diff --git "a/checkpoint-3000/trainer_state.json" "b/checkpoint-3000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-3000/trainer_state.json"
@@ -0,0 +1,3694 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.4,
+  "eval_steps": 100,
+  "global_step": 3000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 3.1496359556913376,
+      "epoch": 0.008,
+      "grad_norm": 298.0,
+      "learning_rate": 1.8e-07,
+      "loss": 7.8456,
+      "mean_token_accuracy": 0.18150889375247062,
+      "num_tokens": 103189.0,
+      "step": 10
+    },
+    {
+      "entropy": 3.2677677035331727,
+      "epoch": 0.016,
+      "grad_norm": 274.0,
+      "learning_rate": 3.8e-07,
+      "loss": 7.8029,
+      "mean_token_accuracy": 0.17940039467066526,
+      "num_tokens": 195510.0,
+      "step": 20
+    },
+    {
+      "entropy": 3.4018275901675223,
+      "epoch": 0.024,
+      "grad_norm": 221.0,
+      "learning_rate": 5.800000000000001e-07,
+      "loss": 7.5603,
+      "mean_token_accuracy": 0.18447871217504144,
+      "num_tokens": 297075.0,
+      "step": 30
+    },
+    {
+      "entropy": 3.904408423602581,
+      "epoch": 0.032,
+      "grad_norm": 183.0,
+      "learning_rate": 7.8e-07,
+      "loss": 7.0974,
+      "mean_token_accuracy": 0.1905162119306624,
+      "num_tokens": 391008.0,
+      "step": 40
+    },
+    {
+      "entropy": 4.471200747787952,
+      "epoch": 0.04,
+      "grad_norm": 116.5,
+      "learning_rate": 9.800000000000001e-07,
+      "loss": 6.474,
+      "mean_token_accuracy": 0.19736560098826886,
+      "num_tokens": 488721.0,
+      "step": 50
+    },
+    {
+      "entropy": 4.936916092038155,
+      "epoch": 0.048,
+      "grad_norm": 61.75,
+      "learning_rate": 1.1800000000000001e-06,
+      "loss": 5.8305,
+      "mean_token_accuracy": 0.2095636417157948,
+      "num_tokens": 590136.0,
+      "step": 60
+    },
+    {
+      "entropy": 5.065159273147583,
+      "epoch": 0.056,
+      "grad_norm": 40.75,
+      "learning_rate": 1.3800000000000001e-06,
+      "loss": 5.3279,
+      "mean_token_accuracy": 0.23331616390496493,
+      "num_tokens": 691017.0,
+      "step": 70
+    },
+    {
+      "entropy": 4.822714896500111,
+      "epoch": 0.064,
+      "grad_norm": 30.625,
+      "learning_rate": 1.5800000000000001e-06,
+      "loss": 4.93,
+      "mean_token_accuracy": 0.2618246308527887,
+      "num_tokens": 791432.0,
+      "step": 80
+    },
+    {
+      "entropy": 4.613001684844494,
+      "epoch": 0.072,
+      "grad_norm": 30.5,
+      "learning_rate": 1.7800000000000001e-06,
+      "loss": 4.6595,
+      "mean_token_accuracy": 0.28834385639056564,
+      "num_tokens": 890431.0,
+      "step": 90
+    },
+    {
+      "entropy": 4.42924614995718,
+      "epoch": 0.08,
+      "grad_norm": 25.0,
+      "learning_rate": 1.98e-06,
+      "loss": 4.4406,
+      "mean_token_accuracy": 0.30723202889785173,
+      "num_tokens": 990342.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.08,
+      "eval_coding_entropy": 2.019406507015228,
+      "eval_coding_loss": 2.2580299377441406,
+      "eval_coding_mean_token_accuracy": 0.49473730635643004,
+      "eval_coding_num_tokens": 990342.0,
+      "eval_coding_runtime": 81.2027,
+      "eval_coding_samples_per_second": 6.157,
+      "eval_coding_steps_per_second": 3.079,
+      "step": 100
+    },
+    {
+      "epoch": 0.08,
+      "eval_biology_entropy": 4.3503482704162595,
+      "eval_biology_loss": 4.414401531219482,
+      "eval_biology_mean_token_accuracy": 0.3150999557375908,
+      "eval_biology_num_tokens": 990342.0,
+      "eval_biology_runtime": 34.5542,
+      "eval_biology_samples_per_second": 14.47,
+      "eval_biology_steps_per_second": 7.235,
+      "step": 100
+    },
+    {
+      "entropy": 4.333127842843533,
+      "epoch": 0.088,
+      "grad_norm": 24.75,
+      "learning_rate": 2.1800000000000003e-06,
+      "loss": 4.33,
+      "mean_token_accuracy": 0.3205883141607046,
+      "num_tokens": 1087616.0,
+      "step": 110
+    },
+    {
+      "entropy": 4.196488052606583,
+      "epoch": 0.096,
+      "grad_norm": 21.25,
+      "learning_rate": 2.38e-06,
+      "loss": 4.1941,
+      "mean_token_accuracy": 0.3275640026666224,
+      "num_tokens": 1182339.0,
+      "step": 120
+    },
+    {
+      "entropy": 4.199820376932621,
+      "epoch": 0.104,
+      "grad_norm": 22.25,
+      "learning_rate": 2.5800000000000003e-06,
+      "loss": 4.1709,
+      "mean_token_accuracy": 0.33598021846264603,
+      "num_tokens": 1284449.0,
+      "step": 130
+    },
+    {
+      "entropy": 4.0606200903654095,
+      "epoch": 0.112,
+      "grad_norm": 22.375,
+      "learning_rate": 2.7800000000000005e-06,
+      "loss": 4.0429,
+      "mean_token_accuracy": 0.34365201108157634,
+      "num_tokens": 1386193.0,
+      "step": 140
+    },
+    {
+      "entropy": 3.976654475927353,
+      "epoch": 0.12,
+      "grad_norm": 25.125,
+      "learning_rate": 2.9800000000000003e-06,
+      "loss": 3.9428,
+      "mean_token_accuracy": 0.35409670416265726,
+      "num_tokens": 1484528.0,
+      "step": 150
+    },
+    {
+      "entropy": 3.9824412897229196,
+      "epoch": 0.128,
+      "grad_norm": 26.5,
+      "learning_rate": 3.1800000000000005e-06,
+      "loss": 3.9225,
+      "mean_token_accuracy": 0.35787415150552987,
+      "num_tokens": 1583261.0,
+      "step": 160
+    },
+    {
+      "entropy": 3.90925519913435,
+      "epoch": 0.136,
+      "grad_norm": 27.75,
+      "learning_rate": 3.3800000000000007e-06,
+      "loss": 3.8876,
+      "mean_token_accuracy": 0.36578934341669084,
+      "num_tokens": 1678138.0,
+      "step": 170
+    },
+    {
+      "entropy": 3.846222287416458,
+      "epoch": 0.144,
+      "grad_norm": 22.5,
+      "learning_rate": 3.58e-06,
+      "loss": 3.7638,
+      "mean_token_accuracy": 0.37533934097737076,
+      "num_tokens": 1774826.0,
+      "step": 180
+    },
+    {
+      "entropy": 3.7961452692747115,
+      "epoch": 0.152,
+      "grad_norm": 20.625,
+      "learning_rate": 3.7800000000000002e-06,
+      "loss": 3.7673,
+      "mean_token_accuracy": 0.3747658669948578,
+      "num_tokens": 1874332.0,
+      "step": 190
+    },
+    {
+      "entropy": 3.8614122182130814,
+      "epoch": 0.16,
+      "grad_norm": 23.5,
+      "learning_rate": 3.980000000000001e-06,
+      "loss": 3.7987,
+      "mean_token_accuracy": 0.3740364293567836,
+      "num_tokens": 1971310.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.16,
+      "eval_coding_entropy": 1.94223806142807,
+      "eval_coding_loss": 2.1925477981567383,
+      "eval_coding_mean_token_accuracy": 0.5086279969215393,
+      "eval_coding_num_tokens": 1971310.0,
+      "eval_coding_runtime": 80.6501,
+      "eval_coding_samples_per_second": 6.2,
+      "eval_coding_steps_per_second": 3.1,
+      "step": 200
+    },
+    {
+      "epoch": 0.16,
+      "eval_biology_entropy": 3.7338000392913817,
+      "eval_biology_loss": 3.729203939437866,
+      "eval_biology_mean_token_accuracy": 0.38531523847579957,
+      "eval_biology_num_tokens": 1971310.0,
+      "eval_biology_runtime": 34.5728,
+      "eval_biology_samples_per_second": 14.462,
+      "eval_biology_steps_per_second": 7.231,
+      "step": 200
+    },
+    {
+      "entropy": 3.7424988850951193,
+      "epoch": 0.168,
+      "grad_norm": 20.0,
+      "learning_rate": 4.18e-06,
+      "loss": 3.6677,
+      "mean_token_accuracy": 0.38609125521034005,
+      "num_tokens": 2068010.0,
+      "step": 210
+    },
+    {
+      "entropy": 3.7136047944426536,
+      "epoch": 0.176,
+      "grad_norm": 24.5,
+      "learning_rate": 4.38e-06,
+      "loss": 3.6613,
+      "mean_token_accuracy": 0.38672342915087937,
+      "num_tokens": 2169453.0,
+      "step": 220
+    },
+    {
+      "entropy": 3.6596395045518877,
+      "epoch": 0.184,
+      "grad_norm": 23.875,
+      "learning_rate": 4.58e-06,
+      "loss": 3.6248,
+      "mean_token_accuracy": 0.38720098566263916,
+      "num_tokens": 2271716.0,
+      "step": 230
+    },
+    {
+      "entropy": 3.696573592722416,
+      "epoch": 0.192,
+      "grad_norm": 21.25,
+      "learning_rate": 4.78e-06,
+      "loss": 3.6063,
+      "mean_token_accuracy": 0.3891244037076831,
+      "num_tokens": 2366844.0,
+      "step": 240
+    },
+    {
+      "entropy": 3.589663599431515,
+      "epoch": 0.2,
+      "grad_norm": 24.5,
+      "learning_rate": 4.980000000000001e-06,
+      "loss": 3.5393,
+      "mean_token_accuracy": 0.3962410934269428,
+      "num_tokens": 2460881.0,
+      "step": 250
+    },
+    {
+      "entropy": 3.576136104762554,
+      "epoch": 0.208,
+      "grad_norm": 21.5,
+      "learning_rate": 5.18e-06,
+      "loss": 3.5108,
+      "mean_token_accuracy": 0.40014535561203957,
+      "num_tokens": 2559708.0,
+      "step": 260
+    },
+    {
+      "entropy": 3.5436186090111734,
+      "epoch": 0.216,
+      "grad_norm": 19.0,
+      "learning_rate": 5.380000000000001e-06,
+      "loss": 3.4656,
+      "mean_token_accuracy": 0.40589779019355776,
+      "num_tokens": 2655858.0,
+      "step": 270
+    },
+    {
+      "entropy": 3.516196443140507,
+      "epoch": 0.224,
+      "grad_norm": 27.125,
+      "learning_rate": 5.580000000000001e-06,
+      "loss": 3.4628,
+      "mean_token_accuracy": 0.40603156946599483,
+      "num_tokens": 2756533.0,
+      "step": 280
+    },
+    {
+      "entropy": 3.474540175497532,
+      "epoch": 0.232,
+      "grad_norm": 20.75,
+      "learning_rate": 5.78e-06,
+      "loss": 3.3983,
+      "mean_token_accuracy": 0.410215818695724,
+      "num_tokens": 2853451.0,
+      "step": 290
+    },
+    {
+      "entropy": 3.386892372369766,
+      "epoch": 0.24,
+      "grad_norm": 23.125,
+      "learning_rate": 5.98e-06,
+      "loss": 3.3306,
+      "mean_token_accuracy": 0.41694455314427614,
+      "num_tokens": 2948600.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.24,
+      "eval_coding_entropy": 1.9961137461662293,
+      "eval_coding_loss": 2.172008752822876,
+      "eval_coding_mean_token_accuracy": 0.5127531417608261,
+      "eval_coding_num_tokens": 2948600.0,
+      "eval_coding_runtime": 80.7019,
+      "eval_coding_samples_per_second": 6.196,
+      "eval_coding_steps_per_second": 3.098,
+      "step": 300
+    },
+    {
+      "epoch": 0.24,
+      "eval_biology_entropy": 3.484758809089661,
+      "eval_biology_loss": 3.406900644302368,
+      "eval_biology_mean_token_accuracy": 0.41982992458343504,
+      "eval_biology_num_tokens": 2948600.0,
+      "eval_biology_runtime": 34.5611,
+      "eval_biology_samples_per_second": 14.467,
+      "eval_biology_steps_per_second": 7.234,
+      "step": 300
+    },
+    {
+      "entropy": 3.495779898762703,
+      "epoch": 0.248,
+      "grad_norm": 21.75,
+      "learning_rate": 6.18e-06,
+      "loss": 3.3982,
+      "mean_token_accuracy": 0.41361558344215155,
+      "num_tokens": 3044471.0,
+      "step": 310
+    },
+    {
+      "entropy": 3.3510731413960455,
+      "epoch": 0.256,
+      "grad_norm": 26.375,
+      "learning_rate": 6.380000000000001e-06,
+      "loss": 3.3239,
+      "mean_token_accuracy": 0.41568741872906684,
+      "num_tokens": 3141473.0,
+      "step": 320
+    },
+    {
+      "entropy": 3.40442940890789,
+      "epoch": 0.264,
+      "grad_norm": 25.5,
+      "learning_rate": 6.5800000000000005e-06,
+      "loss": 3.3236,
+      "mean_token_accuracy": 0.41933552622795106,
+      "num_tokens": 3233662.0,
+      "step": 330
+    },
+    {
+      "entropy": 3.321935287117958,
+      "epoch": 0.272,
+      "grad_norm": 26.75,
+      "learning_rate": 6.780000000000001e-06,
+      "loss": 3.2912,
+      "mean_token_accuracy": 0.42016132585704324,
+      "num_tokens": 3337285.0,
+      "step": 340
+    },
+    {
+      "entropy": 3.317689270526171,
+      "epoch": 0.28,
+      "grad_norm": 30.125,
+      "learning_rate": 6.98e-06,
+      "loss": 3.2355,
+      "mean_token_accuracy": 0.42757263109087945,
+      "num_tokens": 3438579.0,
+      "step": 350
+    },
+    {
+      "entropy": 3.284139819443226,
+      "epoch": 0.288,
+      "grad_norm": 25.0,
+      "learning_rate": 7.180000000000001e-06,
+      "loss": 3.2066,
+      "mean_token_accuracy": 0.42989938110113146,
+      "num_tokens": 3541221.0,
+      "step": 360
+    },
+    {
+      "entropy": 3.3082104876637457,
+      "epoch": 0.296,
+      "grad_norm": 23.875,
+      "learning_rate": 7.3800000000000005e-06,
+      "loss": 3.2212,
+      "mean_token_accuracy": 0.43165368046611546,
+      "num_tokens": 3634910.0,
+      "step": 370
+    },
+    {
+      "entropy": 3.1951300844550135,
+      "epoch": 0.304,
+      "grad_norm": 23.25,
+      "learning_rate": 7.58e-06,
+      "loss": 3.1248,
