diff --git "a/checkpoint-6500/trainer_state.json" "b/checkpoint-6500/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-6500/trainer_state.json"
@@ -0,0 +1,7964 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.783397318199522,
+  "eval_steps": 100,
+  "global_step": 6500,
+  "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
+    },
+    {
+      "entropy": 0.6745331380516291,
+      "epoch": 2.6783508596227676,
+      "grad_norm": 5.53125,
+      "learning_rate": 1.553555555555556e-05,
+      "loss": 0.6292,
+      "mean_token_accuracy": 0.8121513605117798,
+      "num_tokens": 40279074.0,
+      "step": 3010
+    },
+    {
+      "entropy": 0.6798911440186203,
+      "epoch": 2.6872531018750347,
+      "grad_norm": 6.21875,
+      "learning_rate": 1.5513333333333333e-05,
+      "loss": 0.6399,
+      "mean_token_accuracy": 0.8098079148679972,
+      "num_tokens": 40406812.0,
+      "step": 3020
+    },
+    {
+      "entropy": 0.7015228177420795,
+      "epoch": 2.696155344127302,
+      "grad_norm": 4.84375,
+      "learning_rate": 1.5491111111111114e-05,
+      "loss": 0.6698,
+      "mean_token_accuracy": 0.8044562671333552,
+      "num_tokens": 40537421.0,
+      "step": 3030
+    },
+    {
+      "entropy": 0.6907086078077554,
+      "epoch": 2.7050575863795694,
+      "grad_norm": 5.375,
+      "learning_rate": 1.546888888888889e-05,
+      "loss": 0.6583,
+      "mean_token_accuracy": 0.8069438233971595,
+      "num_tokens": 40668543.0,
+      "step": 3040
+    },
+    {
+      "entropy": 0.6553088827058673,
+      "epoch": 2.7139598286318365,
+      "grad_norm": 5.25,
+      "learning_rate": 1.544666666666667e-05,
+      "loss": 0.618,
+      "mean_token_accuracy": 0.8174055136740208,
+      "num_tokens": 40808027.0,
+      "step": 3050
+    },
+    {
+      "entropy": 0.6936577260494232,
+      "epoch": 2.722862070884104,
+      "grad_norm": 5.875,
+      "learning_rate": 1.5424444444444447e-05,
+      "loss": 0.6454,
+      "mean_token_accuracy": 0.8049722142517567,
+      "num_tokens": 40942603.0,
+      "step": 3060
+    },
+    {
+      "entropy": 0.6946257062256336,
+      "epoch": 2.731764313136371,
+      "grad_norm": 5.46875,
+      "learning_rate": 1.5402222222222224e-05,
+      "loss": 0.661,
+      "mean_token_accuracy": 0.8087360549718141,
+      "num_tokens": 41074577.0,
+      "step": 3070
+    },
+    {
+      "entropy": 0.6601529682986438,
+      "epoch": 2.7406665553886382,
+      "grad_norm": 5.84375,
+      "learning_rate": 1.5380000000000002e-05,
+      "loss": 0.624,
+      "mean_token_accuracy": 0.816022364795208,
+      "num_tokens": 41208093.0,
+      "step": 3080
+    },
+    {
+      "entropy": 0.6539908060804009,
+      "epoch": 2.749568797640906,
+      "grad_norm": 5.3125,
+      "learning_rate": 1.535777777777778e-05,
+      "loss": 0.644,
+      "mean_token_accuracy": 0.8129688374698162,
+      "num_tokens": 41341670.0,
+      "step": 3090
+    },
+    {
+      "entropy": 0.6638065639883279,
+      "epoch": 2.7584710398931733,
+      "grad_norm": 5.6875,
+      "learning_rate": 1.5335555555555557e-05,
+      "loss": 0.6345,
+      "mean_token_accuracy": 0.8132123481482267,
+      "num_tokens": 41477562.0,
+      "step": 3100
+    },
+    {
+      "epoch": 2.7584710398931733,
+      "eval_coding_entropy": 1.3336818788051605,
+      "eval_coding_loss": 2.5668866634368896,
+      "eval_coding_mean_token_accuracy": 0.5014687118530273,
+      "eval_coding_num_tokens": 41477562.0,
+      "eval_coding_runtime": 80.6524,
+      "eval_coding_samples_per_second": 6.199,
+      "eval_coding_steps_per_second": 3.1,
+      "step": 3100
+    },
+    {
+      "epoch": 2.7584710398931733,
+      "eval_chemistry_entropy": 0.7200761045217514,
+      "eval_chemistry_loss": 0.9990389347076416,
+      "eval_chemistry_mean_token_accuracy": 0.7580972800254822,
+      "eval_chemistry_num_tokens": 41477562.0,
+      "eval_chemistry_runtime": 44.3239,
+      "eval_chemistry_samples_per_second": 11.281,
+      "eval_chemistry_steps_per_second": 5.64,
+      "step": 3100
+    },
+    {
+      "entropy": 0.6416423838585615,
+      "epoch": 2.7673732821454404,
+      "grad_norm": 5.78125,
+      "learning_rate": 1.5313333333333335e-05,
+      "loss": 0.6136,
+      "mean_token_accuracy": 0.8184031523764134,
+      "num_tokens": 41615276.0,
+      "step": 3110
+    },
+    {
+      "entropy": 0.6809855399653315,
+      "epoch": 2.7762755243977075,
+      "grad_norm": 5.625,
+      "learning_rate": 1.5291111111111112e-05,
+      "loss": 0.653,
+      "mean_token_accuracy": 0.8096095111221076,
+      "num_tokens": 41746571.0,
+      "step": 3120
+    },
+    {
+      "entropy": 0.6584030732512474,
+      "epoch": 2.785177766649975,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.526888888888889e-05,
+      "loss": 0.629,
+      "mean_token_accuracy": 0.815122715011239,
+      "num_tokens": 41875775.0,
+      "step": 3130
+    },
+    {
+      "entropy": 0.6617810314521193,
+      "epoch": 2.794080008902242,
+      "grad_norm": 6.0,
+      "learning_rate": 1.5246666666666668e-05,
+      "loss": 0.6224,
+      "mean_token_accuracy": 0.8135030146688222,
+      "num_tokens": 42011633.0,
+      "step": 3140
+    },
+    {
+      "entropy": 0.694177679810673,
+      "epoch": 2.8029822511545097,
+      "grad_norm": 5.71875,
+      "learning_rate": 1.5224444444444447e-05,
+      "loss": 0.6607,
+      "mean_token_accuracy": 0.8052948355674744,
+      "num_tokens": 42143308.0,
+      "step": 3150
+    },
+    {
+      "entropy": 0.6584483465179801,
+      "epoch": 2.811884493406777,
+      "grad_norm": 5.34375,
+      "learning_rate": 1.5202222222222223e-05,
+      "loss": 0.634,
+      "mean_token_accuracy": 0.814565795660019,
+      "num_tokens": 42283087.0,
+      "step": 3160
+    },
+    {
+      "entropy": 0.6723427005112171,
+      "epoch": 2.820786735659044,
+      "grad_norm": 5.625,
+      "learning_rate": 1.5180000000000002e-05,
+      "loss": 0.6285,
+      "mean_token_accuracy": 0.8106303729116917,
+      "num_tokens": 42420865.0,
+      "step": 3170
+    },
+    {
+      "entropy": 0.6260177625343204,
+      "epoch": 2.8296889779113115,
+      "grad_norm": 4.90625,
+      "learning_rate": 1.5157777777777778e-05,
+      "loss": 0.5945,
+      "mean_token_accuracy": 0.8242932174354791,
+      "num_tokens": 42561534.0,
+      "step": 3180
+    },
+    {
+      "entropy": 0.675075697619468,
+      "epoch": 2.8385912201635786,
+      "grad_norm": 5.46875,
+      "learning_rate": 1.5135555555555557e-05,
+      "loss": 0.6494,
+      "mean_token_accuracy": 0.8105576701462269,
+      "num_tokens": 42693832.0,
+      "step": 3190
+    },
+    {
+      "entropy": 0.6673068619333208,
+      "epoch": 2.847493462415846,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.5113333333333335e-05,
+      "loss": 0.6347,
+      "mean_token_accuracy": 0.8138904709368944,
+      "num_tokens": 42834266.0,
+      "step": 3200
+    },
+    {
+      "epoch": 2.847493462415846,
+      "eval_coding_entropy": 1.3546104590892791,
+      "eval_coding_loss": 2.544434070587158,
+      "eval_coding_mean_token_accuracy": 0.5036582814455033,
+      "eval_coding_num_tokens": 42834266.0,
+      "eval_coding_runtime": 80.6432,
+      "eval_coding_samples_per_second": 6.2,
+      "eval_coding_steps_per_second": 3.1,
+      "step": 3200
+    },
+    {
+      "epoch": 2.847493462415846,
+      "eval_chemistry_entropy": 0.7308159035444259,
+      "eval_chemistry_loss": 0.9910622835159302,
+      "eval_chemistry_mean_token_accuracy": 0.7586741526126861,
+      "eval_chemistry_num_tokens": 42834266.0,
+      "eval_chemistry_runtime": 44.3134,
+      "eval_chemistry_samples_per_second": 11.283,
+      "eval_chemistry_steps_per_second": 5.642,
+      "step": 3200
+    },
+    {
+      "entropy": 0.7081947403028608,
+      "epoch": 2.856395704668113,
+      "grad_norm": 5.53125,
+      "learning_rate": 1.5091111111111113e-05,
+      "loss": 0.647,
+      "mean_token_accuracy": 0.8072695821523667,
+      "num_tokens": 42967494.0,
+      "step": 3210
+    },
+    {
+      "entropy": 0.6875266904011369,
+      "epoch": 2.8652979469203803,
+      "grad_norm": 6.125,
+      "learning_rate": 1.506888888888889e-05,
+      "loss": 0.6712,
+      "mean_token_accuracy": 0.8070441871881485,
+      "num_tokens": 43100513.0,
+      "step": 3220
+    },
+    {
+      "entropy": 0.6705147823318839,
+      "epoch": 2.874200189172648,
+      "grad_norm": 5.5625,
+      "learning_rate": 1.5046666666666668e-05,
+      "loss": 0.6458,
+      "mean_token_accuracy": 0.8108558312058449,
+      "num_tokens": 43240035.0,
+      "step": 3230
+    },
+    {
+      "entropy": 0.6661259103566408,
+      "epoch": 2.8831024314249154,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.5024444444444445e-05,
+      "loss": 0.6249,
+      "mean_token_accuracy": 0.815429400652647,
+      "num_tokens": 43375632.0,
+      "step": 3240
+    },
+    {
+      "entropy": 0.6251379268243908,
+      "epoch": 2.8920046736771825,
+      "grad_norm": 5.125,
+      "learning_rate": 1.5002222222222223e-05,
+      "loss": 0.6021,
+      "mean_token_accuracy": 0.8213659271597862,
+      "num_tokens": 43514400.0,
+      "step": 3250
+    },
+    {
+      "entropy": 0.6793911940418184,
+      "epoch": 2.9009069159294496,
+      "grad_norm": 5.78125,
+      "learning_rate": 1.498e-05,
+      "loss": 0.6521,
+      "mean_token_accuracy": 0.8136643651872874,
+      "num_tokens": 43644341.0,
+      "step": 3260
+    },
+    {
+      "entropy": 0.6614615533500909,
+      "epoch": 2.909809158181717,
+      "grad_norm": 5.6875,
+      "learning_rate": 1.495777777777778e-05,
+      "loss": 0.6333,
+      "mean_token_accuracy": 0.8129747349768877,
+      "num_tokens": 43781318.0,
+      "step": 3270
+    },
+    {
+      "entropy": 0.6454314555041492,
+      "epoch": 2.9187114004339842,
+      "grad_norm": 5.28125,
+      "learning_rate": 1.4935555555555556e-05,
+      "loss": 0.5978,
+      "mean_token_accuracy": 0.8210929311811924,
+      "num_tokens": 43917580.0,
+      "step": 3280
+    },
+    {
+      "entropy": 0.6484104382805527,
+      "epoch": 2.927613642686252,
+      "grad_norm": 5.8125,
+      "learning_rate": 1.4913333333333335e-05,
+      "loss": 0.6214,
+      "mean_token_accuracy": 0.816330011188984,
+      "num_tokens": 44048559.0,
+      "step": 3290
+    },
+    {
+      "entropy": 0.6625750293955207,
+      "epoch": 2.936515884938519,
+      "grad_norm": 5.65625,
+      "learning_rate": 1.4891111111111111e-05,
+      "loss": 0.6249,
+      "mean_token_accuracy": 0.8133521389216185,
+      "num_tokens": 44177325.0,
+      "step": 3300
+    },
+    {
+      "epoch": 2.936515884938519,
+      "eval_coding_entropy": 1.340244551897049,
+      "eval_coding_loss": 2.546386480331421,
+      "eval_coding_mean_token_accuracy": 0.5042038741111755,
+      "eval_coding_num_tokens": 44177325.0,
+      "eval_coding_runtime": 80.6967,
+      "eval_coding_samples_per_second": 6.196,
+      "eval_coding_steps_per_second": 3.098,
+      "step": 3300
+    },
+    {
+      "epoch": 2.936515884938519,
+      "eval_chemistry_entropy": 0.7220234467983245,
+      "eval_chemistry_loss": 0.9898947477340698,
+      "eval_chemistry_mean_token_accuracy": 0.758931795835495,
+      "eval_chemistry_num_tokens": 44177325.0,
+      "eval_chemistry_runtime": 44.3614,
+      "eval_chemistry_samples_per_second": 11.271,
+      "eval_chemistry_steps_per_second": 5.636,
+      "step": 3300
+    },
+    {
+      "entropy": 0.6771549770608545,
+      "epoch": 2.945418127190786,
+      "grad_norm": 5.3125,
+      "learning_rate": 1.486888888888889e-05,
+      "loss": 0.642,
+      "mean_token_accuracy": 0.8109464462846517,
+      "num_tokens": 44311351.0,
+      "step": 3310
+    },
+    {
+      "entropy": 0.6604766797274351,
+      "epoch": 2.9543203694430535,
+      "grad_norm": 5.34375,
+      "learning_rate": 1.4846666666666666e-05,
+      "loss": 0.6346,
+      "mean_token_accuracy": 0.8145252376794815,
+      "num_tokens": 44442947.0,
+      "step": 3320
+    },
+    {
+      "entropy": 0.6826975420117378,
+      "epoch": 2.9632226116953206,
+      "grad_norm": 6.53125,
+      "learning_rate": 1.4824444444444446e-05,
+      "loss": 0.645,
+      "mean_token_accuracy": 0.8113802324980497,
+      "num_tokens": 44571899.0,
+      "step": 3330
+    },
+    {
+      "entropy": 0.6759714109823107,
+      "epoch": 2.972124853947588,
+      "grad_norm": 5.96875,
+      "learning_rate": 1.4802222222222225e-05,
+      "loss": 0.6459,
+      "mean_token_accuracy": 0.8092896059155464,
+      "num_tokens": 44700601.0,
+      "step": 3340
+    },
+    {
+      "entropy": 0.6581103198230267,
+      "epoch": 2.9810270961998553,
+      "grad_norm": 6.03125,
+      "learning_rate": 1.478e-05,
+      "loss": 0.6321,
+      "mean_token_accuracy": 0.8154440209269523,
+      "num_tokens": 44828600.0,
+      "step": 3350
+    },
+    {
+      "entropy": 0.6576101597398519,
+      "epoch": 2.9899293384521224,
+      "grad_norm": 5.46875,
+      "learning_rate": 1.475777777777778e-05,
+      "loss": 0.6166,
+      "mean_token_accuracy": 0.818338543176651,
+      "num_tokens": 44971479.0,
+      "step": 3360
+    },
+    {
+      "entropy": 0.6600229930132627,
+      "epoch": 2.99883158070439,
+      "grad_norm": 5.625,
+      "learning_rate": 1.4735555555555556e-05,
+      "loss": 0.6117,
+      "mean_token_accuracy": 0.8145487550646067,
+      "num_tokens": 45101238.0,
+      "step": 3370
+    },
+    {
+      "entropy": 0.49347520294605485,
+      "epoch": 3.0071217938018138,
+      "grad_norm": 6.15625,
+      "learning_rate": 1.4713333333333335e-05,
+      "loss": 0.4202,
+      "mean_token_accuracy": 0.8740264201324258,
+      "num_tokens": 45231796.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.4081104312092066,
+      "epoch": 3.0160240360540813,
+      "grad_norm": 5.5625,
+      "learning_rate": 1.4691111111111111e-05,
+      "loss": 0.3834,
+      "mean_token_accuracy": 0.8790286026895047,
+      "num_tokens": 45365444.0,
+      "step": 3390
+    },
+    {
+      "entropy": 0.41986756222322585,
+      "epoch": 3.0249262783063484,
+      "grad_norm": 6.375,
+      "learning_rate": 1.466888888888889e-05,
+      "loss": 0.3647,
+      "mean_token_accuracy": 0.8841583203524351,
+      "num_tokens": 45500317.0,
+      "step": 3400
+    },
+    {
+      "epoch": 3.0249262783063484,
+      "eval_coding_entropy": 1.000330668926239,
+      "eval_coding_loss": 3.113379955291748,
+      "eval_coding_mean_token_accuracy": 0.48467672312259674,
+      "eval_coding_num_tokens": 45500317.0,
+      "eval_coding_runtime": 80.8769,
+      "eval_coding_samples_per_second": 6.182,
+      "eval_coding_steps_per_second": 3.091,
+      "step": 3400
+    },
+    {
+      "epoch": 3.0249262783063484,
+      "eval_chemistry_entropy": 0.5033477848768234,
+      "eval_chemistry_loss": 1.2048485279083252,
+      "eval_chemistry_mean_token_accuracy": 0.7507229104042054,
+      "eval_chemistry_num_tokens": 45500317.0,
+      "eval_chemistry_runtime": 44.3578,
+      "eval_chemistry_samples_per_second": 11.272,
+      "eval_chemistry_steps_per_second": 5.636,
+      "step": 3400
+    },
+    {
+      "entropy": 0.3812009177170694,
+      "epoch": 3.0338285205586155,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.4646666666666666e-05,
+      "loss": 0.3336,
+      "mean_token_accuracy": 0.8897557239979506,
+      "num_tokens": 45627891.0,
+      "step": 3410
+    },
+    {
+      "entropy": 0.3608863881789148,
+      "epoch": 3.042730762810883,
+      "grad_norm": 5.46875,
+      "learning_rate": 1.4624444444444446e-05,
+      "loss": 0.3279,
+      "mean_token_accuracy": 0.8938877809792757,
+      "num_tokens": 45762680.0,
+      "step": 3420
+    },
+    {
+      "entropy": 0.3777980367653072,
+      "epoch": 3.05163300506315,
+      "grad_norm": 6.09375,
+      "learning_rate": 1.4602222222222225e-05,
+      "loss": 0.3369,
+      "mean_token_accuracy": 0.8900566104799509,
+      "num_tokens": 45896651.0,
+      "step": 3430
+    },
+    {
+      "entropy": 0.374933801824227,
+      "epoch": 3.0605352473154177,
+      "grad_norm": 5.84375,
+      "learning_rate": 1.4580000000000001e-05,
+      "loss": 0.334,
+      "mean_token_accuracy": 0.8923791516572237,
+      "num_tokens": 46029923.0,
