1260
+    },
+    {
+      "entropy": 1.1004639966413379,
+      "epoch": 1.1299727368831025,
+      "grad_norm": 8.25,
+      "learning_rate": 1.9402222222222223e-05,
+      "loss": 1.0983,
+      "mean_token_accuracy": 0.7219324804842472,
+      "num_tokens": 17011127.0,
+      "step": 1270
+    },
+    {
+      "entropy": 1.0730026822537184,
+      "epoch": 1.1388749791353696,
+      "grad_norm": 8.5625,
+      "learning_rate": 1.938e-05,
+      "loss": 1.0242,
+      "mean_token_accuracy": 0.7301186516880989,
+      "num_tokens": 17145775.0,
+      "step": 1280
+    },
+    {
+      "entropy": 1.0746742237359286,
+      "epoch": 1.147777221387637,
+      "grad_norm": 6.4375,
+      "learning_rate": 1.935777777777778e-05,
+      "loss": 1.0389,
+      "mean_token_accuracy": 0.7312157262116671,
+      "num_tokens": 17278708.0,
+      "step": 1290
+    },
+    {
+      "entropy": 1.1036461044102908,
+      "epoch": 1.1566794636399043,
+      "grad_norm": 6.90625,
+      "learning_rate": 1.9335555555555556e-05,
+      "loss": 1.0666,
+      "mean_token_accuracy": 0.7227642200887203,
+      "num_tokens": 17410775.0,
+      "step": 1300
+    },
+    {
+      "epoch": 1.1566794636399043,
+      "eval_coding_entropy": 1.82445157289505,
+      "eval_coding_loss": 2.3140883445739746,
+      "eval_coding_mean_token_accuracy": 0.5025496450662613,
+      "eval_coding_num_tokens": 17410775.0,
+      "eval_coding_runtime": 80.8109,
+      "eval_coding_samples_per_second": 6.187,
+      "eval_coding_steps_per_second": 3.094,
+      "step": 1300
+    },
+    {
+      "epoch": 1.1566794636399043,
+      "eval_chemistry_entropy": 1.1369108176231384,
+      "eval_chemistry_loss": 1.1869932413101196,
+      "eval_chemistry_mean_token_accuracy": 0.7125645990371704,
+      "eval_chemistry_num_tokens": 17410775.0,
+      "eval_chemistry_runtime": 44.3736,
+      "eval_chemistry_samples_per_second": 11.268,
+      "eval_chemistry_steps_per_second": 5.634,
+      "step": 1300
+    },
+    {
+      "entropy": 1.098898689262569,
+      "epoch": 1.1655817058921716,
+      "grad_norm": 7.96875,
+      "learning_rate": 1.9313333333333334e-05,
+      "loss": 1.0895,
+      "mean_token_accuracy": 0.7260249227285385,
+      "num_tokens": 17543395.0,
+      "step": 1310
+    },
+    {
+      "entropy": 1.1586046267300845,
+      "epoch": 1.174483948144439,
+      "grad_norm": 7.15625,
+      "learning_rate": 1.9291111111111115e-05,
+      "loss": 1.0871,
+      "mean_token_accuracy": 0.7148085344582796,
+      "num_tokens": 17677791.0,
+      "step": 1320
+    },
+    {
+      "entropy": 1.0978492671623825,
+      "epoch": 1.1833861903967062,
+      "grad_norm": 7.8125,
+      "learning_rate": 1.926888888888889e-05,
+      "loss": 1.0462,
+      "mean_token_accuracy": 0.730253504216671,
+      "num_tokens": 17812608.0,
+      "step": 1330
+    },
+    {
+      "entropy": 1.0760172510519623,
+      "epoch": 1.1922884326489736,
+      "grad_norm": 7.1875,
+      "learning_rate": 1.924666666666667e-05,
+      "loss": 1.0303,
+      "mean_token_accuracy": 0.7288852859288454,
+      "num_tokens": 17949441.0,
+      "step": 1340
+    },
+    {
+      "entropy": 1.063178063556552,
+      "epoch": 1.2011906749012407,
+      "grad_norm": 7.125,
+      "learning_rate": 1.9224444444444444e-05,
+      "loss": 1.0237,
+      "mean_token_accuracy": 0.7320365833118558,
+      "num_tokens": 18079231.0,
+      "step": 1350
+    },
+    {
+      "entropy": 1.074139853566885,
+      "epoch": 1.210092917153508,
+      "grad_norm": 7.875,
+      "learning_rate": 1.9202222222222225e-05,
+      "loss": 1.0405,
+      "mean_token_accuracy": 0.7328159822151065,
+      "num_tokens": 18217742.0,
+      "step": 1360
+    },
+    {
+      "entropy": 1.075910792313516,
+      "epoch": 1.2189951594057753,
+      "grad_norm": 7.0625,
+      "learning_rate": 1.918e-05,
+      "loss": 1.0279,
+      "mean_token_accuracy": 0.7312528677284718,
+      "num_tokens": 18360535.0,
+      "step": 1370
+    },
+    {
+      "entropy": 1.0907096166163683,
+      "epoch": 1.2278974016580426,
+      "grad_norm": 7.15625,
+      "learning_rate": 1.915777777777778e-05,
+      "loss": 1.0552,
+      "mean_token_accuracy": 0.7264303725212813,
+      "num_tokens": 18487769.0,
+      "step": 1380
+    },
+    {
+      "entropy": 1.081585726328194,
+      "epoch": 1.23679964391031,
+      "grad_norm": 6.78125,
+      "learning_rate": 1.9135555555555555e-05,
+      "loss": 1.0514,
+      "mean_token_accuracy": 0.727237294241786,
+      "num_tokens": 18622690.0,
+      "step": 1390
+    },
+    {
+      "entropy": 1.0638965075835585,
+      "epoch": 1.2457018861625773,
+      "grad_norm": 6.9375,
+      "learning_rate": 1.9113333333333336e-05,
+      "loss": 1.0176,
+      "mean_token_accuracy": 0.7334342382848262,
+      "num_tokens": 18760296.0,
+      "step": 1400
+    },
+    {
+      "epoch": 1.2457018861625773,
+      "eval_coding_entropy": 1.8175329451560973,
+      "eval_coding_loss": 2.3160665035247803,
+      "eval_coding_mean_token_accuracy": 0.5024578042030334,
+      "eval_coding_num_tokens": 18760296.0,
+      "eval_coding_runtime": 80.7542,
+      "eval_coding_samples_per_second": 6.192,
+      "eval_coding_steps_per_second": 3.096,
+      "step": 1400
+    },
+    {
+      "epoch": 1.2457018861625773,
+      "eval_chemistry_entropy": 1.1010388153791428,
+      "eval_chemistry_loss": 1.154791235923767,
+      "eval_chemistry_mean_token_accuracy": 0.7180610251426697,
+      "eval_chemistry_num_tokens": 18760296.0,
+      "eval_chemistry_runtime": 44.4534,
+      "eval_chemistry_samples_per_second": 11.248,
+      "eval_chemistry_steps_per_second": 5.624,
+      "step": 1400
+    },
+    {
+      "entropy": 1.0695431407541036,
+      "epoch": 1.2546041284148446,
+      "grad_norm": 6.5,
+      "learning_rate": 1.9091111111111113e-05,
+      "loss": 1.0238,
+      "mean_token_accuracy": 0.7328258238732814,
+      "num_tokens": 18902929.0,
+      "step": 1410
+    },
+    {
+      "entropy": 1.0441692693158984,
+      "epoch": 1.2635063706671117,
+      "grad_norm": 6.625,
+      "learning_rate": 1.906888888888889e-05,
+      "loss": 1.0063,
+      "mean_token_accuracy": 0.7369453951716423,
+      "num_tokens": 19035733.0,
+      "step": 1420
+    },
+    {
+      "entropy": 1.0926606407389046,
+      "epoch": 1.272408612919379,
+      "grad_norm": 6.875,
+      "learning_rate": 1.904666666666667e-05,
+      "loss": 1.0584,
+      "mean_token_accuracy": 0.7249008229002356,
+      "num_tokens": 19176688.0,
+      "step": 1430
+    },
+    {
+      "entropy": 1.0925894405692815,
+      "epoch": 1.2813108551716463,
+      "grad_norm": 5.84375,
+      "learning_rate": 1.9024444444444446e-05,
+      "loss": 1.0454,
+      "mean_token_accuracy": 0.7300722308456897,
+      "num_tokens": 19311432.0,
+      "step": 1440
+    },
+    {
+      "entropy": 1.0371008031070232,
+      "epoch": 1.2902130974239137,
+      "grad_norm": 6.9375,
+      "learning_rate": 1.9002222222222224e-05,
+      "loss": 1.0135,
+      "mean_token_accuracy": 0.7359844289720059,
+      "num_tokens": 19455868.0,
+      "step": 1450
+    },
+    {
+      "entropy": 1.0112120743840933,
+      "epoch": 1.299115339676181,
+      "grad_norm": 6.8125,
+      "learning_rate": 1.898e-05,
+      "loss": 0.9746,
+      "mean_token_accuracy": 0.7447358928620815,
+      "num_tokens": 19593016.0,
+      "step": 1460
+    },
+    {
+      "entropy": 1.0387776872143148,
+      "epoch": 1.3080175819284483,
+      "grad_norm": 6.5,
+      "learning_rate": 1.895777777777778e-05,
+      "loss": 0.9989,
+      "mean_token_accuracy": 0.736772421002388,
+      "num_tokens": 19729273.0,
+      "step": 1470
+    },
+    {
+      "entropy": 1.047124046459794,
+      "epoch": 1.3169198241807156,
+      "grad_norm": 5.71875,
+      "learning_rate": 1.8935555555555556e-05,
+      "loss": 1.0033,
+      "mean_token_accuracy": 0.7367634609341621,
+      "num_tokens": 19860476.0,
+      "step": 1480
+    },
+    {
+      "entropy": 1.0483765492215753,
+      "epoch": 1.3258220664329827,
+      "grad_norm": 6.46875,
+      "learning_rate": 1.8913333333333334e-05,
+      "loss": 0.9927,