+      "mean_token_accuracy": 0.43907446004450323,
+      "num_tokens": 3734952.0,
+      "step": 380
+    },
+    {
+      "entropy": 3.2739955455064775,
+      "epoch": 0.312,
+      "grad_norm": 30.25,
+      "learning_rate": 7.78e-06,
+      "loss": 3.1763,
+      "mean_token_accuracy": 0.42868839371949435,
+      "num_tokens": 3826750.0,
+      "step": 390
+    },
+    {
+      "entropy": 3.265377101302147,
+      "epoch": 0.32,
+      "grad_norm": 21.75,
+      "learning_rate": 7.980000000000002e-06,
+      "loss": 3.1998,
+      "mean_token_accuracy": 0.4301817687228322,
+      "num_tokens": 3922209.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.32,
+      "eval_coding_entropy": 2.0029401898384096,
+      "eval_coding_loss": 2.171020030975342,
+      "eval_coding_mean_token_accuracy": 0.5134941293001175,
+      "eval_coding_num_tokens": 3922209.0,
+      "eval_coding_runtime": 80.6648,
+      "eval_coding_samples_per_second": 6.198,
+      "eval_coding_steps_per_second": 3.099,
+      "step": 400
+    },
+    {
+      "epoch": 0.32,
+      "eval_biology_entropy": 3.2002078475952147,
+      "eval_biology_loss": 3.1781411170959473,
+      "eval_biology_mean_token_accuracy": 0.440620893239975,
+      "eval_biology_num_tokens": 3922209.0,
+      "eval_biology_runtime": 34.4923,
+      "eval_biology_samples_per_second": 14.496,
+      "eval_biology_steps_per_second": 7.248,
+      "step": 400
+    },
+    {
+      "entropy": 3.150459985435009,
+      "epoch": 0.328,
+      "grad_norm": 24.5,
+      "learning_rate": 8.18e-06,
+      "loss": 3.0904,
+      "mean_token_accuracy": 0.44143912978470323,
+      "num_tokens": 4018384.0,
+      "step": 410
+    },
+    {
+      "entropy": 3.1615461230278017,
+      "epoch": 0.336,
+      "grad_norm": 24.75,
+      "learning_rate": 8.380000000000001e-06,
+      "loss": 3.1239,
+      "mean_token_accuracy": 0.44076176043599846,
+      "num_tokens": 4115941.0,
+      "step": 420
+    },
+    {
+      "entropy": 3.1629561871290206,
+      "epoch": 0.344,
+      "grad_norm": 23.625,
+      "learning_rate": 8.580000000000001e-06,
+      "loss": 3.1118,
+      "mean_token_accuracy": 0.4393952677026391,
+      "num_tokens": 4212212.0,
+      "step": 430
+    },
+    {
+      "entropy": 3.173071324825287,
+      "epoch": 0.352,
+      "grad_norm": 23.375,
+      "learning_rate": 8.78e-06,
+      "loss": 3.0404,
+      "mean_token_accuracy": 0.4486429067328572,
+      "num_tokens": 4308755.0,
+      "step": 440
+    },
+    {
+      "entropy": 3.11899806112051,
+      "epoch": 0.36,
+      "grad_norm": 21.0,
+      "learning_rate": 8.98e-06,
+      "loss": 3.0634,
+      "mean_token_accuracy": 0.4478369506075978,
+      "num_tokens": 4407923.0,
+      "step": 450
+    },
+    {
+      "entropy": 3.0697645485401153,
+      "epoch": 0.368,
+      "grad_norm": 23.375,
+      "learning_rate": 9.180000000000002e-06,
+      "loss": 2.9971,
+      "mean_token_accuracy": 0.4510502349585295,
+      "num_tokens": 4505910.0,
+      "step": 460
+    },
+    {
+      "entropy": 3.0845467180013655,
+      "epoch": 0.376,
+      "grad_norm": 22.5,
+      "learning_rate": 9.38e-06,
+      "loss": 2.9916,
+      "mean_token_accuracy": 0.45388368237763643,
+      "num_tokens": 4599609.0,
+      "step": 470
+    },
+    {
+      "entropy": 3.0334658533334733,
+      "epoch": 0.384,
+      "grad_norm": 26.375,
+      "learning_rate": 9.58e-06,
+      "loss": 2.965,
+      "mean_token_accuracy": 0.4527210883796215,
+      "num_tokens": 4696101.0,
+      "step": 480
+    },
+    {
+      "entropy": 3.0365326166152955,
+      "epoch": 0.392,
+      "grad_norm": 24.75,
+      "learning_rate": 9.780000000000001e-06,
+      "loss": 2.9627,
+      "mean_token_accuracy": 0.45046965070068834,
+      "num_tokens": 4793234.0,
+      "step": 490
+    },
+    {
+      "entropy": 3.022104251384735,
+      "epoch": 0.4,
+      "grad_norm": 26.625,
+      "learning_rate": 9.980000000000001e-06,
+      "loss": 2.9389,
+      "mean_token_accuracy": 0.4577130895107985,
+      "num_tokens": 4889445.0,
+      "step": 500
+    },
+    {
+      "epoch": 0.4,
+      "eval_coding_entropy": 2.023589345932007,
+      "eval_coding_loss": 2.1741902828216553,
+      "eval_coding_mean_token_accuracy": 0.5137698065042495,
+      "eval_coding_num_tokens": 4889445.0,
+      "eval_coding_runtime": 80.6183,
+      "eval_coding_samples_per_second": 6.202,
+      "eval_coding_steps_per_second": 3.101,
+      "step": 500
+    },
+    {
+      "epoch": 0.4,
+      "eval_biology_entropy": 3.0067848119735716,
+      "eval_biology_loss": 2.9910171031951904,
+      "eval_biology_mean_token_accuracy": 0.4611110507249832,
+      "eval_biology_num_tokens": 4889445.0,
+      "eval_biology_runtime": 34.2723,
+      "eval_biology_samples_per_second": 14.589,
+      "eval_biology_steps_per_second": 7.295,
+      "step": 500
+    },
+    {
+      "entropy": 3.043118818104267,
+      "epoch": 0.408,
+      "grad_norm": 21.875,
+      "learning_rate": 1.018e-05,
+      "loss": 2.9591,
+      "mean_token_accuracy": 0.4550719428807497,
+      "num_tokens": 4987724.0,
+      "step": 510
+    },
+    {
+      "entropy": 3.051181846857071,
+      "epoch": 0.416,
+      "grad_norm": 23.0,
+      "learning_rate": 1.038e-05,
+      "loss": 2.9554,
+      "mean_token_accuracy": 0.45290321484208107,
+      "num_tokens": 5087624.0,
+      "step": 520
+    },
+    {
+      "entropy": 3.023839880526066,
+      "epoch": 0.424,
+      "grad_norm": 37.25,
+      "learning_rate": 1.0580000000000002e-05,
+      "loss": 2.9739,
+      "mean_token_accuracy": 0.45466579012572766,
+      "num_tokens": 5178478.0,
+      "step": 530
+    },
+    {
+      "entropy": 2.9455194897949695,
+      "epoch": 0.432,
+      "grad_norm": 24.875,
+      "learning_rate": 1.0780000000000002e-05,
+      "loss": 2.8345,
+      "mean_token_accuracy": 0.4659645749256015,
+      "num_tokens": 5273794.0,
+      "step": 540
+    },
+    {
+      "entropy": 3.00012748837471,
+      "epoch": 0.44,
+      "grad_norm": 27.5,
+      "learning_rate": 1.0980000000000002e-05,
+      "loss": 2.9082,
+      "mean_token_accuracy": 0.46043862625956533,
+      "num_tokens": 5365232.0,
+      "step": 550
+    },
+    {
+      "entropy": 2.951145689189434,
+      "epoch": 0.448,
+      "grad_norm": 25.75,
+      "learning_rate": 1.1180000000000001e-05,
+      "loss": 2.8758,
+      "mean_token_accuracy": 0.4649818094447255,
+      "num_tokens": 5463288.0,
+      "step": 560
+    },
+    {
+      "entropy": 2.9195503376424314,
+      "epoch": 0.456,
+      "grad_norm": 24.5,
+      "learning_rate": 1.138e-05,
+      "loss": 2.8283,
+      "mean_token_accuracy": 0.46871429681777954,
+      "num_tokens": 5561317.0,
+      "step": 570
+    },
+    {
+      "entropy": 2.895768219232559,
+      "epoch": 0.464,
+      "grad_norm": 23.0,
+      "learning_rate": 1.1580000000000001e-05,
+      "loss": 2.8347,
+      "mean_token_accuracy": 0.4678240044042468,
+      "num_tokens": 5656827.0,
+      "step": 580
+    },
+    {
+      "entropy": 2.919021227955818,
+      "epoch": 0.472,
+      "grad_norm": 21.875,
+      "learning_rate": 1.178e-05,
+      "loss": 2.8239,
+      "mean_token_accuracy": 0.46788553427904844,
+      "num_tokens": 5754593.0,
+      "step": 590
+    },
+    {
+      "entropy": 2.9184277661144735,
+      "epoch": 0.48,
+      "grad_norm": 19.375,
+      "learning_rate": 1.198e-05,
+      "loss": 2.8284,
+      "mean_token_accuracy": 0.46977027915418146,
+      "num_tokens": 5853311.0,
+      "step": 600
+    },
+    {
+      "epoch": 0.48,
+      "eval_coding_entropy": 1.9910136852264404,
+      "eval_coding_loss": 2.1933281421661377,
+      "eval_coding_mean_token_accuracy": 0.5116564213037491,
+      "eval_coding_num_tokens": 5853311.0,
+      "eval_coding_runtime": 80.6686,
+      "eval_coding_samples_per_second": 6.198,
+      "eval_coding_steps_per_second": 3.099,
+      "step": 600
+    },
+    {
+      "epoch": 0.48,
+      "eval_biology_entropy": 2.7489197278022766,
+      "eval_biology_loss": 2.839024305343628,
+      "eval_biology_mean_token_accuracy": 0.478046000957489,
+      "eval_biology_num_tokens": 5853311.0,
+      "eval_biology_runtime": 34.524,
+      "eval_biology_samples_per_second": 14.483,
+      "eval_biology_steps_per_second": 7.241,
+      "step": 600
+    },
+    {
+      "entropy": 2.761022947728634,
+      "epoch": 0.488,
+      "grad_norm": 23.125,
+      "learning_rate": 1.218e-05,
+      "loss": 2.7114,
+      "mean_token_accuracy": 0.48137810807675124,
+      "num_tokens": 5952079.0,
+      "step": 610
+    },
+    {
+      "entropy": 2.802033407241106,
+      "epoch": 0.496,
+      "grad_norm": 24.875,
+      "learning_rate": 1.2380000000000002e-05,
+      "loss": 2.7172,
+      "mean_token_accuracy": 0.480432684533298,
+      "num_tokens": 6049788.0,
+      "step": 620
+    },
+    {
+      "entropy": 2.8570902943611145,
+      "epoch": 0.504,
+      "grad_norm": 24.375,
+      "learning_rate": 1.2580000000000002e-05,
+      "loss": 2.7682,
+      "mean_token_accuracy": 0.4740563785657287,
+      "num_tokens": 6142124.0,
+      "step": 630
+    },
+    {
+      "entropy": 2.803235235065222,
+      "epoch": 0.512,
+      "grad_norm": 19.25,
+      "learning_rate": 1.2780000000000001e-05,
+      "loss": 2.7612,
+      "mean_token_accuracy": 0.48143365047872066,
+      "num_tokens": 6243825.0,
+      "step": 640
+    },
+    {
+      "entropy": 2.8179355546832086,
+      "epoch": 0.52,
+      "grad_norm": 24.375,
+      "learning_rate": 1.2980000000000001e-05,
+      "loss": 2.7405,
+      "mean_token_accuracy": 0.47674270141869785,
+      "num_tokens": 6337416.0,
+      "step": 650
+    },
+    {
+      "entropy": 2.783262898027897,
+      "epoch": 0.528,
+      "grad_norm": 23.5,
+      "learning_rate": 1.3180000000000001e-05,
+      "loss": 2.6905,
+      "mean_token_accuracy": 0.48337399754673244,
+      "num_tokens": 6434188.0,
+      "step": 660
+    },
+    {
+      "entropy": 2.7997302234172823,
+      "epoch": 0.536,
+      "grad_norm": 18.25,
+      "learning_rate": 1.3380000000000002e-05,
+      "loss": 2.7293,
+      "mean_token_accuracy": 0.4824898593127728,
+      "num_tokens": 6536879.0,
+      "step": 670
+    },
+    {
+      "entropy": 2.8198735982179643,
+      "epoch": 0.544,
+      "grad_norm": 23.75,
+      "learning_rate": 1.3580000000000002e-05,
+      "loss": 2.7448,
+      "mean_token_accuracy": 0.48089918941259385,
+      "num_tokens": 6629832.0,
+      "step": 680
+    },
+    {
+      "entropy": 2.7923016041517257,
+      "epoch": 0.552,
+      "grad_norm": 21.125,
+      "learning_rate": 1.378e-05,
+      "loss": 2.7259,
+      "mean_token_accuracy": 0.4828041112050414,
+      "num_tokens": 6723756.0,
+      "step": 690
+    },
+    {
+      "entropy": 2.6786952562630177,
+      "epoch": 0.56,
+      "grad_norm": 17.625,
+      "learning_rate": 1.398e-05,
+      "loss": 2.6311,
+      "mean_token_accuracy": 0.49184530321508646,
+      "num_tokens": 6824668.0,
+      "step": 700
+    },
+    {
+      "epoch": 0.56,
+      "eval_coding_entropy": 2.0318734068870543,
+      "eval_coding_loss": 2.1769535541534424,
+      "eval_coding_mean_token_accuracy": 0.5146223112344742,
+      "eval_coding_num_tokens": 6824668.0,
+      "eval_coding_runtime": 80.6202,
+      "eval_coding_samples_per_second": 6.202,
+      "eval_coding_steps_per_second": 3.101,
+      "step": 700
+    },
+    {
+      "epoch": 0.56,
+      "eval_biology_entropy": 2.7879175729751586,
+      "eval_biology_loss": 2.7142910957336426,
+      "eval_biology_mean_token_accuracy": 0.49161617302894595,
+      "eval_biology_num_tokens": 6824668.0,
+      "eval_biology_runtime": 34.5513,
+      "eval_biology_samples_per_second": 14.471,
+      "eval_biology_steps_per_second": 7.236,
+      "step": 700
+    },
+    {
+      "entropy": 2.731781490147114,
+      "epoch": 0.568,
+      "grad_norm": 22.0,
+      "learning_rate": 1.418e-05,
+      "loss": 2.6484,
+      "mean_token_accuracy": 0.4882859470322728,
+      "num_tokens": 6928939.0,
+      "step": 710
+    },
+    {
+      "entropy": 2.7027835391461847,
+      "epoch": 0.576,
+      "grad_norm": 20.375,
+      "learning_rate": 1.4380000000000001e-05,
+      "loss": 2.6258,
+      "mean_token_accuracy": 0.48718241080641744,
+      "num_tokens": 7025923.0,
+      "step": 720
+    },
+    {
+      "entropy": 2.697667721658945,
+      "epoch": 0.584,
+      "grad_norm": 24.25,
+      "learning_rate": 1.4580000000000001e-05,
+      "loss": 2.6369,
+      "mean_token_accuracy": 0.4901735274121165,
+      "num_tokens": 7122614.0,
+      "step": 730
+    },
+    {