+      "step": 3440
+    },
+    {
+      "entropy": 0.379786391556263,
+      "epoch": 3.069437489567685,
+      "grad_norm": 5.78125,
+      "learning_rate": 1.455777777777778e-05,
+      "loss": 0.3445,
+      "mean_token_accuracy": 0.8881295207887888,
+      "num_tokens": 46173467.0,
+      "step": 3450
+    },
+    {
+      "entropy": 0.37924998281523586,
+      "epoch": 3.0783397318199524,
+      "grad_norm": 5.78125,
+      "learning_rate": 1.4535555555555556e-05,
+      "loss": 0.3406,
+      "mean_token_accuracy": 0.8885490503162146,
+      "num_tokens": 46305156.0,
+      "step": 3460
+    },
+    {
+      "entropy": 0.3563411055132747,
+      "epoch": 3.0872419740722195,
+      "grad_norm": 6.15625,
+      "learning_rate": 1.4513333333333336e-05,
+      "loss": 0.3265,
+      "mean_token_accuracy": 0.8946562297642231,
+      "num_tokens": 46446562.0,
+      "step": 3470
+    },
+    {
+      "entropy": 0.38276689909398554,
+      "epoch": 3.0961442163244866,
+      "grad_norm": 5.75,
+      "learning_rate": 1.4491111111111111e-05,
+      "loss": 0.3449,
+      "mean_token_accuracy": 0.8880472250282765,
+      "num_tokens": 46582874.0,
+      "step": 3480
+    },
+    {
+      "entropy": 0.3743845520541072,
+      "epoch": 3.105046458576754,
+      "grad_norm": 5.875,
+      "learning_rate": 1.446888888888889e-05,
+      "loss": 0.3447,
+      "mean_token_accuracy": 0.8908812377601862,
+      "num_tokens": 46715265.0,
+      "step": 3490
+    },
+    {
+      "entropy": 0.397824642714113,
+      "epoch": 3.113948700829021,
+      "grad_norm": 7.71875,
+      "learning_rate": 1.4446666666666668e-05,
+      "loss": 0.3672,
+      "mean_token_accuracy": 0.8840487454086542,
+      "num_tokens": 46850175.0,
+      "step": 3500
+    },
+    {
+      "epoch": 3.113948700829021,
+      "eval_coding_entropy": 0.977915655374527,
+      "eval_coding_loss": 3.1338014602661133,
+      "eval_coding_mean_token_accuracy": 0.48520465755462644,
+      "eval_coding_num_tokens": 46850175.0,
+      "eval_coding_runtime": 80.6689,
+      "eval_coding_samples_per_second": 6.198,
+      "eval_coding_steps_per_second": 3.099,
+      "step": 3500
+    },
+    {
+      "epoch": 3.113948700829021,
+      "eval_chemistry_entropy": 0.49352808421850203,
+      "eval_chemistry_loss": 1.2194854021072388,
+      "eval_chemistry_mean_token_accuracy": 0.7510249562263489,
+      "eval_chemistry_num_tokens": 46850175.0,
+      "eval_chemistry_runtime": 44.1305,
+      "eval_chemistry_samples_per_second": 11.33,
+      "eval_chemistry_steps_per_second": 5.665,
+      "step": 3500
+    },
+    {
+      "entropy": 0.3744693847373128,
+      "epoch": 3.1228509430812887,
+      "grad_norm": 6.5625,
+      "learning_rate": 1.4424444444444446e-05,
+      "loss": 0.341,
+      "mean_token_accuracy": 0.8903376907110214,
+      "num_tokens": 46985661.0,
+      "step": 3510
+    },
+    {
+      "entropy": 0.3711514542810619,
+      "epoch": 3.131753185333556,
+      "grad_norm": 6.15625,
+      "learning_rate": 1.4402222222222224e-05,
+      "loss": 0.3367,
+      "mean_token_accuracy": 0.8904415719211102,
+      "num_tokens": 47125899.0,
+      "step": 3520
+    },
+    {
+      "entropy": 0.3872178150340915,
+      "epoch": 3.1406554275858234,
+      "grad_norm": 6.25,
+      "learning_rate": 1.4380000000000001e-05,
+      "loss": 0.344,
+      "mean_token_accuracy": 0.8865831214934587,
+      "num_tokens": 47260838.0,
+      "step": 3530
+    },
+    {
+      "entropy": 0.39034450743347404,
+      "epoch": 3.1495576698380905,
+      "grad_norm": 6.125,
+      "learning_rate": 1.4357777777777779e-05,
+      "loss": 0.3536,
+      "mean_token_accuracy": 0.8871991705149412,
+      "num_tokens": 47390194.0,
+      "step": 3540
+    },
+    {
+      "entropy": 0.3801939754746854,
+      "epoch": 3.1584599120903576,
+      "grad_norm": 6.25,
+      "learning_rate": 1.4335555555555556e-05,
+      "loss": 0.347,
+      "mean_token_accuracy": 0.888665160536766,
+      "num_tokens": 47521475.0,
+      "step": 3550
+    },
+    {
+      "entropy": 0.38943441603332757,
+      "epoch": 3.167362154342625,
+      "grad_norm": 5.71875,
+      "learning_rate": 1.4313333333333334e-05,
+      "loss": 0.3578,
+      "mean_token_accuracy": 0.8856595657765866,
+      "num_tokens": 47654784.0,
+      "step": 3560
+    },
+    {
+      "entropy": 0.3890751742757857,
+      "epoch": 3.1762643965948922,
+      "grad_norm": 6.28125,
+      "learning_rate": 1.4291111111111113e-05,
+      "loss": 0.3453,
+      "mean_token_accuracy": 0.886602982133627,
+      "num_tokens": 47789404.0,
+      "step": 3570
+    },
+    {
+      "entropy": 0.37524048453196884,
+      "epoch": 3.18516663884716,
+      "grad_norm": 7.34375,
+      "learning_rate": 1.426888888888889e-05,
+      "loss": 0.3419,
+      "mean_token_accuracy": 0.890496239438653,
+      "num_tokens": 47919400.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.39707792829722166,
+      "epoch": 3.194068881099427,
+      "grad_norm": 6.0,
+      "learning_rate": 1.4246666666666669e-05,
+      "loss": 0.3625,
+      "mean_token_accuracy": 0.8839322429150342,
+      "num_tokens": 48054367.0,
+      "step": 3590
+    },
+    {
+      "entropy": 0.3782722636125982,
+      "epoch": 3.202971123351694,
+      "grad_norm": 5.96875,
+      "learning_rate": 1.4224444444444445e-05,
+      "loss": 0.3495,
+      "mean_token_accuracy": 0.888066741079092,
+      "num_tokens": 48187796.0,
+      "step": 3600
+    },
+    {
+      "epoch": 3.202971123351694,
+      "eval_coding_entropy": 0.9752645392417908,
+      "eval_coding_loss": 3.160203218460083,
+      "eval_coding_mean_token_accuracy": 0.4847764061689377,
+      "eval_coding_num_tokens": 48187796.0,
+      "eval_coding_runtime": 80.8158,
+      "eval_coding_samples_per_second": 6.187,
+      "eval_coding_steps_per_second": 3.093,
+      "step": 3600
+    },
+    {
+      "epoch": 3.202971123351694,
+      "eval_chemistry_entropy": 0.49633402544260025,
+      "eval_chemistry_loss": 1.2185009717941284,
+      "eval_chemistry_mean_token_accuracy": 0.7517516205310821,
+      "eval_chemistry_num_tokens": 48187796.0,
+      "eval_chemistry_runtime": 44.3971,
+      "eval_chemistry_samples_per_second": 11.262,
+      "eval_chemistry_steps_per_second": 5.631,
+      "step": 3600
+    },
+    {
+      "entropy": 0.3712208589538932,
+      "epoch": 3.2118733656039615,
+      "grad_norm": 6.28125,
+      "learning_rate": 1.4202222222222224e-05,
+      "loss": 0.3353,
+      "mean_token_accuracy": 0.8915024094283581,
+      "num_tokens": 48325104.0,
+      "step": 3610
+    },
+    {
+      "entropy": 0.38365742797032,
+      "epoch": 3.2207756078562286,
+      "grad_norm": 6.28125,
+      "learning_rate": 1.418e-05,
+      "loss": 0.3507,
+      "mean_token_accuracy": 0.8865523029118776,
+      "num_tokens": 48457428.0,
+      "step": 3620
+    },
+    {
+      "entropy": 0.41866554114967586,
+      "epoch": 3.229677850108496,
+      "grad_norm": 6.5,
+      "learning_rate": 1.4157777777777779e-05,
+      "loss": 0.3641,
+      "mean_token_accuracy": 0.8780390486121178,
+      "num_tokens": 48586505.0,
+      "step": 3630
+    },
+    {
+      "entropy": 0.40869386503472926,
+      "epoch": 3.2385800923607633,
+      "grad_norm": 6.5,
+      "learning_rate": 1.4135555555555555e-05,
+      "loss": 0.3605,
+      "mean_token_accuracy": 0.8820974063128233,
+      "num_tokens": 48716417.0,
+      "step": 3640
+    },
+    {
+      "entropy": 0.37882136218249796,
+      "epoch": 3.247482334613031,
+      "grad_norm": 5.3125,
+      "learning_rate": 1.4113333333333334e-05,
+      "loss": 0.3427,
+      "mean_token_accuracy": 0.8887066043913364,
+      "num_tokens": 48850508.0,
+      "step": 3650
+    },
+    {
+      "entropy": 0.40716419126838443,
+      "epoch": 3.256384576865298,
+      "grad_norm": 7.4375,
+      "learning_rate": 1.4091111111111114e-05,
+      "loss": 0.3608,
+      "mean_token_accuracy": 0.8827173482626677,
+      "num_tokens": 48977306.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.3971007606945932,
+      "epoch": 3.2652868191175655,
+      "grad_norm": 5.84375,
+      "learning_rate": 1.406888888888889e-05,
+      "loss": 0.3531,
+      "mean_token_accuracy": 0.8846440050750971,
+      "num_tokens": 49110796.0,
+      "step": 3670
+    },
+    {
+      "entropy": 0.36585476556792856,
+      "epoch": 3.2741890613698326,
+      "grad_norm": 5.9375,
+      "learning_rate": 1.4046666666666669e-05,
+      "loss": 0.3371,
+      "mean_token_accuracy": 0.8907744135707617,
+      "num_tokens": 49249667.0,
+      "step": 3680
+    },
+    {
+      "entropy": 0.3798886900767684,
+      "epoch": 3.2830913036220997,
+      "grad_norm": 5.78125,
+      "learning_rate": 1.4024444444444445e-05,
+      "loss": 0.3477,
+      "mean_token_accuracy": 0.8869309145957232,
+      "num_tokens": 49386582.0,
+      "step": 3690
+    },
+    {
+      "entropy": 0.3772418413311243,
+      "epoch": 3.291993545874367,
+      "grad_norm": 6.21875,
+      "learning_rate": 1.4002222222222224e-05,
+      "loss": 0.3434,
+      "mean_token_accuracy": 0.8884297959506512,
+      "num_tokens": 49520900.0,
+      "step": 3700
+    },
+    {
+      "epoch": 3.291993545874367,
+      "eval_coding_entropy": 0.9666137053966523,
+      "eval_coding_loss": 3.1675589084625244,
+      "eval_coding_mean_token_accuracy": 0.48544125580787656,
+      "eval_coding_num_tokens": 49520900.0,
+      "eval_coding_runtime": 81.1383,
+      "eval_coding_samples_per_second": 6.162,
+      "eval_coding_steps_per_second": 3.081,
+      "step": 3700
+    },
+    {
+      "epoch": 3.291993545874367,
+      "eval_chemistry_entropy": 0.49393032884597776,
+      "eval_chemistry_loss": 1.2216311693191528,
+      "eval_chemistry_mean_token_accuracy": 0.7508244645595551,
+      "eval_chemistry_num_tokens": 49520900.0,
+      "eval_chemistry_runtime": 44.8333,
+      "eval_chemistry_samples_per_second": 11.152,
+      "eval_chemistry_steps_per_second": 5.576,
+      "step": 3700
+    },
+    {
+      "entropy": 0.40607556868344546,
+      "epoch": 3.3008957881266343,
+      "grad_norm": 6.25,
+      "learning_rate": 1.398e-05,
+      "loss": 0.3706,
+      "mean_token_accuracy": 0.8812401454895735,
+      "num_tokens": 49651961.0,
+      "step": 3710
+    },
+    {
+      "entropy": 0.38907681507989766,
+      "epoch": 3.309798030378902,
+      "grad_norm": 6.6875,
+      "learning_rate": 1.395777777777778e-05,
+      "loss": 0.3513,
+      "mean_token_accuracy": 0.8861382525414229,
+      "num_tokens": 49786411.0,
+      "step": 3720
+    },
+    {
+      "entropy": 0.38017242583446204,
+      "epoch": 3.318700272631169,
+      "grad_norm": 6.46875,
+      "learning_rate": 1.3935555555555557e-05,
+      "loss": 0.3414,
+      "mean_token_accuracy": 0.8890792105346919,
+      "num_tokens": 49922597.0,
+      "step": 3730
+    },
+    {
+      "entropy": 0.3877768457867205,
+      "epoch": 3.327602514883436,
+      "grad_norm": 6.4375,
+      "learning_rate": 1.3913333333333335e-05,
+      "loss": 0.3537,
+      "mean_token_accuracy": 0.8852655585855246,
+      "num_tokens": 50056207.0,
+      "step": 3740
+    },
+    {
+      "entropy": 0.3718561602756381,
+      "epoch": 3.3365047571357036,
+      "grad_norm": 6.3125,
+      "learning_rate": 1.3891111111111114e-05,
+      "loss": 0.3399,
+      "mean_token_accuracy": 0.8905533041805029,
+      "num_tokens": 50197267.0,
+      "step": 3750
+    },
+    {
+      "entropy": 0.40157550387084484,
+      "epoch": 3.3454069993879707,
+      "grad_norm": 5.9375,
+      "learning_rate": 1.386888888888889e-05,
+      "loss": 0.3624,
+      "mean_token_accuracy": 0.8817423343658447,
+      "num_tokens": 50330876.0,
+      "step": 3760
+    },
+    {
+      "entropy": 0.3865166649222374,
+      "epoch": 3.3543092416402382,
+      "grad_norm": 5.4375,
+      "learning_rate": 1.3846666666666669e-05,
+      "loss": 0.3516,
+      "mean_token_accuracy": 0.8865749329328537,
+      "num_tokens": 50470644.0,
+      "step": 3770
+    },
+    {
+      "entropy": 0.39948349092155694,
+      "epoch": 3.3632114838925053,
+      "grad_norm": 6.4375,
+      "learning_rate": 1.3824444444444445e-05,
+      "loss": 0.3562,
+      "mean_token_accuracy": 0.8841321788728237,
+      "num_tokens": 50602004.0,
+      "step": 3780
+    },
+    {
+      "entropy": 0.3734517255797982,
+      "epoch": 3.372113726144773,
+      "grad_norm": 6.28125,
+      "learning_rate": 1.3802222222222224e-05,
+      "loss": 0.3351,
+      "mean_token_accuracy": 0.88988156914711,
+      "num_tokens": 50738870.0,
+      "step": 3790
+    },
+    {
+      "entropy": 0.3977280559018254,
+      "epoch": 3.38101596839704,
+      "grad_norm": 6.28125,
+      "learning_rate": 1.378e-05,
+      "loss": 0.3614,
+      "mean_token_accuracy": 0.8829979106783867,
+      "num_tokens": 50873444.0,
+      "step": 3800
+    },
+    {
+      "epoch": 3.38101596839704,
+      "eval_coding_entropy": 0.9476412839889526,
+      "eval_coding_loss": 3.217289447784424,
+      "eval_coding_mean_token_accuracy": 0.4827978718280792,
+      "eval_coding_num_tokens": 50873444.0,
+      "eval_coding_runtime": 81.1382,
+      "eval_coding_samples_per_second": 6.162,
+      "eval_coding_steps_per_second": 3.081,
+      "step": 3800
+    },
+    {
+      "epoch": 3.38101596839704,
+      "eval_chemistry_entropy": 0.48563810282945635,
+      "eval_chemistry_loss": 1.2338879108428955,
+      "eval_chemistry_mean_token_accuracy": 0.7510707828998565,
+      "eval_chemistry_num_tokens": 50873444.0,
+      "eval_chemistry_runtime": 44.8497,
+      "eval_chemistry_samples_per_second": 11.148,
+      "eval_chemistry_steps_per_second": 5.574,
+      "step": 3800
+    },
+    {
+      "entropy": 0.3870542447082698,
+      "epoch": 3.3899182106493075,
+      "grad_norm": 6.5625,
+      "learning_rate": 1.375777777777778e-05,
+      "loss": 0.3594,
+      "mean_token_accuracy": 0.8847264006733895,
+      "num_tokens": 51006529.0,
+      "step": 3810
+    },
+    {
+      "entropy": 0.3994511841796339,
+      "epoch": 3.3988204529015746,
+      "grad_norm": 6.5625,
+      "learning_rate": 1.3735555555555557e-05,
+      "loss": 0.357,
+      "mean_token_accuracy": 0.8831588868051767,
+      "num_tokens": 51141077.0,
+      "step": 3820
+    },
+    {
+      "entropy": 0.3977972437627614,
+      "epoch": 3.4077226951538417,
+      "grad_norm": 6.65625,
+      "learning_rate": 1.3713333333333335e-05,
+      "loss": 0.3585,
+      "mean_token_accuracy": 0.8825859725475311,
+      "num_tokens": 51275677.0,
+      "step": 3830
+    },
+    {
+      "entropy": 0.39362509176135063,
+      "epoch": 3.4166249374061093,
+      "grad_norm": 6.25,
+      "learning_rate": 1.3691111111111112e-05,
+      "loss": 0.3552,
+      "mean_token_accuracy": 0.8844130784273148,
+      "num_tokens": 51406483.0,
+      "step": 3840
+    },
+    {
+      "entropy": 0.41689663371071217,
+      "epoch": 3.4255271796583764,
+      "grad_norm": 6.84375,
+      "learning_rate": 1.366888888888889e-05,
+      "loss": 0.361,
+      "mean_token_accuracy": 0.8807850524783134,
+      "num_tokens": 51536468.0,
+      "step": 3850
+    },
+    {
+      "entropy": 0.38266281606629493,
+      "epoch": 3.434429421910644,
+      "grad_norm": 6.1875,
+      "learning_rate": 1.3646666666666668e-05,
+      "loss": 0.3415,
+      "mean_token_accuracy": 0.8877913266420364,
+      "num_tokens": 51668458.0,
+      "step": 3860
+    },
+    {
+      "entropy": 0.38727299189195036,
+      "epoch": 3.443331664162911,
+      "grad_norm": 7.3125,
+      "learning_rate": 1.3624444444444445e-05,
+      "loss": 0.3504,
+      "mean_token_accuracy": 0.8869855593889951,
+      "num_tokens": 51804123.0,
+      "step": 3870
+    },
+    {
+      "entropy": 0.405052235070616,
+      "epoch": 3.452233906415178,
+      "grad_norm": 6.8125,
+      "learning_rate": 1.3602222222222223e-05,
+      "loss": 0.3702,
+      "mean_token_accuracy": 0.8813161548227072,
+      "num_tokens": 51928199.0,
+      "step": 3880
+    },
+    {
+      "entropy": 0.3869968270882964,
+      "epoch": 3.4611361486674457,
+      "grad_norm": 5.96875,
+      "learning_rate": 1.3580000000000002e-05,
+      "loss": 0.35,
+      "mean_token_accuracy": 0.8862483523786068,
+      "num_tokens": 52061534.0,