+      "mean_token_accuracy": 0.7344137353822588,
+      "num_tokens": 19993681.0,
+      "step": 1490
+    },
+    {
+      "entropy": 1.0336147954687476,
+      "epoch": 1.33472430868525,
+      "grad_norm": 6.625,
+      "learning_rate": 1.8891111111111115e-05,
+      "loss": 0.9945,
+      "mean_token_accuracy": 0.7402323342859745,
+      "num_tokens": 20124436.0,
+      "step": 1500
+    },
+    {
+      "epoch": 1.33472430868525,
+      "eval_coding_entropy": 1.817141619682312,
+      "eval_coding_loss": 2.247633218765259,
+      "eval_coding_mean_token_accuracy": 0.5122213376760483,
+      "eval_coding_num_tokens": 20124436.0,
+      "eval_coding_runtime": 80.8023,
+      "eval_coding_samples_per_second": 6.188,
+      "eval_coding_steps_per_second": 3.094,
+      "step": 1500
+    },
+    {
+      "epoch": 1.33472430868525,
+      "eval_chemistry_entropy": 1.099737853527069,
+      "eval_chemistry_loss": 1.11974036693573,
+      "eval_chemistry_mean_token_accuracy": 0.7246058889627457,
+      "eval_chemistry_num_tokens": 20124436.0,
+      "eval_chemistry_runtime": 44.0197,
+      "eval_chemistry_samples_per_second": 11.359,
+      "eval_chemistry_steps_per_second": 5.679,
+      "step": 1500
+    },
+    {
+      "entropy": 1.0875422598794102,
+      "epoch": 1.3436265509375174,
+      "grad_norm": 6.75,
+      "learning_rate": 1.886888888888889e-05,
+      "loss": 1.0491,
+      "mean_token_accuracy": 0.7286154057830572,
+      "num_tokens": 20256101.0,
+      "step": 1510
+    },
+    {
+      "entropy": 1.0377698469907046,
+      "epoch": 1.3525287931897847,
+      "grad_norm": 6.84375,
+      "learning_rate": 1.884666666666667e-05,
+      "loss": 1.0071,
+      "mean_token_accuracy": 0.7385351993143559,
+      "num_tokens": 20388027.0,
+      "step": 1520
+    },
+    {
+      "entropy": 1.04939345382154,
+      "epoch": 1.361431035442052,
+      "grad_norm": 6.4375,
+      "learning_rate": 1.8824444444444445e-05,
+      "loss": 0.9965,
+      "mean_token_accuracy": 0.7372564105316997,
+      "num_tokens": 20519030.0,
+      "step": 1530
+    },
+    {
+      "entropy": 1.0302043357864021,
+      "epoch": 1.3703332776943191,
+      "grad_norm": 6.375,
+      "learning_rate": 1.8802222222222226e-05,
+      "loss": 0.9984,
+      "mean_token_accuracy": 0.7377668373286724,
+      "num_tokens": 20656076.0,
+      "step": 1540
+    },
+    {
+      "entropy": 1.092260554805398,
+      "epoch": 1.3792355199465867,
+      "grad_norm": 6.4375,
+      "learning_rate": 1.878e-05,
+      "loss": 1.0758,
+      "mean_token_accuracy": 0.7272779047489166,
+      "num_tokens": 20781429.0,
+      "step": 1550
+    },
+    {
+      "entropy": 1.0867322444915772,
+      "epoch": 1.3881377621988538,
+      "grad_norm": 11.375,
+      "learning_rate": 1.875777777777778e-05,
+      "loss": 1.0343,
+      "mean_token_accuracy": 0.7299443442374468,
+      "num_tokens": 20911633.0,
+      "step": 1560
+    },
+    {
+      "entropy": 1.0433940727263689,
+      "epoch": 1.397040004451121,
+      "grad_norm": 7.125,
+      "learning_rate": 1.873555555555556e-05,
+      "loss": 1.0004,
+      "mean_token_accuracy": 0.7396705735474824,
+      "num_tokens": 21044481.0,
+      "step": 1570
+    },
+    {
+      "entropy": 1.073896124958992,
+      "epoch": 1.4059422467033884,
+      "grad_norm": 6.5625,
+      "learning_rate": 1.8713333333333336e-05,
+      "loss": 1.0542,
+      "mean_token_accuracy": 0.7299106653779746,
+      "num_tokens": 21176227.0,
+      "step": 1580
+    },
+    {
+      "entropy": 1.054144686087966,
+      "epoch": 1.4148444889556557,
+      "grad_norm": 5.875,
+      "learning_rate": 1.8691111111111114e-05,
+      "loss": 1.008,
+      "mean_token_accuracy": 0.736966310814023,
+      "num_tokens": 21308809.0,
+      "step": 1590
+    },
+    {
+      "entropy": 1.0131638515740633,
+      "epoch": 1.423746731207923,
+      "grad_norm": 6.09375,
+      "learning_rate": 1.866888888888889e-05,
+      "loss": 0.9884,
+      "mean_token_accuracy": 0.7402045801281929,
+      "num_tokens": 21445242.0,
+      "step": 1600
+    },
+    {
+      "epoch": 1.423746731207923,
+      "eval_coding_entropy": 1.7055041899681092,
+      "eval_coding_loss": 2.283527374267578,
+      "eval_coding_mean_token_accuracy": 0.5110386753082276,
+      "eval_coding_num_tokens": 21445242.0,
+      "eval_coding_runtime": 80.8292,
+      "eval_coding_samples_per_second": 6.186,
+      "eval_coding_steps_per_second": 3.093,
+      "step": 1600
+    },
+    {
+      "epoch": 1.423746731207923,
+      "eval_chemistry_entropy": 1.0189639842510223,
+      "eval_chemistry_loss": 1.0948059558868408,
+      "eval_chemistry_mean_token_accuracy": 0.7295370472669601,
+      "eval_chemistry_num_tokens": 21445242.0,
+      "eval_chemistry_runtime": 44.4303,
+      "eval_chemistry_samples_per_second": 11.254,
+      "eval_chemistry_steps_per_second": 5.627,
+      "step": 1600
+    },
+    {
+      "entropy": 1.0424250226467848,
+      "epoch": 1.4326489734601902,
+      "grad_norm": 6.0,
+      "learning_rate": 1.864666666666667e-05,
+      "loss": 1.0037,
+      "mean_token_accuracy": 0.738079284876585,
+      "num_tokens": 21580828.0,
+      "step": 1610
+    },
+    {
+      "entropy": 1.0000621777027845,
+      "epoch": 1.4415512157124577,
+      "grad_norm": 6.71875,
+      "learning_rate": 1.8624444444444446e-05,
+      "loss": 0.9756,
+      "mean_token_accuracy": 0.7446792133152484,
+      "num_tokens": 21718455.0,
+      "step": 1620
+    },
+    {
+      "entropy": 1.0381389891728758,
+      "epoch": 1.4504534579647248,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.8602222222222224e-05,
+      "loss": 0.9908,
+      "mean_token_accuracy": 0.7371132237836718,
+      "num_tokens": 21843858.0,
+      "step": 1630
+    },
+    {
+      "entropy": 1.0006099876016379,
+      "epoch": 1.4593557002169921,
+      "grad_norm": 5.8125,
+      "learning_rate": 1.858e-05,
+      "loss": 0.9719,
+      "mean_token_accuracy": 0.7490028616040945,
+      "num_tokens": 21974076.0,
+      "step": 1640
+    },
+    {
+      "entropy": 1.0176434211432934,
+      "epoch": 1.4682579424692594,
+      "grad_norm": 8.0,
+      "learning_rate": 1.855777777777778e-05,
+      "loss": 0.988,
+      "mean_token_accuracy": 0.7422546174377203,
+      "num_tokens": 22105208.0,
+      "step": 1650
+    },
+    {
+      "entropy": 1.051332799717784,
+      "epoch": 1.4771601847215268,
+      "grad_norm": 5.9375,
+      "learning_rate": 1.8535555555555557e-05,
+      "loss": 1.0022,
+      "mean_token_accuracy": 0.7370473204180599,
+      "num_tokens": 22241599.0,
+      "step": 1660
+    },
+    {
+      "entropy": 1.047087043710053,
+      "epoch": 1.486062426973794,
+      "grad_norm": 7.40625,
+      "learning_rate": 1.8513333333333335e-05,
+      "loss": 1.0112,
+      "mean_token_accuracy": 0.7353191252797842,
+      "num_tokens": 22376331.0,
+      "step": 1670
+    },
+    {
+      "entropy": 0.998528147302568,
+      "epoch": 1.4949646692260612,
+      "grad_norm": 6.6875,
+      "learning_rate": 1.8491111111111112e-05,
+      "loss": 0.962,
+      "mean_token_accuracy": 0.7432398173958064,
+      "num_tokens": 22510725.0,
+      "step": 1680
+    },
+    {
+      "entropy": 1.0193962009623647,
+      "epoch": 1.5038669114783287,
+      "grad_norm": 6.71875,
+      "learning_rate": 1.846888888888889e-05,
+      "loss": 0.9954,
+      "mean_token_accuracy": 0.7430227894335986,
+      "num_tokens": 22647287.0,
+      "step": 1690
+    },
+    {
+      "entropy": 0.9964680068194867,
+      "epoch": 1.5127691537305958,
+      "grad_norm": 5.65625,
+      "learning_rate": 1.8446666666666667e-05,
+      "loss": 0.9422,
+      "mean_token_accuracy": 0.7483414333313704,
+      "num_tokens": 22778514.0,
+      "step": 1700
+    },
+    {
+      "epoch": 1.5127691537305958,
+      "eval_coding_entropy": 1.7789310560226441,
+      "eval_coding_loss": 2.2901668548583984,
+      "eval_coding_mean_token_accuracy": 0.5082898802757263,
+      "eval_coding_num_tokens": 22778514.0,
+      "eval_coding_runtime": 80.8117,
+      "eval_coding_samples_per_second": 6.187,
+      "eval_coding_steps_per_second": 3.094,