+      "entropy": 2.6893031299114227,
+      "epoch": 0.592,
+      "grad_norm": 20.625,
+      "learning_rate": 1.478e-05,
+      "loss": 2.5922,
+      "mean_token_accuracy": 0.49478939212858675,
+      "num_tokens": 7218224.0,
+      "step": 740
+    },
+    {
+      "entropy": 2.7066168434917928,
+      "epoch": 0.6,
+      "grad_norm": 31.375,
+      "learning_rate": 1.498e-05,
+      "loss": 2.6573,
+      "mean_token_accuracy": 0.4925660764798522,
+      "num_tokens": 7322313.0,
+      "step": 750
+    },
+    {
+      "entropy": 2.69825324267149,
+      "epoch": 0.608,
+      "grad_norm": 21.25,
+      "learning_rate": 1.5180000000000002e-05,
+      "loss": 2.6359,
+      "mean_token_accuracy": 0.48993428722023963,
+      "num_tokens": 7421216.0,
+      "step": 760
+    },
+    {
+      "entropy": 2.649106864631176,
+      "epoch": 0.616,
+      "grad_norm": 22.375,
+      "learning_rate": 1.5380000000000002e-05,
+      "loss": 2.5961,
+      "mean_token_accuracy": 0.4944638149812818,
+      "num_tokens": 7518453.0,
+      "step": 770
+    },
+    {
+      "entropy": 2.6871195189654826,
+      "epoch": 0.624,
+      "grad_norm": 25.0,
+      "learning_rate": 1.5580000000000003e-05,
+      "loss": 2.5837,
+      "mean_token_accuracy": 0.4975929221138358,
+      "num_tokens": 7617258.0,
+      "step": 780
+    },
+    {
+      "entropy": 2.6687942549586294,
+      "epoch": 0.632,
+      "grad_norm": 20.25,
+      "learning_rate": 1.578e-05,
+      "loss": 2.607,
+      "mean_token_accuracy": 0.4943814314901829,
+      "num_tokens": 7714799.0,
+      "step": 790
+    },
+    {
+      "entropy": 2.679428345710039,
+      "epoch": 0.64,
+      "grad_norm": 22.125,
+      "learning_rate": 1.5980000000000003e-05,
+      "loss": 2.5938,
+      "mean_token_accuracy": 0.4952407155185938,
+      "num_tokens": 7806173.0,
+      "step": 800
+    },
+    {
+      "epoch": 0.64,
+      "eval_coding_entropy": 1.9768212442398072,
+      "eval_coding_loss": 2.224097967147827,
+      "eval_coding_mean_token_accuracy": 0.5107745778560638,
+      "eval_coding_num_tokens": 7806173.0,
+      "eval_coding_runtime": 80.6252,
+      "eval_coding_samples_per_second": 6.202,
+      "eval_coding_steps_per_second": 3.101,
+      "step": 800
+    },
+    {
+      "epoch": 0.64,
+      "eval_biology_entropy": 2.545660098552704,
+      "eval_biology_loss": 2.5940585136413574,
+      "eval_biology_mean_token_accuracy": 0.5062511616945267,
+      "eval_biology_num_tokens": 7806173.0,
+      "eval_biology_runtime": 34.513,
+      "eval_biology_samples_per_second": 14.487,
+      "eval_biology_steps_per_second": 7.244,
+      "step": 800
+    },
+    {
+      "entropy": 2.6028101302683355,
+      "epoch": 0.648,
+      "grad_norm": 21.125,
+      "learning_rate": 1.618e-05,
+      "loss": 2.5534,
+      "mean_token_accuracy": 0.49852073900401594,
+      "num_tokens": 7903614.0,
+      "step": 810
+    },
+    {
+      "entropy": 2.665095742046833,
+      "epoch": 0.656,
+      "grad_norm": 20.5,
+      "learning_rate": 1.638e-05,
+      "loss": 2.6054,
+      "mean_token_accuracy": 0.49321497566998007,
+      "num_tokens": 8000514.0,
+      "step": 820
+    },
+    {
+      "entropy": 2.5793248772621156,
+      "epoch": 0.664,
+      "grad_norm": 17.875,
+      "learning_rate": 1.658e-05,
+      "loss": 2.5159,
+      "mean_token_accuracy": 0.5033486964181065,
+      "num_tokens": 8096743.0,
+      "step": 830
+    },
+    {
+      "entropy": 2.641881722956896,
+      "epoch": 0.672,
+      "grad_norm": 23.875,
+      "learning_rate": 1.6780000000000002e-05,
+      "loss": 2.5868,
+      "mean_token_accuracy": 0.4976224903017282,
+      "num_tokens": 8187884.0,
+      "step": 840
+    },
+    {
+      "entropy": 2.5881785310804846,
+      "epoch": 0.68,
+      "grad_norm": 15.3125,
+      "learning_rate": 1.698e-05,
+      "loss": 2.5281,
+      "mean_token_accuracy": 0.504815262183547,
+      "num_tokens": 8288387.0,
+      "step": 850
+    },
+    {
+      "entropy": 2.5754016518592833,
+      "epoch": 0.688,
+      "grad_norm": 16.25,
+      "learning_rate": 1.718e-05,
+      "loss": 2.499,
+      "mean_token_accuracy": 0.5069523887708783,
+      "num_tokens": 8390643.0,
+      "step": 860
+    },
+    {
+      "entropy": 2.6078361086547375,
+      "epoch": 0.696,
+      "grad_norm": 20.625,
+      "learning_rate": 1.7380000000000003e-05,
+      "loss": 2.5317,
+      "mean_token_accuracy": 0.5026106856763363,
+      "num_tokens": 8484952.0,
+      "step": 870
+    },
+    {
+      "entropy": 2.5303950123488903,
+      "epoch": 0.704,
+      "grad_norm": 21.5,
+      "learning_rate": 1.758e-05,
+      "loss": 2.4753,
+      "mean_token_accuracy": 0.5120382806286216,
+      "num_tokens": 8583266.0,
+      "step": 880
+    },
+    {
+      "entropy": 2.4733798749744893,
+      "epoch": 0.712,
+      "grad_norm": 18.875,
+      "learning_rate": 1.7780000000000003e-05,
+      "loss": 2.3787,
+      "mean_token_accuracy": 0.5168206626549363,
+      "num_tokens": 8684764.0,
+      "step": 890
+    },
+    {
+      "entropy": 2.4695195190608503,
+      "epoch": 0.72,
+      "grad_norm": 19.5,
+      "learning_rate": 1.798e-05,
+      "loss": 2.4147,
+      "mean_token_accuracy": 0.5180861415341497,
+      "num_tokens": 8786233.0,
+      "step": 900
+    },
+    {
+      "epoch": 0.72,
+      "eval_coding_entropy": 1.9488533411026001,
+      "eval_coding_loss": 2.2393529415130615,
+      "eval_coding_mean_token_accuracy": 0.5095039163827896,
+      "eval_coding_num_tokens": 8786233.0,
+      "eval_coding_runtime": 80.6302,
+      "eval_coding_samples_per_second": 6.201,
+      "eval_coding_steps_per_second": 3.101,
+      "step": 900
+    },
+    {
+      "epoch": 0.72,
+      "eval_biology_entropy": 2.453756254196167,
+      "eval_biology_loss": 2.5033159255981445,
+      "eval_biology_mean_token_accuracy": 0.5165660305023193,
+      "eval_biology_num_tokens": 8786233.0,
+      "eval_biology_runtime": 34.5573,
+      "eval_biology_samples_per_second": 14.469,
+      "eval_biology_steps_per_second": 7.234,
+      "step": 900
+    },
+    {
+      "entropy": 2.4767140075564384,
+      "epoch": 0.728,
+      "grad_norm": 16.625,
+      "learning_rate": 1.8180000000000002e-05,
+      "loss": 2.4331,
+      "mean_token_accuracy": 0.5123517647385597,
+      "num_tokens": 8881233.0,
+      "step": 910
+    },
+    {
+      "entropy": 2.475684958696365,
+      "epoch": 0.736,
+      "grad_norm": 22.5,
+      "learning_rate": 1.8380000000000004e-05,
+      "loss": 2.3936,
+      "mean_token_accuracy": 0.5183122556656599,
+      "num_tokens": 8982898.0,
+      "step": 920
+    },
+    {
+      "entropy": 2.495747657865286,
+      "epoch": 0.744,
+      "grad_norm": 17.5,
+      "learning_rate": 1.858e-05,
+      "loss": 2.3898,
+      "mean_token_accuracy": 0.5161861320957541,
+      "num_tokens": 9078921.0,
+      "step": 930
+    },
+    {
+      "entropy": 2.4943869531154634,
+      "epoch": 0.752,
+      "grad_norm": 17.0,
+      "learning_rate": 1.878e-05,
+      "loss": 2.4483,
+      "mean_token_accuracy": 0.5111985813826323,
+      "num_tokens": 9174361.0,
+      "step": 940
+    },
+    {
+      "entropy": 2.4575762204825877,
+      "epoch": 0.76,
+      "grad_norm": 16.0,
+      "learning_rate": 1.898e-05,
+      "loss": 2.4013,
+      "mean_token_accuracy": 0.5162310445681214,
+      "num_tokens": 9272571.0,
+      "step": 950
+    },
+    {
+      "entropy": 2.431351166218519,
+      "epoch": 0.768,
+      "grad_norm": 16.875,
+      "learning_rate": 1.918e-05,
+      "loss": 2.3378,
+      "mean_token_accuracy": 0.521942549943924,
+      "num_tokens": 9373430.0,
+      "step": 960
+    },
+    {
+      "entropy": 2.4587918162345885,
+      "epoch": 0.776,
+      "grad_norm": 15.125,
+      "learning_rate": 1.938e-05,
+      "loss": 2.4059,
+      "mean_token_accuracy": 0.5153441710397602,
+      "num_tokens": 9468903.0,
+      "step": 970
+    },
+    {
+      "entropy": 2.425259476900101,
+      "epoch": 0.784,
+      "grad_norm": 16.875,
+      "learning_rate": 1.9580000000000002e-05,
+      "loss": 2.3285,
+      "mean_token_accuracy": 0.5206136658787728,
+      "num_tokens": 9570409.0,
+      "step": 980
+    },
+    {
+      "entropy": 2.42026736587286,
+      "epoch": 0.792,
+      "grad_norm": 15.9375,
+      "learning_rate": 1.978e-05,
+      "loss": 2.3534,
+      "mean_token_accuracy": 0.5174617141485214,
+      "num_tokens": 9669925.0,
+      "step": 990
+    },
+    {
+      "entropy": 2.4630366936326027,
+      "epoch": 0.8,
+      "grad_norm": 22.125,
+      "learning_rate": 1.9980000000000002e-05,
+      "loss": 2.426,
+      "mean_token_accuracy": 0.5160385722294449,
+      "num_tokens": 9766360.0,
+      "step": 1000
+    },
+    {
+      "epoch": 0.8,
+      "eval_coding_entropy": 2.0347842559814455,
+      "eval_coding_loss": 2.295905351638794,
+      "eval_coding_mean_token_accuracy": 0.5001156101226807,
+      "eval_coding_num_tokens": 9766360.0,
+      "eval_coding_runtime": 80.6454,
+      "eval_coding_samples_per_second": 6.2,
+      "eval_coding_steps_per_second": 3.1,
+      "step": 1000
+    },
+    {
+      "epoch": 0.8,
+      "eval_biology_entropy": 2.336130635738373,
+      "eval_biology_loss": 2.4337234497070312,
+      "eval_biology_mean_token_accuracy": 0.5239072642326355,
+      "eval_biology_num_tokens": 9766360.0,
+      "eval_biology_runtime": 34.2599,
+      "eval_biology_samples_per_second": 14.594,
+      "eval_biology_steps_per_second": 7.297,
+      "step": 1000
+    },
+    {
+      "entropy": 2.3636431239545344,
+      "epoch": 0.808,
+      "grad_norm": 18.125,
+      "learning_rate": 1.9980000000000002e-05,
+      "loss": 2.3056,
+      "mean_token_accuracy": 0.5313040925189852,
+      "num_tokens": 9868563.0,
+      "step": 1010
+    },
+    {
+      "entropy": 2.395121469348669,
+      "epoch": 0.816,
+      "grad_norm": 19.25,
+      "learning_rate": 1.995777777777778e-05,
+      "loss": 2.3484,
+      "mean_token_accuracy": 0.5250098271295428,
+      "num_tokens": 9963834.0,
+      "step": 1020
+    },
+    {
+      "entropy": 2.4034232169389727,
+      "epoch": 0.824,
+      "grad_norm": 15.75,
+      "learning_rate": 1.9935555555555557e-05,
+      "loss": 2.3392,
+      "mean_token_accuracy": 0.5230668963864445,
+      "num_tokens": 10064222.0,
+      "step": 1030
+    },
+    {
+      "entropy": 2.351021596044302,
+      "epoch": 0.832,
+      "grad_norm": 25.5,
+      "learning_rate": 1.9913333333333335e-05,
+      "loss": 2.2799,
+      "mean_token_accuracy": 0.5308501249179244,
+      "num_tokens": 10160950.0,
+      "step": 1040
+    },
+    {
+      "entropy": 2.3506353706121446,
+      "epoch": 0.84,
+      "grad_norm": 17.375,
+      "learning_rate": 1.9891111111111112e-05,
+      "loss": 2.2904,
+      "mean_token_accuracy": 0.5276814445853233,
+      "num_tokens": 10261255.0,
+      "step": 1050
+    },
+    {
+      "entropy": 2.330299597978592,
+      "epoch": 0.848,
+      "grad_norm": 14.625,
+      "learning_rate": 1.986888888888889e-05,
+      "loss": 2.2807,
+      "mean_token_accuracy": 0.533601225912571,
+      "num_tokens": 10363131.0,
+      "step": 1060
+    },
+    {
+      "entropy": 2.310487486422062,
+      "epoch": 0.856,
+      "grad_norm": 12.6875,
+      "learning_rate": 1.9846666666666668e-05,
+      "loss": 2.2289,
+      "mean_token_accuracy": 0.5375016804784536,
+      "num_tokens": 10459821.0,
+      "step": 1070
+    },
+    {
+      "entropy": 2.3588565312325955,
+      "epoch": 0.864,
+      "grad_norm": 17.5,
+      "learning_rate": 1.9824444444444445e-05,
+      "loss": 2.3075,
+      "mean_token_accuracy": 0.5270110530778765,
+      "num_tokens": 10558003.0,
+      "step": 1080
+    },
+    {
+      "entropy": 2.3483712531626226,
+      "epoch": 0.872,
+      "grad_norm": 15.6875,
+      "learning_rate": 1.9802222222222226e-05,
+      "loss": 2.2938,
+      "mean_token_accuracy": 0.5279141725972295,
+      "num_tokens": 10657009.0,
+      "step": 1090
+    },
+    {
+      "entropy": 2.301337753236294,
+      "epoch": 0.88,
+      "grad_norm": 16.125,
+      "learning_rate": 1.978e-05,
+      "loss": 2.2388,
+      "mean_token_accuracy": 0.5370641218498349,
+      "num_tokens": 10754470.0,
+      "step": 1100
+    },
+    {
+      "epoch": 0.88,
+      "eval_coding_entropy": 2.0812352204322817,
+      "eval_coding_loss": 2.270843744277954,
+      "eval_coding_mean_token_accuracy": 0.501986554980278,
+      "eval_coding_num_tokens": 10754470.0,
+      "eval_coding_runtime": 80.6786,
+      "eval_coding_samples_per_second": 6.197,
+      "eval_coding_steps_per_second": 3.099,
+      "step": 1100
+    },
+    {