+      "step": 3890
+    },
+    {
+      "entropy": 0.38898652140051126,
+      "epoch": 3.470038390919713,
+      "grad_norm": 5.8125,
+      "learning_rate": 1.3557777777777778e-05,
+      "loss": 0.3521,
+      "mean_token_accuracy": 0.8868940342217684,
+      "num_tokens": 52199244.0,
+      "step": 3900
+    },
+    {
+      "epoch": 3.470038390919713,
+      "eval_coding_entropy": 0.9591941924095154,
+      "eval_coding_loss": 3.206862449645996,
+      "eval_coding_mean_token_accuracy": 0.48201393818855287,
+      "eval_coding_num_tokens": 52199244.0,
+      "eval_coding_runtime": 81.0999,
+      "eval_coding_samples_per_second": 6.165,
+      "eval_coding_steps_per_second": 3.083,
+      "step": 3900
+    },
+    {
+      "epoch": 3.470038390919713,
+      "eval_chemistry_entropy": 0.49060807263851164,
+      "eval_chemistry_loss": 1.2261359691619873,
+      "eval_chemistry_mean_token_accuracy": 0.7514483702182769,
+      "eval_chemistry_num_tokens": 52199244.0,
+      "eval_chemistry_runtime": 44.8149,
+      "eval_chemistry_samples_per_second": 11.157,
+      "eval_chemistry_steps_per_second": 5.578,
+      "step": 3900
+    },
+    {
+      "entropy": 0.3775276933796704,
+      "epoch": 3.4789406331719803,
+      "grad_norm": 6.1875,
+      "learning_rate": 1.3535555555555557e-05,
+      "loss": 0.3513,
+      "mean_token_accuracy": 0.8891081947833299,
+      "num_tokens": 52336598.0,
+      "step": 3910
+    },
+    {
+      "entropy": 0.39280238198116424,
+      "epoch": 3.4878428754242474,
+      "grad_norm": 5.84375,
+      "learning_rate": 1.3513333333333333e-05,
+      "loss": 0.3524,
+      "mean_token_accuracy": 0.8826503712683916,
+      "num_tokens": 52473476.0,
+      "step": 3920
+    },
+    {
+      "entropy": 0.40133273908868433,
+      "epoch": 3.4967451176765145,
+      "grad_norm": 6.21875,
+      "learning_rate": 1.3491111111111113e-05,
+      "loss": 0.3591,
+      "mean_token_accuracy": 0.8842132449150085,
+      "num_tokens": 52603068.0,
+      "step": 3930
+    },
+    {
+      "entropy": 0.3991576720029116,
+      "epoch": 3.505647359928782,
+      "grad_norm": 5.25,
+      "learning_rate": 1.3468888888888888e-05,
+      "loss": 0.3643,
+      "mean_token_accuracy": 0.88127959407866,
+      "num_tokens": 52743275.0,
+      "step": 3940
+    },
+    {
+      "entropy": 0.38993287677876654,
+      "epoch": 3.5145496021810496,
+      "grad_norm": 6.625,
+      "learning_rate": 1.3446666666666668e-05,
+      "loss": 0.3552,
+      "mean_token_accuracy": 0.8855182223021985,
+      "num_tokens": 52881050.0,
+      "step": 3950
+    },
+    {
+      "entropy": 0.38315050583332777,
+      "epoch": 3.5234518444333167,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.3424444444444447e-05,
+      "loss": 0.3482,
+      "mean_token_accuracy": 0.8880773987621069,
+      "num_tokens": 53014827.0,
+      "step": 3960
+    },
+    {
+      "entropy": 0.38314321227371695,
+      "epoch": 3.532354086685584,
+      "grad_norm": 6.71875,
+      "learning_rate": 1.3402222222222223e-05,
+      "loss": 0.353,
+      "mean_token_accuracy": 0.8859844785183668,
+      "num_tokens": 53147376.0,
+      "step": 3970
+    },
+    {
+      "entropy": 0.40427553560584784,
+      "epoch": 3.5412563289378514,
+      "grad_norm": 6.84375,
+      "learning_rate": 1.3380000000000002e-05,
+      "loss": 0.3697,
+      "mean_token_accuracy": 0.8827211875468493,
+      "num_tokens": 53285577.0,
+      "step": 3980
+    },
+    {
+      "entropy": 0.384245518874377,
+      "epoch": 3.5501585711901185,
+      "grad_norm": 6.1875,
+      "learning_rate": 1.3357777777777778e-05,
+      "loss": 0.3474,
+      "mean_token_accuracy": 0.8869146913290024,
+      "num_tokens": 53417896.0,
+      "step": 3990
+    },
+    {
+      "entropy": 0.36794486418366434,
+      "epoch": 3.559060813442386,
+      "grad_norm": 6.09375,
+      "learning_rate": 1.3335555555555558e-05,
+      "loss": 0.3464,
+      "mean_token_accuracy": 0.8904211200773716,
+      "num_tokens": 53553391.0,
+      "step": 4000
+    },
+    {
+      "epoch": 3.559060813442386,
+      "eval_coding_entropy": 0.9717776861190796,
+      "eval_coding_loss": 3.1887309551239014,
+      "eval_coding_mean_token_accuracy": 0.48341964721679687,
+      "eval_coding_num_tokens": 53553391.0,
+      "eval_coding_runtime": 81.3132,
+      "eval_coding_samples_per_second": 6.149,
+      "eval_coding_steps_per_second": 3.075,
+      "step": 4000
+    },
+    {
+      "epoch": 3.559060813442386,
+      "eval_chemistry_entropy": 0.49389336627721786,
+      "eval_chemistry_loss": 1.2218774557113647,
+      "eval_chemistry_mean_token_accuracy": 0.7522559967041016,
+      "eval_chemistry_num_tokens": 53553391.0,
+      "eval_chemistry_runtime": 44.0263,
+      "eval_chemistry_samples_per_second": 11.357,
+      "eval_chemistry_steps_per_second": 5.678,
+      "step": 4000
+    },
+    {
+      "entropy": 0.38230774868279693,
+      "epoch": 3.567963055694653,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.3313333333333333e-05,
+      "loss": 0.3439,
+      "mean_token_accuracy": 0.8890045057982207,
+      "num_tokens": 53686764.0,
+      "step": 4010
+    },
+    {
+      "entropy": 0.3961940433830023,
+      "epoch": 3.57686529794692,
+      "grad_norm": 6.34375,
+      "learning_rate": 1.3291111111111113e-05,
+      "loss": 0.3602,
+      "mean_token_accuracy": 0.882722993567586,
+      "num_tokens": 53826074.0,
+      "step": 4020
+    },
+    {
+      "entropy": 0.38231923077255486,
+      "epoch": 3.5857675401991878,
+      "grad_norm": 5.9375,
+      "learning_rate": 1.3268888888888889e-05,
+      "loss": 0.3558,
+      "mean_token_accuracy": 0.8863572716712952,
+      "num_tokens": 53959151.0,
+      "step": 4030
+    },
+    {
+      "entropy": 0.390913794003427,
+      "epoch": 3.594669782451455,
+      "grad_norm": 6.625,
+      "learning_rate": 1.3246666666666668e-05,
+      "loss": 0.3514,
+      "mean_token_accuracy": 0.8858473252505064,
+      "num_tokens": 54090804.0,
+      "step": 4040
+    },
+    {
+      "entropy": 0.3786553875543177,
+      "epoch": 3.6035720247037224,
+      "grad_norm": 5.875,
+      "learning_rate": 1.3224444444444446e-05,
+      "loss": 0.3488,
+      "mean_token_accuracy": 0.8884392313659191,
+      "num_tokens": 54223304.0,
+      "step": 4050
+    },
+    {
+      "entropy": 0.3775904984213412,
+      "epoch": 3.6124742669559895,
+      "grad_norm": 5.875,
+      "learning_rate": 1.3202222222222223e-05,
+      "loss": 0.3428,
+      "mean_token_accuracy": 0.8887510970234871,
+      "num_tokens": 54350633.0,
+      "step": 4060
+    },
+    {
+      "entropy": 0.4148494116030633,
+      "epoch": 3.6213765092082566,
+      "grad_norm": 6.90625,
+      "learning_rate": 1.3180000000000001e-05,
+      "loss": 0.3785,
+      "mean_token_accuracy": 0.8792279571294784,
+      "num_tokens": 54483733.0,
+      "step": 4070
+    },
+    {
+      "entropy": 0.39638197449967266,
+      "epoch": 3.630278751460524,
+      "grad_norm": 6.71875,
+      "learning_rate": 1.3157777777777778e-05,
+      "loss": 0.3576,
+      "mean_token_accuracy": 0.8833456009626388,
+      "num_tokens": 54623090.0,
+      "step": 4080
+    },
+    {
+      "entropy": 0.3640109305270016,
+      "epoch": 3.6391809937127912,
+      "grad_norm": 6.15625,
+      "learning_rate": 1.3135555555555558e-05,
+      "loss": 0.3384,
+      "mean_token_accuracy": 0.8904370553791523,
+      "num_tokens": 54760901.0,
+      "step": 4090
+    },
+    {
+      "entropy": 0.3825213139876723,
+      "epoch": 3.648083235965059,
+      "grad_norm": 6.0,
+      "learning_rate": 1.3113333333333334e-05,
+      "loss": 0.3501,
+      "mean_token_accuracy": 0.8870889421552419,
+      "num_tokens": 54891961.0,
+      "step": 4100
+    },
+    {
+      "epoch": 3.648083235965059,
+      "eval_coding_entropy": 0.9538664617538453,
+      "eval_coding_loss": 3.235166072845459,
+      "eval_coding_mean_token_accuracy": 0.4814978255033493,
+      "eval_coding_num_tokens": 54891961.0,
+      "eval_coding_runtime": 81.099,
+      "eval_coding_samples_per_second": 6.165,
+      "eval_coding_steps_per_second": 3.083,
+      "step": 4100
+    },
+    {
+      "epoch": 3.648083235965059,
+      "eval_chemistry_entropy": 0.49036789387464524,
+      "eval_chemistry_loss": 1.227230429649353,
+      "eval_chemistry_mean_token_accuracy": 0.752094425201416,
+      "eval_chemistry_num_tokens": 54891961.0,
+      "eval_chemistry_runtime": 44.7448,
+      "eval_chemistry_samples_per_second": 11.174,
+      "eval_chemistry_steps_per_second": 5.587,
+      "step": 4100
+    },
+    {
+      "entropy": 0.3809140153229237,
+      "epoch": 3.656985478217326,
+      "grad_norm": 6.1875,
+      "learning_rate": 1.3091111111111113e-05,
+      "loss": 0.3483,
+      "mean_token_accuracy": 0.8866299767047167,
+      "num_tokens": 55025274.0,
+      "step": 4110
+    },
+    {
+      "entropy": 0.39286829279735686,
+      "epoch": 3.6658877204695934,
+      "grad_norm": 5.9375,
+      "learning_rate": 1.306888888888889e-05,
+      "loss": 0.356,
+      "mean_token_accuracy": 0.8851433273404836,
+      "num_tokens": 55160443.0,
+      "step": 4120
+    },
+    {
+      "entropy": 0.3894599758088589,
+      "epoch": 3.6747899627218605,
+      "grad_norm": 6.0625,
+      "learning_rate": 1.3046666666666668e-05,
+      "loss": 0.3526,
+      "mean_token_accuracy": 0.885522336140275,
+      "num_tokens": 55294408.0,
+      "step": 4130
+    },
+    {
+      "entropy": 0.3929037579335272,
+      "epoch": 3.683692204974128,
+      "grad_norm": 7.1875,
+      "learning_rate": 1.3024444444444446e-05,
+      "loss": 0.3583,
+      "mean_token_accuracy": 0.884577476605773,
+      "num_tokens": 55425148.0,
+      "step": 4140
+    },
+    {
+      "entropy": 0.38234278140589595,
+      "epoch": 3.692594447226395,
+      "grad_norm": 6.84375,
+      "learning_rate": 1.3002222222222223e-05,
+      "loss": 0.3483,
+      "mean_token_accuracy": 0.8881897337734699,
+      "num_tokens": 55562110.0,
+      "step": 4150
+    },
+    {
+      "entropy": 0.38905579233542087,
+      "epoch": 3.7014966894786623,
+      "grad_norm": 7.3125,
+      "learning_rate": 1.2980000000000001e-05,
+      "loss": 0.3628,
+      "mean_token_accuracy": 0.8837911289185285,
+      "num_tokens": 55701877.0,
+      "step": 4160
+    },
+    {
+      "entropy": 0.4067420202307403,
+      "epoch": 3.71039893173093,
+      "grad_norm": 5.875,
+      "learning_rate": 1.2957777777777779e-05,
+      "loss": 0.3774,
+      "mean_token_accuracy": 0.8823860224336386,
+      "num_tokens": 55835925.0,
+      "step": 4170
+    },
+    {
+      "entropy": 0.3821092516183853,
+      "epoch": 3.719301173983197,
+      "grad_norm": 6.03125,
+      "learning_rate": 1.2935555555555556e-05,
+      "loss": 0.3539,
+      "mean_token_accuracy": 0.8878915555775165,
+      "num_tokens": 55969756.0,
+      "step": 4180
+    },
+    {
+      "entropy": 0.38974099773913623,
+      "epoch": 3.7282034162354645,
+      "grad_norm": 6.125,
+      "learning_rate": 1.2913333333333336e-05,
+      "loss": 0.3623,
+      "mean_token_accuracy": 0.884148221090436,
+      "num_tokens": 56104262.0,
+      "step": 4190
+    },
+    {
+      "entropy": 0.3784426558762789,
+      "epoch": 3.7371056584877316,
+      "grad_norm": 5.90625,
+      "learning_rate": 1.2891111111111112e-05,
+      "loss": 0.3365,
+      "mean_token_accuracy": 0.8892817750573159,
+      "num_tokens": 56238576.0,
+      "step": 4200
+    },
+    {
+      "epoch": 3.7371056584877316,
+      "eval_coding_entropy": 0.9571860041618347,
+      "eval_coding_loss": 3.2195303440093994,
+      "eval_coding_mean_token_accuracy": 0.48232095313072204,
+      "eval_coding_num_tokens": 56238576.0,
+      "eval_coding_runtime": 80.9869,
+      "eval_coding_samples_per_second": 6.174,
+      "eval_coding_steps_per_second": 3.087,
+      "step": 4200
+    },
+    {
+      "epoch": 3.7371056584877316,
+      "eval_chemistry_entropy": 0.48772701364755633,
+      "eval_chemistry_loss": 1.2261513471603394,
+      "eval_chemistry_mean_token_accuracy": 0.7523908896446228,
+      "eval_chemistry_num_tokens": 56238576.0,
+      "eval_chemistry_runtime": 44.7162,
+      "eval_chemistry_samples_per_second": 11.182,
+      "eval_chemistry_steps_per_second": 5.591,
+      "step": 4200
+    },
+    {
+      "entropy": 0.3820441760122776,
+      "epoch": 3.7460079007399987,
+      "grad_norm": 6.34375,
+      "learning_rate": 1.2868888888888891e-05,
+      "loss": 0.3519,
+      "mean_token_accuracy": 0.8867039252072573,
+      "num_tokens": 56368273.0,
+      "step": 4210
+    },
+    {
+      "entropy": 0.3825038241222501,
+      "epoch": 3.754910142992266,
+      "grad_norm": 7.21875,
+      "learning_rate": 1.2846666666666667e-05,
+      "loss": 0.3553,
+      "mean_token_accuracy": 0.8876116342842579,
+      "num_tokens": 56497905.0,
+      "step": 4220
+    },
+    {
+      "entropy": 0.3849992061033845,
+      "epoch": 3.7638123852445333,
+      "grad_norm": 5.625,
+      "learning_rate": 1.2824444444444446e-05,
+      "loss": 0.3552,
+      "mean_token_accuracy": 0.8870672207325697,
+      "num_tokens": 56631077.0,
+      "step": 4230
+    },
+    {
+      "entropy": 0.40312140723690393,
+      "epoch": 3.772714627496801,
+      "grad_norm": 7.25,
+      "learning_rate": 1.2802222222222222e-05,
+      "loss": 0.3599,
+      "mean_token_accuracy": 0.8827943548560142,
+      "num_tokens": 56757407.0,
+      "step": 4240
+    },
+    {
+      "entropy": 0.3670321382582188,
+      "epoch": 3.781616869749068,
+      "grad_norm": 6.78125,
+      "learning_rate": 1.2780000000000001e-05,
+      "loss": 0.3371,
+      "mean_token_accuracy": 0.8919087298214435,
+      "num_tokens": 56888016.0,
+      "step": 4250
+    },
+    {
+      "entropy": 0.3870931029319763,
+      "epoch": 3.790519112001335,
+      "grad_norm": 6.0,
+      "learning_rate": 1.2757777777777777e-05,
+      "loss": 0.3601,
+      "mean_token_accuracy": 0.8861101109534502,
+      "num_tokens": 57019213.0,
+      "step": 4260
+    },
+    {
+      "entropy": 0.38339523756876587,
+      "epoch": 3.7994213542536026,
+      "grad_norm": 7.125,
+      "learning_rate": 1.2735555555555557e-05,
+      "loss": 0.3441,
+      "mean_token_accuracy": 0.8890860129147768,
+      "num_tokens": 57143973.0,
+      "step": 4270
+    },
+    {
+      "entropy": 0.3919250200502574,
+      "epoch": 3.80832359650587,
+      "grad_norm": 6.28125,
+      "learning_rate": 1.2713333333333336e-05,
+      "loss": 0.3519,
+      "mean_token_accuracy": 0.8847432967275382,
+      "num_tokens": 57274629.0,
+      "step": 4280
+    },
+    {
+      "entropy": 0.38726737154647706,
+      "epoch": 3.8172258387581373,
+      "grad_norm": 6.8125,
+      "learning_rate": 1.2691111111111112e-05,
+      "loss": 0.3476,
+      "mean_token_accuracy": 0.8858231928199529,
+      "num_tokens": 57411493.0,
+      "step": 4290
+    },
+    {
+      "entropy": 0.3880287946201861,
+      "epoch": 3.8261280810104044,
+      "grad_norm": 6.75,
+      "learning_rate": 1.2668888888888891e-05,
+      "loss": 0.3509,
+      "mean_token_accuracy": 0.8834996774792672,
+      "num_tokens": 57549791.0,
+      "step": 4300
+    },
+    {
+      "epoch": 3.8261280810104044,
+      "eval_coding_entropy": 0.9566324682235717,
+      "eval_coding_loss": 3.211818218231201,
+      "eval_coding_mean_token_accuracy": 0.4824616641998291,
+      "eval_coding_num_tokens": 57549791.0,
+      "eval_coding_runtime": 81.1078,
+      "eval_coding_samples_per_second": 6.165,
+      "eval_coding_steps_per_second": 3.082,
+      "step": 4300
+    },
+    {
+      "epoch": 3.8261280810104044,
+      "eval_chemistry_entropy": 0.4877074437141418,
+      "eval_chemistry_loss": 1.221612811088562,
+      "eval_chemistry_mean_token_accuracy": 0.7530347707271576,
+      "eval_chemistry_num_tokens": 57549791.0,
+      "eval_chemistry_runtime": 44.7623,
+      "eval_chemistry_samples_per_second": 11.17,
+      "eval_chemistry_steps_per_second": 5.585,
+      "step": 4300
+    },
+    {
+      "entropy": 0.3739363309927285,
+      "epoch": 3.835030323262672,
+      "grad_norm": 6.15625,
+      "learning_rate": 1.2646666666666667e-05,
+      "loss": 0.3433,
+      "mean_token_accuracy": 0.8885865934193135,
+      "num_tokens": 57680894.0,