+      "step": 1700
+    },
+    {
+      "epoch": 1.5127691537305958,
+      "eval_chemistry_entropy": 1.0228302110433578,
+      "eval_chemistry_loss": 1.0728719234466553,
+      "eval_chemistry_mean_token_accuracy": 0.7340625635385514,
+      "eval_chemistry_num_tokens": 22778514.0,
+      "eval_chemistry_runtime": 44.3932,
+      "eval_chemistry_samples_per_second": 11.263,
+      "eval_chemistry_steps_per_second": 5.631,
+      "step": 1700
+    },
+    {
+      "entropy": 1.018488030694425,
+      "epoch": 1.5216713959828632,
+      "grad_norm": 5.96875,
+      "learning_rate": 1.842444444444445e-05,
+      "loss": 0.9724,
+      "mean_token_accuracy": 0.741841921582818,
+      "num_tokens": 22913183.0,
+      "step": 1710
+    },
+    {
+      "entropy": 0.9667788481339812,
+      "epoch": 1.5305736382351305,
+      "grad_norm": 6.34375,
+      "learning_rate": 1.8402222222222223e-05,
+      "loss": 0.927,
+      "mean_token_accuracy": 0.7500551588833332,
+      "num_tokens": 23051230.0,
+      "step": 1720
+    },
+    {
+      "entropy": 1.0106308363378047,
+      "epoch": 1.5394758804873978,
+      "grad_norm": 6.125,
+      "learning_rate": 1.8380000000000004e-05,
+      "loss": 0.9704,
+      "mean_token_accuracy": 0.7423111679032445,
+      "num_tokens": 23179208.0,
+      "step": 1730
+    },
+    {
+      "entropy": 1.0059592995792628,
+      "epoch": 1.5483781227396651,
+      "grad_norm": 5.84375,
+      "learning_rate": 1.8357777777777778e-05,
+      "loss": 0.9573,
+      "mean_token_accuracy": 0.7479906413704157,
+      "num_tokens": 23306327.0,
+      "step": 1740
+    },
+    {
+      "entropy": 0.9837975237518549,
+      "epoch": 1.5572803649919322,
+      "grad_norm": 6.125,
+      "learning_rate": 1.833555555555556e-05,
+      "loss": 0.9434,
+      "mean_token_accuracy": 0.7474345922470093,
+      "num_tokens": 23435151.0,
+      "step": 1750
+    },
+    {
+      "entropy": 1.0287323439493776,
+      "epoch": 1.5661826072441998,
+      "grad_norm": 5.96875,
+      "learning_rate": 1.8313333333333333e-05,
+      "loss": 0.9897,
+      "mean_token_accuracy": 0.7407072335481644,
+      "num_tokens": 23573127.0,
+      "step": 1760
+    },
+    {
+      "entropy": 1.0248231099918486,
+      "epoch": 1.5750848494964669,
+      "grad_norm": 5.28125,
+      "learning_rate": 1.8291111111111114e-05,
+      "loss": 0.9815,
+      "mean_token_accuracy": 0.7394947454333305,
+      "num_tokens": 23704824.0,
+      "step": 1770
+    },
+    {
+      "entropy": 0.9785062935203314,
+      "epoch": 1.5839870917487342,
+      "grad_norm": 5.78125,
+      "learning_rate": 1.8268888888888888e-05,
+      "loss": 0.9392,
+      "mean_token_accuracy": 0.7487792737782002,
+      "num_tokens": 23841914.0,
+      "step": 1780
+    },
+    {
+      "entropy": 0.9828181255608797,
+      "epoch": 1.5928893340010015,
+      "grad_norm": 5.0625,
+      "learning_rate": 1.824666666666667e-05,
+      "loss": 0.9569,
+      "mean_token_accuracy": 0.7477491162717342,
+      "num_tokens": 23976648.0,
+      "step": 1790
+    },
+    {
+      "entropy": 1.0069462068378925,
+      "epoch": 1.6017915762532688,
+      "grad_norm": 6.75,
+      "learning_rate": 1.8224444444444447e-05,
+      "loss": 0.9675,
+      "mean_token_accuracy": 0.7469035653397441,
+      "num_tokens": 24105114.0,
+      "step": 1800
+    },
+    {
+      "epoch": 1.6017915762532688,
+      "eval_coding_entropy": 1.7382126560211182,
+      "eval_coding_loss": 2.2909605503082275,
+      "eval_coding_mean_token_accuracy": 0.5080101939439774,
+      "eval_coding_num_tokens": 24105114.0,
+      "eval_coding_runtime": 80.8719,
+      "eval_coding_samples_per_second": 6.183,
+      "eval_coding_steps_per_second": 3.091,
+      "step": 1800
+    },
+    {
+      "epoch": 1.6017915762532688,
+      "eval_chemistry_entropy": 1.03293852186203,
+      "eval_chemistry_loss": 1.056942343711853,
+      "eval_chemistry_mean_token_accuracy": 0.7363584907054901,
+      "eval_chemistry_num_tokens": 24105114.0,
+      "eval_chemistry_runtime": 44.5514,
+      "eval_chemistry_samples_per_second": 11.223,
+      "eval_chemistry_steps_per_second": 5.611,
+      "step": 1800
+    },
+    {
+      "entropy": 0.9643349962309002,
+      "epoch": 1.6106938185055362,
+      "grad_norm": 4.59375,
+      "learning_rate": 1.8202222222222225e-05,
+      "loss": 0.9435,
+      "mean_token_accuracy": 0.7507723122835159,
+      "num_tokens": 24242945.0,
+      "step": 1810
+    },
+    {
+      "entropy": 0.9434997234493494,
+      "epoch": 1.6195960607578033,
+      "grad_norm": 6.40625,
+      "learning_rate": 1.8180000000000002e-05,
+      "loss": 0.9229,
+      "mean_token_accuracy": 0.7551070984452963,
+      "num_tokens": 24384008.0,
+      "step": 1820
+    },
+    {
+      "entropy": 1.0243266159668565,
+      "epoch": 1.6284983030100708,
+      "grad_norm": 5.96875,
+      "learning_rate": 1.815777777777778e-05,
+      "loss": 0.987,
+      "mean_token_accuracy": 0.7420570751652121,
+      "num_tokens": 24510350.0,
+      "step": 1830
+    },
+    {
+      "entropy": 0.959172617085278,
+      "epoch": 1.637400545262338,
+      "grad_norm": 6.84375,
+      "learning_rate": 1.8135555555555557e-05,
+      "loss": 0.9155,
+      "mean_token_accuracy": 0.7518640130758285,
+      "num_tokens": 24638205.0,
+      "step": 1840
+    },
+    {
+      "entropy": 0.9474278407171368,
+      "epoch": 1.6463027875146052,
+      "grad_norm": 5.53125,
+      "learning_rate": 1.8113333333333335e-05,
+      "loss": 0.8923,
+      "mean_token_accuracy": 0.7558273140341043,
+      "num_tokens": 24773805.0,
+      "step": 1850
+    },
+    {
+      "entropy": 0.9222735103219748,
+      "epoch": 1.6552050297668726,
+      "grad_norm": 6.03125,
+      "learning_rate": 1.8091111111111113e-05,
+      "loss": 0.8927,
+      "mean_token_accuracy": 0.7610101442784071,
+      "num_tokens": 24910104.0,
+      "step": 1860
+    },
+    {
+      "entropy": 0.9674294136464596,
+      "epoch": 1.6641072720191397,
+      "grad_norm": 5.46875,
+      "learning_rate": 1.806888888888889e-05,
+      "loss": 0.9182,
+      "mean_token_accuracy": 0.7508355010300874,
+      "num_tokens": 25047786.0,
+      "step": 1870
+    },
+    {
+      "entropy": 0.9484452249482274,
+      "epoch": 1.6730095142714072,
+      "grad_norm": 5.125,
+      "learning_rate": 1.8046666666666668e-05,
+      "loss": 0.9237,
+      "mean_token_accuracy": 0.7539303492754698,
+      "num_tokens": 25181306.0,
+      "step": 1880
+    },
+    {
+      "entropy": 1.0244407325983047,
+      "epoch": 1.6819117565236743,
+      "grad_norm": 6.09375,
+      "learning_rate": 1.8024444444444445e-05,
+      "loss": 0.9899,
+      "mean_token_accuracy": 0.7406658142805099,
+      "num_tokens": 25313859.0,
+      "step": 1890
+    },
+    {
+      "entropy": 0.89578603412956,
+      "epoch": 1.6908139987759419,
+      "grad_norm": 5.625,
+      "learning_rate": 1.8002222222222223e-05,
+      "loss": 0.8741,
+      "mean_token_accuracy": 0.7654941972345114,
+      "num_tokens": 25449138.0,
+      "step": 1900
+    },
+    {
+      "epoch": 1.6908139987759419,
+      "eval_coding_entropy": 1.7772144932746887,
+      "eval_coding_loss": 2.3101391792297363,
+      "eval_coding_mean_token_accuracy": 0.507433453321457,
+      "eval_coding_num_tokens": 25449138.0,
+      "eval_coding_runtime": 80.6897,
+      "eval_coding_samples_per_second": 6.197,
+      "eval_coding_steps_per_second": 3.098,
+      "step": 1900
+    },
+    {
+      "epoch": 1.6908139987759419,
+      "eval_chemistry_entropy": 1.0048357552289964,
+      "eval_chemistry_loss": 1.0314446687698364,
+      "eval_chemistry_mean_token_accuracy": 0.7415237789154052,
+      "eval_chemistry_num_tokens": 25449138.0,
+      "eval_chemistry_runtime": 44.3145,
+      "eval_chemistry_samples_per_second": 11.283,
+      "eval_chemistry_steps_per_second": 5.641,
+      "step": 1900
+    },
+    {
+      "entropy": 0.9545569067820907,
+      "epoch": 1.699716241028209,
+      "grad_norm": 6.125,
+      "learning_rate": 1.798e-05,
+      "loss": 0.9039,
+      "mean_token_accuracy": 0.7564696136862039,
+      "num_tokens": 25583745.0,