+      "epoch": 0.88,
+      "eval_biology_entropy": 2.3388767604827883,
+      "eval_biology_loss": 2.3078601360321045,
+      "eval_biology_mean_token_accuracy": 0.5390308572053909,
+      "eval_biology_num_tokens": 10754470.0,
+      "eval_biology_runtime": 34.486,
+      "eval_biology_samples_per_second": 14.499,
+      "eval_biology_steps_per_second": 7.249,
+      "step": 1100
+    },
+    {
+      "entropy": 2.318739663809538,
+      "epoch": 0.888,
+      "grad_norm": 14.0,
+      "learning_rate": 1.975777777777778e-05,
+      "loss": 2.2605,
+      "mean_token_accuracy": 0.5366495896130801,
+      "num_tokens": 10858512.0,
+      "step": 1110
+    },
+    {
+      "entropy": 2.345344864577055,
+      "epoch": 0.896,
+      "grad_norm": 19.25,
+      "learning_rate": 1.9735555555555556e-05,
+      "loss": 2.3028,
+      "mean_token_accuracy": 0.5301774218678474,
+      "num_tokens": 10948943.0,
+      "step": 1120
+    },
+    {
+      "entropy": 2.305903819203377,
+      "epoch": 0.904,
+      "grad_norm": 13.6875,
+      "learning_rate": 1.9713333333333337e-05,
+      "loss": 2.2382,
+      "mean_token_accuracy": 0.5397843992337584,
+      "num_tokens": 11043458.0,
+      "step": 1130
+    },
+    {
+      "entropy": 2.2818943060934544,
+      "epoch": 0.912,
+      "grad_norm": 14.0625,
+      "learning_rate": 1.969111111111111e-05,
+      "loss": 2.193,
+      "mean_token_accuracy": 0.5407467754557729,
+      "num_tokens": 11142613.0,
+      "step": 1140
+    },
+    {
+      "entropy": 2.3112077616155147,
+      "epoch": 0.92,
+      "grad_norm": 13.0,
+      "learning_rate": 1.9668888888888892e-05,
+      "loss": 2.2325,
+      "mean_token_accuracy": 0.5370767971500754,
+      "num_tokens": 11243618.0,
+      "step": 1150
+    },
+    {
+      "entropy": 2.293438446521759,
+      "epoch": 0.928,
+      "grad_norm": 13.75,
+      "learning_rate": 1.9646666666666666e-05,
+      "loss": 2.2228,
+      "mean_token_accuracy": 0.5370372215285897,
+      "num_tokens": 11336779.0,
+      "step": 1160
+    },
+    {
+      "entropy": 2.2055763468146323,
+      "epoch": 0.936,
+      "grad_norm": 11.5,
+      "learning_rate": 1.9624444444444447e-05,
+      "loss": 2.1619,
+      "mean_token_accuracy": 0.5508569484576583,
+      "num_tokens": 11441397.0,
+      "step": 1170
+    },
+    {
+      "entropy": 2.2443367682397364,
+      "epoch": 0.944,
+      "grad_norm": 14.0,
+      "learning_rate": 1.9602222222222225e-05,
+      "loss": 2.1774,
+      "mean_token_accuracy": 0.5465354908257722,
+      "num_tokens": 11536381.0,
+      "step": 1180
+    },
+    {
+      "entropy": 2.220568550378084,
+      "epoch": 0.952,
+      "grad_norm": 12.3125,
+      "learning_rate": 1.9580000000000002e-05,
+      "loss": 2.143,
+      "mean_token_accuracy": 0.5484613511711359,
+      "num_tokens": 11634931.0,
+      "step": 1190
+    },
+    {
+      "entropy": 2.252422795444727,
+      "epoch": 0.96,
+      "grad_norm": 15.0,
+      "learning_rate": 1.955777777777778e-05,
+      "loss": 2.1688,
+      "mean_token_accuracy": 0.5444858035072684,
+      "num_tokens": 11735527.0,
+      "step": 1200
+    },
+    {
+      "epoch": 0.96,
+      "eval_coding_entropy": 2.028546452522278,
+      "eval_coding_loss": 2.313232898712158,
+      "eval_coding_mean_token_accuracy": 0.4952474715709686,
+      "eval_coding_num_tokens": 11735527.0,
+      "eval_coding_runtime": 80.6645,
+      "eval_coding_samples_per_second": 6.199,
+      "eval_coding_steps_per_second": 3.099,
+      "step": 1200
+    },
+    {
+      "epoch": 0.96,
+      "eval_biology_entropy": 2.2140285110473634,
+      "eval_biology_loss": 2.2470810413360596,
+      "eval_biology_mean_token_accuracy": 0.5471095204353332,
+      "eval_biology_num_tokens": 11735527.0,
+      "eval_biology_runtime": 34.5057,
+      "eval_biology_samples_per_second": 14.49,
+      "eval_biology_steps_per_second": 7.245,
+      "step": 1200
+    },
+    {
+      "entropy": 2.200479466468096,
+      "epoch": 0.968,
+      "grad_norm": 12.0,
+      "learning_rate": 1.9535555555555557e-05,
+      "loss": 2.1626,
+      "mean_token_accuracy": 0.5465732211247086,
+      "num_tokens": 11829912.0,
+      "step": 1210
+    },
+    {
+      "entropy": 2.210404744744301,
+      "epoch": 0.976,
+      "grad_norm": 12.875,
+      "learning_rate": 1.9513333333333335e-05,
+      "loss": 2.1098,
+      "mean_token_accuracy": 0.5487013891339302,
+      "num_tokens": 11930935.0,
+      "step": 1220
+    },
+    {
+      "entropy": 2.2029160290956495,
+      "epoch": 0.984,
+      "grad_norm": 11.5625,
+      "learning_rate": 1.9491111111111113e-05,
+      "loss": 2.1556,
+      "mean_token_accuracy": 0.5468338681384921,
+      "num_tokens": 12028616.0,
+      "step": 1230
+    },
+    {
+      "entropy": 2.2255366176366804,
+      "epoch": 0.992,
+      "grad_norm": 12.0,
+      "learning_rate": 1.946888888888889e-05,
+      "loss": 2.1711,
+      "mean_token_accuracy": 0.547148628346622,
+      "num_tokens": 12129585.0,
+      "step": 1240
+    },
+    {
+      "entropy": 2.2456682115793227,
+      "epoch": 1.0,
+      "grad_norm": 11.5,
+      "learning_rate": 1.9446666666666668e-05,
+      "loss": 2.1789,
+      "mean_token_accuracy": 0.544199176505208,
+      "num_tokens": 12227772.0,
+      "step": 1250
+    },
+    {
+      "entropy": 2.04770105406642,
+      "epoch": 1.008,
+      "grad_norm": 10.5,
+      "learning_rate": 1.9424444444444446e-05,
+      "loss": 1.9599,
+      "mean_token_accuracy": 0.5732083506882191,
+      "num_tokens": 12325847.0,
+      "step": 1260
+    },
+    {
+      "entropy": 1.9857938677072524,
+      "epoch": 1.016,
+      "grad_norm": 10.6875,
+      "learning_rate": 1.9402222222222223e-05,
+      "loss": 1.9223,
+      "mean_token_accuracy": 0.5803239606320858,
+      "num_tokens": 12428864.0,
+      "step": 1270
+    },
+    {
+      "entropy": 2.0265946321189405,
+      "epoch": 1.024,
+      "grad_norm": 14.0625,
+      "learning_rate": 1.938e-05,
+      "loss": 1.9706,
+      "mean_token_accuracy": 0.5730046048760414,
+      "num_tokens": 12530820.0,
+      "step": 1280
+    },
+    {
+      "entropy": 2.033829855918884,
+      "epoch": 1.032,
+      "grad_norm": 12.125,
+      "learning_rate": 1.935777777777778e-05,
+      "loss": 1.9723,
+      "mean_token_accuracy": 0.5693843860179186,
+      "num_tokens": 12628246.0,
+      "step": 1290
+    },
+    {
+      "entropy": 2.1004927776753903,
+      "epoch": 1.04,
+      "grad_norm": 11.6875,
+      "learning_rate": 1.9335555555555556e-05,
+      "loss": 2.0295,
+      "mean_token_accuracy": 0.5636359706521035,
+      "num_tokens": 12725384.0,
+      "step": 1300
+    },
+    {
+      "epoch": 1.04,
+      "eval_coding_entropy": 1.9919977383613587,
+      "eval_coding_loss": 2.314964532852173,
+      "eval_coding_mean_token_accuracy": 0.49944881129264834,
+      "eval_coding_num_tokens": 12725384.0,
+      "eval_coding_runtime": 80.772,
+      "eval_coding_samples_per_second": 6.19,
+      "eval_coding_steps_per_second": 3.095,
+      "step": 1300
+    },
+    {
+      "epoch": 1.04,
+      "eval_biology_entropy": 2.0739205145835875,
+      "eval_biology_loss": 2.1763668060302734,
+      "eval_biology_mean_token_accuracy": 0.5573548142910003,
+      "eval_biology_num_tokens": 12725384.0,
+      "eval_biology_runtime": 34.5814,
+      "eval_biology_samples_per_second": 14.459,
+      "eval_biology_steps_per_second": 7.229,
+      "step": 1300
+    },
+    {
+      "entropy": 2.0259235687553883,
+      "epoch": 1.048,
+      "grad_norm": 11.0625,
+      "learning_rate": 1.9313333333333334e-05,
+      "loss": 1.9459,
+      "mean_token_accuracy": 0.5764682356268167,
+      "num_tokens": 12824389.0,
+      "step": 1310
+    },
+    {
+      "entropy": 1.9896535106003284,
+      "epoch": 1.056,
+      "grad_norm": 10.9375,
+      "learning_rate": 1.9291111111111115e-05,
+      "loss": 1.9418,
+      "mean_token_accuracy": 0.5765218947082758,
+      "num_tokens": 12924592.0,
+      "step": 1320
+    },
+    {
+      "entropy": 2.024248744547367,
+      "epoch": 1.064,
+      "grad_norm": 14.0,
+      "learning_rate": 1.926888888888889e-05,
+      "loss": 1.9566,
+      "mean_token_accuracy": 0.5732226436957717,
+      "num_tokens": 13019864.0,
+      "step": 1330
+    },
+    {
+      "entropy": 2.0493085965514184,
+      "epoch": 1.072,
+      "grad_norm": 11.6875,
+      "learning_rate": 1.924666666666667e-05,
+      "loss": 1.9728,
+      "mean_token_accuracy": 0.5697184054180979,
+      "num_tokens": 13112848.0,
+      "step": 1340
+    },
+    {
+      "entropy": 2.009561914205551,
+      "epoch": 1.08,
+      "grad_norm": 12.5625,
+      "learning_rate": 1.9224444444444444e-05,
+      "loss": 1.9588,
+      "mean_token_accuracy": 0.5739864658564329,
+      "num_tokens": 13208593.0,
+      "step": 1350
+    },
+    {
+      "entropy": 2.0082264699041845,
+      "epoch": 1.088,
+      "grad_norm": 12.3125,
+      "learning_rate": 1.9202222222222225e-05,
+      "loss": 1.9356,
+      "mean_token_accuracy": 0.5787085667252541,
+      "num_tokens": 13305725.0,
+      "step": 1360
+    },
+    {
+      "entropy": 2.050854667276144,
+      "epoch": 1.096,
+      "grad_norm": 13.25,
+      "learning_rate": 1.918e-05,
+      "loss": 2.0031,
+      "mean_token_accuracy": 0.568497353605926,
+      "num_tokens": 13397615.0,
+      "step": 1370
+    },
+    {
+      "entropy": 1.9909487955272198,
+      "epoch": 1.104,
+      "grad_norm": 11.5,
+      "learning_rate": 1.915777777777778e-05,
+      "loss": 1.9246,
+      "mean_token_accuracy": 0.5775162002071739,
+      "num_tokens": 13498685.0,
+      "step": 1380
+    },
+    {
+      "entropy": 2.022812781482935,
+      "epoch": 1.112,
+      "grad_norm": 10.4375,
+      "learning_rate": 1.9135555555555555e-05,
+      "loss": 1.9621,
+      "mean_token_accuracy": 0.5735303290188313,
+      "num_tokens": 13587398.0,
+      "step": 1390
+    },
+    {
+      "entropy": 1.9650254637002944,
+      "epoch": 1.12,
+      "grad_norm": 12.1875,
+      "learning_rate": 1.9113333333333336e-05,
+      "loss": 1.8716,
+      "mean_token_accuracy": 0.5869596699252725,
+      "num_tokens": 13682146.0,
+      "step": 1400
+    },
+    {
+      "epoch": 1.12,
+      "eval_coding_entropy": 1.8874374437332153,
+      "eval_coding_loss": 2.3130838871002197,
+      "eval_coding_mean_token_accuracy": 0.4985515549182892,
+      "eval_coding_num_tokens": 13682146.0,
+      "eval_coding_runtime": 80.6107,
+      "eval_coding_samples_per_second": 6.203,
+      "eval_coding_steps_per_second": 3.101,
+      "step": 1400
+    },
+    {
+      "epoch": 1.12,
+      "eval_biology_entropy": 1.9534716877937317,
+      "eval_biology_loss": 2.117032051086426,
+      "eval_biology_mean_token_accuracy": 0.5663710038661957,
+      "eval_biology_num_tokens": 13682146.0,
+      "eval_biology_runtime": 34.5491,
+      "eval_biology_samples_per_second": 14.472,
+      "eval_biology_steps_per_second": 7.236,
+      "step": 1400
+    },
+    {
+      "entropy": 1.9430220112204553,
+      "epoch": 1.1280000000000001,
+      "grad_norm": 12.5625,
+      "learning_rate": 1.9091111111111113e-05,
+      "loss": 1.9125,
+      "mean_token_accuracy": 0.5838606007397175,
+      "num_tokens": 13781647.0,
+      "step": 1410
+    },
+    {
+      "entropy": 1.9991029907017945,
+      "epoch": 1.1360000000000001,
+      "grad_norm": 13.25,
+      "learning_rate": 1.906888888888889e-05,
+      "loss": 1.9323,
+      "mean_token_accuracy": 0.5773349234834313,
+      "num_tokens": 13876656.0,
+      "step": 1420
+    },
+    {
+      "entropy": 1.9905764922499656,
+      "epoch": 1.144,
+      "grad_norm": 11.875,
+      "learning_rate": 1.904666666666667e-05,
+      "loss": 1.9036,
+      "mean_token_accuracy": 0.5791543116793036,
+      "num_tokens": 13981024.0,
+      "step": 1430
+    },
+    {
+      "entropy": 1.938994688540697,
+      "epoch": 1.152,
+      "grad_norm": 11.4375,
+      "learning_rate": 1.9024444444444446e-05,
+      "loss": 1.8861,
+      "mean_token_accuracy": 0.5848018407821656,
+      "num_tokens": 14086104.0,
+      "step": 1440
+    },
+    {
+      "entropy": 2.0158420264720918,
+      "epoch": 1.16,
+      "grad_norm": 12.8125,
+      "learning_rate": 1.9002222222222224e-05,
+      "loss": 1.9323,
+      "mean_token_accuracy": 0.5754542848095298,
+      "num_tokens": 14179152.0,
+      "step": 1450
+    },
+    {
+      "entropy": 1.9549302212893962,
+      "epoch": 1.168,
+      "grad_norm": 11.5,
+      "learning_rate": 1.898e-05,
+      "loss": 1.9101,
+      "mean_token_accuracy": 0.582378326356411,
+      "num_tokens": 14274520.0,
+      "step": 1460
+    },