+      "step": 4310
+    },
+    {
+      "entropy": 0.3911302207037807,
+      "epoch": 3.843932565514939,
+      "grad_norm": 6.1875,
+      "learning_rate": 1.2624444444444446e-05,
+      "loss": 0.3596,
+      "mean_token_accuracy": 0.8836597688496113,
+      "num_tokens": 57815454.0,
+      "step": 4320
+    },
+    {
+      "entropy": 0.3965963849797845,
+      "epoch": 3.8528348077672065,
+      "grad_norm": 6.59375,
+      "learning_rate": 1.2602222222222222e-05,
+      "loss": 0.3678,
+      "mean_token_accuracy": 0.8839665818959475,
+      "num_tokens": 57948688.0,
+      "step": 4330
+    },
+    {
+      "entropy": 0.3907907651271671,
+      "epoch": 3.8617370500194736,
+      "grad_norm": 5.65625,
+      "learning_rate": 1.2580000000000002e-05,
+      "loss": 0.3634,
+      "mean_token_accuracy": 0.8861332602798939,
+      "num_tokens": 58076430.0,
+      "step": 4340
+    },
+    {
+      "entropy": 0.39657945642247794,
+      "epoch": 3.8706392922717408,
+      "grad_norm": 6.34375,
+      "learning_rate": 1.2557777777777779e-05,
+      "loss": 0.3547,
+      "mean_token_accuracy": 0.8846510373055935,
+      "num_tokens": 58209649.0,
+      "step": 4350
+    },
+    {
+      "entropy": 0.3810889968648553,
+      "epoch": 3.8795415345240083,
+      "grad_norm": 6.8125,
+      "learning_rate": 1.2535555555555557e-05,
+      "loss": 0.35,
+      "mean_token_accuracy": 0.8848568014800549,
+      "num_tokens": 58346307.0,
+      "step": 4360
+    },
+    {
+      "entropy": 0.3942238797433674,
+      "epoch": 3.8884437767762754,
+      "grad_norm": 6.1875,
+      "learning_rate": 1.2513333333333334e-05,
+      "loss": 0.3573,
+      "mean_token_accuracy": 0.8848738487809896,
+      "num_tokens": 58481503.0,
+      "step": 4370
+    },
+    {
+      "entropy": 0.4030935447663069,
+      "epoch": 3.897346019028543,
+      "grad_norm": 6.75,
+      "learning_rate": 1.2491111111111112e-05,
+      "loss": 0.3714,
+      "mean_token_accuracy": 0.8816877316683531,
+      "num_tokens": 58621624.0,
+      "step": 4380
+    },
+    {
+      "entropy": 0.38910891832783817,
+      "epoch": 3.90624826128081,
+      "grad_norm": 6.40625,
+      "learning_rate": 1.246888888888889e-05,
+      "loss": 0.3456,
+      "mean_token_accuracy": 0.8864168729633093,
+      "num_tokens": 58752170.0,
+      "step": 4390
+    },
+    {
+      "entropy": 0.39874236946925523,
+      "epoch": 3.915150503533077,
+      "grad_norm": 6.4375,
+      "learning_rate": 1.2446666666666667e-05,
+      "loss": 0.3677,
+      "mean_token_accuracy": 0.8836867261677981,
+      "num_tokens": 58874383.0,
+      "step": 4400
+    },
+    {
+      "epoch": 3.915150503533077,
+      "eval_coding_entropy": 0.9587757647037506,
+      "eval_coding_loss": 3.219923257827759,
+      "eval_coding_mean_token_accuracy": 0.48205505073070526,
+      "eval_coding_num_tokens": 58874383.0,
+      "eval_coding_runtime": 81.1784,
+      "eval_coding_samples_per_second": 6.159,
+      "eval_coding_steps_per_second": 3.08,
+      "step": 4400
+    },
+    {
+      "epoch": 3.915150503533077,
+      "eval_chemistry_entropy": 0.48978844755887985,
+      "eval_chemistry_loss": 1.2214304208755493,
+      "eval_chemistry_mean_token_accuracy": 0.7529387938976287,
+      "eval_chemistry_num_tokens": 58874383.0,
+      "eval_chemistry_runtime": 45.1227,
+      "eval_chemistry_samples_per_second": 11.081,
+      "eval_chemistry_steps_per_second": 5.54,
+      "step": 4400
+    },
+    {
+      "entropy": 0.39951314544305205,
+      "epoch": 3.9240527457853447,
+      "grad_norm": 6.46875,
+      "learning_rate": 1.2424444444444445e-05,
+      "loss": 0.3653,
+      "mean_token_accuracy": 0.8839349821209908,
+      "num_tokens": 59007866.0,
+      "step": 4410
+    },
+    {
+      "entropy": 0.38370390040799973,
+      "epoch": 3.9329549880376122,
+      "grad_norm": 6.0,
+      "learning_rate": 1.2402222222222222e-05,
+      "loss": 0.3457,
+      "mean_token_accuracy": 0.8883906628936529,
+      "num_tokens": 59149972.0,
+      "step": 4420
+    },
+    {
+      "entropy": 0.3748090685345232,
+      "epoch": 3.9418572302898793,
+      "grad_norm": 6.625,
+      "learning_rate": 1.2380000000000002e-05,
+      "loss": 0.3411,
+      "mean_token_accuracy": 0.8877852894365788,
+      "num_tokens": 59288798.0,
+      "step": 4430
+    },
+    {
+      "entropy": 0.4002852192148566,
+      "epoch": 3.9507594725421464,
+      "grad_norm": 6.0625,
+      "learning_rate": 1.235777777777778e-05,
+      "loss": 0.3724,
+      "mean_token_accuracy": 0.883031913638115,
+      "num_tokens": 59418591.0,
+      "step": 4440
+    },
+    {
+      "entropy": 0.3742801400832832,
+      "epoch": 3.959661714794414,
+      "grad_norm": 6.15625,
+      "learning_rate": 1.2335555555555557e-05,
+      "loss": 0.3436,
+      "mean_token_accuracy": 0.8887698747217655,
+      "num_tokens": 59558179.0,
+      "step": 4450
+    },
+    {
+      "entropy": 0.4168287088163197,
+      "epoch": 3.968563957046681,
+      "grad_norm": 6.90625,
+      "learning_rate": 1.2313333333333335e-05,
+      "loss": 0.3787,
+      "mean_token_accuracy": 0.8788019310683012,
+      "num_tokens": 59691322.0,
+      "step": 4460
+    },
+    {
+      "entropy": 0.37275569383054974,
+      "epoch": 3.9774661992989486,
+      "grad_norm": 6.125,
+      "learning_rate": 1.2291111111111112e-05,
+      "loss": 0.3407,
+      "mean_token_accuracy": 0.8905260745435953,
+      "num_tokens": 59823160.0,
+      "step": 4470
+    },
+    {
+      "entropy": 0.38167881648987534,
+      "epoch": 3.9863684415512157,
+      "grad_norm": 6.71875,
+      "learning_rate": 1.226888888888889e-05,
+      "loss": 0.3473,
+      "mean_token_accuracy": 0.8874802011996508,
+      "num_tokens": 59953838.0,
+      "step": 4480
+    },
+    {
+      "entropy": 0.3941557636484504,
+      "epoch": 3.995270683803483,
+      "grad_norm": 5.625,
+      "learning_rate": 1.2246666666666667e-05,
+      "loss": 0.3515,
+      "mean_token_accuracy": 0.8860646814107895,
+      "num_tokens": 60088284.0,
+      "step": 4490
+    },
+    {
+      "entropy": 0.34514831326551887,
+      "epoch": 4.003560896900907,
+      "grad_norm": 4.15625,
+      "learning_rate": 1.2224444444444445e-05,
+      "loss": 0.2758,
+      "mean_token_accuracy": 0.9133024559725051,
+      "num_tokens": 60216049.0,
+      "step": 4500
+    },
+    {
+      "epoch": 4.003560896900907,
+      "eval_coding_entropy": 0.8768186528682709,
+      "eval_coding_loss": 3.4144227504730225,
+      "eval_coding_mean_token_accuracy": 0.47889174830913545,
+      "eval_coding_num_tokens": 60216049.0,
+      "eval_coding_runtime": 81.0838,
+      "eval_coding_samples_per_second": 6.166,
+      "eval_coding_steps_per_second": 3.083,
+      "step": 4500
+    },
+    {
+      "epoch": 4.003560896900907,
+      "eval_chemistry_entropy": 0.43171330517530443,
+      "eval_chemistry_loss": 1.3168725967407227,
+      "eval_chemistry_mean_token_accuracy": 0.7501466166973114,
+      "eval_chemistry_num_tokens": 60216049.0,
+      "eval_chemistry_runtime": 44.4505,
+      "eval_chemistry_samples_per_second": 11.248,
+      "eval_chemistry_steps_per_second": 5.624,
+      "step": 4500
+    },
+    {
+      "entropy": 0.24242392512969674,
+      "epoch": 4.012463139153174,
+      "grad_norm": 5.0625,
+      "learning_rate": 1.2202222222222224e-05,
+      "loss": 0.1779,
+      "mean_token_accuracy": 0.9447803523391485,
+      "num_tokens": 60354075.0,
+      "step": 4510
+    },
+    {
+      "entropy": 0.22955687125213445,
+      "epoch": 4.021365381405442,
+      "grad_norm": 4.4375,
+      "learning_rate": 1.218e-05,
+      "loss": 0.1579,
+      "mean_token_accuracy": 0.9492106229066849,
+      "num_tokens": 60490937.0,
+      "step": 4520
+    },
+    {
+      "entropy": 0.23463427713140844,
+      "epoch": 4.030267623657709,
+      "grad_norm": 4.96875,
+      "learning_rate": 1.215777777777778e-05,
+      "loss": 0.164,
+      "mean_token_accuracy": 0.9488690692931414,
+      "num_tokens": 60623552.0,
+      "step": 4530
+    },
+    {
+      "entropy": 0.23421558788977564,
+      "epoch": 4.039169865909976,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.2135555555555556e-05,
+      "loss": 0.1672,
+      "mean_token_accuracy": 0.9476096056401729,
+      "num_tokens": 60755892.0,
+      "step": 4540
+    },
+    {
+      "entropy": 0.2241153002716601,
+      "epoch": 4.048072108162243,
+      "grad_norm": 5.8125,
+      "learning_rate": 1.2113333333333335e-05,
+      "loss": 0.1611,
+      "mean_token_accuracy": 0.9504038102924823,
+      "num_tokens": 60891882.0,
+      "step": 4550
+    },
+    {
+      "entropy": 0.22649794803000986,
+      "epoch": 4.056974350414511,
+      "grad_norm": 5.75,
+      "learning_rate": 1.209111111111111e-05,
+      "loss": 0.1623,
+      "mean_token_accuracy": 0.9500045076012611,
+      "num_tokens": 61030893.0,
+      "step": 4560
+    },
+    {
+      "entropy": 0.21756527186371385,
+      "epoch": 4.065876592666778,
+      "grad_norm": 4.625,
+      "learning_rate": 1.206888888888889e-05,
+      "loss": 0.162,
+      "mean_token_accuracy": 0.9519479867070914,
+      "num_tokens": 61161776.0,
+      "step": 4570
+    },
+    {
+      "entropy": 0.21714273458346725,
+      "epoch": 4.074778834919045,
+      "grad_norm": 4.5625,
+      "learning_rate": 1.204666666666667e-05,
+      "loss": 0.1612,
+      "mean_token_accuracy": 0.9507524479180575,
+      "num_tokens": 61301458.0,
+      "step": 4580
+    },
+    {
+      "entropy": 0.23145420481450857,
+      "epoch": 4.083681077171312,
+      "grad_norm": 4.8125,
+      "learning_rate": 1.2024444444444445e-05,
+      "loss": 0.1633,
+      "mean_token_accuracy": 0.9495500519871711,
+      "num_tokens": 61436163.0,
+      "step": 4590
+    },
+    {
+      "entropy": 0.22256430764682592,
+      "epoch": 4.0925833194235794,
+      "grad_norm": 5.3125,
+      "learning_rate": 1.2002222222222225e-05,
+      "loss": 0.1635,
+      "mean_token_accuracy": 0.9500260829925538,
+      "num_tokens": 61574717.0,
+      "step": 4600
+    },
+    {
+      "epoch": 4.0925833194235794,
+      "eval_coding_entropy": 0.8724180943965912,
+      "eval_coding_loss": 3.5647330284118652,
+      "eval_coding_mean_token_accuracy": 0.4710871074199677,
+      "eval_coding_num_tokens": 61574717.0,
+      "eval_coding_runtime": 81.3933,
+      "eval_coding_samples_per_second": 6.143,
+      "eval_coding_steps_per_second": 3.072,
+      "step": 4600
+    },
+    {
+      "epoch": 4.0925833194235794,
+      "eval_chemistry_entropy": 0.4042525665163994,
+      "eval_chemistry_loss": 1.4055781364440918,
+      "eval_chemistry_mean_token_accuracy": 0.746590149641037,
+      "eval_chemistry_num_tokens": 61574717.0,
+      "eval_chemistry_runtime": 44.7876,
+      "eval_chemistry_samples_per_second": 11.164,
+      "eval_chemistry_steps_per_second": 5.582,
+      "step": 4600
+    },
+    {
+      "entropy": 0.2261083422228694,
+      "epoch": 4.101485561675847,
+      "grad_norm": 5.0625,
+      "learning_rate": 1.198e-05,
+      "loss": 0.1628,
+      "mean_token_accuracy": 0.9489869005978108,
+      "num_tokens": 61707807.0,
+      "step": 4610
+    },
+    {
+      "entropy": 0.22845622207969427,
+      "epoch": 4.1103878039281145,
+      "grad_norm": 5.65625,
+      "learning_rate": 1.195777777777778e-05,
+      "loss": 0.1702,
+      "mean_token_accuracy": 0.9484166529029607,
+      "num_tokens": 61840186.0,
+      "step": 4620
+    },
+    {
+      "entropy": 0.23903292729519307,
+      "epoch": 4.119290046180382,
+      "grad_norm": 6.1875,
+      "learning_rate": 1.1935555555555556e-05,
+      "loss": 0.1764,
+      "mean_token_accuracy": 0.9459130361676216,
+      "num_tokens": 61976105.0,
+      "step": 4630
+    },
+    {
+      "entropy": 0.23321412005461753,
+      "epoch": 4.128192288432649,
+      "grad_norm": 5.4375,
+      "learning_rate": 1.1913333333333335e-05,
+      "loss": 0.1648,
+      "mean_token_accuracy": 0.9471546486020088,
+      "num_tokens": 62112844.0,
+      "step": 4640
+    },
+    {
+      "entropy": 0.21454251632094384,
+      "epoch": 4.137094530684916,
+      "grad_norm": 5.0625,
+      "learning_rate": 1.1891111111111111e-05,
+      "loss": 0.1561,
+      "mean_token_accuracy": 0.9513641458004714,
+      "num_tokens": 62248742.0,
+      "step": 4650
+    },
+    {
+      "entropy": 0.2344472697004676,
+      "epoch": 4.145996772937184,
+      "grad_norm": 5.21875,
+      "learning_rate": 1.186888888888889e-05,
+      "loss": 0.1673,
+      "mean_token_accuracy": 0.947455882281065,
+      "num_tokens": 62382390.0,
+      "step": 4660
+    },
+    {
+      "entropy": 0.2293095056898892,
+      "epoch": 4.154899015189451,
+      "grad_norm": 5.96875,
+      "learning_rate": 1.1846666666666668e-05,
+      "loss": 0.1703,
+      "mean_token_accuracy": 0.9481649126857519,
+      "num_tokens": 62514047.0,
+      "step": 4670
+    },
+    {
+      "entropy": 0.22974966512992978,
+      "epoch": 4.163801257441718,
+      "grad_norm": 4.46875,
+      "learning_rate": 1.1824444444444445e-05,
+      "loss": 0.1685,
+      "mean_token_accuracy": 0.947289840131998,
+      "num_tokens": 62658026.0,
+      "step": 4680
+    },
+    {
+      "entropy": 0.2325438741594553,
+      "epoch": 4.172703499693985,
+      "grad_norm": 6.03125,
+      "learning_rate": 1.1802222222222223e-05,
+      "loss": 0.1708,
+      "mean_token_accuracy": 0.9471368182450532,
+      "num_tokens": 62789713.0,
+      "step": 4690
+    },
+    {
+      "entropy": 0.25073454147204755,
+      "epoch": 4.181605741946253,
+      "grad_norm": 5.75,
+      "learning_rate": 1.178e-05,
+      "loss": 0.1834,
+      "mean_token_accuracy": 0.9429597809910775,
+      "num_tokens": 62916724.0,
+      "step": 4700
+    },
+    {
+      "epoch": 4.181605741946253,
+      "eval_coding_entropy": 0.8541329336166382,
+      "eval_coding_loss": 3.5756771564483643,
+      "eval_coding_mean_token_accuracy": 0.4727411354780197,
+      "eval_coding_num_tokens": 62916724.0,
+      "eval_coding_runtime": 81.3963,
+      "eval_coding_samples_per_second": 6.143,
+      "eval_coding_steps_per_second": 3.071,
+      "step": 4700
+    },
+    {
+      "epoch": 4.181605741946253,
+      "eval_chemistry_entropy": 0.39880191755294797,
+      "eval_chemistry_loss": 1.4111989736557007,
+      "eval_chemistry_mean_token_accuracy": 0.7473588895797729,
+      "eval_chemistry_num_tokens": 62916724.0,
+      "eval_chemistry_runtime": 44.6474,
+      "eval_chemistry_samples_per_second": 11.199,
+      "eval_chemistry_steps_per_second": 5.599,
+      "step": 4700
+    },
+    {
+      "entropy": 0.23026872095651923,
+      "epoch": 4.19050798419852,
+      "grad_norm": 6.40625,
+      "learning_rate": 1.1757777777777778e-05,
+      "loss": 0.1676,
+      "mean_token_accuracy": 0.9474555592983961,
+      "num_tokens": 63058552.0,
+      "step": 4710
+    },
+    {
+      "entropy": 0.2302445410285145,
+      "epoch": 4.199410226450787,
+      "grad_norm": 4.90625,
+      "learning_rate": 1.1735555555555556e-05,
+      "loss": 0.1754,
+      "mean_token_accuracy": 0.9457339525222779,
+      "num_tokens": 63195271.0,
+      "step": 4720
+    },
+    {
+      "entropy": 0.22676812605932356,
+      "epoch": 4.208312468703054,
+      "grad_norm": 5.875,
+      "learning_rate": 1.1713333333333334e-05,
+      "loss": 0.17,
+      "mean_token_accuracy": 0.9480170357972384,
+      "num_tokens": 63334835.0,
+      "step": 4730
+    },
+    {
+      "entropy": 0.2265602413099259,
+      "epoch": 4.2172147109553215,
+      "grad_norm": 5.21875,
+      "learning_rate": 1.1691111111111113e-05,
+      "loss": 0.168,
+      "mean_token_accuracy": 0.9489620171487332,
+      "num_tokens": 63466066.0,
+      "step": 4740
+    },
+    {
+      "entropy": 0.2227494718041271,
+      "epoch": 4.2261169532075895,
+      "grad_norm": 4.59375,
+      "learning_rate": 1.1668888888888889e-05,
+      "loss": 0.1652,
+      "mean_token_accuracy": 0.9485546950250864,
+      "num_tokens": 63605471.0,
+      "step": 4750
+    },
+    {
+      "entropy": 0.23669234062545003,
+      "epoch": 4.235019195459857,
+      "grad_norm": 5.625,
+      "learning_rate": 1.1646666666666668e-05,
+      "loss": 0.1765,
+      "mean_token_accuracy": 0.9448253709822894,
+      "num_tokens": 63739678.0,