+      "step": 1910
+    },
+    {
+      "entropy": 0.9878080926835537,
+      "epoch": 1.7086184832804763,
+      "grad_norm": 7.25,
+      "learning_rate": 1.7957777777777778e-05,
+      "loss": 0.9657,
+      "mean_token_accuracy": 0.7482136309146881,
+      "num_tokens": 25709030.0,
+      "step": 1920
+    },
+    {
+      "entropy": 1.0241941774263978,
+      "epoch": 1.7175207255327436,
+      "grad_norm": 4.8125,
+      "learning_rate": 1.7935555555555556e-05,
+      "loss": 0.9613,
+      "mean_token_accuracy": 0.7427621867507697,
+      "num_tokens": 25842094.0,
+      "step": 1930
+    },
+    {
+      "entropy": 0.969705024920404,
+      "epoch": 1.7264229677850107,
+      "grad_norm": 5.28125,
+      "learning_rate": 1.7913333333333337e-05,
+      "loss": 0.9264,
+      "mean_token_accuracy": 0.748832119628787,
+      "num_tokens": 25972807.0,
+      "step": 1940
+    },
+    {
+      "entropy": 0.9519479189068079,
+      "epoch": 1.7353252100372782,
+      "grad_norm": 5.3125,
+      "learning_rate": 1.789111111111111e-05,
+      "loss": 0.9243,
+      "mean_token_accuracy": 0.7541633494198322,
+      "num_tokens": 26111702.0,
+      "step": 1950
+    },
+    {
+      "entropy": 0.9655054366216064,
+      "epoch": 1.7442274522895453,
+      "grad_norm": 5.75,
+      "learning_rate": 1.7868888888888892e-05,
+      "loss": 0.9167,
+      "mean_token_accuracy": 0.7538340780884027,
+      "num_tokens": 26244955.0,
+      "step": 1960
+    },
+    {
+      "entropy": 0.9491505783051253,
+      "epoch": 1.7531296945418127,
+      "grad_norm": 5.4375,
+      "learning_rate": 1.7846666666666666e-05,
+      "loss": 0.9042,
+      "mean_token_accuracy": 0.7555052500218153,
+      "num_tokens": 26383177.0,
+      "step": 1970
+    },
+    {
+      "entropy": 0.876483421958983,
+      "epoch": 1.76203193679408,
+      "grad_norm": 5.15625,
+      "learning_rate": 1.7824444444444447e-05,
+      "loss": 0.8621,
+      "mean_token_accuracy": 0.768435088545084,
+      "num_tokens": 26515548.0,
+      "step": 1980
+    },
+    {
+      "entropy": 0.9604039745405316,
+      "epoch": 1.7709341790463473,
+      "grad_norm": 4.75,
+      "learning_rate": 1.780222222222222e-05,
+      "loss": 0.9117,
+      "mean_token_accuracy": 0.7546427011489868,
+      "num_tokens": 26652194.0,
+      "step": 1990
+    },
+    {
+      "entropy": 0.944821952097118,
+      "epoch": 1.7798364212986146,
+      "grad_norm": 5.09375,
+      "learning_rate": 1.7780000000000003e-05,
+      "loss": 0.9035,
+      "mean_token_accuracy": 0.7559604255482555,
+      "num_tokens": 26782376.0,
+      "step": 2000
+    },
+    {
+      "epoch": 1.7798364212986146,
+      "eval_coding_entropy": 1.712769726753235,
+      "eval_coding_loss": 2.3114633560180664,
+      "eval_coding_mean_token_accuracy": 0.5093689965009689,
+      "eval_coding_num_tokens": 26782376.0,
+      "eval_coding_runtime": 80.6697,
+      "eval_coding_samples_per_second": 6.198,
+      "eval_coding_steps_per_second": 3.099,
+      "step": 2000
+    },
+    {
+      "epoch": 1.7798364212986146,
+      "eval_chemistry_entropy": 0.9783082293272019,
+      "eval_chemistry_loss": 1.0181097984313965,
+      "eval_chemistry_mean_token_accuracy": 0.7442418551445007,
+      "eval_chemistry_num_tokens": 26782376.0,
+      "eval_chemistry_runtime": 44.0501,
+      "eval_chemistry_samples_per_second": 11.351,
+      "eval_chemistry_steps_per_second": 5.675,
+      "step": 2000
+    },
+    {
+      "entropy": 0.9411061625927687,
+      "epoch": 1.7887386635508817,
+      "grad_norm": 5.84375,
+      "learning_rate": 1.7757777777777777e-05,
+      "loss": 0.9013,
+      "mean_token_accuracy": 0.7582196421921253,
+      "num_tokens": 26913674.0,
+      "step": 2010
+    },
+    {
+      "entropy": 0.9466140553355217,
+      "epoch": 1.7976409058031493,
+      "grad_norm": 4.6875,
+      "learning_rate": 1.7735555555555558e-05,
+      "loss": 0.9284,
+      "mean_token_accuracy": 0.7540723484009504,
+      "num_tokens": 27049648.0,
+      "step": 2020
+    },
+    {
+      "entropy": 0.9803621105849742,
+      "epoch": 1.8065431480554164,
+      "grad_norm": 5.15625,
+      "learning_rate": 1.7713333333333335e-05,
+      "loss": 0.9306,
+      "mean_token_accuracy": 0.7513340801000595,
+      "num_tokens": 27179178.0,
+      "step": 2030
+    },
+    {
+      "entropy": 0.9712000517174602,
+      "epoch": 1.8154453903076837,
+      "grad_norm": 5.9375,
+      "learning_rate": 1.7691111111111113e-05,
+      "loss": 0.9478,
+      "mean_token_accuracy": 0.752249525859952,
+      "num_tokens": 27315061.0,
+      "step": 2040
+    },
+    {
+      "entropy": 0.9317782137542963,
+      "epoch": 1.824347632559951,
+      "grad_norm": 5.28125,
+      "learning_rate": 1.766888888888889e-05,
+      "loss": 0.8958,
+      "mean_token_accuracy": 0.7610440488904715,
+      "num_tokens": 27451473.0,
+      "step": 2050
+    },
+    {
+      "entropy": 0.9378326464444399,
+      "epoch": 1.8332498748122183,
+      "grad_norm": 5.71875,
+      "learning_rate": 1.7646666666666668e-05,
+      "loss": 0.8917,
+      "mean_token_accuracy": 0.7592362996190787,
+      "num_tokens": 27587650.0,
+      "step": 2060
+    },
+    {
+      "entropy": 0.9810647163540125,
+      "epoch": 1.8421521170644857,
+      "grad_norm": 5.375,
+      "learning_rate": 1.7624444444444446e-05,
+      "loss": 0.9384,
+      "mean_token_accuracy": 0.749565614387393,
+      "num_tokens": 27718059.0,
+      "step": 2070
+    },
+    {
+      "entropy": 0.9164110349491239,
+      "epoch": 1.8510543593167528,
+      "grad_norm": 5.0625,
+      "learning_rate": 1.7602222222222223e-05,
+      "loss": 0.8921,
+      "mean_token_accuracy": 0.7612464178353548,
+      "num_tokens": 27855458.0,
+      "step": 2080
+    },
+    {
+      "entropy": 0.983701690286398,
+      "epoch": 1.8599566015690203,
+      "grad_norm": 5.84375,
+      "learning_rate": 1.758e-05,
+      "loss": 0.93,
+      "mean_token_accuracy": 0.7531817588955164,
+      "num_tokens": 27989381.0,
+      "step": 2090
+    },
+    {
+      "entropy": 0.916964913904667,
+      "epoch": 1.8688588438212874,
+      "grad_norm": 5.625,
+      "learning_rate": 1.755777777777778e-05,
+      "loss": 0.885,
+      "mean_token_accuracy": 0.7609612088650465,
+      "num_tokens": 28127703.0,
+      "step": 2100
+    },
+    {
+      "epoch": 1.8688588438212874,
+      "eval_coding_entropy": 1.7431006593704224,
+      "eval_coding_loss": 2.2912614345550537,
+      "eval_coding_mean_token_accuracy": 0.5093610838651658,
+      "eval_coding_num_tokens": 28127703.0,
+      "eval_coding_runtime": 80.8352,
+      "eval_coding_samples_per_second": 6.185,
+      "eval_coding_steps_per_second": 3.093,
+      "step": 2100
+    },
+    {
+      "epoch": 1.8688588438212874,
+      "eval_chemistry_entropy": 0.9440541378259659,
+      "eval_chemistry_loss": 1.0031005144119263,
+      "eval_chemistry_mean_token_accuracy": 0.7476753370761872,
+      "eval_chemistry_num_tokens": 28127703.0,
+      "eval_chemistry_runtime": 44.4802,
+      "eval_chemistry_samples_per_second": 11.241,
+      "eval_chemistry_steps_per_second": 5.62,
+      "step": 2100
+    },
+    {
+      "entropy": 0.9171810772269964,
+      "epoch": 1.8777610860735547,
+      "grad_norm": 5.0625,
+      "learning_rate": 1.7535555555555556e-05,
+      "loss": 0.8602,
+      "mean_token_accuracy": 0.7633361782878637,
+      "num_tokens": 28261081.0,
+      "step": 2110
+    },
+    {
+      "entropy": 0.8876332893967629,
+      "epoch": 1.886663328325822,
+      "grad_norm": 4.875,
+      "learning_rate": 1.7513333333333334e-05,
+      "loss": 0.8625,
+      "mean_token_accuracy": 0.767698946967721,
+      "num_tokens": 28395128.0,
+      "step": 2120
+    },
+    {
+      "entropy": 0.926989870518446,
+      "epoch": 1.8955655705780894,
+      "grad_norm": 5.21875,
+      "learning_rate": 1.749111111111111e-05,
+      "loss": 0.8888,
+      "mean_token_accuracy": 0.7574160646647214,
+      "num_tokens": 28526017.0,
+      "step": 2130
+    },
+    {
+      "entropy": 0.9683940343558788,
+      "epoch": 1.9044678128303567,
+      "grad_norm": 5.4375,