+    {
+      "entropy": 1.9402282141149043,
+      "epoch": 1.176,
+      "grad_norm": 10.875,
+      "learning_rate": 1.895777777777778e-05,
+      "loss": 1.8703,
+      "mean_token_accuracy": 0.5832702025771141,
+      "num_tokens": 14378506.0,
+      "step": 1470
+    },
+    {
+      "entropy": 2.0116791941225527,
+      "epoch": 1.184,
+      "grad_norm": 13.4375,
+      "learning_rate": 1.8935555555555556e-05,
+      "loss": 1.9544,
+      "mean_token_accuracy": 0.5769317414611578,
+      "num_tokens": 14473623.0,
+      "step": 1480
+    },
+    {
+      "entropy": 1.9328050471842289,
+      "epoch": 1.192,
+      "grad_norm": 12.4375,
+      "learning_rate": 1.8913333333333334e-05,
+      "loss": 1.8569,
+      "mean_token_accuracy": 0.5865075359120965,
+      "num_tokens": 14567216.0,
+      "step": 1490
+    },
+    {
+      "entropy": 1.984081008285284,
+      "epoch": 1.2,
+      "grad_norm": 12.125,
+      "learning_rate": 1.8891111111111115e-05,
+      "loss": 1.9149,
+      "mean_token_accuracy": 0.578367224521935,
+      "num_tokens": 14661238.0,
+      "step": 1500
+    },
+    {
+      "epoch": 1.2,
+      "eval_coding_entropy": 1.9068009638786316,
+      "eval_coding_loss": 2.345546245574951,
+      "eval_coding_mean_token_accuracy": 0.49546057629585266,
+      "eval_coding_num_tokens": 14661238.0,
+      "eval_coding_runtime": 80.6257,
+      "eval_coding_samples_per_second": 6.201,
+      "eval_coding_steps_per_second": 3.101,
+      "step": 1500
+    },
+    {
+      "epoch": 1.2,
+      "eval_biology_entropy": 1.9247408218383788,
+      "eval_biology_loss": 2.0717103481292725,
+      "eval_biology_mean_token_accuracy": 0.5721474689245224,
+      "eval_biology_num_tokens": 14661238.0,
+      "eval_biology_runtime": 34.3306,
+      "eval_biology_samples_per_second": 14.564,
+      "eval_biology_steps_per_second": 7.282,
+      "step": 1500
+    },
+    {
+      "entropy": 1.8859902337193488,
+      "epoch": 1.208,
+      "grad_norm": 9.625,
+      "learning_rate": 1.886888888888889e-05,
+      "loss": 1.8198,
+      "mean_token_accuracy": 0.5924539837986231,
+      "num_tokens": 14764963.0,
+      "step": 1510
+    },
+    {
+      "entropy": 1.9147443607449532,
+      "epoch": 1.216,
+      "grad_norm": 10.0,
+      "learning_rate": 1.884666666666667e-05,
+      "loss": 1.8428,
+      "mean_token_accuracy": 0.5875054152682424,
+      "num_tokens": 14865915.0,
+      "step": 1520
+    },
+    {
+      "entropy": 1.9541442558169364,
+      "epoch": 1.224,
+      "grad_norm": 12.6875,
+      "learning_rate": 1.8824444444444445e-05,
+      "loss": 1.8888,
+      "mean_token_accuracy": 0.5863493517041206,
+      "num_tokens": 14958678.0,
+      "step": 1530
+    },
+    {
+      "entropy": 1.8855828166007995,
+      "epoch": 1.232,
+      "grad_norm": 10.0,
+      "learning_rate": 1.8802222222222226e-05,
+      "loss": 1.8133,
+      "mean_token_accuracy": 0.5939298838376998,
+      "num_tokens": 15055682.0,
+      "step": 1540
+    },
+    {
+      "entropy": 1.9235595107078551,
+      "epoch": 1.24,
+      "grad_norm": 11.3125,
+      "learning_rate": 1.878e-05,
+      "loss": 1.8509,
+      "mean_token_accuracy": 0.5880285549908877,
+      "num_tokens": 15150956.0,
+      "step": 1550
+    },
+    {
+      "entropy": 1.9634107306599617,
+      "epoch": 1.248,
+      "grad_norm": 10.3125,
+      "learning_rate": 1.875777777777778e-05,
+      "loss": 1.8895,
+      "mean_token_accuracy": 0.5849297363311052,
+      "num_tokens": 15249576.0,
+      "step": 1560
+    },
+    {
+      "entropy": 1.9873026944696903,
+      "epoch": 1.256,
+      "grad_norm": 12.375,
+      "learning_rate": 1.873555555555556e-05,
+      "loss": 1.9352,
+      "mean_token_accuracy": 0.5761945569887758,
+      "num_tokens": 15344436.0,
+      "step": 1570
+    },
+    {
+      "entropy": 1.8877964243292809,
+      "epoch": 1.264,
+      "grad_norm": 10.0,
+      "learning_rate": 1.8713333333333336e-05,
+      "loss": 1.8079,
+      "mean_token_accuracy": 0.5952812194824219,
+      "num_tokens": 15446149.0,
+      "step": 1580
+    },
+    {
+      "entropy": 1.9913021348416806,
+      "epoch": 1.272,
+      "grad_norm": 10.375,
+      "learning_rate": 1.8691111111111114e-05,
+      "loss": 1.9326,
+      "mean_token_accuracy": 0.5766056586056948,
+      "num_tokens": 15539268.0,
+      "step": 1590
+    },
+    {
+      "entropy": 1.9224685184657573,
+      "epoch": 1.28,
+      "grad_norm": 12.3125,
+      "learning_rate": 1.866888888888889e-05,
+      "loss": 1.8535,
+      "mean_token_accuracy": 0.5897443380206824,
+      "num_tokens": 15639317.0,
+      "step": 1600
+    },
+    {
+      "epoch": 1.28,
+      "eval_coding_entropy": 1.8912849707603454,
+      "eval_coding_loss": 2.344475269317627,
+      "eval_coding_mean_token_accuracy": 0.4952514443397522,
+      "eval_coding_num_tokens": 15639317.0,
+      "eval_coding_runtime": 80.7231,
+      "eval_coding_samples_per_second": 6.194,
+      "eval_coding_steps_per_second": 3.097,
+      "step": 1600
+    },
+    {
+      "epoch": 1.28,
+      "eval_biology_entropy": 1.916876890182495,
+      "eval_biology_loss": 2.0352823734283447,
+      "eval_biology_mean_token_accuracy": 0.576624665260315,
+      "eval_biology_num_tokens": 15639317.0,
+      "eval_biology_runtime": 34.5704,
+      "eval_biology_samples_per_second": 14.463,
+      "eval_biology_steps_per_second": 7.232,
+      "step": 1600
+    },
+    {
+      "entropy": 1.9578663676977157,
+      "epoch": 1.288,
+      "grad_norm": 11.0,
+      "learning_rate": 1.864666666666667e-05,
+      "loss": 1.8945,
+      "mean_token_accuracy": 0.5816468941047788,
+      "num_tokens": 15734071.0,
+      "step": 1610
+    },
+    {
+      "entropy": 1.9439551319926978,
+      "epoch": 1.296,
+      "grad_norm": 11.0625,
+      "learning_rate": 1.8624444444444446e-05,
+      "loss": 1.8571,
+      "mean_token_accuracy": 0.585198700428009,
+      "num_tokens": 15835805.0,
+      "step": 1620
+    },
+    {
+      "entropy": 1.9310760222375394,
+      "epoch": 1.304,
+      "grad_norm": 10.5,
+      "learning_rate": 1.8602222222222224e-05,
+      "loss": 1.8625,
+      "mean_token_accuracy": 0.5881968166679143,
+      "num_tokens": 15932611.0,
+      "step": 1630
+    },
+    {
+      "entropy": 1.9141236551105976,
+      "epoch": 1.312,
+      "grad_norm": 10.8125,
+      "learning_rate": 1.858e-05,
+      "loss": 1.8295,
+      "mean_token_accuracy": 0.58838891685009,
+      "num_tokens": 16028375.0,
+      "step": 1640
+    },
+    {
+      "entropy": 1.970909407734871,
+      "epoch": 1.32,
+      "grad_norm": 11.125,
+      "learning_rate": 1.855777777777778e-05,
+      "loss": 1.8766,
+      "mean_token_accuracy": 0.5829805953428149,
+      "num_tokens": 16124235.0,
+      "step": 1650
+    },
+    {
+      "entropy": 1.9246645115315915,
+      "epoch": 1.328,
+      "grad_norm": 10.875,
+      "learning_rate": 1.8535555555555557e-05,
+      "loss": 1.8779,
+      "mean_token_accuracy": 0.5896998945623636,
+      "num_tokens": 16230555.0,
+      "step": 1660
+    },
+    {
+      "entropy": 1.8431354731321334,
+      "epoch": 1.336,
+      "grad_norm": 10.0625,
+      "learning_rate": 1.8513333333333335e-05,
+      "loss": 1.8193,
+      "mean_token_accuracy": 0.5972616789862514,
+      "num_tokens": 16331599.0,
+      "step": 1670
+    },
+    {
+      "entropy": 1.9220659539103508,
+      "epoch": 1.3439999999999999,
+      "grad_norm": 11.375,
+      "learning_rate": 1.8491111111111112e-05,
+      "loss": 1.8508,
+      "mean_token_accuracy": 0.5920726301148533,
+      "num_tokens": 16436297.0,
+      "step": 1680
+    },
+    {
+      "entropy": 1.895934323593974,
+      "epoch": 1.3519999999999999,
+      "grad_norm": 11.5625,
+      "learning_rate": 1.846888888888889e-05,
+      "loss": 1.8257,
+      "mean_token_accuracy": 0.5953480795025825,
+      "num_tokens": 16536195.0,
+      "step": 1690
+    },
+    {
+      "entropy": 1.8695914663374424,
+      "epoch": 1.3599999999999999,
+      "grad_norm": 12.125,
+      "learning_rate": 1.8446666666666667e-05,
+      "loss": 1.79,
+      "mean_token_accuracy": 0.5937470164149999,
+      "num_tokens": 16630563.0,
+      "step": 1700
+    },
+    {
+      "epoch": 1.3599999999999999,
+      "eval_coding_entropy": 1.9096598467826844,
+      "eval_coding_loss": 2.3258213996887207,
+      "eval_coding_mean_token_accuracy": 0.49758183360099795,
+      "eval_coding_num_tokens": 16630563.0,
+      "eval_coding_runtime": 80.6822,
+      "eval_coding_samples_per_second": 6.197,
+      "eval_coding_steps_per_second": 3.099,
+      "step": 1700
+    },
+    {
+      "epoch": 1.3599999999999999,
+      "eval_biology_entropy": 1.9426390035152434,
+      "eval_biology_loss": 1.9884332418441772,
+      "eval_biology_mean_token_accuracy": 0.5829362199306488,
+      "eval_biology_num_tokens": 16630563.0,
+      "eval_biology_runtime": 34.538,
+      "eval_biology_samples_per_second": 14.477,
+      "eval_biology_steps_per_second": 7.238,
+      "step": 1700
+    },
+    {
+      "entropy": 1.8714322451502086,
+      "epoch": 1.3679999999999999,
+      "grad_norm": 11.4375,
+      "learning_rate": 1.842444444444445e-05,
+      "loss": 1.8211,
+      "mean_token_accuracy": 0.5943556617945432,
+      "num_tokens": 16727975.0,
+      "step": 1710
+    },
+    {
+      "entropy": 1.9365650050342083,
+      "epoch": 1.376,
+      "grad_norm": 9.4375,
+      "learning_rate": 1.8402222222222223e-05,
+      "loss": 1.8462,
+      "mean_token_accuracy": 0.5863204646855593,
+      "num_tokens": 16822602.0,
+      "step": 1720
+    },
+    {
+      "entropy": 1.8875491715967656,
+      "epoch": 1.384,
+      "grad_norm": 11.3125,
+      "learning_rate": 1.8380000000000004e-05,
+      "loss": 1.8205,
+      "mean_token_accuracy": 0.5902999579906464,
+      "num_tokens": 16917777.0,
+      "step": 1730
+    },
+    {
+      "entropy": 1.9061699729412793,
+      "epoch": 1.392,
+      "grad_norm": 10.5625,
+      "learning_rate": 1.8357777777777778e-05,
+      "loss": 1.8472,
+      "mean_token_accuracy": 0.5923036687076092,
+      "num_tokens": 17015750.0,
+      "step": 1740
+    },
+    {
+      "entropy": 1.8518050290644168,
+      "epoch": 1.4,
+      "grad_norm": 11.0625,
+      "learning_rate": 1.833555555555556e-05,
+      "loss": 1.7891,
+      "mean_token_accuracy": 0.5978840438649058,
+      "num_tokens": 17107360.0,
+      "step": 1750
+    },
+    {
+      "entropy": 1.8917289264500141,
+      "epoch": 1.408,
+      "grad_norm": 12.375,
+      "learning_rate": 1.8313333333333333e-05,
+      "loss": 1.8269,
+      "mean_token_accuracy": 0.5924701662734151,
+      "num_tokens": 17207248.0,
+      "step": 1760
+    },
+    {
+      "entropy": 1.9043795056641102,
+      "epoch": 1.416,
+      "grad_norm": 10.1875,
+      "learning_rate": 1.8291111111111114e-05,
+      "loss": 1.8403,
+      "mean_token_accuracy": 0.5912218499928713,
+      "num_tokens": 17306295.0,
+      "step": 1770
+    },
+    {
+      "entropy": 1.879479529708624,
+      "epoch": 1.424,
+      "grad_norm": 10.4375,
+      "learning_rate": 1.8268888888888888e-05,
+      "loss": 1.8073,
+      "mean_token_accuracy": 0.5946688577532768,
+      "num_tokens": 17403869.0,
+      "step": 1780
+    },
+    {
+      "entropy": 1.8418946724385024,
+      "epoch": 1.432,
+      "grad_norm": 8.625,
+      "learning_rate": 1.824666666666667e-05,
+      "loss": 1.7774,
+      "mean_token_accuracy": 0.5993103161454201,
+      "num_tokens": 17507714.0,
+      "step": 1790
+    },
+    {
+      "entropy": 1.8919135071337223,
+      "epoch": 1.44,
+      "grad_norm": 11.625,
+      "learning_rate": 1.8224444444444447e-05,
+      "loss": 1.8237,
+      "mean_token_accuracy": 0.5932244326919317,
+      "num_tokens": 17597428.0,
+      "step": 1800
+    },
+    {
+      "epoch": 1.44,
+      "eval_coding_entropy": 1.8839851083755492,
+      "eval_coding_loss": 2.3147287368774414,
+      "eval_coding_mean_token_accuracy": 0.5010540502071381,
+      "eval_coding_num_tokens": 17597428.0,
+      "eval_coding_runtime": 80.7535,
+      "eval_coding_samples_per_second": 6.192,
+      "eval_coding_steps_per_second": 3.096,
+      "step": 1800
+    },
+    {
+      "epoch": 1.44,
+      "eval_biology_entropy": 1.87712366604805,
+      "eval_biology_loss": 1.9589980840682983,
+      "eval_biology_mean_token_accuracy": 0.5894684064388275,
+      "eval_biology_num_tokens": 17597428.0,
+      "eval_biology_runtime": 34.6184,
+      "eval_biology_samples_per_second": 14.443,
+      "eval_biology_steps_per_second": 7.222,
+      "step": 1800
+    },
+    {
+      "entropy": 1.8756542146205901,
+      "epoch": 1.448,