+      "step": 4760
+    },
+    {
+      "entropy": 0.2399829567875713,
+      "epoch": 4.243921437712124,
+      "grad_norm": 5.03125,
+      "learning_rate": 1.1624444444444446e-05,
+      "loss": 0.1808,
+      "mean_token_accuracy": 0.9445577647536993,
+      "num_tokens": 63866206.0,
+      "step": 4770
+    },
+    {
+      "entropy": 0.23024522429332137,
+      "epoch": 4.252823679964391,
+      "grad_norm": 4.84375,
+      "learning_rate": 1.1602222222222223e-05,
+      "loss": 0.168,
+      "mean_token_accuracy": 0.9469419229775667,
+      "num_tokens": 63999364.0,
+      "step": 4780
+    },
+    {
+      "entropy": 0.22945892037823795,
+      "epoch": 4.261725922216658,
+      "grad_norm": 4.625,
+      "learning_rate": 1.1580000000000001e-05,
+      "loss": 0.1718,
+      "mean_token_accuracy": 0.9469772640615701,
+      "num_tokens": 64139391.0,
+      "step": 4790
+    },
+    {
+      "entropy": 0.23268290692940355,
+      "epoch": 4.270628164468926,
+      "grad_norm": 5.34375,
+      "learning_rate": 1.1557777777777779e-05,
+      "loss": 0.1699,
+      "mean_token_accuracy": 0.9466451004147529,
+      "num_tokens": 64270501.0,
+      "step": 4800
+    },
+    {
+      "epoch": 4.270628164468926,
+      "eval_coding_entropy": 0.8582223451137543,
+      "eval_coding_loss": 3.58956241607666,
+      "eval_coding_mean_token_accuracy": 0.47166978716850283,
+      "eval_coding_num_tokens": 64270501.0,
+      "eval_coding_runtime": 80.8418,
+      "eval_coding_samples_per_second": 6.185,
+      "eval_coding_steps_per_second": 3.092,
+      "step": 4800
+    },
+    {
+      "epoch": 4.270628164468926,
+      "eval_chemistry_entropy": 0.39922611087560655,
+      "eval_chemistry_loss": 1.408676028251648,
+      "eval_chemistry_mean_token_accuracy": 0.7473593473434448,
+      "eval_chemistry_num_tokens": 64270501.0,
+      "eval_chemistry_runtime": 44.4471,
+      "eval_chemistry_samples_per_second": 11.249,
+      "eval_chemistry_steps_per_second": 5.625,
+      "step": 4800
+    },
+    {
+      "entropy": 0.23174612070433795,
+      "epoch": 4.279530406721193,
+      "grad_norm": 5.84375,
+      "learning_rate": 1.1535555555555556e-05,
+      "loss": 0.173,
+      "mean_token_accuracy": 0.946338752284646,
+      "num_tokens": 64400244.0,
+      "step": 4810
+    },
+    {
+      "entropy": 0.22706288844347,
+      "epoch": 4.28843264897346,
+      "grad_norm": 6.0625,
+      "learning_rate": 1.1513333333333334e-05,
+      "loss": 0.1669,
+      "mean_token_accuracy": 0.9481809731572867,
+      "num_tokens": 64533223.0,
+      "step": 4820
+    },
+    {
+      "entropy": 0.23001772416755556,
+      "epoch": 4.297334891225727,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.1491111111111113e-05,
+      "loss": 0.1699,
+      "mean_token_accuracy": 0.9476369392126799,
+      "num_tokens": 64659556.0,
+      "step": 4830
+    },
+    {
+      "entropy": 0.2293076674453914,
+      "epoch": 4.306237133477994,
+      "grad_norm": 5.75,
+      "learning_rate": 1.1468888888888889e-05,
+      "loss": 0.1651,
+      "mean_token_accuracy": 0.946686502546072,
+      "num_tokens": 64799113.0,
+      "step": 4840
+    },
+    {
+      "entropy": 0.22725293152034282,
+      "epoch": 4.315139375730262,
+      "grad_norm": 5.65625,
+      "learning_rate": 1.1446666666666668e-05,
+      "loss": 0.1722,
+      "mean_token_accuracy": 0.9452566973865032,
+      "num_tokens": 64936178.0,
+      "step": 4850
+    },
+    {
+      "entropy": 0.2256384169217199,
+      "epoch": 4.324041617982529,
+      "grad_norm": 5.4375,
+      "learning_rate": 1.1424444444444444e-05,
+      "loss": 0.169,
+      "mean_token_accuracy": 0.9475756708532572,
+      "num_tokens": 65075459.0,
+      "step": 4860
+    },
+    {
+      "entropy": 0.2308055377099663,
+      "epoch": 4.3329438602347965,
+      "grad_norm": 6.4375,
+      "learning_rate": 1.1402222222222224e-05,
+      "loss": 0.1738,
+      "mean_token_accuracy": 0.9462615545839071,
+      "num_tokens": 65209855.0,
+      "step": 4870
+    },
+    {
+      "entropy": 0.23068277714774013,
+      "epoch": 4.341846102487064,
+      "grad_norm": 5.34375,
+      "learning_rate": 1.138e-05,
+      "loss": 0.1718,
+      "mean_token_accuracy": 0.9457338221371174,
+      "num_tokens": 65341896.0,
+      "step": 4880
+    },
+    {
+      "entropy": 0.22801067023538052,
+      "epoch": 4.350748344739332,
+      "grad_norm": 6.375,
+      "learning_rate": 1.1357777777777779e-05,
+      "loss": 0.1687,
+      "mean_token_accuracy": 0.9477497901767492,
+      "num_tokens": 65474356.0,
+      "step": 4890
+    },
+    {
+      "entropy": 0.22734114117920398,
+      "epoch": 4.359650586991599,
+      "grad_norm": 4.53125,
+      "learning_rate": 1.1335555555555558e-05,
+      "loss": 0.17,
+      "mean_token_accuracy": 0.9471019744873047,
+      "num_tokens": 65614487.0,
+      "step": 4900
+    },
+    {
+      "epoch": 4.359650586991599,
+      "eval_coding_entropy": 0.8466500201225281,
+      "eval_coding_loss": 3.6008620262145996,
+      "eval_coding_mean_token_accuracy": 0.4710108668804169,
+      "eval_coding_num_tokens": 65614487.0,
+      "eval_coding_runtime": 80.491,
+      "eval_coding_samples_per_second": 6.212,
+      "eval_coding_steps_per_second": 3.106,
+      "step": 4900
+    },
+    {
+      "epoch": 4.359650586991599,
+      "eval_chemistry_entropy": 0.3944016572833061,
+      "eval_chemistry_loss": 1.4170645475387573,
+      "eval_chemistry_mean_token_accuracy": 0.7483508133888245,
+      "eval_chemistry_num_tokens": 65614487.0,
+      "eval_chemistry_runtime": 44.2366,
+      "eval_chemistry_samples_per_second": 11.303,
+      "eval_chemistry_steps_per_second": 5.651,
+      "step": 4900
+    },
+    {
+      "entropy": 0.2389337553177029,
+      "epoch": 4.368552829243866,
+      "grad_norm": 5.53125,
+      "learning_rate": 1.1313333333333334e-05,
+      "loss": 0.1785,
+      "mean_token_accuracy": 0.944093668460846,
+      "num_tokens": 65751413.0,
+      "step": 4910
+    },
+    {
+      "entropy": 0.23988082823343576,
+      "epoch": 4.377455071496133,
+      "grad_norm": 5.375,
+      "learning_rate": 1.1291111111111113e-05,
+      "loss": 0.1802,
+      "mean_token_accuracy": 0.9449889577925206,
+      "num_tokens": 65887311.0,
+      "step": 4920
+    },
+    {
+      "entropy": 0.2304832440800965,
+      "epoch": 4.3863573137484,
+      "grad_norm": 5.03125,
+      "learning_rate": 1.126888888888889e-05,
+      "loss": 0.1692,
+      "mean_token_accuracy": 0.9467207115143538,
+      "num_tokens": 66013524.0,
+      "step": 4930
+    },
+    {
+      "entropy": 0.2243732753675431,
+      "epoch": 4.395259556000668,
+      "grad_norm": 5.125,
+      "learning_rate": 1.1246666666666669e-05,
+      "loss": 0.1676,
+      "mean_token_accuracy": 0.9478620443493128,
+      "num_tokens": 66151098.0,
+      "step": 4940
+    },
+    {
+      "entropy": 0.2197400763630867,
+      "epoch": 4.404161798252935,
+      "grad_norm": 5.625,
+      "learning_rate": 1.1224444444444444e-05,
+      "loss": 0.1662,
+      "mean_token_accuracy": 0.9483846206218004,
+      "num_tokens": 66286869.0,
+      "step": 4950
+    },
+    {
+      "entropy": 0.22291607735678554,
+      "epoch": 4.413064040505202,
+      "grad_norm": 5.5,
+      "learning_rate": 1.1202222222222224e-05,
+      "loss": 0.1674,
+      "mean_token_accuracy": 0.9477259434759617,
+      "num_tokens": 66417928.0,
+      "step": 4960
+    },
+    {
+      "entropy": 0.2266538655385375,
+      "epoch": 4.421966282757469,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.1180000000000001e-05,
+      "loss": 0.1685,
+      "mean_token_accuracy": 0.946687414124608,
+      "num_tokens": 66552806.0,
+      "step": 4970
+    },
+    {
+      "entropy": 0.23183473013341427,
+      "epoch": 4.430868525009737,
+      "grad_norm": 6.0625,
+      "learning_rate": 1.1157777777777779e-05,
+      "loss": 0.171,
+      "mean_token_accuracy": 0.9465353973209858,
+      "num_tokens": 66684481.0,
+      "step": 4980
+    },
+    {
+      "entropy": 0.2375920951832086,
+      "epoch": 4.439770767262004,
+      "grad_norm": 5.3125,
+      "learning_rate": 1.1135555555555557e-05,
+      "loss": 0.1723,
+      "mean_token_accuracy": 0.9453356441110372,
+      "num_tokens": 66813292.0,
+      "step": 4990
+    },
+    {
+      "entropy": 0.23693633913062512,
+      "epoch": 4.4486730095142715,
+      "grad_norm": 5.65625,
+      "learning_rate": 1.1113333333333334e-05,
+      "loss": 0.1761,
+      "mean_token_accuracy": 0.9444756794720888,
+      "num_tokens": 66940215.0,
+      "step": 5000
+    },
+    {
+      "epoch": 4.4486730095142715,
+      "eval_coding_entropy": 0.8510654926300049,
+      "eval_coding_loss": 3.5973262786865234,
+      "eval_coding_mean_token_accuracy": 0.4716501851081848,
+      "eval_coding_num_tokens": 66940215.0,
+      "eval_coding_runtime": 80.6032,
+      "eval_coding_samples_per_second": 6.203,
+      "eval_coding_steps_per_second": 3.102,
+      "step": 5000
+    },
+    {
+      "epoch": 4.4486730095142715,
+      "eval_chemistry_entropy": 0.3941673068404198,
+      "eval_chemistry_loss": 1.4187324047088623,
+      "eval_chemistry_mean_token_accuracy": 0.7475662422180176,
+      "eval_chemistry_num_tokens": 66940215.0,
+      "eval_chemistry_runtime": 43.8414,
+      "eval_chemistry_samples_per_second": 11.405,
+      "eval_chemistry_steps_per_second": 5.702,
+      "step": 5000
+    },
+    {
+      "entropy": 0.23550806432031096,
+      "epoch": 4.457575251766539,
+      "grad_norm": 5.4375,
+      "learning_rate": 1.1091111111111112e-05,
+      "loss": 0.179,
+      "mean_token_accuracy": 0.9438197817653418,
+      "num_tokens": 67074190.0,
+      "step": 5010
+    },
+    {
+      "entropy": 0.2376973264850676,
+      "epoch": 4.466477494018806,
+      "grad_norm": 5.125,
+      "learning_rate": 1.106888888888889e-05,
+      "loss": 0.1803,
+      "mean_token_accuracy": 0.9438081528991461,
+      "num_tokens": 67207177.0,
+      "step": 5020
+    },
+    {
+      "entropy": 0.24397737681865692,
+      "epoch": 4.475379736271074,
+      "grad_norm": 5.4375,
+      "learning_rate": 1.1046666666666667e-05,
+      "loss": 0.1794,
+      "mean_token_accuracy": 0.9431049011647701,
+      "num_tokens": 67340640.0,
+      "step": 5030
+    },
+    {
+      "entropy": 0.2208402134012431,
+      "epoch": 4.484281978523341,
+      "grad_norm": 6.1875,
+      "learning_rate": 1.1024444444444445e-05,
+      "loss": 0.1674,
+      "mean_token_accuracy": 0.9483952656388283,
+      "num_tokens": 67477558.0,
+      "step": 5040
+    },
+    {
+      "entropy": 0.2281324910465628,
+      "epoch": 4.493184220775608,
+      "grad_norm": 5.78125,
+      "learning_rate": 1.1002222222222222e-05,
+      "loss": 0.17,
+      "mean_token_accuracy": 0.9458577323704958,
+      "num_tokens": 67608262.0,
+      "step": 5050
+    },
+    {
+      "entropy": 0.22787021417170764,
+      "epoch": 4.502086463027875,
+      "grad_norm": 5.3125,
+      "learning_rate": 1.0980000000000002e-05,
+      "loss": 0.1754,
+      "mean_token_accuracy": 0.9469763696193695,
+      "num_tokens": 67741434.0,
+      "step": 5060
+    },
+    {
+      "entropy": 0.2348407140467316,
+      "epoch": 4.510988705280142,
+      "grad_norm": 5.46875,
+      "learning_rate": 1.0957777777777778e-05,
+      "loss": 0.178,
+      "mean_token_accuracy": 0.9435417033731938,
+      "num_tokens": 67872007.0,
+      "step": 5070
+    },
+    {
+      "entropy": 0.23434097403660417,
+      "epoch": 4.51989094753241,
+      "grad_norm": 5.4375,
+      "learning_rate": 1.0935555555555557e-05,
+      "loss": 0.1745,
+      "mean_token_accuracy": 0.9457706995308399,
+      "num_tokens": 68004497.0,
+      "step": 5080
+    },
+    {
+      "entropy": 0.22781363227404655,
+      "epoch": 4.528793189784677,
+      "grad_norm": 6.0,
+      "learning_rate": 1.0913333333333333e-05,
+      "loss": 0.1726,
+      "mean_token_accuracy": 0.9464398104697466,
+      "num_tokens": 68140220.0,
+      "step": 5090
+    },
+    {
+      "entropy": 0.2327739880885929,
+      "epoch": 4.537695432036944,
+      "grad_norm": 5.375,
+      "learning_rate": 1.0891111111111112e-05,
+      "loss": 0.1756,
+      "mean_token_accuracy": 0.9442274782806634,
+      "num_tokens": 68276413.0,
+      "step": 5100
+    },
+    {
+      "epoch": 4.537695432036944,
+      "eval_coding_entropy": 0.850239022731781,
+      "eval_coding_loss": 3.5939643383026123,
+      "eval_coding_mean_token_accuracy": 0.4712141501903534,
+      "eval_coding_num_tokens": 68276413.0,
+      "eval_coding_runtime": 80.8121,
+      "eval_coding_samples_per_second": 6.187,
+      "eval_coding_steps_per_second": 3.094,
+      "step": 5100
+    },
+    {
+      "epoch": 4.537695432036944,
+      "eval_chemistry_entropy": 0.3966505128145218,
+      "eval_chemistry_loss": 1.4176384210586548,
+      "eval_chemistry_mean_token_accuracy": 0.7478301131725311,
+      "eval_chemistry_num_tokens": 68276413.0,
+      "eval_chemistry_runtime": 44.3914,
+      "eval_chemistry_samples_per_second": 11.263,
+      "eval_chemistry_steps_per_second": 5.632,
+      "step": 5100
+    },
+    {
+      "entropy": 0.2254376716911793,
+      "epoch": 4.546597674289211,
+      "grad_norm": 5.0,
+      "learning_rate": 1.0868888888888888e-05,
+      "loss": 0.1683,
+      "mean_token_accuracy": 0.9474512230604887,
+      "num_tokens": 68416870.0,
+      "step": 5110
+    },
+    {
+      "entropy": 0.2281053819693625,
+      "epoch": 4.5554999165414785,
+      "grad_norm": 5.1875,
+      "learning_rate": 1.0846666666666667e-05,
+      "loss": 0.1724,
+      "mean_token_accuracy": 0.9459663175046444,
+      "num_tokens": 68552989.0,
+      "step": 5120
+    },
+    {
+      "entropy": 0.22092577163130045,
+      "epoch": 4.5644021587937464,
+      "grad_norm": 5.34375,
+      "learning_rate": 1.0824444444444447e-05,
+      "loss": 0.1657,
+      "mean_token_accuracy": 0.9477611124515534,
+      "num_tokens": 68693758.0,
+      "step": 5130
+    },
+    {
+      "entropy": 0.23709081965498627,
+      "epoch": 4.5733044010460135,
+      "grad_norm": 6.3125,
+      "learning_rate": 1.0802222222222223e-05,
+      "loss": 0.1747,
+      "mean_token_accuracy": 0.9453907024115324,
+      "num_tokens": 68816878.0,
+      "step": 5140
+    },
+    {
+      "entropy": 0.23196564675308765,
+      "epoch": 4.582206643298281,
+      "grad_norm": 5.9375,
+      "learning_rate": 1.0780000000000002e-05,
+      "loss": 0.1728,
+      "mean_token_accuracy": 0.9467667568475008,
+      "num_tokens": 68948279.0,
+      "step": 5150
+    },
+    {
+      "entropy": 0.23262699088081717,
+      "epoch": 4.591108885550548,
+      "grad_norm": 5.71875,
+      "learning_rate": 1.0757777777777778e-05,
+      "loss": 0.1787,
+      "mean_token_accuracy": 0.9441255994141102,
+      "num_tokens": 69078503.0,
+      "step": 5160
+    },
+    {
+      "entropy": 0.21952104708179832,
+      "epoch": 4.600011127802816,
+      "grad_norm": 5.09375,
+      "learning_rate": 1.0735555555555557e-05,
+      "loss": 0.1654,
+      "mean_token_accuracy": 0.9484395917505026,
+      "num_tokens": 69223197.0,
+      "step": 5170
+    },
+    {
+      "entropy": 0.22129605412483216,
+      "epoch": 4.608913370055083,
+      "grad_norm": 5.3125,
+      "learning_rate": 1.0713333333333333e-05,
+      "loss": 0.163,
+      "mean_token_accuracy": 0.9480540484189988,
+      "num_tokens": 69357964.0,
+      "step": 5180
+    },
+    {
+      "entropy": 0.2382539676502347,
+      "epoch": 4.61781561230735,
+      "grad_norm": 5.78125,
+      "learning_rate": 1.0691111111111112e-05,
+      "loss": 0.1759,
+      "mean_token_accuracy": 0.9441524092108011,
+      "num_tokens": 69489473.0,
+      "step": 5190
+    },
+    {
+      "entropy": 0.23668308607302607,
+      "epoch": 4.626717854559617,
+      "grad_norm": 5.96875,
+      "learning_rate": 1.0668888888888892e-05,
+      "loss": 0.1759,
+      "mean_token_accuracy": 0.9445841103792191,
+      "num_tokens": 69615568.0,
+      "step": 5200
+    },
+    {
+      "epoch": 4.626717854559617,
+      "eval_coding_entropy": 0.8395522680282593,
+      "eval_coding_loss": 3.6133320331573486,
+      "eval_coding_mean_token_accuracy": 0.4710421931743622,
+      "eval_coding_num_tokens": 69615568.0,