+      "learning_rate": 1.746888888888889e-05,
+      "loss": 0.905,
+      "mean_token_accuracy": 0.7561913870275021,
+      "num_tokens": 28662344.0,
+      "step": 2140
+    },
+    {
+      "entropy": 0.8912612002342939,
+      "epoch": 1.9133700550826238,
+      "grad_norm": 5.5625,
+      "learning_rate": 1.7446666666666667e-05,
+      "loss": 0.866,
+      "mean_token_accuracy": 0.7663097100332379,
+      "num_tokens": 28798041.0,
+      "step": 2150
+    },
+    {
+      "entropy": 0.9882675973698497,
+      "epoch": 1.9222722973348914,
+      "grad_norm": 5.4375,
+      "learning_rate": 1.7424444444444444e-05,
+      "loss": 0.9295,
+      "mean_token_accuracy": 0.7492683254182338,
+      "num_tokens": 28928769.0,
+      "step": 2160
+    },
+    {
+      "entropy": 0.9113500153645873,
+      "epoch": 1.9311745395871585,
+      "grad_norm": 5.53125,
+      "learning_rate": 1.7402222222222222e-05,
+      "loss": 0.8739,
+      "mean_token_accuracy": 0.7638713479042053,
+      "num_tokens": 29058479.0,
+      "step": 2170
+    },
+    {
+      "entropy": 0.9551330221816897,
+      "epoch": 1.9400767818394258,
+      "grad_norm": 5.125,
+      "learning_rate": 1.7380000000000003e-05,
+      "loss": 0.916,
+      "mean_token_accuracy": 0.7549599496647715,
+      "num_tokens": 29186214.0,
+      "step": 2180
+    },
+    {
+      "entropy": 0.9826798997819424,
+      "epoch": 1.948979024091693,
+      "grad_norm": 5.65625,
+      "learning_rate": 1.735777777777778e-05,
+      "loss": 0.9479,
+      "mean_token_accuracy": 0.7481921702623368,
+      "num_tokens": 29311083.0,
+      "step": 2190
+    },
+    {
+      "entropy": 0.9114339074119926,
+      "epoch": 1.9578812663439604,
+      "grad_norm": 5.71875,
+      "learning_rate": 1.7335555555555558e-05,
+      "loss": 0.8971,
+      "mean_token_accuracy": 0.7607169698923826,
+      "num_tokens": 29446445.0,
+      "step": 2200
+    },
+    {
+      "epoch": 1.9578812663439604,
+      "eval_coding_entropy": 1.758779797077179,
+      "eval_coding_loss": 2.3063414096832275,
+      "eval_coding_mean_token_accuracy": 0.5077208431959153,
+      "eval_coding_num_tokens": 29446445.0,
+      "eval_coding_runtime": 80.7301,
+      "eval_coding_samples_per_second": 6.193,
+      "eval_coding_steps_per_second": 3.097,
+      "step": 2200
+    },
+    {
+      "epoch": 1.9578812663439604,
+      "eval_chemistry_entropy": 0.9747029945850373,
+      "eval_chemistry_loss": 0.9925632476806641,
+      "eval_chemistry_mean_token_accuracy": 0.7497075235843659,
+      "eval_chemistry_num_tokens": 29446445.0,
+      "eval_chemistry_runtime": 44.3289,
+      "eval_chemistry_samples_per_second": 11.279,
+      "eval_chemistry_steps_per_second": 5.64,
+      "step": 2200
+    },
+    {
+      "entropy": 0.9837614057585597,
+      "epoch": 1.9667835085962277,
+      "grad_norm": 5.0625,
+      "learning_rate": 1.7313333333333336e-05,
+      "loss": 0.9202,
+      "mean_token_accuracy": 0.7511004183441401,
+      "num_tokens": 29577117.0,
+      "step": 2210
+    },
+    {
+      "entropy": 0.8847709085792304,
+      "epoch": 1.9756857508484948,
+      "grad_norm": 5.53125,
+      "learning_rate": 1.7291111111111113e-05,
+      "loss": 0.8448,
+      "mean_token_accuracy": 0.7678694428876043,
+      "num_tokens": 29711771.0,
+      "step": 2220
+    },
+    {
+      "entropy": 0.9394596133381128,
+      "epoch": 1.9845879931007624,
+      "grad_norm": 4.6875,
+      "learning_rate": 1.726888888888889e-05,
+      "loss": 0.8788,
+      "mean_token_accuracy": 0.7633742790669202,
+      "num_tokens": 29846056.0,
+      "step": 2230
+    },
+    {
+      "entropy": 0.9131581572815776,
+      "epoch": 1.9934902353530295,
+      "grad_norm": 5.46875,
+      "learning_rate": 1.724666666666667e-05,
+      "loss": 0.8999,
+      "mean_token_accuracy": 0.7595547553151846,
+      "num_tokens": 29979994.0,
+      "step": 2240
+    },
+    {
+      "entropy": 0.9380778378288218,
+      "epoch": 2.0017804484504533,
+      "grad_norm": 5.96875,
+      "learning_rate": 1.7224444444444446e-05,
+      "loss": 0.8632,
+      "mean_token_accuracy": 0.7650105865209694,
+      "num_tokens": 30104339.0,
+      "step": 2250
+    },
+    {
+      "entropy": 0.6751446353271604,
+      "epoch": 2.010682690702721,
+      "grad_norm": 5.34375,
+      "learning_rate": 1.7202222222222224e-05,
+      "loss": 0.6607,
+      "mean_token_accuracy": 0.8066760376095772,
+      "num_tokens": 30240793.0,
+      "step": 2260
+    },
+    {
+      "entropy": 0.7322767108678818,
+      "epoch": 2.019584932954988,
+      "grad_norm": 4.3125,
+      "learning_rate": 1.718e-05,
+      "loss": 0.671,
+      "mean_token_accuracy": 0.8030222978442907,
+      "num_tokens": 30374084.0,
+      "step": 2270
+    },
+    {
+      "entropy": 0.6310971546918154,
+      "epoch": 2.0284871752072555,
+      "grad_norm": 5.5625,
+      "learning_rate": 1.715777777777778e-05,
+      "loss": 0.6053,
+      "mean_token_accuracy": 0.8167277801781893,
+      "num_tokens": 30503988.0,
+      "step": 2280
+    },
+    {
+      "entropy": 0.7133538842201232,
+      "epoch": 2.0373894174595226,
+      "grad_norm": 5.28125,
+      "learning_rate": 1.7135555555555557e-05,
+      "loss": 0.6663,
+      "mean_token_accuracy": 0.8047899607568979,
+      "num_tokens": 30639178.0,
+      "step": 2290
+    },
+    {
+      "entropy": 0.673244996368885,
+      "epoch": 2.0462916597117897,
+      "grad_norm": 5.15625,
+      "learning_rate": 1.7113333333333334e-05,
+      "loss": 0.624,
+      "mean_token_accuracy": 0.8107521183788776,
+      "num_tokens": 30771841.0,
+      "step": 2300
+    },
+    {
+      "epoch": 2.0462916597117897,
+      "eval_coding_entropy": 1.3595808396339417,
+      "eval_coding_loss": 2.503028392791748,
+      "eval_coding_mean_token_accuracy": 0.5042585589885712,
+      "eval_coding_num_tokens": 30771841.0,
+      "eval_coding_runtime": 80.7845,
+      "eval_coding_samples_per_second": 6.189,
+      "eval_coding_steps_per_second": 3.095,
+      "step": 2300
+    },
+    {
+      "epoch": 2.0462916597117897,
+      "eval_chemistry_entropy": 0.7220290039777756,
+      "eval_chemistry_loss": 1.044023871421814,
+      "eval_chemistry_mean_token_accuracy": 0.7492555620670318,
+      "eval_chemistry_num_tokens": 30771841.0,
+      "eval_chemistry_runtime": 44.4082,
+      "eval_chemistry_samples_per_second": 11.259,
+      "eval_chemistry_steps_per_second": 5.63,
+      "step": 2300
+    },
+    {
+      "entropy": 0.650447968672961,
+      "epoch": 2.0551939019640573,
+      "grad_norm": 4.84375,
+      "learning_rate": 1.7091111111111112e-05,
+      "loss": 0.6149,
+      "mean_token_accuracy": 0.8162594970315695,
+      "num_tokens": 30911385.0,
+      "step": 2310
+    },
+    {
+      "entropy": 0.6901814254000783,
+      "epoch": 2.0640961442163244,
+      "grad_norm": 5.65625,
+      "learning_rate": 1.706888888888889e-05,
+      "loss": 0.6577,
+      "mean_token_accuracy": 0.807989289984107,
+      "num_tokens": 31045037.0,
+      "step": 2320
+    },
+    {
+      "entropy": 0.7224311189725995,
+      "epoch": 2.072998386468592,
+      "grad_norm": 5.8125,
+      "learning_rate": 1.704666666666667e-05,
+      "loss": 0.6841,
+      "mean_token_accuracy": 0.7975382342934608,
+      "num_tokens": 31171463.0,
+      "step": 2330
+    },
+    {
+      "entropy": 0.7207206780090928,
+      "epoch": 2.081900628720859,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.7024444444444445e-05,
+      "loss": 0.6787,
+      "mean_token_accuracy": 0.8005865413695574,
+      "num_tokens": 31299575.0,
+      "step": 2340
+    },
+    {
+      "entropy": 0.7077139405533671,
+      "epoch": 2.0908028709731266,
+      "grad_norm": 5.75,
+      "learning_rate": 1.7002222222222226e-05,
+      "loss": 0.6649,
+      "mean_token_accuracy": 0.8006750706583261,
+      "num_tokens": 31433410.0,
+      "step": 2350
+    },
+    {
+      "entropy": 0.7052157809957862,
+      "epoch": 2.0997051132253937,
+      "grad_norm": 5.96875,
+      "learning_rate": 1.698e-05,
+      "loss": 0.6952,
+      "mean_token_accuracy": 0.7997228655964136,
+      "num_tokens": 31565963.0,
+      "step": 2360