+      "grad_norm": 8.875,
+      "learning_rate": 1.8202222222222225e-05,
+      "loss": 1.8173,
+      "mean_token_accuracy": 0.5969021037220955,
+      "num_tokens": 17698329.0,
+      "step": 1810
+    },
+    {
+      "entropy": 1.8537494942545891,
+      "epoch": 1.456,
+      "grad_norm": 8.5,
+      "learning_rate": 1.8180000000000002e-05,
+      "loss": 1.7835,
+      "mean_token_accuracy": 0.5985489264130592,
+      "num_tokens": 17798083.0,
+      "step": 1820
+    },
+    {
+      "entropy": 1.846832074970007,
+      "epoch": 1.464,
+      "grad_norm": 9.9375,
+      "learning_rate": 1.815777777777778e-05,
+      "loss": 1.7661,
+      "mean_token_accuracy": 0.5986342612653971,
+      "num_tokens": 17892361.0,
+      "step": 1830
+    },
+    {
+      "entropy": 1.8376515425741673,
+      "epoch": 1.472,
+      "grad_norm": 10.0625,
+      "learning_rate": 1.8135555555555557e-05,
+      "loss": 1.7883,
+      "mean_token_accuracy": 0.6009603329002857,
+      "num_tokens": 17993802.0,
+      "step": 1840
+    },
+    {
+      "entropy": 1.8447901919484138,
+      "epoch": 1.48,
+      "grad_norm": 11.5625,
+      "learning_rate": 1.8113333333333335e-05,
+      "loss": 1.797,
+      "mean_token_accuracy": 0.5993185771629215,
+      "num_tokens": 18090617.0,
+      "step": 1850
+    },
+    {
+      "entropy": 1.8464039027690888,
+      "epoch": 1.488,
+      "grad_norm": 10.0625,
+      "learning_rate": 1.8091111111111113e-05,
+      "loss": 1.7906,
+      "mean_token_accuracy": 0.5992585981264711,
+      "num_tokens": 18185570.0,
+      "step": 1860
+    },
+    {
+      "entropy": 1.8622294574975968,
+      "epoch": 1.496,
+      "grad_norm": 11.375,
+      "learning_rate": 1.806888888888889e-05,
+      "loss": 1.799,
+      "mean_token_accuracy": 0.597082006931305,
+      "num_tokens": 18278908.0,
+      "step": 1870
+    },
+    {
+      "entropy": 1.8341825246810912,
+      "epoch": 1.504,
+      "grad_norm": 9.5625,
+      "learning_rate": 1.8046666666666668e-05,
+      "loss": 1.765,
+      "mean_token_accuracy": 0.6009232444688678,
+      "num_tokens": 18377903.0,
+      "step": 1880
+    },
+    {
+      "entropy": 1.8509348623454571,
+      "epoch": 1.512,
+      "grad_norm": 9.375,
+      "learning_rate": 1.8024444444444445e-05,
+      "loss": 1.7705,
+      "mean_token_accuracy": 0.5985264440998435,
+      "num_tokens": 18479107.0,
+      "step": 1890
+    },
+    {
+      "entropy": 1.8879661358892919,
+      "epoch": 1.52,
+      "grad_norm": 9.625,
+      "learning_rate": 1.8002222222222223e-05,
+      "loss": 1.8164,
+      "mean_token_accuracy": 0.5940329255536199,
+      "num_tokens": 18579284.0,
+      "step": 1900
+    },
+    {
+      "epoch": 1.52,
+      "eval_coding_entropy": 1.8767889976501464,
+      "eval_coding_loss": 2.3431949615478516,
+      "eval_coding_mean_token_accuracy": 0.4959298541545868,
+      "eval_coding_num_tokens": 18579284.0,
+      "eval_coding_runtime": 80.7542,
+      "eval_coding_samples_per_second": 6.192,
+      "eval_coding_steps_per_second": 3.096,
+      "step": 1900
+    },
+    {
+      "epoch": 1.52,
+      "eval_biology_entropy": 1.8529948596954346,
+      "eval_biology_loss": 1.913954734802246,
+      "eval_biology_mean_token_accuracy": 0.5940197825431823,
+      "eval_biology_num_tokens": 18579284.0,
+      "eval_biology_runtime": 34.5204,
+      "eval_biology_samples_per_second": 14.484,
+      "eval_biology_steps_per_second": 7.242,
+      "step": 1900
+    },
+    {
+      "entropy": 1.8475319847464562,
+      "epoch": 1.528,
+      "grad_norm": 10.125,
+      "learning_rate": 1.798e-05,
+      "loss": 1.7722,
+      "mean_token_accuracy": 0.5960186466574668,
+      "num_tokens": 18671484.0,
+      "step": 1910
+    },
+    {
+      "entropy": 1.846816373616457,
+      "epoch": 1.536,
+      "grad_norm": 12.0,
+      "learning_rate": 1.7957777777777778e-05,
+      "loss": 1.7605,
+      "mean_token_accuracy": 0.5992849035188555,
+      "num_tokens": 18765324.0,
+      "step": 1920
+    },
+    {
+      "entropy": 1.8267472043633461,
+      "epoch": 1.544,
+      "grad_norm": 9.625,
+      "learning_rate": 1.7935555555555556e-05,
+      "loss": 1.7549,
+      "mean_token_accuracy": 0.6032338388264179,
+      "num_tokens": 18865316.0,
+      "step": 1930
+    },
+    {
+      "entropy": 1.8062134094536304,
+      "epoch": 1.552,
+      "grad_norm": 9.0625,
+      "learning_rate": 1.7913333333333337e-05,
+      "loss": 1.7421,
+      "mean_token_accuracy": 0.604615181311965,
+      "num_tokens": 18966382.0,
+      "step": 1940
+    },
+    {
+      "entropy": 1.8020479552447797,
+      "epoch": 1.56,
+      "grad_norm": 9.0,
+      "learning_rate": 1.789111111111111e-05,
+      "loss": 1.7317,
+      "mean_token_accuracy": 0.608651296235621,
+      "num_tokens": 19062128.0,
+      "step": 1950
+    },
+    {
+      "entropy": 1.8492420889437198,
+      "epoch": 1.568,
+      "grad_norm": 10.1875,
+      "learning_rate": 1.7868888888888892e-05,
+      "loss": 1.7824,
+      "mean_token_accuracy": 0.6007449336349964,
+      "num_tokens": 19160509.0,
+      "step": 1960
+    },
+    {
+      "entropy": 1.8319963850080967,
+      "epoch": 1.576,
+      "grad_norm": 10.6875,
+      "learning_rate": 1.7846666666666666e-05,
+      "loss": 1.7687,
+      "mean_token_accuracy": 0.6003297364339233,
+      "num_tokens": 19258305.0,
+      "step": 1970
+    },
+    {
+      "entropy": 1.7432583637535572,
+      "epoch": 1.584,
+      "grad_norm": 9.125,
+      "learning_rate": 1.7824444444444447e-05,
+      "loss": 1.6755,
+      "mean_token_accuracy": 0.613487584143877,
+      "num_tokens": 19354662.0,
+      "step": 1980
+    },
+    {
+      "entropy": 1.849979268014431,
+      "epoch": 1.592,
+      "grad_norm": 9.5625,
+      "learning_rate": 1.780222222222222e-05,
+      "loss": 1.7805,
+      "mean_token_accuracy": 0.6017373757436871,
+      "num_tokens": 19446577.0,
+      "step": 1990
+    },
+    {
+      "entropy": 1.848842553049326,
+      "epoch": 1.6,
+      "grad_norm": 9.1875,
+      "learning_rate": 1.7780000000000003e-05,
+      "loss": 1.7833,
+      "mean_token_accuracy": 0.5976102851331234,
+      "num_tokens": 19543088.0,
+      "step": 2000
+    },
+    {
+      "epoch": 1.6,
+      "eval_coding_entropy": 1.8642783060073853,
+      "eval_coding_loss": 2.3345677852630615,
+      "eval_coding_mean_token_accuracy": 0.49641851997375486,
+      "eval_coding_num_tokens": 19543088.0,
+      "eval_coding_runtime": 80.931,
+      "eval_coding_samples_per_second": 6.178,
+      "eval_coding_steps_per_second": 3.089,
+      "step": 2000
+    },
+    {
+      "epoch": 1.6,
+      "eval_biology_entropy": 1.8082833728790284,
+      "eval_biology_loss": 1.8987895250320435,
+      "eval_biology_mean_token_accuracy": 0.5959605065584183,
+      "eval_biology_num_tokens": 19543088.0,
+      "eval_biology_runtime": 34.2382,
+      "eval_biology_samples_per_second": 14.604,
+      "eval_biology_steps_per_second": 7.302,
+      "step": 2000
+    },
+    {
+      "entropy": 1.8170477211475373,
+      "epoch": 1.608,
+      "grad_norm": 9.0625,
+      "learning_rate": 1.7757777777777777e-05,
+      "loss": 1.7425,
+      "mean_token_accuracy": 0.6050298389047384,
+      "num_tokens": 19642298.0,
+      "step": 2010
+    },
+    {
+      "entropy": 1.7922820329666138,
+      "epoch": 1.616,
+      "grad_norm": 9.125,
+      "learning_rate": 1.7735555555555558e-05,
+      "loss": 1.718,
+      "mean_token_accuracy": 0.6069137429818511,
+      "num_tokens": 19743506.0,
+      "step": 2020
+    },
+    {
+      "entropy": 1.834612490236759,
+      "epoch": 1.624,
+      "grad_norm": 13.1875,
+      "learning_rate": 1.7713333333333335e-05,
+      "loss": 1.7703,
+      "mean_token_accuracy": 0.6007826989516616,
+      "num_tokens": 19845228.0,
+      "step": 2030
+    },
+    {
+      "entropy": 1.8345312468707562,
+      "epoch": 1.6320000000000001,
+      "grad_norm": 11.8125,
+      "learning_rate": 1.7691111111111113e-05,
+      "loss": 1.7904,
+      "mean_token_accuracy": 0.6026085142046214,
+      "num_tokens": 19937567.0,
+      "step": 2040
+    },
+    {
+      "entropy": 1.8599664010107517,
+      "epoch": 1.6400000000000001,
+      "grad_norm": 10.375,
+      "learning_rate": 1.766888888888889e-05,
+      "loss": 1.7666,
+      "mean_token_accuracy": 0.596138310432434,
+      "num_tokens": 20028494.0,
+      "step": 2050
+    },
+    {
+      "entropy": 1.7672480165958404,
+      "epoch": 1.6480000000000001,
+      "grad_norm": 8.5,
+      "learning_rate": 1.7646666666666668e-05,
+      "loss": 1.7025,
+      "mean_token_accuracy": 0.6117627980187536,
+      "num_tokens": 20129743.0,
+      "step": 2060
+    },
+    {
+      "entropy": 1.7967971567064525,
+      "epoch": 1.6560000000000001,
+      "grad_norm": 9.875,
+      "learning_rate": 1.7624444444444446e-05,
+      "loss": 1.7408,
+      "mean_token_accuracy": 0.607316411100328,
+      "num_tokens": 20230156.0,
+      "step": 2070
+    },
+    {
+      "entropy": 1.7646208584308625,
+      "epoch": 1.6640000000000001,
+      "grad_norm": 9.375,
+      "learning_rate": 1.7602222222222223e-05,
+      "loss": 1.6996,
+      "mean_token_accuracy": 0.6130116602405906,
+      "num_tokens": 20331548.0,
+      "step": 2080
+    },
+    {
+      "entropy": 1.8086679019033909,
+      "epoch": 1.6720000000000002,
+      "grad_norm": 9.75,
+      "learning_rate": 1.758e-05,
+      "loss": 1.7253,
+      "mean_token_accuracy": 0.6075613485649228,
+      "num_tokens": 20425782.0,
+      "step": 2090
+    },
+    {
+      "entropy": 1.8103238429874182,
+      "epoch": 1.6800000000000002,
+      "grad_norm": 10.8125,
+      "learning_rate": 1.755777777777778e-05,
+      "loss": 1.7715,
+      "mean_token_accuracy": 0.6018931455910206,
+      "num_tokens": 20524247.0,
+      "step": 2100
+    },
+    {
+      "epoch": 1.6800000000000002,
+      "eval_coding_entropy": 1.9248092899322509,
+      "eval_coding_loss": 2.3239312171936035,
+      "eval_coding_mean_token_accuracy": 0.4967528132200241,
+      "eval_coding_num_tokens": 20524247.0,
+      "eval_coding_runtime": 80.5845,
+      "eval_coding_samples_per_second": 6.205,
+      "eval_coding_steps_per_second": 3.102,
+      "step": 2100
+    },
+    {
+      "epoch": 1.6800000000000002,
+      "eval_biology_entropy": 1.8323220880031585,
+      "eval_biology_loss": 1.8671725988388062,
+      "eval_biology_mean_token_accuracy": 0.5994496657848358,
+      "eval_biology_num_tokens": 20524247.0,
+      "eval_biology_runtime": 34.4548,
+      "eval_biology_samples_per_second": 14.512,
+      "eval_biology_steps_per_second": 7.256,
+      "step": 2100
+    },
+    {
+      "entropy": 1.8471850778907537,
+      "epoch": 1.688,
+      "grad_norm": 12.6875,
+      "learning_rate": 1.7535555555555556e-05,
+      "loss": 1.7729,
+      "mean_token_accuracy": 0.6012401515617967,
+      "num_tokens": 20615831.0,
+      "step": 2110
+    },
+    {
+      "entropy": 1.743248750269413,
+      "epoch": 1.696,
+      "grad_norm": 10.375,
+      "learning_rate": 1.7513333333333334e-05,
+      "loss": 1.6782,
+      "mean_token_accuracy": 0.6167684741318226,
+      "num_tokens": 20718121.0,
+      "step": 2120
+    },
+    {
+      "entropy": 1.7873006261885167,
+      "epoch": 1.704,
+      "grad_norm": 9.6875,
+      "learning_rate": 1.749111111111111e-05,
+      "loss": 1.7311,
+      "mean_token_accuracy": 0.6056294169276952,
+      "num_tokens": 20823099.0,
+      "step": 2130
+    },
+    {
+      "entropy": 1.81577892228961,
+      "epoch": 1.712,
+      "grad_norm": 9.5,
+      "learning_rate": 1.746888888888889e-05,
+      "loss": 1.7519,
+      "mean_token_accuracy": 0.6086800323799253,
+      "num_tokens": 20921419.0,
+      "step": 2140
+    },
+    {
+      "entropy": 1.7905849933624267,
+      "epoch": 1.72,
+      "grad_norm": 9.0,
+      "learning_rate": 1.7446666666666667e-05,
+      "loss": 1.708,
+      "mean_token_accuracy": 0.6048404024913907,
+      "num_tokens": 21020164.0,
+      "step": 2150
+    },
+    {
+      "entropy": 1.7664998650550843,
+      "epoch": 1.728,
+      "grad_norm": 8.75,
+      "learning_rate": 1.7424444444444444e-05,
+      "loss": 1.6999,
+      "mean_token_accuracy": 0.6104172091931105,
+      "num_tokens": 21118581.0,
+      "step": 2160
+    },
+    {
+      "entropy": 1.7681288361549377,
+      "epoch": 1.736,
+      "grad_norm": 9.5,
+      "learning_rate": 1.7402222222222222e-05,
+      "loss": 1.6913,
+      "mean_token_accuracy": 0.6129325466230512,
+      "num_tokens": 21216158.0,
+      "step": 2170
+    },
+    {
+      "entropy": 1.7116123758256436,
+      "epoch": 1.744,