+      "eval_coding_runtime": 80.8264,
+      "eval_coding_samples_per_second": 6.186,
+      "eval_coding_steps_per_second": 3.093,
+      "step": 5200
+    },
+    {
+      "epoch": 4.626717854559617,
+      "eval_chemistry_entropy": 0.38993376117944717,
+      "eval_chemistry_loss": 1.4204314947128296,
+      "eval_chemistry_mean_token_accuracy": 0.7483351209163666,
+      "eval_chemistry_num_tokens": 69615568.0,
+      "eval_chemistry_runtime": 44.6183,
+      "eval_chemistry_samples_per_second": 11.206,
+      "eval_chemistry_steps_per_second": 5.603,
+      "step": 5200
+    },
+    {
+      "entropy": 0.2235158104915172,
+      "epoch": 4.635620096811884,
+      "grad_norm": 5.5,
+      "learning_rate": 1.0646666666666668e-05,
+      "loss": 0.1723,
+      "mean_token_accuracy": 0.9476559136062861,
+      "num_tokens": 69745101.0,
+      "step": 5210
+    },
+    {
+      "entropy": 0.23570836717262864,
+      "epoch": 4.644522339064152,
+      "grad_norm": 5.40625,
+      "learning_rate": 1.0624444444444447e-05,
+      "loss": 0.1774,
+      "mean_token_accuracy": 0.9456980671733618,
+      "num_tokens": 69882933.0,
+      "step": 5220
+    },
+    {
+      "entropy": 0.24335786448791624,
+      "epoch": 4.653424581316419,
+      "grad_norm": 5.09375,
+      "learning_rate": 1.0602222222222223e-05,
+      "loss": 0.1812,
+      "mean_token_accuracy": 0.9432407166808844,
+      "num_tokens": 70010514.0,
+      "step": 5230
+    },
+    {
+      "entropy": 0.22568360017612576,
+      "epoch": 4.662326823568686,
+      "grad_norm": 5.8125,
+      "learning_rate": 1.0580000000000002e-05,
+      "loss": 0.1751,
+      "mean_token_accuracy": 0.945692765340209,
+      "num_tokens": 70147891.0,
+      "step": 5240
+    },
+    {
+      "entropy": 0.2358994423877448,
+      "epoch": 4.671229065820953,
+      "grad_norm": 5.84375,
+      "learning_rate": 1.0557777777777778e-05,
+      "loss": 0.1776,
+      "mean_token_accuracy": 0.9435206431895494,
+      "num_tokens": 70282273.0,
+      "step": 5250
+    },
+    {
+      "entropy": 0.2258037155959755,
+      "epoch": 4.680131308073221,
+      "grad_norm": 6.25,
+      "learning_rate": 1.0535555555555557e-05,
+      "loss": 0.1669,
+      "mean_token_accuracy": 0.9477230921387673,
+      "num_tokens": 70411454.0,
+      "step": 5260
+    },
+    {
+      "entropy": 0.23818841939792038,
+      "epoch": 4.6890335503254885,
+      "grad_norm": 5.96875,
+      "learning_rate": 1.0513333333333333e-05,
+      "loss": 0.18,
+      "mean_token_accuracy": 0.9439577654004097,
+      "num_tokens": 70543597.0,
+      "step": 5270
+    },
+    {
+      "entropy": 0.23046916308812798,
+      "epoch": 4.697935792577756,
+      "grad_norm": 5.03125,
+      "learning_rate": 1.0491111111111112e-05,
+      "loss": 0.1751,
+      "mean_token_accuracy": 0.9454480644315482,
+      "num_tokens": 70673950.0,
+      "step": 5280
+    },
+    {
+      "entropy": 0.2336379162967205,
+      "epoch": 4.706838034830023,
+      "grad_norm": 5.625,
+      "learning_rate": 1.046888888888889e-05,
+      "loss": 0.1776,
+      "mean_token_accuracy": 0.9437473889440298,
+      "num_tokens": 70807799.0,
+      "step": 5290
+    },
+    {
+      "entropy": 0.24410929200239478,
+      "epoch": 4.71574027708229,
+      "grad_norm": 5.84375,
+      "learning_rate": 1.0446666666666668e-05,
+      "loss": 0.1813,
+      "mean_token_accuracy": 0.9425773676484823,
+      "num_tokens": 70933220.0,
+      "step": 5300
+    },
+    {
+      "epoch": 4.71574027708229,
+      "eval_coding_entropy": 0.8398437478542328,
+      "eval_coding_loss": 3.604057550430298,
+      "eval_coding_mean_token_accuracy": 0.47171269488334655,
+      "eval_coding_num_tokens": 70933220.0,
+      "eval_coding_runtime": 80.7193,
+      "eval_coding_samples_per_second": 6.194,
+      "eval_coding_steps_per_second": 3.097,
+      "step": 5300
+    },
+    {
+      "epoch": 4.71574027708229,
+      "eval_chemistry_entropy": 0.3937647603750229,
+      "eval_chemistry_loss": 1.4195643663406372,
+      "eval_chemistry_mean_token_accuracy": 0.7482435362339019,
+      "eval_chemistry_num_tokens": 70933220.0,
+      "eval_chemistry_runtime": 44.6231,
+      "eval_chemistry_samples_per_second": 11.205,
+      "eval_chemistry_steps_per_second": 5.602,
+      "step": 5300
+    },
+    {
+      "entropy": 0.2266634378116578,
+      "epoch": 4.724642519334557,
+      "grad_norm": 5.0,
+      "learning_rate": 1.0424444444444445e-05,
+      "loss": 0.1697,
+      "mean_token_accuracy": 0.948122626543045,
+      "num_tokens": 71068354.0,
+      "step": 5310
+    },
+    {
+      "entropy": 0.2253299684263766,
+      "epoch": 4.733544761586825,
+      "grad_norm": 5.375,
+      "learning_rate": 1.0402222222222223e-05,
+      "loss": 0.1696,
+      "mean_token_accuracy": 0.9465535543859005,
+      "num_tokens": 71202429.0,
+      "step": 5320
+    },
+    {
+      "entropy": 0.2279558821581304,
+      "epoch": 4.742447003839092,
+      "grad_norm": 4.75,
+      "learning_rate": 1.038e-05,
+      "loss": 0.175,
+      "mean_token_accuracy": 0.9465553410351276,
+      "num_tokens": 71334993.0,
+      "step": 5330
+    },
+    {
+      "entropy": 0.23041798770427704,
+      "epoch": 4.751349246091359,
+      "grad_norm": 5.96875,
+      "learning_rate": 1.0357777777777778e-05,
+      "loss": 0.173,
+      "mean_token_accuracy": 0.9459192682057619,
+      "num_tokens": 71469322.0,
+      "step": 5340
+    },
+    {
+      "entropy": 0.22485785973258315,
+      "epoch": 4.760251488343626,
+      "grad_norm": 5.5,
+      "learning_rate": 1.0335555555555556e-05,
+      "loss": 0.1696,
+      "mean_token_accuracy": 0.9471837174147367,
+      "num_tokens": 71602799.0,
+      "step": 5350
+    },
+    {
+      "entropy": 0.22710144040174782,
+      "epoch": 4.769153730595894,
+      "grad_norm": 5.375,
+      "learning_rate": 1.0313333333333335e-05,
+      "loss": 0.1719,
+      "mean_token_accuracy": 0.9457287844270468,
+      "num_tokens": 71736522.0,
+      "step": 5360
+    },
+    {
+      "entropy": 0.23323910706676543,
+      "epoch": 4.778055972848161,
+      "grad_norm": 5.75,
+      "learning_rate": 1.0291111111111111e-05,
+      "loss": 0.1705,
+      "mean_token_accuracy": 0.9452896058559418,
+      "num_tokens": 71868562.0,
+      "step": 5370
+    },
+    {
+      "entropy": 0.23475208664312958,
+      "epoch": 4.786958215100428,
+      "grad_norm": 6.46875,
+      "learning_rate": 1.026888888888889e-05,
+      "loss": 0.18,
+      "mean_token_accuracy": 0.9427293427288532,
+      "num_tokens": 71999537.0,
+      "step": 5380
+    },
+    {
+      "entropy": 0.22713543763384222,
+      "epoch": 4.7958604573526955,
+      "grad_norm": 5.1875,
+      "learning_rate": 1.0246666666666666e-05,
+      "loss": 0.1696,
+      "mean_token_accuracy": 0.9457174211740493,
+      "num_tokens": 72136428.0,
+      "step": 5390
+    },
+    {
+      "entropy": 0.22952594058588147,
+      "epoch": 4.804762699604963,
+      "grad_norm": 5.84375,
+      "learning_rate": 1.0224444444444446e-05,
+      "loss": 0.1745,
+      "mean_token_accuracy": 0.9462833143770695,
+      "num_tokens": 72269735.0,
+      "step": 5400
+    },
+    {
+      "epoch": 4.804762699604963,
+      "eval_coding_entropy": 0.8455443418025971,
+      "eval_coding_loss": 3.607527494430542,
+      "eval_coding_mean_token_accuracy": 0.47096783232688905,
+      "eval_coding_num_tokens": 72269735.0,
+      "eval_coding_runtime": 80.778,
+      "eval_coding_samples_per_second": 6.19,
+      "eval_coding_steps_per_second": 3.095,
+      "step": 5400
+    },
+    {
+      "epoch": 4.804762699604963,
+      "eval_chemistry_entropy": 0.3952053197026253,
+      "eval_chemistry_loss": 1.4210429191589355,
+      "eval_chemistry_mean_token_accuracy": 0.7480710239410401,
+      "eval_chemistry_num_tokens": 72269735.0,
+      "eval_chemistry_runtime": 44.334,
+      "eval_chemistry_samples_per_second": 11.278,
+      "eval_chemistry_steps_per_second": 5.639,
+      "step": 5400
+    },
+    {
+      "entropy": 0.22755389823578298,
+      "epoch": 4.813664941857231,
+      "grad_norm": 5.28125,
+      "learning_rate": 1.0202222222222221e-05,
+      "loss": 0.1692,
+      "mean_token_accuracy": 0.9476055391132832,
+      "num_tokens": 72402491.0,
+      "step": 5410
+    },
+    {
+      "entropy": 0.22932352414354681,
+      "epoch": 4.822567184109498,
+      "grad_norm": 5.25,
+      "learning_rate": 1.018e-05,
+      "loss": 0.1707,
+      "mean_token_accuracy": 0.9456290531903505,
+      "num_tokens": 72537744.0,
+      "step": 5420
+    },
+    {
+      "entropy": 0.23575032642111182,
+      "epoch": 4.831469426361765,
+      "grad_norm": 6.125,
+      "learning_rate": 1.0157777777777777e-05,
+      "loss": 0.1783,
+      "mean_token_accuracy": 0.9442883662879467,
+      "num_tokens": 72669928.0,
+      "step": 5430
+    },
+    {
+      "entropy": 0.2392411462496966,
+      "epoch": 4.840371668614032,
+      "grad_norm": 5.875,
+      "learning_rate": 1.0135555555555556e-05,
+      "loss": 0.1775,
+      "mean_token_accuracy": 0.943459227308631,
+      "num_tokens": 72801656.0,
+      "step": 5440
+    },
+    {
+      "entropy": 0.22705278350040317,
+      "epoch": 4.8492739108663,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.0113333333333335e-05,
+      "loss": 0.1716,
+      "mean_token_accuracy": 0.946533314883709,
+      "num_tokens": 72935287.0,
+      "step": 5450
+    },
+    {
+      "entropy": 0.2225622243247926,
+      "epoch": 4.858176153118567,
+      "grad_norm": 5.15625,
+      "learning_rate": 1.0091111111111111e-05,
+      "loss": 0.1707,
+      "mean_token_accuracy": 0.9471365500241518,
+      "num_tokens": 73075773.0,
+      "step": 5460
+    },
+    {
+      "entropy": 0.23216438828967512,
+      "epoch": 4.867078395370834,
+      "grad_norm": 5.5,
+      "learning_rate": 1.006888888888889e-05,
+      "loss": 0.1751,
+      "mean_token_accuracy": 0.944920876249671,
+      "num_tokens": 73209459.0,
+      "step": 5470
+    },
+    {
+      "entropy": 0.21709224614314734,
+      "epoch": 4.875980637623101,
+      "grad_norm": 5.03125,
+      "learning_rate": 1.0046666666666666e-05,
+      "loss": 0.1652,
+      "mean_token_accuracy": 0.9484371032565833,
+      "num_tokens": 73353692.0,
+      "step": 5480
+    },
+    {
+      "entropy": 0.24180672229267658,
+      "epoch": 4.884882879875368,
+      "grad_norm": 6.0,
+      "learning_rate": 1.0024444444444446e-05,
+      "loss": 0.1797,
+      "mean_token_accuracy": 0.9426640078425408,
+      "num_tokens": 73487424.0,
+      "step": 5490
+    },
+    {
+      "entropy": 0.22669697673991324,
+      "epoch": 4.893785122127636,
+      "grad_norm": 6.0625,
+      "learning_rate": 1.0002222222222222e-05,
+      "loss": 0.1744,
+      "mean_token_accuracy": 0.9454805295914411,
+      "num_tokens": 73620577.0,
+      "step": 5500
+    },
+    {
+      "epoch": 4.893785122127636,
+      "eval_coding_entropy": 0.84176673579216,
+      "eval_coding_loss": 3.6000828742980957,
+      "eval_coding_mean_token_accuracy": 0.4712278873920441,
+      "eval_coding_num_tokens": 73620577.0,
+      "eval_coding_runtime": 80.7685,
+      "eval_coding_samples_per_second": 6.191,
+      "eval_coding_steps_per_second": 3.095,
+      "step": 5500
+    },
+    {
+      "epoch": 4.893785122127636,
+      "eval_chemistry_entropy": 0.39245396077632905,
+      "eval_chemistry_loss": 1.419095516204834,
+      "eval_chemistry_mean_token_accuracy": 0.7478463299274445,
+      "eval_chemistry_num_tokens": 73620577.0,
+      "eval_chemistry_runtime": 44.0546,
+      "eval_chemistry_samples_per_second": 11.35,
+      "eval_chemistry_steps_per_second": 5.675,
+      "step": 5500
+    },
+    {
+      "entropy": 0.22467689230106772,
+      "epoch": 4.902687364379903,
+      "grad_norm": 5.78125,
+      "learning_rate": 9.980000000000001e-06,
+      "loss": 0.169,
+      "mean_token_accuracy": 0.9462432086467742,
+      "num_tokens": 73753015.0,
+      "step": 5510
+    },
+    {
+      "entropy": 0.22226891443133354,
+      "epoch": 4.9115896066321705,
+      "grad_norm": 5.40625,
+      "learning_rate": 9.957777777777779e-06,
+      "loss": 0.1657,
+      "mean_token_accuracy": 0.9475815407931805,
+      "num_tokens": 73888689.0,
+      "step": 5520
+    },
+    {
+      "entropy": 0.23495133672840893,
+      "epoch": 4.920491848884438,
+      "grad_norm": 5.5625,
+      "learning_rate": 9.935555555555556e-06,
+      "loss": 0.1758,
+      "mean_token_accuracy": 0.9449925906956196,
+      "num_tokens": 74019326.0,
+      "step": 5530
+    },
+    {
+      "entropy": 0.2331462860573083,
+      "epoch": 4.929394091136705,
+      "grad_norm": 5.28125,
+      "learning_rate": 9.913333333333334e-06,
+      "loss": 0.1738,
+      "mean_token_accuracy": 0.9450077880173922,
+      "num_tokens": 74150672.0,
+      "step": 5540
+    },
+    {
+      "entropy": 0.23217929382808505,
+      "epoch": 4.938296333388973,
+      "grad_norm": 5.5625,
+      "learning_rate": 9.891111111111113e-06,
+      "loss": 0.1786,
+      "mean_token_accuracy": 0.9443294260650873,
+      "num_tokens": 74282146.0,
+      "step": 5550
+    },
+    {
+      "entropy": 0.22361933514475824,
+      "epoch": 4.94719857564124,
+      "grad_norm": 5.0625,
+      "learning_rate": 9.86888888888889e-06,
+      "loss": 0.1698,
+      "mean_token_accuracy": 0.9464732237160206,
+      "num_tokens": 74419554.0,
+      "step": 5560
+    },
+    {
+      "entropy": 0.22855324661359191,
+      "epoch": 4.956100817893507,
+      "grad_norm": 5.625,
+      "learning_rate": 9.846666666666668e-06,
+      "loss": 0.1689,
+      "mean_token_accuracy": 0.9461436871439218,
+      "num_tokens": 74554353.0,
+      "step": 5570
+    },
+    {
+      "entropy": 0.23716661310754716,
+      "epoch": 4.965003060145774,
+      "grad_norm": 5.3125,
+      "learning_rate": 9.824444444444446e-06,
+      "loss": 0.1806,
+      "mean_token_accuracy": 0.9444054517894983,
+      "num_tokens": 74686704.0,
+      "step": 5580
+    },
+    {
+      "entropy": 0.22816496598534286,
+      "epoch": 4.973905302398041,
+      "grad_norm": 6.5625,
+      "learning_rate": 9.802222222222224e-06,
+      "loss": 0.173,
+      "mean_token_accuracy": 0.9461337625980377,
+      "num_tokens": 74818267.0,
+      "step": 5590
+    },
+    {
+      "entropy": 0.23673083679750562,
+      "epoch": 4.982807544650309,
+      "grad_norm": 5.96875,
+      "learning_rate": 9.780000000000001e-06,
+      "loss": 0.1747,
+      "mean_token_accuracy": 0.9443794980645179,
+      "num_tokens": 74947391.0,
+      "step": 5600
+    },
+    {
+      "epoch": 4.982807544650309,
+      "eval_coding_entropy": 0.8414379541873932,
+      "eval_coding_loss": 3.611063003540039,
+      "eval_coding_mean_token_accuracy": 0.4710089634656906,
+      "eval_coding_num_tokens": 74947391.0,
+      "eval_coding_runtime": 80.7745,
+      "eval_coding_samples_per_second": 6.19,
+      "eval_coding_steps_per_second": 3.095,
+      "step": 5600
+    },
+    {
+      "epoch": 4.982807544650309,
+      "eval_chemistry_entropy": 0.3931359322667122,
+      "eval_chemistry_loss": 1.417233943939209,
+      "eval_chemistry_mean_token_accuracy": 0.7482943406105041,
+      "eval_chemistry_num_tokens": 74947391.0,
+      "eval_chemistry_runtime": 44.3759,
+      "eval_chemistry_samples_per_second": 11.267,
+      "eval_chemistry_steps_per_second": 5.634,
+      "step": 5600
+    },
+    {
+      "entropy": 0.23637565835379065,
+      "epoch": 4.991709786902576,
+      "grad_norm": 5.53125,
+      "learning_rate": 9.757777777777779e-06,
+      "loss": 0.177,
+      "mean_token_accuracy": 0.9439791545271874,
+      "num_tokens": 75076596.0,
+      "step": 5610
+    },
+    {
+      "entropy": 0.22718585097549746,
+      "epoch": 5.0,
+      "grad_norm": 13.0,
+      "learning_rate": 9.735555555555556e-06,
+      "loss": 0.1738,
+      "mean_token_accuracy": 0.9462867783220023,
+      "num_tokens": 75199650.0,
+      "step": 5620
+    },
+    {
+      "entropy": 0.17867366704158485,
+      "epoch": 5.008902242252267,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.713333333333334e-06,
+      "loss": 0.0964,
+      "mean_token_accuracy": 0.9741215713322162,
+      "num_tokens": 75325245.0,
+      "step": 5630
+    },
+    {
+      "entropy": 0.1610317365732044,