+    },
+    {
+      "entropy": 0.6745271515101194,
+      "epoch": 2.1086073554776608,
+      "grad_norm": 5.46875,
+      "learning_rate": 1.695777777777778e-05,
+      "loss": 0.6383,
+      "mean_token_accuracy": 0.813958278298378,
+      "num_tokens": 31700651.0,
+      "step": 2370
+    },
+    {
+      "entropy": 0.681758170388639,
+      "epoch": 2.1175095977299283,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.6935555555555555e-05,
+      "loss": 0.6577,
+      "mean_token_accuracy": 0.8078952629119158,
+      "num_tokens": 31834212.0,
+      "step": 2380
+    },
+    {
+      "entropy": 0.6909406213089824,
+      "epoch": 2.1264118399821954,
+      "grad_norm": 5.125,
+      "learning_rate": 1.6913333333333336e-05,
+      "loss": 0.6573,
+      "mean_token_accuracy": 0.8060090221464634,
+      "num_tokens": 31973291.0,
+      "step": 2390
+    },
+    {
+      "entropy": 0.7074554307386279,
+      "epoch": 2.135314082234463,
+      "grad_norm": 5.5,
+      "learning_rate": 1.689111111111111e-05,
+      "loss": 0.6642,
+      "mean_token_accuracy": 0.8029129397124052,
+      "num_tokens": 32103356.0,
+      "step": 2400
+    },
+    {
+      "epoch": 2.135314082234463,
+      "eval_coding_entropy": 1.3474301302433014,
+      "eval_coding_loss": 2.5246565341949463,
+      "eval_coding_mean_token_accuracy": 0.5042781318426133,
+      "eval_coding_num_tokens": 32103356.0,
+      "eval_coding_runtime": 80.7668,
+      "eval_coding_samples_per_second": 6.191,
+      "eval_coding_steps_per_second": 3.095,
+      "step": 2400
+    },
+    {
+      "epoch": 2.135314082234463,
+      "eval_chemistry_entropy": 0.7277124656438827,
+      "eval_chemistry_loss": 1.0443603992462158,
+      "eval_chemistry_mean_token_accuracy": 0.7491718275547028,
+      "eval_chemistry_num_tokens": 32103356.0,
+      "eval_chemistry_runtime": 44.6557,
+      "eval_chemistry_samples_per_second": 11.197,
+      "eval_chemistry_steps_per_second": 5.598,
+      "step": 2400
+    },
+    {
+      "entropy": 0.7237576061859727,
+      "epoch": 2.14421632448673,
+      "grad_norm": 5.28125,
+      "learning_rate": 1.686888888888889e-05,
+      "loss": 0.6829,
+      "mean_token_accuracy": 0.8009689141064882,
+      "num_tokens": 32238918.0,
+      "step": 2410
+    },
+    {
+      "entropy": 0.6664205349981784,
+      "epoch": 2.153118566738997,
+      "grad_norm": 5.28125,
+      "learning_rate": 1.684666666666667e-05,
+      "loss": 0.6276,
+      "mean_token_accuracy": 0.8129349954426288,
+      "num_tokens": 32374830.0,
+      "step": 2420
+    },
+    {
+      "entropy": 0.6417453872039914,
+      "epoch": 2.1620208089912647,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.6824444444444447e-05,
+      "loss": 0.6145,
+      "mean_token_accuracy": 0.8182108178734779,
+      "num_tokens": 32510032.0,
+      "step": 2430
+    },
+    {
+      "entropy": 0.6808106200769544,
+      "epoch": 2.170923051243532,
+      "grad_norm": 5.34375,
+      "learning_rate": 1.6802222222222224e-05,
+      "loss": 0.6513,
+      "mean_token_accuracy": 0.8094353195279836,
+      "num_tokens": 32643479.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.67868527546525,
+      "epoch": 2.1798252934957993,
+      "grad_norm": 5.03125,
+      "learning_rate": 1.6780000000000002e-05,
+      "loss": 0.6431,
+      "mean_token_accuracy": 0.8081219315528869,
+      "num_tokens": 32776239.0,
+      "step": 2450
+    },
+    {
+      "entropy": 0.6879025629721582,
+      "epoch": 2.1887275357480664,
+      "grad_norm": 4.9375,
+      "learning_rate": 1.675777777777778e-05,
+      "loss": 0.6155,
+      "mean_token_accuracy": 0.8085521880537272,
+      "num_tokens": 32908761.0,
+      "step": 2460
+    },
+    {
+      "entropy": 0.7213755307719112,
+      "epoch": 2.197629778000334,
+      "grad_norm": 5.15625,
+      "learning_rate": 1.6735555555555557e-05,
+      "loss": 0.6681,
+      "mean_token_accuracy": 0.8011581059545279,
+      "num_tokens": 33030976.0,
+      "step": 2470
+    },
+    {
+      "entropy": 0.6821736327372492,
+      "epoch": 2.206532020252601,
+      "grad_norm": 5.15625,
+      "learning_rate": 1.6713333333333335e-05,
+      "loss": 0.6526,
+      "mean_token_accuracy": 0.8078901413828135,
+      "num_tokens": 33166413.0,
+      "step": 2480
+    },
+    {
+      "entropy": 0.6923310263082385,
+      "epoch": 2.2154342625048686,
+      "grad_norm": 5.9375,
+      "learning_rate": 1.6691111111111112e-05,
+      "loss": 0.6706,
+      "mean_token_accuracy": 0.8054627288132906,
+      "num_tokens": 33297834.0,
+      "step": 2490
+    },
+    {
+      "entropy": 0.6578195100650192,
+      "epoch": 2.2243365047571357,
+      "grad_norm": 4.875,
+      "learning_rate": 1.666888888888889e-05,
+      "loss": 0.6292,
+      "mean_token_accuracy": 0.8126751393079757,
+      "num_tokens": 33432686.0,
+      "step": 2500
+    },
+    {
+      "epoch": 2.2243365047571357,
+      "eval_coding_entropy": 1.359846888065338,
+      "eval_coding_loss": 2.5393831729888916,
+      "eval_coding_mean_token_accuracy": 0.5013983782529831,
+      "eval_coding_num_tokens": 33432686.0,
+      "eval_coding_runtime": 80.7592,
+      "eval_coding_samples_per_second": 6.191,
+      "eval_coding_steps_per_second": 3.096,
+      "step": 2500
+    },
+    {
+      "epoch": 2.2243365047571357,
+      "eval_chemistry_entropy": 0.7504983623027801,
+      "eval_chemistry_loss": 1.0311698913574219,
+      "eval_chemistry_mean_token_accuracy": 0.7517331650257111,
+      "eval_chemistry_num_tokens": 33432686.0,
+      "eval_chemistry_runtime": 44.3538,
+      "eval_chemistry_samples_per_second": 11.273,
+      "eval_chemistry_steps_per_second": 5.636,
+      "step": 2500
+    },
+    {
+      "entropy": 0.7050275545567274,
+      "epoch": 2.233238747009403,
+      "grad_norm": 4.71875,
+      "learning_rate": 1.6646666666666668e-05,
+      "loss": 0.668,
+      "mean_token_accuracy": 0.8031698003411293,
+      "num_tokens": 33563511.0,
+      "step": 2510
+    },
+    {
+      "entropy": 0.6760105773806572,
+      "epoch": 2.2421409892616704,
+      "grad_norm": 5.53125,
+      "learning_rate": 1.6624444444444445e-05,
+      "loss": 0.6429,
+      "mean_token_accuracy": 0.8107778910547495,
+      "num_tokens": 33693242.0,
+      "step": 2520
+    },
+    {
+      "entropy": 0.6887492289766669,
+      "epoch": 2.2510432315139375,
+      "grad_norm": 5.78125,
+      "learning_rate": 1.6602222222222223e-05,
+      "loss": 0.6424,
+      "mean_token_accuracy": 0.8106433190405369,
+      "num_tokens": 33831513.0,
+      "step": 2530
+    },
+    {
+      "entropy": 0.6382455182261765,
+      "epoch": 2.259945473766205,
+      "grad_norm": 5.125,
+      "learning_rate": 1.658e-05,
+      "loss": 0.6039,
+      "mean_token_accuracy": 0.8192764405161143,
+      "num_tokens": 33970820.0,
+      "step": 2540
+    },
+    {
+      "entropy": 0.641087737120688,
+      "epoch": 2.268847716018472,
+      "grad_norm": 5.8125,
+      "learning_rate": 1.6557777777777778e-05,
+      "loss": 0.6166,
+      "mean_token_accuracy": 0.8172270767390728,
+      "num_tokens": 34107209.0,
+      "step": 2550
+    },
+    {
+      "entropy": 0.7278640048578382,
+      "epoch": 2.2777499582707392,
+      "grad_norm": 5.6875,
+      "learning_rate": 1.6535555555555556e-05,
+      "loss": 0.6825,
+      "mean_token_accuracy": 0.7987151812762022,
+      "num_tokens": 34243042.0,
+      "step": 2560
+    },
+    {
+      "entropy": 0.6859472109004855,
+      "epoch": 2.2866522005230068,
+      "grad_norm": 5.21875,
+      "learning_rate": 1.6513333333333333e-05,
+      "loss": 0.6676,
+      "mean_token_accuracy": 0.8081459250301123,
+      "num_tokens": 34378803.0,
+      "step": 2570
+    },
+    {
+      "entropy": 0.682016589678824,
+      "epoch": 2.295554442775274,
+      "grad_norm": 4.71875,
+      "learning_rate": 1.6491111111111114e-05,
+      "loss": 0.6406,
+      "mean_token_accuracy": 0.8095824711024762,
+      "num_tokens": 34515236.0,
+      "step": 2580
+    },
+    {
+      "entropy": 0.6486262887716293,
+      "epoch": 2.3044566850275414,
+      "grad_norm": 5.375,
+      "learning_rate": 1.646888888888889e-05,