+      "grad_norm": 10.5,
+      "learning_rate": 1.7380000000000003e-05,
+      "loss": 1.638,
+      "mean_token_accuracy": 0.6149178206920624,
+      "num_tokens": 21315128.0,
+      "step": 2180
+    },
+    {
+      "entropy": 1.767675194516778,
+      "epoch": 1.752,
+      "grad_norm": 9.1875,
+      "learning_rate": 1.735777777777778e-05,
+      "loss": 1.7082,
+      "mean_token_accuracy": 0.610173468105495,
+      "num_tokens": 21420182.0,
+      "step": 2190
+    },
+    {
+      "entropy": 1.7936776638031007,
+      "epoch": 1.76,
+      "grad_norm": 10.5625,
+      "learning_rate": 1.7335555555555558e-05,
+      "loss": 1.7213,
+      "mean_token_accuracy": 0.6062763595953584,
+      "num_tokens": 21511066.0,
+      "step": 2200
+    },
+    {
+      "epoch": 1.76,
+      "eval_coding_entropy": 1.9285395607948304,
+      "eval_coding_loss": 2.3115410804748535,
+      "eval_coding_mean_token_accuracy": 0.49883099663257596,
+      "eval_coding_num_tokens": 21511066.0,
+      "eval_coding_runtime": 80.7188,
+      "eval_coding_samples_per_second": 6.194,
+      "eval_coding_steps_per_second": 3.097,
+      "step": 2200
+    },
+    {
+      "epoch": 1.76,
+      "eval_biology_entropy": 1.815982218503952,
+      "eval_biology_loss": 1.84175443649292,
+      "eval_biology_mean_token_accuracy": 0.6039358637332917,
+      "eval_biology_num_tokens": 21511066.0,
+      "eval_biology_runtime": 34.5383,
+      "eval_biology_samples_per_second": 14.477,
+      "eval_biology_steps_per_second": 7.238,
+      "step": 2200
+    },
+    {
+      "entropy": 1.7788202807307243,
+      "epoch": 1.768,
+      "grad_norm": 8.0625,
+      "learning_rate": 1.7313333333333336e-05,
+      "loss": 1.6969,
+      "mean_token_accuracy": 0.6084614660590887,
+      "num_tokens": 21607794.0,
+      "step": 2210
+    },
+    {
+      "entropy": 1.7595535799860955,
+      "epoch": 1.776,
+      "grad_norm": 9.125,
+      "learning_rate": 1.7291111111111113e-05,
+      "loss": 1.7076,
+      "mean_token_accuracy": 0.6141810776665807,
+      "num_tokens": 21704368.0,
+      "step": 2220
+    },
+    {
+      "entropy": 1.7047814682126046,
+      "epoch": 1.784,
+      "grad_norm": 8.5625,
+      "learning_rate": 1.726888888888889e-05,
+      "loss": 1.6273,
+      "mean_token_accuracy": 0.6201400419697165,
+      "num_tokens": 21807887.0,
+      "step": 2230
+    },
+    {
+      "entropy": 1.7654159277677537,
+      "epoch": 1.792,
+      "grad_norm": 8.75,
+      "learning_rate": 1.724666666666667e-05,
+      "loss": 1.6976,
+      "mean_token_accuracy": 0.6115840144455433,
+      "num_tokens": 21905784.0,
+      "step": 2240
+    },
+    {
+      "entropy": 1.7922681115567685,
+      "epoch": 1.8,
+      "grad_norm": 10.1875,
+      "learning_rate": 1.7224444444444446e-05,
+      "loss": 1.7383,
+      "mean_token_accuracy": 0.6069337302818895,
+      "num_tokens": 21997410.0,
+      "step": 2250
+    },
+    {
+      "entropy": 1.741399647668004,
+      "epoch": 1.808,
+      "grad_norm": 9.8125,
+      "learning_rate": 1.7202222222222224e-05,
+      "loss": 1.6652,
+      "mean_token_accuracy": 0.6142366178333759,
+      "num_tokens": 22094159.0,
+      "step": 2260
+    },
+    {
+      "entropy": 1.7799841571599244,
+      "epoch": 1.8159999999999998,
+      "grad_norm": 8.25,
+      "learning_rate": 1.718e-05,
+      "loss": 1.7144,
+      "mean_token_accuracy": 0.6076404828578233,
+      "num_tokens": 22192305.0,
+      "step": 2270
+    },
+    {
+      "entropy": 1.7761202543973922,
+      "epoch": 1.8239999999999998,
+      "grad_norm": 9.75,
+      "learning_rate": 1.715777777777778e-05,
+      "loss": 1.7016,
+      "mean_token_accuracy": 0.6089693499729038,
+      "num_tokens": 22290654.0,
+      "step": 2280
+    },
+    {
+      "entropy": 1.775102337449789,
+      "epoch": 1.8319999999999999,
+      "grad_norm": 7.90625,
+      "learning_rate": 1.7135555555555557e-05,
+      "loss": 1.7129,
+      "mean_token_accuracy": 0.6085843306034804,
+      "num_tokens": 22389707.0,
+      "step": 2290
+    },
+    {
+      "entropy": 1.8077950671315193,
+      "epoch": 1.8399999999999999,
+      "grad_norm": 10.5,
+      "learning_rate": 1.7113333333333334e-05,
+      "loss": 1.7412,
+      "mean_token_accuracy": 0.6040071191266179,
+      "num_tokens": 22485675.0,
+      "step": 2300
+    },
+    {
+      "epoch": 1.8399999999999999,
+      "eval_coding_entropy": 1.8626320977211,
+      "eval_coding_loss": 2.318260669708252,
+      "eval_coding_mean_token_accuracy": 0.4991067146062851,
+      "eval_coding_num_tokens": 22485675.0,
+      "eval_coding_runtime": 80.6837,
+      "eval_coding_samples_per_second": 6.197,
+      "eval_coding_steps_per_second": 3.099,
+      "step": 2300
+    },
+    {
+      "epoch": 1.8399999999999999,
+      "eval_biology_entropy": 1.7646894090175629,
+      "eval_biology_loss": 1.8258318901062012,
+      "eval_biology_mean_token_accuracy": 0.6067340680360794,
+      "eval_biology_num_tokens": 22485675.0,
+      "eval_biology_runtime": 34.5007,
+      "eval_biology_samples_per_second": 14.492,
+      "eval_biology_steps_per_second": 7.246,
+      "step": 2300
+    },
+    {
+      "entropy": 1.7107921369373797,
+      "epoch": 1.8479999999999999,
+      "grad_norm": 8.75,
+      "learning_rate": 1.7091111111111112e-05,
+      "loss": 1.6406,
+      "mean_token_accuracy": 0.6187074469402433,
+      "num_tokens": 22594136.0,
+      "step": 2310
+    },
+    {
+      "entropy": 1.7815124906599522,
+      "epoch": 1.8559999999999999,
+      "grad_norm": 9.6875,
+      "learning_rate": 1.706888888888889e-05,
+      "loss": 1.7306,
+      "mean_token_accuracy": 0.6066237287595868,
+      "num_tokens": 22689900.0,
+      "step": 2320
+    },
+    {
+      "entropy": 1.7289459429681302,
+      "epoch": 1.8639999999999999,
+      "grad_norm": 9.3125,
+      "learning_rate": 1.704666666666667e-05,
+      "loss": 1.6782,
+      "mean_token_accuracy": 0.6185855753719807,
+      "num_tokens": 22787887.0,
+      "step": 2330
+    },
+    {
+      "entropy": 1.6866063613444566,
+      "epoch": 1.8719999999999999,
+      "grad_norm": 9.25,
+      "learning_rate": 1.7024444444444445e-05,
+      "loss": 1.6238,
+      "mean_token_accuracy": 0.6253292145207524,
+      "num_tokens": 22887275.0,
+      "step": 2340
+    },
+    {
+      "entropy": 1.774533098563552,
+      "epoch": 1.88,
+      "grad_norm": 11.5,
+      "learning_rate": 1.7002222222222226e-05,
+      "loss": 1.7142,
+      "mean_token_accuracy": 0.6074549812823534,
+      "num_tokens": 22984975.0,
+      "step": 2350
+    },
+    {
+      "entropy": 1.7709153965115547,
+      "epoch": 1.888,
+      "grad_norm": 10.8125,
+      "learning_rate": 1.698e-05,
+      "loss": 1.7091,
+      "mean_token_accuracy": 0.6101190822198987,
+      "num_tokens": 23081241.0,
+      "step": 2360
+    },
+    {
+      "entropy": 1.7443819627165795,
+      "epoch": 1.896,
+      "grad_norm": 9.5625,
+      "learning_rate": 1.695777777777778e-05,
+      "loss": 1.6749,
+      "mean_token_accuracy": 0.6157219119369983,
+      "num_tokens": 23181355.0,
+      "step": 2370
+    },
+    {
+      "entropy": 1.8135411240160466,
+      "epoch": 1.904,
+      "grad_norm": 10.1875,
+      "learning_rate": 1.6935555555555555e-05,
+      "loss": 1.7442,
+      "mean_token_accuracy": 0.6022866329178214,
+      "num_tokens": 23276417.0,
+      "step": 2380
+    },
+    {
+      "entropy": 1.7852804265916347,
+      "epoch": 1.912,
+      "grad_norm": 8.125,
+      "learning_rate": 1.6913333333333336e-05,
+      "loss": 1.7054,
+      "mean_token_accuracy": 0.6094424337148666,
+      "num_tokens": 23374460.0,
+      "step": 2390
+    },
+    {
+      "entropy": 1.7223038360476495,
+      "epoch": 1.92,
+      "grad_norm": 9.0625,
+      "learning_rate": 1.689111111111111e-05,
+      "loss": 1.6464,
+      "mean_token_accuracy": 0.6168824147433043,
+      "num_tokens": 23476340.0,
+      "step": 2400
+    },
+    {
+      "epoch": 1.92,
+      "eval_coding_entropy": 1.9000126180648804,
+      "eval_coding_loss": 2.3312573432922363,
+      "eval_coding_mean_token_accuracy": 0.49816246128082275,
+      "eval_coding_num_tokens": 23476340.0,
+      "eval_coding_runtime": 80.8263,
+      "eval_coding_samples_per_second": 6.186,
+      "eval_coding_steps_per_second": 3.093,
+      "step": 2400
+    },
+    {
+      "epoch": 1.92,
+      "eval_biology_entropy": 1.74286589884758,
+      "eval_biology_loss": 1.8049228191375732,
+      "eval_biology_mean_token_accuracy": 0.608675300359726,
+      "eval_biology_num_tokens": 23476340.0,
+      "eval_biology_runtime": 34.581,
+      "eval_biology_samples_per_second": 14.459,
+      "eval_biology_steps_per_second": 7.229,
+      "step": 2400
+    },
+    {
+      "entropy": 1.7121219135820867,
+      "epoch": 1.928,
+      "grad_norm": 7.625,
+      "learning_rate": 1.686888888888889e-05,
+      "loss": 1.6408,
+      "mean_token_accuracy": 0.621227914094925,
+      "num_tokens": 23570764.0,
+      "step": 2410
+    },
+    {
+      "entropy": 1.700171983987093,
+      "epoch": 1.936,
+      "grad_norm": 8.5625,
+      "learning_rate": 1.684666666666667e-05,
+      "loss": 1.6153,
+      "mean_token_accuracy": 0.6197861280292273,
+      "num_tokens": 23673208.0,
+      "step": 2420
+    },
+    {
+      "entropy": 1.7056194875389337,
+      "epoch": 1.944,
+      "grad_norm": 9.0,
+      "learning_rate": 1.6824444444444447e-05,
+      "loss": 1.6393,
+      "mean_token_accuracy": 0.6218485841527581,
+      "num_tokens": 23777687.0,
+      "step": 2430
+    },
+    {
+      "entropy": 1.7795246351510285,
+      "epoch": 1.952,
+      "grad_norm": 7.53125,
+      "learning_rate": 1.6802222222222224e-05,
+      "loss": 1.7116,
+      "mean_token_accuracy": 0.6097978493198752,
+      "num_tokens": 23870993.0,
+      "step": 2440
+    },
+    {
+      "entropy": 1.7756722070276738,
+      "epoch": 1.96,
+      "grad_norm": 8.875,
+      "learning_rate": 1.6780000000000002e-05,
+      "loss": 1.6959,
+      "mean_token_accuracy": 0.6106678569689393,
+      "num_tokens": 23970557.0,
+      "step": 2450
+    },
+    {
+      "entropy": 1.7476738385856152,
+      "epoch": 1.968,
+      "grad_norm": 8.375,
+      "learning_rate": 1.675777777777778e-05,
+      "loss": 1.6952,
+      "mean_token_accuracy": 0.6131310626864434,
+      "num_tokens": 24066234.0,
+      "step": 2460
+    },
+    {
+      "entropy": 1.7443486232310534,
+      "epoch": 1.976,
+      "grad_norm": 8.6875,
+      "learning_rate": 1.6735555555555557e-05,
+      "loss": 1.6747,
+      "mean_token_accuracy": 0.6122729379683733,
+      "num_tokens": 24164350.0,
+      "step": 2470
+    },
+    {
+      "entropy": 1.743413707241416,
+      "epoch": 1.984,
+      "grad_norm": 8.375,
+      "learning_rate": 1.6713333333333335e-05,
+      "loss": 1.6735,
+      "mean_token_accuracy": 0.616204534098506,
+      "num_tokens": 24262389.0,
+      "step": 2480
+    },
+    {
+      "entropy": 1.745465162023902,
+      "epoch": 1.992,
+      "grad_norm": 9.6875,
+      "learning_rate": 1.6691111111111112e-05,
+      "loss": 1.6884,
+      "mean_token_accuracy": 0.6156047407537699,
+      "num_tokens": 24359792.0,
+      "step": 2490
+    },
+    {
+      "entropy": 1.7558744188398123,
+      "epoch": 2.0,
+      "grad_norm": 10.375,
+      "learning_rate": 1.666888888888889e-05,
+      "loss": 1.6838,
+      "mean_token_accuracy": 0.6169366030022502,
+      "num_tokens": 24455544.0,
+      "step": 2500
+    },
+    {
+      "epoch": 2.0,
+      "eval_coding_entropy": 1.809644660949707,
+      "eval_coding_loss": 2.3500161170959473,
+      "eval_coding_mean_token_accuracy": 0.49760334622859953,
+      "eval_coding_num_tokens": 24455544.0,
+      "eval_coding_runtime": 81.1577,
+      "eval_coding_samples_per_second": 6.161,
+      "eval_coding_steps_per_second": 3.08,
+      "step": 2500
+    },
+    {
+      "epoch": 2.0,
+      "eval_biology_entropy": 1.665675255537033,
+      "eval_biology_loss": 1.7844414710998535,
+      "eval_biology_mean_token_accuracy": 0.6123956483602524,
+      "eval_biology_num_tokens": 24455544.0,
+      "eval_biology_runtime": 34.3075,
+      "eval_biology_samples_per_second": 14.574,
+      "eval_biology_steps_per_second": 7.287,
+      "step": 2500
+    },
+    {
+      "entropy": 1.3941724233329296,
+      "epoch": 2.008,
+      "grad_norm": 10.625,
+      "learning_rate": 1.6646666666666668e-05,
+      "loss": 1.3194,
+      "mean_token_accuracy": 0.669821274280548,
+      "num_tokens": 24549381.0,
+      "step": 2510
+    },
+    {
+      "entropy": 1.3493203684687614,
+      "epoch": 2.016,
+      "grad_norm": 9.4375,