+      "epoch": 5.017804484504534,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.691111111111112e-06,
+      "loss": 0.0887,
+      "mean_token_accuracy": 0.9765503745526075,
+      "num_tokens": 75457508.0,
+      "step": 5640
+    },
+    {
+      "entropy": 0.16381550594232977,
+      "epoch": 5.026706726756802,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.66888888888889e-06,
+      "loss": 0.0958,
+      "mean_token_accuracy": 0.9741655349731445,
+      "num_tokens": 75593491.0,
+      "step": 5650
+    },
+    {
+      "entropy": 0.1561744086444378,
+      "epoch": 5.035608969009069,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.646666666666667e-06,
+      "loss": 0.0888,
+      "mean_token_accuracy": 0.9759612791240215,
+      "num_tokens": 75731578.0,
+      "step": 5660
+    },
+    {
+      "entropy": 0.16033569290302693,
+      "epoch": 5.044511211261336,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.624444444444445e-06,
+      "loss": 0.0902,
+      "mean_token_accuracy": 0.9765744600445032,
+      "num_tokens": 75872634.0,
+      "step": 5670
+    },
+    {
+      "entropy": 0.15865014377050102,
+      "epoch": 5.0534134535136035,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.602222222222222e-06,
+      "loss": 0.0937,
+      "mean_token_accuracy": 0.9746300250291824,
+      "num_tokens": 76008875.0,
+      "step": 5680
+    },
+    {
+      "entropy": 0.16327327778562903,
+      "epoch": 5.062315695765871,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.58e-06,
+      "loss": 0.0903,
+      "mean_token_accuracy": 0.9756045900285244,
+      "num_tokens": 76143227.0,
+      "step": 5690
+    },
+    {
+      "entropy": 0.1525247700046748,
+      "epoch": 5.071217938018139,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.557777777777777e-06,
+      "loss": 0.0874,
+      "mean_token_accuracy": 0.9762872893363237,
+      "num_tokens": 76281818.0,
+      "step": 5700
+    },
+    {
+      "epoch": 5.071217938018139,
+      "eval_coding_entropy": 0.827942444562912,
+      "eval_coding_loss": 3.7542102336883545,
+      "eval_coding_mean_token_accuracy": 0.4682268179655075,
+      "eval_coding_num_tokens": 76281818.0,
+      "eval_coding_runtime": 80.8197,
+      "eval_coding_samples_per_second": 6.187,
+      "eval_coding_steps_per_second": 3.093,
+      "step": 5700
+    },
+    {
+      "epoch": 5.071217938018139,
+      "eval_chemistry_entropy": 0.37002195131778715,
+      "eval_chemistry_loss": 1.4995241165161133,
+      "eval_chemistry_mean_token_accuracy": 0.7467469420433045,
+      "eval_chemistry_num_tokens": 76281818.0,
+      "eval_chemistry_runtime": 44.4104,
+      "eval_chemistry_samples_per_second": 11.259,
+      "eval_chemistry_steps_per_second": 5.629,
+      "step": 5700
+    },
+    {
+      "entropy": 0.1614917192608118,
+      "epoch": 5.080120180270406,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.535555555555557e-06,
+      "loss": 0.0959,
+      "mean_token_accuracy": 0.9737134899944067,
+      "num_tokens": 76416947.0,
+      "step": 5710
+    },
+    {
+      "entropy": 0.15788423281628639,
+      "epoch": 5.089022422522673,
+      "grad_norm": 3.71875,
+      "learning_rate": 9.513333333333334e-06,
+      "loss": 0.0905,
+      "mean_token_accuracy": 0.9744159411638975,
+      "num_tokens": 76549100.0,
+      "step": 5720
+    },
+    {
+      "entropy": 0.16304625156335534,
+      "epoch": 5.09792466477494,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.491111111111112e-06,
+      "loss": 0.0949,
+      "mean_token_accuracy": 0.9744466677308082,
+      "num_tokens": 76680483.0,
+      "step": 5730
+    },
+    {
+      "entropy": 0.1644583241082728,
+      "epoch": 5.106826907027208,
+      "grad_norm": 4.25,
+      "learning_rate": 9.46888888888889e-06,
+      "loss": 0.0951,
+      "mean_token_accuracy": 0.9743630006909371,
+      "num_tokens": 76815070.0,
+      "step": 5740
+    },
+    {
+      "entropy": 0.1535755860619247,
+      "epoch": 5.115729149279475,
+      "grad_norm": 3.765625,
+      "learning_rate": 9.446666666666667e-06,
+      "loss": 0.0893,
+      "mean_token_accuracy": 0.9762954819947481,
+      "num_tokens": 76953049.0,
+      "step": 5750
+    },
+    {
+      "entropy": 0.15686267567798495,
+      "epoch": 5.124631391531742,
+      "grad_norm": 3.796875,
+      "learning_rate": 9.424444444444445e-06,
+      "loss": 0.0905,
+      "mean_token_accuracy": 0.9752937160432339,
+      "num_tokens": 77090788.0,
+      "step": 5760
+    },
+    {
+      "entropy": 0.16354108229279518,
+      "epoch": 5.133533633784009,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.402222222222222e-06,
+      "loss": 0.095,
+      "mean_token_accuracy": 0.97445203140378,
+      "num_tokens": 77222819.0,
+      "step": 5770
+    },
+    {
+      "entropy": 0.16132679376751186,
+      "epoch": 5.142435876036276,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.38e-06,
+      "loss": 0.0963,
+      "mean_token_accuracy": 0.9740243505686521,
+      "num_tokens": 77356917.0,
+      "step": 5780
+    },
+    {
+      "entropy": 0.16132226085755974,
+      "epoch": 5.151338118288544,
+      "grad_norm": 3.90625,
+      "learning_rate": 9.35777777777778e-06,
+      "loss": 0.0944,
+      "mean_token_accuracy": 0.9737773440778256,
+      "num_tokens": 77483469.0,
+      "step": 5790
+    },
+    {
+      "entropy": 0.16071695541031658,
+      "epoch": 5.160240360540811,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.335555555555557e-06,
+      "loss": 0.0962,
+      "mean_token_accuracy": 0.9731524262577296,
+      "num_tokens": 77619491.0,
+      "step": 5800
+    },
+    {
+      "epoch": 5.160240360540811,
+      "eval_coding_entropy": 0.811889109134674,
+      "eval_coding_loss": 3.784849166870117,
+      "eval_coding_mean_token_accuracy": 0.46791209757328034,
+      "eval_coding_num_tokens": 77619491.0,
+      "eval_coding_runtime": 80.8434,
+      "eval_coding_samples_per_second": 6.185,
+      "eval_coding_steps_per_second": 3.092,
+      "step": 5800
+    },
+    {
+      "epoch": 5.160240360540811,
+      "eval_chemistry_entropy": 0.3668853663802147,
+      "eval_chemistry_loss": 1.5098819732666016,
+      "eval_chemistry_mean_token_accuracy": 0.7461963883638382,
+      "eval_chemistry_num_tokens": 77619491.0,
+      "eval_chemistry_runtime": 44.3902,
+      "eval_chemistry_samples_per_second": 11.264,
+      "eval_chemistry_steps_per_second": 5.632,
+      "step": 5800
+    },
+    {
+      "entropy": 0.16483651311136782,
+      "epoch": 5.1691426027930785,
+      "grad_norm": 3.796875,
+      "learning_rate": 9.313333333333335e-06,
+      "loss": 0.0999,
+      "mean_token_accuracy": 0.9726654328405857,
+      "num_tokens": 77752370.0,
+      "step": 5810
+    },
+    {
+      "entropy": 0.15356866563670338,
+      "epoch": 5.178044845045346,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.291111111111112e-06,
+      "loss": 0.0905,
+      "mean_token_accuracy": 0.9752234745770693,
+      "num_tokens": 77885239.0,
+      "step": 5820
+    },
+    {
+      "entropy": 0.16171410414390266,
+      "epoch": 5.186947087297613,
+      "grad_norm": 3.640625,
+      "learning_rate": 9.26888888888889e-06,
+      "loss": 0.0952,
+      "mean_token_accuracy": 0.9739616516977548,
+      "num_tokens": 78023267.0,
+      "step": 5830
+    },
+    {
+      "entropy": 0.15476314998231827,
+      "epoch": 5.195849329549881,
+      "grad_norm": 3.90625,
+      "learning_rate": 9.246666666666667e-06,
+      "loss": 0.091,
+      "mean_token_accuracy": 0.9749835442751646,
+      "num_tokens": 78154475.0,
+      "step": 5840
+    },
+    {
+      "entropy": 0.15230524041689933,
+      "epoch": 5.204751571802148,
+      "grad_norm": 3.953125,
+      "learning_rate": 9.224444444444445e-06,
+      "loss": 0.09,
+      "mean_token_accuracy": 0.9757960099726916,
+      "num_tokens": 78292588.0,
+      "step": 5850
+    },
+    {
+      "entropy": 0.16375803463160993,
+      "epoch": 5.213653814054415,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.202222222222224e-06,
+      "loss": 0.0974,
+      "mean_token_accuracy": 0.9736481428146362,
+      "num_tokens": 78418518.0,
+      "step": 5860
+    },
+    {
+      "entropy": 0.15657724894117564,
+      "epoch": 5.222556056306682,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.180000000000002e-06,
+      "loss": 0.0904,
+      "mean_token_accuracy": 0.9752040814608336,
+      "num_tokens": 78549427.0,
+      "step": 5870
+    },
+    {
+      "entropy": 0.15152168162167073,
+      "epoch": 5.23145829855895,
+      "grad_norm": 3.703125,
+      "learning_rate": 9.15777777777778e-06,
+      "loss": 0.0889,
+      "mean_token_accuracy": 0.9761624101549387,
+      "num_tokens": 78687867.0,
+      "step": 5880
+    },
+    {
+      "entropy": 0.15705924143549055,
+      "epoch": 5.240360540811217,
+      "grad_norm": 3.984375,
+      "learning_rate": 9.135555555555557e-06,
+      "loss": 0.0914,
+      "mean_token_accuracy": 0.9746583215892315,
+      "num_tokens": 78817892.0,
+      "step": 5890
+    },
+    {
+      "entropy": 0.15385546805337072,
+      "epoch": 5.249262783063484,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.113333333333335e-06,
+      "loss": 0.0924,
+      "mean_token_accuracy": 0.9741925753653049,
+      "num_tokens": 78954512.0,
+      "step": 5900
+    },
+    {
+      "epoch": 5.249262783063484,
+      "eval_coding_entropy": 0.8082081110477447,
+      "eval_coding_loss": 3.7952687740325928,
+      "eval_coding_mean_token_accuracy": 0.4675855830907822,
+      "eval_coding_num_tokens": 78954512.0,
+      "eval_coding_runtime": 80.8004,
+      "eval_coding_samples_per_second": 6.188,
+      "eval_coding_steps_per_second": 3.094,
+      "step": 5900
+    },
+    {
+      "epoch": 5.249262783063484,
+      "eval_chemistry_entropy": 0.3629044607281685,
+      "eval_chemistry_loss": 1.5144990682601929,
+      "eval_chemistry_mean_token_accuracy": 0.7466348993778229,
+      "eval_chemistry_num_tokens": 78954512.0,
+      "eval_chemistry_runtime": 44.5863,
+      "eval_chemistry_samples_per_second": 11.214,
+      "eval_chemistry_steps_per_second": 5.607,
+      "step": 5900
+    },
+    {
+      "entropy": 0.15517175211571158,
+      "epoch": 5.258165025315751,
+      "grad_norm": 3.765625,
+      "learning_rate": 9.091111111111112e-06,
+      "loss": 0.092,
+      "mean_token_accuracy": 0.9751035623252392,
+      "num_tokens": 79088271.0,
+      "step": 5910
+    },
+    {
+      "entropy": 0.15760833849199116,
+      "epoch": 5.267067267568018,
+      "grad_norm": 3.96875,
+      "learning_rate": 9.06888888888889e-06,
+      "loss": 0.0915,
+      "mean_token_accuracy": 0.9744772966951132,
+      "num_tokens": 79219529.0,
+      "step": 5920
+    },
+    {
+      "entropy": 0.15843094624578952,
+      "epoch": 5.275969509820286,
+      "grad_norm": 3.703125,
+      "learning_rate": 9.046666666666668e-06,
+      "loss": 0.0959,
+      "mean_token_accuracy": 0.9734116688370704,
+      "num_tokens": 79356710.0,
+      "step": 5930
+    },
+    {
+      "entropy": 0.15544052980840206,
+      "epoch": 5.284871752072553,
+      "grad_norm": 3.796875,
+      "learning_rate": 9.024444444444445e-06,
+      "loss": 0.0911,
+      "mean_token_accuracy": 0.9753961212933063,
+      "num_tokens": 79494122.0,
+      "step": 5940
+    },
+    {
+      "entropy": 0.1602229298092425,
+      "epoch": 5.2937739943248205,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.002222222222223e-06,
+      "loss": 0.0943,
+      "mean_token_accuracy": 0.9727800093591213,
+      "num_tokens": 79629209.0,
+      "step": 5950
+    },
+    {
+      "entropy": 0.16046311233658345,
+      "epoch": 5.302676236577088,
+      "grad_norm": 3.625,
+      "learning_rate": 8.98e-06,
+      "loss": 0.0911,
+      "mean_token_accuracy": 0.9738341730087996,
+      "num_tokens": 79766616.0,
+      "step": 5960
+    },
+    {
+      "entropy": 0.1569967006566003,
+      "epoch": 5.311578478829355,
+      "grad_norm": 3.609375,
+      "learning_rate": 8.957777777777778e-06,
+      "loss": 0.0935,
+      "mean_token_accuracy": 0.9740807063877582,
+      "num_tokens": 79896973.0,
+      "step": 5970
+    },
+    {
+      "entropy": 0.1632675497792661,
+      "epoch": 5.320480721081623,
+      "grad_norm": 4.25,
+      "learning_rate": 8.935555555555556e-06,
+      "loss": 0.0951,
+      "mean_token_accuracy": 0.9732080519199371,
+      "num_tokens": 80028355.0,
+      "step": 5980
+    },
+    {
+      "entropy": 0.15404832367785276,
+      "epoch": 5.32938296333389,
+      "grad_norm": 4.0625,
+      "learning_rate": 8.913333333333333e-06,
+      "loss": 0.0911,
+      "mean_token_accuracy": 0.9750418324023485,
+      "num_tokens": 80161917.0,
+      "step": 5990
+    },
+    {
+      "entropy": 0.15460194665938615,
+      "epoch": 5.338285205586157,
+      "grad_norm": 3.40625,
+      "learning_rate": 8.891111111111111e-06,
+      "loss": 0.0919,
+      "mean_token_accuracy": 0.9743941921740771,
+      "num_tokens": 80296531.0,
+      "step": 6000
+    },
+    {
+      "epoch": 5.338285205586157,
+      "eval_coding_entropy": 0.8038165469169617,
+      "eval_coding_loss": 3.797417163848877,
+      "eval_coding_mean_token_accuracy": 0.46754062867164614,
+      "eval_coding_num_tokens": 80296531.0,
+      "eval_coding_runtime": 80.7926,
+      "eval_coding_samples_per_second": 6.189,
+      "eval_coding_steps_per_second": 3.094,
+      "step": 6000
+    },
+    {
+      "epoch": 5.338285205586157,
+      "eval_chemistry_entropy": 0.36199628439545634,
+      "eval_chemistry_loss": 1.5164880752563477,
+      "eval_chemistry_mean_token_accuracy": 0.7462306938171387,
+      "eval_chemistry_num_tokens": 80296531.0,
+      "eval_chemistry_runtime": 44.1437,
+      "eval_chemistry_samples_per_second": 11.327,
+      "eval_chemistry_steps_per_second": 5.663,
+      "step": 6000
+    },
+    {
+      "entropy": 0.16026379759423434,
+      "epoch": 5.347187447838424,
+      "grad_norm": 3.625,
+      "learning_rate": 8.868888888888888e-06,
+      "loss": 0.0979,
+      "mean_token_accuracy": 0.9729666762053967,
+      "num_tokens": 80423584.0,
+      "step": 6010
+    },
+    {
+      "entropy": 0.15793778821825982,
+      "epoch": 5.356089690090691,
+      "grad_norm": 4.15625,
+      "learning_rate": 8.846666666666668e-06,
+      "loss": 0.0951,
+      "mean_token_accuracy": 0.9732918571680784,
+      "num_tokens": 80556132.0,
+      "step": 6020
+    },
+    {
+      "entropy": 0.15369412386789919,
+      "epoch": 5.364991932342959,
+      "grad_norm": 3.4375,
+      "learning_rate": 8.824444444444445e-06,
+      "loss": 0.0942,
+      "mean_token_accuracy": 0.9744994219392538,
+      "num_tokens": 80688653.0,
+      "step": 6030
+    },
+    {
+      "entropy": 0.16266587390564383,
+      "epoch": 5.373894174595226,
+      "grad_norm": 3.953125,
+      "learning_rate": 8.802222222222223e-06,
+      "loss": 0.0958,
+      "mean_token_accuracy": 0.9731946781277656,
+      "num_tokens": 80823194.0,
+      "step": 6040
+    },
+    {
+      "entropy": 0.1592084462288767,
+      "epoch": 5.382796416847493,
+      "grad_norm": 4.28125,
+      "learning_rate": 8.78e-06,
+      "loss": 0.0949,
+      "mean_token_accuracy": 0.9737325940281153,
+      "num_tokens": 80963216.0,
+      "step": 6050
+    },
+    {
+      "entropy": 0.15987437926232814,
+      "epoch": 5.39169865909976,
+      "grad_norm": 4.09375,
+      "learning_rate": 8.757777777777778e-06,
+      "loss": 0.0929,
+      "mean_token_accuracy": 0.9736930549144744,
+      "num_tokens": 81092927.0,
+      "step": 6060
+    },
+    {
+      "entropy": 0.159979838039726,
+      "epoch": 5.400600901352028,
+      "grad_norm": 4.0,
+      "learning_rate": 8.735555555555556e-06,
+      "loss": 0.0969,
+      "mean_token_accuracy": 0.9731356892734766,
+      "num_tokens": 81226201.0,
+      "step": 6070
+    },
+    {
+      "entropy": 0.156123501714319,
+      "epoch": 5.4095031436042955,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.713333333333333e-06,
+      "loss": 0.0894,
+      "mean_token_accuracy": 0.975045183673501,
+      "num_tokens": 81365817.0,
+      "step": 6080
+    },
+    {
+      "entropy": 0.16147018638439475,
+      "epoch": 5.418405385856563,