+      "loss": 0.6183,
+      "mean_token_accuracy": 0.8155608825385571,
+      "num_tokens": 34647579.0,
+      "step": 2590
+    },
+    {
+      "entropy": 0.6554163221269846,
+      "epoch": 2.3133589272798085,
+      "grad_norm": 5.625,
+      "learning_rate": 1.644666666666667e-05,
+      "loss": 0.6139,
+      "mean_token_accuracy": 0.8146194528788329,
+      "num_tokens": 34785068.0,
+      "step": 2600
+    },
+    {
+      "epoch": 2.3133589272798085,
+      "eval_coding_entropy": 1.341162795305252,
+      "eval_coding_loss": 2.568235158920288,
+      "eval_coding_mean_token_accuracy": 0.5024716736078262,
+      "eval_coding_num_tokens": 34785068.0,
+      "eval_coding_runtime": 80.9857,
+      "eval_coding_samples_per_second": 6.174,
+      "eval_coding_steps_per_second": 3.087,
+      "step": 2600
+    },
+    {
+      "epoch": 2.3133589272798085,
+      "eval_chemistry_entropy": 0.7452526745796204,
+      "eval_chemistry_loss": 1.027672290802002,
+      "eval_chemistry_mean_token_accuracy": 0.7516778976917267,
+      "eval_chemistry_num_tokens": 34785068.0,
+      "eval_chemistry_runtime": 44.4616,
+      "eval_chemistry_samples_per_second": 11.246,
+      "eval_chemistry_steps_per_second": 5.623,
+      "step": 2600
+    },
+    {
+      "entropy": 0.7047396818175912,
+      "epoch": 2.322261169532076,
+      "grad_norm": 5.40625,
+      "learning_rate": 1.6424444444444444e-05,
+      "loss": 0.6632,
+      "mean_token_accuracy": 0.8033771358430386,
+      "num_tokens": 34919157.0,
+      "step": 2610
+    },
+    {
+      "entropy": 0.6924863770604134,
+      "epoch": 2.331163411784343,
+      "grad_norm": 4.90625,
+      "learning_rate": 1.6402222222222225e-05,
+      "loss": 0.6536,
+      "mean_token_accuracy": 0.8057784728705883,
+      "num_tokens": 35055845.0,
+      "step": 2620
+    },
+    {
+      "entropy": 0.6686609257012606,
+      "epoch": 2.3400656540366107,
+      "grad_norm": 5.3125,
+      "learning_rate": 1.638e-05,
+      "loss": 0.6401,
+      "mean_token_accuracy": 0.8125207956880331,
+      "num_tokens": 35188805.0,
+      "step": 2630
+    },
+    {
+      "entropy": 0.6943664757534862,
+      "epoch": 2.348967896288878,
+      "grad_norm": 5.40625,
+      "learning_rate": 1.635777777777778e-05,
+      "loss": 0.6548,
+      "mean_token_accuracy": 0.8056631416082383,
+      "num_tokens": 35327481.0,
+      "step": 2640
+    },
+    {
+      "entropy": 0.7098184963688254,
+      "epoch": 2.357870138541145,
+      "grad_norm": 5.3125,
+      "learning_rate": 1.6335555555555558e-05,
+      "loss": 0.679,
+      "mean_token_accuracy": 0.8018865004181862,
+      "num_tokens": 35459391.0,
+      "step": 2650
+    },
+    {
+      "entropy": 0.6859694801270961,
+      "epoch": 2.3667723807934125,
+      "grad_norm": 5.34375,
+      "learning_rate": 1.6313333333333335e-05,
+      "loss": 0.6556,
+      "mean_token_accuracy": 0.8075915783643722,
+      "num_tokens": 35596362.0,
+      "step": 2660
+    },
+    {
+      "entropy": 0.6629813998937607,
+      "epoch": 2.3756746230456796,
+      "grad_norm": 5.125,
+      "learning_rate": 1.6291111111111113e-05,
+      "loss": 0.6287,
+      "mean_token_accuracy": 0.8144315045326949,
+      "num_tokens": 35733119.0,
+      "step": 2670
+    },
+    {
+      "entropy": 0.6971998689696193,
+      "epoch": 2.384576865297947,
+      "grad_norm": 5.34375,
+      "learning_rate": 1.626888888888889e-05,
+      "loss": 0.6527,
+      "mean_token_accuracy": 0.8070112053304911,
+      "num_tokens": 35869585.0,
+      "step": 2680
+    },
+    {
+      "entropy": 0.6961546177044511,
+      "epoch": 2.393479107550214,
+      "grad_norm": 4.8125,
+      "learning_rate": 1.6246666666666668e-05,
+      "loss": 0.6466,
+      "mean_token_accuracy": 0.8089884545654058,
+      "num_tokens": 36000516.0,
+      "step": 2690
+    },
+    {
+      "entropy": 0.6580943001434207,
+      "epoch": 2.4023813498024813,
+      "grad_norm": 5.625,
+      "learning_rate": 1.6224444444444446e-05,
+      "loss": 0.6404,
+      "mean_token_accuracy": 0.8133330389857292,
+      "num_tokens": 36134072.0,
+      "step": 2700
+    },
+    {
+      "epoch": 2.4023813498024813,
+      "eval_coding_entropy": 1.3438351492881775,
+      "eval_coding_loss": 2.551229953765869,
+      "eval_coding_mean_token_accuracy": 0.5017013101577759,
+      "eval_coding_num_tokens": 36134072.0,
+      "eval_coding_runtime": 80.8424,
+      "eval_coding_samples_per_second": 6.185,
+      "eval_coding_steps_per_second": 3.092,
+      "step": 2700
+    },
+    {
+      "epoch": 2.4023813498024813,
+      "eval_chemistry_entropy": 0.7395072062015533,
+      "eval_chemistry_loss": 1.0199018716812134,
+      "eval_chemistry_mean_token_accuracy": 0.7541784026622772,
+      "eval_chemistry_num_tokens": 36134072.0,
+      "eval_chemistry_runtime": 44.3917,
+      "eval_chemistry_samples_per_second": 11.263,
+      "eval_chemistry_steps_per_second": 5.632,
+      "step": 2700
+    },
+    {
+      "entropy": 0.6819272680208087,
+      "epoch": 2.411283592054749,
+      "grad_norm": 5.34375,
+      "learning_rate": 1.6202222222222223e-05,
+      "loss": 0.6503,
+      "mean_token_accuracy": 0.8103040024638176,
+      "num_tokens": 36266582.0,
+      "step": 2710
+    },
+    {
+      "entropy": 0.6610682730562984,
+      "epoch": 2.420185834307016,
+      "grad_norm": 5.1875,
+      "learning_rate": 1.618e-05,
+      "loss": 0.644,
+      "mean_token_accuracy": 0.8126811794936657,
+      "num_tokens": 36406263.0,
+      "step": 2720
+    },
+    {
+      "entropy": 0.6747404722496867,
+      "epoch": 2.4290880765592835,
+      "grad_norm": 5.40625,
+      "learning_rate": 1.615777777777778e-05,
+      "loss": 0.6379,
+      "mean_token_accuracy": 0.8103346832096576,
+      "num_tokens": 36544353.0,
+      "step": 2730
+    },
+    {
+      "entropy": 0.6939026135951281,
+      "epoch": 2.4379903188115506,
+      "grad_norm": 6.125,
+      "learning_rate": 1.6135555555555556e-05,
+      "loss": 0.6479,
+      "mean_token_accuracy": 0.8088636647909879,
+      "num_tokens": 36676822.0,
+      "step": 2740
+    },
+    {
+      "entropy": 0.6915046261623502,
+      "epoch": 2.446892561063818,
+      "grad_norm": 5.21875,
+      "learning_rate": 1.6113333333333334e-05,
+      "loss": 0.6531,
+      "mean_token_accuracy": 0.8075901720672846,
+      "num_tokens": 36809542.0,
+      "step": 2750
+    },
+    {
+      "entropy": 0.6678896933794022,
+      "epoch": 2.4557948033160852,
+      "grad_norm": 5.875,
+      "learning_rate": 1.609111111111111e-05,
+      "loss": 0.634,
+      "mean_token_accuracy": 0.8129198741167784,
+      "num_tokens": 36942837.0,
+      "step": 2760
+    },
+    {
+      "entropy": 0.6581751000136137,
+      "epoch": 2.4646970455683523,
+      "grad_norm": 4.8125,
+      "learning_rate": 1.606888888888889e-05,
+      "loss": 0.6243,
+      "mean_token_accuracy": 0.8144248973578214,
+      "num_tokens": 37073852.0,
+      "step": 2770
+    },
+    {
+      "entropy": 0.6735358687117696,
+      "epoch": 2.47359928782062,
+      "grad_norm": 5.84375,
+      "learning_rate": 1.6046666666666667e-05,
+      "loss": 0.6561,
+      "mean_token_accuracy": 0.8086125176399946,
+      "num_tokens": 37210791.0,
+      "step": 2780
+    },
+    {
+      "entropy": 0.6608247993513942,
+      "epoch": 2.482501530072887,
+      "grad_norm": 5.1875,
+      "learning_rate": 1.6024444444444444e-05,
+      "loss": 0.6273,
+      "mean_token_accuracy": 0.8159277431666851,
+      "num_tokens": 37349906.0,
+      "step": 2790
+    },
+    {
+      "entropy": 0.6531393617391587,
+      "epoch": 2.4914037723251545,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.6002222222222222e-05,
+      "loss": 0.6442,
+      "mean_token_accuracy": 0.814842538163066,
+      "num_tokens": 37485822.0,
+      "step": 2800
+    },
+    {
+      "epoch": 2.4914037723251545,
+      "eval_coding_entropy": 1.3514180865287782,
+      "eval_coding_loss": 2.535834550857544,
+      "eval_coding_mean_token_accuracy": 0.5022077556848527,
+      "eval_coding_num_tokens": 37485822.0,
+      "eval_coding_runtime": 80.7305,
+      "eval_coding_samples_per_second": 6.193,