+      "grad_norm": 3.96875,
+      "learning_rate": 8.691111111111111e-06,
+      "loss": 0.0968,
+      "mean_token_accuracy": 0.9731846872717143,
+      "num_tokens": 81497364.0,
+      "step": 6090
+    },
+    {
+      "entropy": 0.16099602580070496,
+      "epoch": 5.42730762810883,
+      "grad_norm": 3.8125,
+      "learning_rate": 8.66888888888889e-06,
+      "loss": 0.0962,
+      "mean_token_accuracy": 0.9732182115316391,
+      "num_tokens": 81634089.0,
+      "step": 6100
+    },
+    {
+      "epoch": 5.42730762810883,
+      "eval_coding_entropy": 0.8083561947345733,
+      "eval_coding_loss": 3.8006105422973633,
+      "eval_coding_mean_token_accuracy": 0.4666549872159958,
+      "eval_coding_num_tokens": 81634089.0,
+      "eval_coding_runtime": 80.7892,
+      "eval_coding_samples_per_second": 6.189,
+      "eval_coding_steps_per_second": 3.094,
+      "step": 6100
+    },
+    {
+      "epoch": 5.42730762810883,
+      "eval_chemistry_entropy": 0.36372234630584716,
+      "eval_chemistry_loss": 1.5172410011291504,
+      "eval_chemistry_mean_token_accuracy": 0.7455400416851043,
+      "eval_chemistry_num_tokens": 81634089.0,
+      "eval_chemistry_runtime": 44.4222,
+      "eval_chemistry_samples_per_second": 11.256,
+      "eval_chemistry_steps_per_second": 5.628,
+      "step": 6100
+    },
+    {
+      "entropy": 0.16010536891408264,
+      "epoch": 5.436209870361097,
+      "grad_norm": 4.65625,
+      "learning_rate": 8.646666666666668e-06,
+      "loss": 0.0946,
+      "mean_token_accuracy": 0.9733412466943264,
+      "num_tokens": 81762459.0,
+      "step": 6110
+    },
+    {
+      "entropy": 0.1571366470772773,
+      "epoch": 5.445112112613365,
+      "grad_norm": 4.84375,
+      "learning_rate": 8.624444444444446e-06,
+      "loss": 0.0951,
+      "mean_token_accuracy": 0.973135146126151,
+      "num_tokens": 81896949.0,
+      "step": 6120
+    },
+    {
+      "entropy": 0.1600673615699634,
+      "epoch": 5.454014354865632,
+      "grad_norm": 3.90625,
+      "learning_rate": 8.602222222222223e-06,
+      "loss": 0.0934,
+      "mean_token_accuracy": 0.9739319559186697,
+      "num_tokens": 82026737.0,
+      "step": 6130
+    },
+    {
+      "entropy": 0.15509645310230552,
+      "epoch": 5.462916597117899,
+      "grad_norm": 3.859375,
+      "learning_rate": 8.580000000000001e-06,
+      "loss": 0.0913,
+      "mean_token_accuracy": 0.974665903672576,
+      "num_tokens": 82163467.0,
+      "step": 6140
+    },
+    {
+      "entropy": 0.1533514441922307,
+      "epoch": 5.471818839370166,
+      "grad_norm": 4.3125,
+      "learning_rate": 8.557777777777778e-06,
+      "loss": 0.0884,
+      "mean_token_accuracy": 0.9754397418349982,
+      "num_tokens": 82299107.0,
+      "step": 6150
+    },
+    {
+      "entropy": 0.15721930435393006,
+      "epoch": 5.480721081622434,
+      "grad_norm": 3.75,
+      "learning_rate": 8.535555555555556e-06,
+      "loss": 0.0923,
+      "mean_token_accuracy": 0.9743526566773653,
+      "num_tokens": 82438582.0,
+      "step": 6160
+    },
+    {
+      "entropy": 0.16244703736156224,
+      "epoch": 5.489623323874701,
+      "grad_norm": 3.625,
+      "learning_rate": 8.513333333333335e-06,
+      "loss": 0.0927,
+      "mean_token_accuracy": 0.9737268473953009,
+      "num_tokens": 82575197.0,
+      "step": 6170
+    },
+    {
+      "entropy": 0.15668164175003768,
+      "epoch": 5.498525566126968,
+      "grad_norm": 3.8125,
+      "learning_rate": 8.491111111111113e-06,
+      "loss": 0.093,
+      "mean_token_accuracy": 0.9744974743574858,
+      "num_tokens": 82709044.0,
+      "step": 6180
+    },
+    {
+      "entropy": 0.16688738060183822,
+      "epoch": 5.507427808379235,
+      "grad_norm": 4.3125,
+      "learning_rate": 8.46888888888889e-06,
+      "loss": 0.0993,
+      "mean_token_accuracy": 0.9718231119215488,
+      "num_tokens": 82840527.0,
+      "step": 6190
+    },
+    {
+      "entropy": 0.16121533289551734,
+      "epoch": 5.5163300506315025,
+      "grad_norm": 4.03125,
+      "learning_rate": 8.446666666666668e-06,
+      "loss": 0.0964,
+      "mean_token_accuracy": 0.972299450263381,
+      "num_tokens": 82971488.0,
+      "step": 6200
+    },
+    {
+      "epoch": 5.5163300506315025,
+      "eval_coding_entropy": 0.8058954026699067,
+      "eval_coding_loss": 3.8047866821289062,
+      "eval_coding_mean_token_accuracy": 0.46677429604530335,
+      "eval_coding_num_tokens": 82971488.0,
+      "eval_coding_runtime": 80.8137,
+      "eval_coding_samples_per_second": 6.187,
+      "eval_coding_steps_per_second": 3.094,
+      "step": 6200
+    },
+    {
+      "epoch": 5.5163300506315025,
+      "eval_chemistry_entropy": 0.3620537666678429,
+      "eval_chemistry_loss": 1.5191386938095093,
+      "eval_chemistry_mean_token_accuracy": 0.7462863502502441,
+      "eval_chemistry_num_tokens": 82971488.0,
+      "eval_chemistry_runtime": 44.3861,
+      "eval_chemistry_samples_per_second": 11.265,
+      "eval_chemistry_steps_per_second": 5.632,
+      "step": 6200
+    },
+    {
+      "entropy": 0.15345031160395592,
+      "epoch": 5.52523229288377,
+      "grad_norm": 4.0,
+      "learning_rate": 8.424444444444446e-06,
+      "loss": 0.0897,
+      "mean_token_accuracy": 0.9746635273098946,
+      "num_tokens": 83108199.0,
+      "step": 6210
+    },
+    {
+      "entropy": 0.15271266987547277,
+      "epoch": 5.534134535136038,
+      "grad_norm": 3.640625,
+      "learning_rate": 8.402222222222223e-06,
+      "loss": 0.0921,
+      "mean_token_accuracy": 0.9745833948254585,
+      "num_tokens": 83243796.0,
+      "step": 6220
+    },
+    {
+      "entropy": 0.1583990196697414,
+      "epoch": 5.543036777388305,
+      "grad_norm": 3.984375,
+      "learning_rate": 8.380000000000001e-06,
+      "loss": 0.0946,
+      "mean_token_accuracy": 0.9743069004267454,
+      "num_tokens": 83382332.0,
+      "step": 6230
+    },
+    {
+      "entropy": 0.15886061848141253,
+      "epoch": 5.551939019640572,
+      "grad_norm": 3.734375,
+      "learning_rate": 8.357777777777779e-06,
+      "loss": 0.0939,
+      "mean_token_accuracy": 0.9740310918539763,
+      "num_tokens": 83525241.0,
+      "step": 6240
+    },
+    {
+      "entropy": 0.15637027248740196,
+      "epoch": 5.560841261892839,
+      "grad_norm": 3.4375,
+      "learning_rate": 8.335555555555556e-06,
+      "loss": 0.0928,
+      "mean_token_accuracy": 0.9739946339279413,
+      "num_tokens": 83659636.0,
+      "step": 6250
+    },
+    {
+      "entropy": 0.15620031617581845,
+      "epoch": 5.569743504145107,
+      "grad_norm": 4.25,
+      "learning_rate": 8.313333333333334e-06,
+      "loss": 0.0936,
+      "mean_token_accuracy": 0.974035232886672,
+      "num_tokens": 83794521.0,
+      "step": 6260
+    },
+    {
+      "entropy": 0.1635045639704913,
+      "epoch": 5.578645746397374,
+      "grad_norm": 4.28125,
+      "learning_rate": 8.291111111111112e-06,
+      "loss": 0.0985,
+      "mean_token_accuracy": 0.9723141383379698,
+      "num_tokens": 83925114.0,
+      "step": 6270
+    },
+    {
+      "entropy": 0.1543718134518713,
+      "epoch": 5.587547988649641,
+      "grad_norm": 4.0625,
+      "learning_rate": 8.268888888888889e-06,
+      "loss": 0.0913,
+      "mean_token_accuracy": 0.974146144464612,
+      "num_tokens": 84053712.0,
+      "step": 6280
+    },
+    {
+      "entropy": 0.16336621679365634,
+      "epoch": 5.596450230901908,
+      "grad_norm": 3.65625,
+      "learning_rate": 8.246666666666667e-06,
+      "loss": 0.099,
+      "mean_token_accuracy": 0.9724921647459268,
+      "num_tokens": 84184139.0,
+      "step": 6290
+    },
+    {
+      "entropy": 0.16260496508330108,
+      "epoch": 5.605352473154175,
+      "grad_norm": 4.09375,
+      "learning_rate": 8.224444444444444e-06,
+      "loss": 0.0949,
+      "mean_token_accuracy": 0.9730042342096568,
+      "num_tokens": 84314508.0,
+      "step": 6300
+    },
+    {
+      "epoch": 5.605352473154175,
+      "eval_coding_entropy": 0.7993298921585082,
+      "eval_coding_loss": 3.8124871253967285,
+      "eval_coding_mean_token_accuracy": 0.46699910748004914,
+      "eval_coding_num_tokens": 84314508.0,
+      "eval_coding_runtime": 80.8134,
+      "eval_coding_samples_per_second": 6.187,
+      "eval_coding_steps_per_second": 3.094,
+      "step": 6300
+    },
+    {
+      "epoch": 5.605352473154175,
+      "eval_chemistry_entropy": 0.35975411379337313,
+      "eval_chemistry_loss": 1.5226541757583618,
+      "eval_chemistry_mean_token_accuracy": 0.7461283633708954,
+      "eval_chemistry_num_tokens": 84314508.0,
+      "eval_chemistry_runtime": 44.5975,
+      "eval_chemistry_samples_per_second": 11.211,
+      "eval_chemistry_steps_per_second": 5.606,
+      "step": 6300
+    },
+    {
+      "entropy": 0.16280582938343285,
+      "epoch": 5.614254715406443,
+      "grad_norm": 3.75,
+      "learning_rate": 8.202222222222222e-06,
+      "loss": 0.0968,
+      "mean_token_accuracy": 0.972305765748024,
+      "num_tokens": 84448459.0,
+      "step": 6310
+    },
+    {
+      "entropy": 0.1566913712769747,
+      "epoch": 5.62315695765871,
+      "grad_norm": 3.734375,
+      "learning_rate": 8.18e-06,
+      "loss": 0.0931,
+      "mean_token_accuracy": 0.9745364040136337,
+      "num_tokens": 84585078.0,
+      "step": 6320
+    },
+    {
+      "entropy": 0.1546374478843063,
+      "epoch": 5.6320591999109775,
+      "grad_norm": 3.71875,
+      "learning_rate": 8.157777777777779e-06,
+      "loss": 0.091,
+      "mean_token_accuracy": 0.9743380308151245,
+      "num_tokens": 84716822.0,
+      "step": 6330
+    },
+    {
+      "entropy": 0.15615240070037545,
+      "epoch": 5.640961442163245,
+      "grad_norm": 4.03125,
+      "learning_rate": 8.135555555555557e-06,
+      "loss": 0.0915,
+      "mean_token_accuracy": 0.9742133468389511,
+      "num_tokens": 84850440.0,
+      "step": 6340
+    },
+    {
+      "entropy": 0.1583402527263388,
+      "epoch": 5.649863684415513,
+      "grad_norm": 3.765625,
+      "learning_rate": 8.113333333333334e-06,
+      "loss": 0.0926,
+      "mean_token_accuracy": 0.9733719509094954,
+      "num_tokens": 84976864.0,
+      "step": 6350
+    },
+    {
+      "entropy": 0.15825757351703942,
+      "epoch": 5.65876592666778,
+      "grad_norm": 4.28125,
+      "learning_rate": 8.091111111111112e-06,
+      "loss": 0.0949,
+      "mean_token_accuracy": 0.9734121549874544,
+      "num_tokens": 85108727.0,
+      "step": 6360
+    },
+    {
+      "entropy": 0.1527583096176386,
+      "epoch": 5.667668168920047,
+      "grad_norm": 3.6875,
+      "learning_rate": 8.06888888888889e-06,
+      "loss": 0.093,
+      "mean_token_accuracy": 0.9747172854840755,
+      "num_tokens": 85247528.0,
+      "step": 6370
+    },
+    {
+      "entropy": 0.15339813930913807,
+      "epoch": 5.676570411172314,
+      "grad_norm": 3.59375,
+      "learning_rate": 8.046666666666667e-06,
+      "loss": 0.0933,
+      "mean_token_accuracy": 0.9737620994448661,
+      "num_tokens": 85375410.0,
+      "step": 6380
+    },
+    {
+      "entropy": 0.15788870016112924,
+      "epoch": 5.685472653424581,
+      "grad_norm": 4.40625,
+      "learning_rate": 8.024444444444445e-06,
+      "loss": 0.0944,
+      "mean_token_accuracy": 0.9735754244029522,
+      "num_tokens": 85504592.0,
+      "step": 6390
+    },
+    {
+      "entropy": 0.15294686027336865,
+      "epoch": 5.694374895676849,
+      "grad_norm": 3.78125,
+      "learning_rate": 8.002222222222222e-06,
+      "loss": 0.0892,
+      "mean_token_accuracy": 0.9750144191086292,
+      "num_tokens": 85644389.0,
+      "step": 6400
+    },
+    {
+      "epoch": 5.694374895676849,
+      "eval_coding_entropy": 0.8001924264431,
+      "eval_coding_loss": 3.8183720111846924,
+      "eval_coding_mean_token_accuracy": 0.46623272287845613,
+      "eval_coding_num_tokens": 85644389.0,
+      "eval_coding_runtime": 80.7565,
+      "eval_coding_samples_per_second": 6.191,
+      "eval_coding_steps_per_second": 3.096,
+      "step": 6400
+    },
+    {
+      "epoch": 5.694374895676849,
+      "eval_chemistry_entropy": 0.358530241549015,
+      "eval_chemistry_loss": 1.5255893468856812,
+      "eval_chemistry_mean_token_accuracy": 0.7458330864906311,
+      "eval_chemistry_num_tokens": 85644389.0,
+      "eval_chemistry_runtime": 44.3383,
+      "eval_chemistry_samples_per_second": 11.277,
+      "eval_chemistry_steps_per_second": 5.638,
+      "step": 6400
+    },
+    {
+      "entropy": 0.15758345648646355,
+      "epoch": 5.703277137929116,
+      "grad_norm": 3.703125,
+      "learning_rate": 7.980000000000002e-06,
+      "loss": 0.0958,
+      "mean_token_accuracy": 0.972952925786376,
+      "num_tokens": 85778153.0,
+      "step": 6410
+    },
+    {
+      "entropy": 0.1567698176484555,
+      "epoch": 5.712179380181383,
+      "grad_norm": 4.21875,
+      "learning_rate": 7.957777777777779e-06,
+      "loss": 0.0939,
+      "mean_token_accuracy": 0.9733938571065665,
+      "num_tokens": 85913924.0,
+      "step": 6420
+    },
+    {
+      "entropy": 0.1524928154889494,
+      "epoch": 5.72108162243365,
+      "grad_norm": 3.75,
+      "learning_rate": 7.935555555555557e-06,
+      "loss": 0.0921,
+      "mean_token_accuracy": 0.973828661814332,
+      "num_tokens": 86049874.0,
+      "step": 6430
+    },
+    {
+      "entropy": 0.157682257052511,
+      "epoch": 5.729983864685918,
+      "grad_norm": 3.671875,
+      "learning_rate": 7.913333333333334e-06,
+      "loss": 0.0972,
+      "mean_token_accuracy": 0.9726391550153494,
+      "num_tokens": 86186499.0,
+      "step": 6440
+    },
+    {
+      "entropy": 0.15490025812759994,
+      "epoch": 5.738886106938185,
+      "grad_norm": 3.953125,
+      "learning_rate": 7.891111111111112e-06,
+      "loss": 0.0968,
+      "mean_token_accuracy": 0.9724067624658346,
+      "num_tokens": 86325069.0,
+      "step": 6450
+    },
+    {
+      "entropy": 0.1545777594903484,
+      "epoch": 5.7477883491904525,
+      "grad_norm": 3.828125,
+      "learning_rate": 7.86888888888889e-06,
+      "loss": 0.0912,
+      "mean_token_accuracy": 0.9741701658815145,
+      "num_tokens": 86457529.0,
+      "step": 6460
+    },
+    {
+      "entropy": 0.1611583491321653,
+      "epoch": 5.7566905914427196,
+      "grad_norm": 4.125,
+      "learning_rate": 7.846666666666667e-06,
+      "loss": 0.0949,
+      "mean_token_accuracy": 0.9737535487860441,
+      "num_tokens": 86589534.0,
+      "step": 6470
+    },
+    {
+      "entropy": 0.16059296927414834,
+      "epoch": 5.765592833694987,
+      "grad_norm": 4.59375,
+      "learning_rate": 7.824444444444445e-06,
+      "loss": 0.0973,
+      "mean_token_accuracy": 0.9716222181916236,
+      "num_tokens": 86721227.0,
+      "step": 6480
+    },
+    {
+      "entropy": 0.1553802145179361,
+      "epoch": 5.774495075947254,
+      "grad_norm": 3.8125,
+      "learning_rate": 7.802222222222222e-06,
+      "loss": 0.0946,
+      "mean_token_accuracy": 0.9732388138771058,
+      "num_tokens": 86854637.0,
+      "step": 6490
+    },
+    {
+      "entropy": 0.14898386783897877,
+      "epoch": 5.783397318199522,
+      "grad_norm": 4.03125,
+      "learning_rate": 7.78e-06,
+      "loss": 0.0912,
+      "mean_token_accuracy": 0.9745634362101555,
+      "num_tokens": 86987700.0,
+      "step": 6500
+    },
+    {
+      "epoch": 5.783397318199522,
+      "eval_coding_entropy": 0.8035327267646789,
+      "eval_coding_loss": 3.8109354972839355,
+      "eval_coding_mean_token_accuracy": 0.4665455397367477,
+      "eval_coding_num_tokens": 86987700.0,
+      "eval_coding_runtime": 82.2511,
+      "eval_coding_samples_per_second": 6.079,
+      "eval_coding_steps_per_second": 3.039,
+      "step": 6500
+    },
+    {
+      "epoch": 5.783397318199522,
+      "eval_chemistry_entropy": 0.35935574513673785,
+      "eval_chemistry_loss": 1.5261948108673096,
+      "eval_chemistry_mean_token_accuracy": 0.7464038653373718,
+      "eval_chemistry_num_tokens": 86987700.0,
+      "eval_chemistry_runtime": 53.8995,
+      "eval_chemistry_samples_per_second": 9.277,
+      "eval_chemistry_steps_per_second": 4.638,
+      "step": 6500
+    }
+  ],
+  "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": 5.586610689575301e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}