+      "eval_coding_steps_per_second": 3.097,
+      "step": 2800
+    },
+    {
+      "epoch": 2.4914037723251545,
+      "eval_chemistry_entropy": 0.734144370675087,
+      "eval_chemistry_loss": 1.0119251012802124,
+      "eval_chemistry_mean_token_accuracy": 0.7545760450363159,
+      "eval_chemistry_num_tokens": 37485822.0,
+      "eval_chemistry_runtime": 44.3631,
+      "eval_chemistry_samples_per_second": 11.271,
+      "eval_chemistry_steps_per_second": 5.635,
+      "step": 2800
+    },
+    {
+      "entropy": 0.7091395361348987,
+      "epoch": 2.5003060145774216,
+      "grad_norm": 5.75,
+      "learning_rate": 1.5980000000000003e-05,
+      "loss": 0.6636,
+      "mean_token_accuracy": 0.8052146650850773,
+      "num_tokens": 37611329.0,
+      "step": 2810
+    },
+    {
+      "entropy": 0.6451047549024225,
+      "epoch": 2.509208256829689,
+      "grad_norm": 5.5625,
+      "learning_rate": 1.5957777777777777e-05,
+      "loss": 0.6169,
+      "mean_token_accuracy": 0.8182801440358162,
+      "num_tokens": 37746150.0,
+      "step": 2820
+    },
+    {
+      "entropy": 0.6536917492747307,
+      "epoch": 2.5181104990819563,
+      "grad_norm": 5.03125,
+      "learning_rate": 1.5935555555555558e-05,
+      "loss": 0.6318,
+      "mean_token_accuracy": 0.8141068946570158,
+      "num_tokens": 37876900.0,
+      "step": 2830
+    },
+    {
+      "entropy": 0.6732046099379658,
+      "epoch": 2.5270127413342234,
+      "grad_norm": 5.71875,
+      "learning_rate": 1.5913333333333332e-05,
+      "loss": 0.6423,
+      "mean_token_accuracy": 0.8091734830290079,
+      "num_tokens": 38016863.0,
+      "step": 2840
+    },
+    {
+      "entropy": 0.6968869367614389,
+      "epoch": 2.535914983586491,
+      "grad_norm": 5.90625,
+      "learning_rate": 1.5891111111111113e-05,
+      "loss": 0.6583,
+      "mean_token_accuracy": 0.8065891288220882,
+      "num_tokens": 38150165.0,
+      "step": 2850
+    },
+    {
+      "entropy": 0.6866594105958939,
+      "epoch": 2.544817225838758,
+      "grad_norm": 6.6875,
+      "learning_rate": 1.5868888888888888e-05,
+      "loss": 0.6418,
+      "mean_token_accuracy": 0.8063335064798594,
+      "num_tokens": 38277754.0,
+      "step": 2860
+    },
+    {
+      "entropy": 0.6670206711627543,
+      "epoch": 2.5537194680910256,
+      "grad_norm": 5.3125,
+      "learning_rate": 1.584666666666667e-05,
+      "loss": 0.6472,
+      "mean_token_accuracy": 0.8101547524333,
+      "num_tokens": 38412875.0,
+      "step": 2870
+    },
+    {
+      "entropy": 0.7017344998195767,
+      "epoch": 2.5626217103432927,
+      "grad_norm": 4.96875,
+      "learning_rate": 1.5824444444444446e-05,
+      "loss": 0.6731,
+      "mean_token_accuracy": 0.808038018643856,
+      "num_tokens": 38542761.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.6877943404018879,
+      "epoch": 2.5715239525955598,
+      "grad_norm": 5.875,
+      "learning_rate": 1.5802222222222224e-05,
+      "loss": 0.6576,
+      "mean_token_accuracy": 0.8062680896371603,
+      "num_tokens": 38681049.0,
+      "step": 2890
+    },
+    {
+      "entropy": 0.6650467852130533,
+      "epoch": 2.5804261948478273,
+      "grad_norm": 5.84375,
+      "learning_rate": 1.578e-05,
+      "loss": 0.6346,
+      "mean_token_accuracy": 0.8112391501665115,
+      "num_tokens": 38820999.0,
+      "step": 2900
+    },
+    {
+      "epoch": 2.5804261948478273,
+      "eval_coding_entropy": 1.349726175546646,
+      "eval_coding_loss": 2.5497632026672363,
+      "eval_coding_mean_token_accuracy": 0.5023152486085892,
+      "eval_coding_num_tokens": 38820999.0,
+      "eval_coding_runtime": 80.8248,
+      "eval_coding_samples_per_second": 6.186,
+      "eval_coding_steps_per_second": 3.093,
+      "step": 2900
+    },
+    {
+      "epoch": 2.5804261948478273,
+      "eval_chemistry_entropy": 0.730859316110611,
+      "eval_chemistry_loss": 1.0118193626403809,
+      "eval_chemistry_mean_token_accuracy": 0.7558274228572845,
+      "eval_chemistry_num_tokens": 38820999.0,
+      "eval_chemistry_runtime": 44.3858,
+      "eval_chemistry_samples_per_second": 11.265,
+      "eval_chemistry_steps_per_second": 5.632,
+      "step": 2900
+    },
+    {
+      "entropy": 0.6655731063336134,
+      "epoch": 2.589328437100095,
+      "grad_norm": 6.5,
+      "learning_rate": 1.575777777777778e-05,
+      "loss": 0.633,
+      "mean_token_accuracy": 0.814475579187274,
+      "num_tokens": 38947566.0,
+      "step": 2910
+    },
+    {
+      "entropy": 0.6725127598270774,
+      "epoch": 2.598230679352362,
+      "grad_norm": 5.53125,
+      "learning_rate": 1.5735555555555557e-05,
+      "loss": 0.6431,
+      "mean_token_accuracy": 0.8106610123068094,
+      "num_tokens": 39080016.0,
+      "step": 2920
+    },
+    {
+      "entropy": 0.6704443937167526,
+      "epoch": 2.607132921604629,
+      "grad_norm": 5.15625,
+      "learning_rate": 1.5713333333333334e-05,
+      "loss": 0.626,
+      "mean_token_accuracy": 0.8133666381239891,
+      "num_tokens": 39211023.0,
+      "step": 2930
+    },
+    {
+      "entropy": 0.6696614004671574,
+      "epoch": 2.6160351638568966,
+      "grad_norm": 5.5,
+      "learning_rate": 1.5691111111111112e-05,
+      "loss": 0.6411,
+      "mean_token_accuracy": 0.8135382037609815,
+      "num_tokens": 39345434.0,
+      "step": 2940
+    },
+    {
+      "entropy": 0.6526558686047792,
+      "epoch": 2.6249374061091637,
+      "grad_norm": 6.0625,
+      "learning_rate": 1.5668888888888893e-05,
+      "loss": 0.6228,
+      "mean_token_accuracy": 0.8154466915875673,
+      "num_tokens": 39477589.0,
+      "step": 2950
+    },
+    {
+      "entropy": 0.6912691283971071,
+      "epoch": 2.6338396483614313,
+      "grad_norm": 6.3125,
+      "learning_rate": 1.5646666666666667e-05,
+      "loss": 0.6538,
+      "mean_token_accuracy": 0.8079862985759974,
+      "num_tokens": 39610041.0,
+      "step": 2960
+    },
+    {
+      "entropy": 0.6524309016764164,
+      "epoch": 2.6427418906136984,
+      "grad_norm": 5.0625,
+      "learning_rate": 1.5624444444444448e-05,
+      "loss": 0.6221,
+      "mean_token_accuracy": 0.8171541560441256,
+      "num_tokens": 39752300.0,
+      "step": 2970
+    },
+    {
+      "entropy": 0.6497844392433763,
+      "epoch": 2.6516441328659655,
+      "grad_norm": 5.28125,
+      "learning_rate": 1.5602222222222222e-05,
+      "loss": 0.6191,
+      "mean_token_accuracy": 0.816034659743309,
+      "num_tokens": 39884185.0,
+      "step": 2980
+    },
+    {
+      "entropy": 0.7120374557562172,
+      "epoch": 2.660546375118233,
+      "grad_norm": 5.40625,
+      "learning_rate": 1.5580000000000003e-05,
+      "loss": 0.6781,
+      "mean_token_accuracy": 0.8020179472863674,
+      "num_tokens": 40013789.0,
+      "step": 2990
+    },
+    {
+      "entropy": 0.6579019205644727,
+      "epoch": 2.6694486173705,
+      "grad_norm": 6.1875,
+      "learning_rate": 1.5557777777777778e-05,
+      "loss": 0.6214,
+      "mean_token_accuracy": 0.8132367212325334,
+      "num_tokens": 40143543.0,
+      "step": 3000
+    },
+    {
+      "epoch": 2.6694486173705,
+      "eval_coding_entropy": 1.352729766368866,
+      "eval_coding_loss": 2.5435996055603027,
+      "eval_coding_mean_token_accuracy": 0.5027551363706588,
+      "eval_coding_num_tokens": 40143543.0,
+      "eval_coding_runtime": 80.891,
+      "eval_coding_samples_per_second": 6.181,
+      "eval_coding_steps_per_second": 3.091,
+      "step": 3000
+    },
+    {
+      "epoch": 2.6694486173705,
+      "eval_chemistry_entropy": 0.7262587877511978,
+      "eval_chemistry_loss": 1.008080244064331,
+      "eval_chemistry_mean_token_accuracy": 0.7569679837226868,
+      "eval_chemistry_num_tokens": 40143543.0,
+      "eval_chemistry_runtime": 44.0719,
+      "eval_chemistry_samples_per_second": 11.345,
+      "eval_chemistry_steps_per_second": 5.673,
+      "step": 3000
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 10000,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 9,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.578483596497185e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}