+      "learning_rate": 1.6624444444444445e-05,
+      "loss": 1.2516,
+      "mean_token_accuracy": 0.6811134446412325,
+      "num_tokens": 24645425.0,
+      "step": 2520
+    },
+    {
+      "entropy": 1.321524141356349,
+      "epoch": 2.024,
+      "grad_norm": 9.3125,
+      "learning_rate": 1.6602222222222223e-05,
+      "loss": 1.2473,
+      "mean_token_accuracy": 0.6803081015124917,
+      "num_tokens": 24741400.0,
+      "step": 2530
+    },
+    {
+      "entropy": 1.3757263593375684,
+      "epoch": 2.032,
+      "grad_norm": 9.875,
+      "learning_rate": 1.658e-05,
+      "loss": 1.3085,
+      "mean_token_accuracy": 0.6724627614021301,
+      "num_tokens": 24835956.0,
+      "step": 2540
+    },
+    {
+      "entropy": 1.3186254695057869,
+      "epoch": 2.04,
+      "grad_norm": 8.8125,
+      "learning_rate": 1.6557777777777778e-05,
+      "loss": 1.2345,
+      "mean_token_accuracy": 0.6815349081531167,
+      "num_tokens": 24935969.0,
+      "step": 2550
+    },
+    {
+      "entropy": 1.3261476553976537,
+      "epoch": 2.048,
+      "grad_norm": 9.75,
+      "learning_rate": 1.6535555555555556e-05,
+      "loss": 1.25,
+      "mean_token_accuracy": 0.6800847489386797,
+      "num_tokens": 25036470.0,
+      "step": 2560
+    },
+    {
+      "entropy": 1.3340820781886578,
+      "epoch": 2.056,
+      "grad_norm": 9.0,
+      "learning_rate": 1.6513333333333333e-05,
+      "loss": 1.2517,
+      "mean_token_accuracy": 0.6790327195078134,
+      "num_tokens": 25133816.0,
+      "step": 2570
+    },
+    {
+      "entropy": 1.353278061747551,
+      "epoch": 2.064,
+      "grad_norm": 8.5625,
+      "learning_rate": 1.6491111111111114e-05,
+      "loss": 1.2811,
+      "mean_token_accuracy": 0.6770677912980319,
+      "num_tokens": 25230959.0,
+      "step": 2580
+    },
+    {
+      "entropy": 1.3099045217037202,
+      "epoch": 2.072,
+      "grad_norm": 9.75,
+      "learning_rate": 1.646888888888889e-05,
+      "loss": 1.2501,
+      "mean_token_accuracy": 0.6826524507254362,
+      "num_tokens": 25325005.0,
+      "step": 2590
+    },
+    {
+      "entropy": 1.3453321371227502,
+      "epoch": 2.08,
+      "grad_norm": 8.875,
+      "learning_rate": 1.644666666666667e-05,
+      "loss": 1.2711,
+      "mean_token_accuracy": 0.6777556100860238,
+      "num_tokens": 25422625.0,
+      "step": 2600
+    },
+    {
+      "epoch": 2.08,
+      "eval_coding_entropy": 1.4200371098518372,
+      "eval_coding_loss": 2.5741419792175293,
+      "eval_coding_mean_token_accuracy": 0.4915895385742188,
+      "eval_coding_num_tokens": 25422625.0,
+      "eval_coding_runtime": 80.5782,
+      "eval_coding_samples_per_second": 6.205,
+      "eval_coding_steps_per_second": 3.103,
+      "step": 2600
+    },
+    {
+      "epoch": 2.08,
+      "eval_biology_entropy": 1.377419537782669,
+      "eval_biology_loss": 1.8940107822418213,
+      "eval_biology_mean_token_accuracy": 0.6067952680587768,
+      "eval_biology_num_tokens": 25422625.0,
+      "eval_biology_runtime": 34.5064,
+      "eval_biology_samples_per_second": 14.49,
+      "eval_biology_steps_per_second": 7.245,
+      "step": 2600
+    },
+    {
+      "entropy": 1.306822381168604,
+      "epoch": 2.088,
+      "grad_norm": 9.5,
+      "learning_rate": 1.6424444444444444e-05,
+      "loss": 1.2552,
+      "mean_token_accuracy": 0.6827932462096215,
+      "num_tokens": 25521950.0,
+      "step": 2610
+    },
+    {
+      "entropy": 1.353487553820014,
+      "epoch": 2.096,
+      "grad_norm": 9.3125,
+      "learning_rate": 1.6402222222222225e-05,
+      "loss": 1.2842,
+      "mean_token_accuracy": 0.6742482740432024,
+      "num_tokens": 25614286.0,
+      "step": 2620
+    },
+    {
+      "entropy": 1.3189049158245325,
+      "epoch": 2.104,
+      "grad_norm": 8.4375,
+      "learning_rate": 1.638e-05,
+      "loss": 1.2223,
+      "mean_token_accuracy": 0.6869060814380645,
+      "num_tokens": 25713850.0,
+      "step": 2630
+    },
+    {
+      "entropy": 1.2969917487353086,
+      "epoch": 2.112,
+      "grad_norm": 7.90625,
+      "learning_rate": 1.635777777777778e-05,
+      "loss": 1.2404,
+      "mean_token_accuracy": 0.6812446774914861,
+      "num_tokens": 25810384.0,
+      "step": 2640
+    },
+    {
+      "entropy": 1.340190590918064,
+      "epoch": 2.12,
+      "grad_norm": 9.0625,
+      "learning_rate": 1.6335555555555558e-05,
+      "loss": 1.2673,
+      "mean_token_accuracy": 0.6766477670520544,
+      "num_tokens": 25906804.0,
+      "step": 2650
+    },
+    {
+      "entropy": 1.3120084423571825,
+      "epoch": 2.128,
+      "grad_norm": 8.6875,
+      "learning_rate": 1.6313333333333335e-05,
+      "loss": 1.2339,
+      "mean_token_accuracy": 0.6853831727057695,
+      "num_tokens": 26005047.0,
+      "step": 2660
+    },
+    {
+      "entropy": 1.3091088984161616,
+      "epoch": 2.136,
+      "grad_norm": 8.0625,
+      "learning_rate": 1.6291111111111113e-05,
+      "loss": 1.228,
+      "mean_token_accuracy": 0.6843484036624432,
+      "num_tokens": 26108999.0,
+      "step": 2670
+    },
+    {
+      "entropy": 1.328462427854538,
+      "epoch": 2.144,
+      "grad_norm": 10.4375,
+      "learning_rate": 1.626888888888889e-05,
+      "loss": 1.2675,
+      "mean_token_accuracy": 0.6777665082365274,
+      "num_tokens": 26207169.0,
+      "step": 2680
+    },
+    {
+      "entropy": 1.354077434912324,
+      "epoch": 2.152,
+      "grad_norm": 11.3125,
+      "learning_rate": 1.6246666666666668e-05,
+      "loss": 1.2786,
+      "mean_token_accuracy": 0.6771173100918532,
+      "num_tokens": 26305937.0,
+      "step": 2690
+    },
+    {
+      "entropy": 1.3734657648950814,
+      "epoch": 2.16,
+      "grad_norm": 9.375,
+      "learning_rate": 1.6224444444444446e-05,
+      "loss": 1.2925,
+      "mean_token_accuracy": 0.6744703065603972,
+      "num_tokens": 26402150.0,
+      "step": 2700
+    },
+    {
+      "epoch": 2.16,
+      "eval_coding_entropy": 1.4100214166641236,
+      "eval_coding_loss": 2.5930116176605225,
+      "eval_coding_mean_token_accuracy": 0.4897049493789673,
+      "eval_coding_num_tokens": 26402150.0,
+      "eval_coding_runtime": 80.7059,
+      "eval_coding_samples_per_second": 6.195,
+      "eval_coding_steps_per_second": 3.098,
+      "step": 2700
+    },
+    {
+      "epoch": 2.16,
+      "eval_biology_entropy": 1.4006369497776032,
+      "eval_biology_loss": 1.8709315061569214,
+      "eval_biology_mean_token_accuracy": 0.6084704607725143,
+      "eval_biology_num_tokens": 26402150.0,
+      "eval_biology_runtime": 34.5439,
+      "eval_biology_samples_per_second": 14.474,
+      "eval_biology_steps_per_second": 7.237,
+      "step": 2700
+    },
+    {
+      "entropy": 1.318521523103118,
+      "epoch": 2.168,
+      "grad_norm": 9.375,
+      "learning_rate": 1.6202222222222223e-05,
+      "loss": 1.2664,
+      "mean_token_accuracy": 0.6791813928633929,
+      "num_tokens": 26499545.0,
+      "step": 2710
+    },
+    {
+      "entropy": 1.3708464104682208,
+      "epoch": 2.176,
+      "grad_norm": 10.375,
+      "learning_rate": 1.618e-05,
+      "loss": 1.3008,
+      "mean_token_accuracy": 0.671920969709754,
+      "num_tokens": 26599464.0,
+      "step": 2720
+    },
+    {
+      "entropy": 1.2873451314866542,
+      "epoch": 2.184,
+      "grad_norm": 10.8125,
+      "learning_rate": 1.615777777777778e-05,
+      "loss": 1.2165,
+      "mean_token_accuracy": 0.6868797164410353,
+      "num_tokens": 26698280.0,
+      "step": 2730
+    },
+    {
+      "entropy": 1.3334475455805659,
+      "epoch": 2.192,
+      "grad_norm": 8.4375,
+      "learning_rate": 1.6135555555555556e-05,
+      "loss": 1.2483,
+      "mean_token_accuracy": 0.6825383372604847,
+      "num_tokens": 26790601.0,
+      "step": 2740
+    },
+    {
+      "entropy": 1.3112976301461459,
+      "epoch": 2.2,
+      "grad_norm": 8.875,
+      "learning_rate": 1.6113333333333334e-05,
+      "loss": 1.2487,
+      "mean_token_accuracy": 0.6837869491428137,
+      "num_tokens": 26890948.0,
+      "step": 2750
+    },
+    {
+      "entropy": 1.361701337993145,
+      "epoch": 2.208,
+      "grad_norm": 8.9375,
+      "learning_rate": 1.609111111111111e-05,
+      "loss": 1.2995,
+      "mean_token_accuracy": 0.6752991519868374,
+      "num_tokens": 26988598.0,
+      "step": 2760
+    },
+    {
+      "entropy": 1.3458108559250832,
+      "epoch": 2.216,
+      "grad_norm": 9.0625,
+      "learning_rate": 1.606888888888889e-05,
+      "loss": 1.269,
+      "mean_token_accuracy": 0.6751867750659585,
+      "num_tokens": 27086447.0,
+      "step": 2770
+    },
+    {
+      "entropy": 1.3709994018077851,
+      "epoch": 2.224,
+      "grad_norm": 9.5,
+      "learning_rate": 1.6046666666666667e-05,
+      "loss": 1.3122,
+      "mean_token_accuracy": 0.6743873700499534,
+      "num_tokens": 27181933.0,
+      "step": 2780
+    },
+    {
+      "entropy": 1.3100591607391834,
+      "epoch": 2.232,
+      "grad_norm": 10.0625,
+      "learning_rate": 1.6024444444444444e-05,
+      "loss": 1.2356,
+      "mean_token_accuracy": 0.6815704194828868,
+      "num_tokens": 27281887.0,
+      "step": 2790
+    },
+    {
+      "entropy": 1.2877999659627677,
+      "epoch": 2.24,
+      "grad_norm": 8.5,
+      "learning_rate": 1.6002222222222222e-05,
+      "loss": 1.2114,
+      "mean_token_accuracy": 0.6887289330363273,
+      "num_tokens": 27378686.0,
+      "step": 2800
+    },
+    {
+      "epoch": 2.24,
+      "eval_coding_entropy": 1.4068603639602661,
+      "eval_coding_loss": 2.6122212409973145,
+      "eval_coding_mean_token_accuracy": 0.4874311695098877,
+      "eval_coding_num_tokens": 27378686.0,
+      "eval_coding_runtime": 80.8152,
+      "eval_coding_samples_per_second": 6.187,
+      "eval_coding_steps_per_second": 3.093,
+      "step": 2800
+    },
+    {
+      "epoch": 2.24,
+      "eval_biology_entropy": 1.379576261997223,
+      "eval_biology_loss": 1.8703761100769043,
+      "eval_biology_mean_token_accuracy": 0.6101479414701462,
+      "eval_biology_num_tokens": 27378686.0,
+      "eval_biology_runtime": 34.5774,
+      "eval_biology_samples_per_second": 14.46,
+      "eval_biology_steps_per_second": 7.23,
+      "step": 2800
+    },
+    {
+      "entropy": 1.3135050516575575,
+      "epoch": 2.248,
+      "grad_norm": 9.125,
+      "learning_rate": 1.5980000000000003e-05,
+      "loss": 1.2502,
+      "mean_token_accuracy": 0.6803241483867168,
+      "num_tokens": 27471374.0,
+      "step": 2810
+    },
+    {
+      "entropy": 1.2820098880678414,
+      "epoch": 2.2560000000000002,
+      "grad_norm": 9.75,
+      "learning_rate": 1.5957777777777777e-05,
+      "loss": 1.2162,
+      "mean_token_accuracy": 0.6899815808981657,
+      "num_tokens": 27571908.0,
+      "step": 2820
+    },
+    {
+      "entropy": 1.2829747460782528,
+      "epoch": 2.2640000000000002,
+      "grad_norm": 8.6875,
+      "learning_rate": 1.5935555555555558e-05,
+      "loss": 1.2215,
+      "mean_token_accuracy": 0.6869070068001747,
+      "num_tokens": 27670576.0,
+      "step": 2830
+    },
+    {
+      "entropy": 1.2975538209080697,
+      "epoch": 2.2720000000000002,
+      "grad_norm": 9.875,
+      "learning_rate": 1.5913333333333332e-05,
+      "loss": 1.2467,
+      "mean_token_accuracy": 0.6818758625537157,
+      "num_tokens": 27770478.0,
+      "step": 2840
+    },
+    {
+      "entropy": 1.3254982884973288,
+      "epoch": 2.2800000000000002,
+      "grad_norm": 9.6875,
+      "learning_rate": 1.5891111111111113e-05,
+      "loss": 1.2547,
+      "mean_token_accuracy": 0.6790332119911909,
+      "num_tokens": 27869124.0,
+      "step": 2850
+    },
+    {
+      "entropy": 1.36055579893291,
+      "epoch": 2.288,
+      "grad_norm": 9.25,
+      "learning_rate": 1.5868888888888888e-05,
+      "loss": 1.2916,
+      "mean_token_accuracy": 0.6745719544589519,
+      "num_tokens": 27965558.0,
+      "step": 2860
+    },
+    {
+      "entropy": 1.3263279289007186,
+      "epoch": 2.296,
+      "grad_norm": 9.5,
+      "learning_rate": 1.584666666666667e-05,
+      "loss": 1.2545,
+      "mean_token_accuracy": 0.6796434124931693,
+      "num_tokens": 28063623.0,
+      "step": 2870
+    },
+    {
+      "entropy": 1.3337733939290046,
+      "epoch": 2.304,
+      "grad_norm": 8.375,
+      "learning_rate": 1.5824444444444446e-05,
+      "loss": 1.278,
+      "mean_token_accuracy": 0.6782555224373936,
+      "num_tokens": 28165624.0,
+      "step": 2880
+    },
+    {
+      "entropy": 1.3085902910679579,
+      "epoch": 2.312,
+      "grad_norm": 10.3125,
+      "learning_rate": 1.5802222222222224e-05,
+      "loss": 1.2231,
+      "mean_token_accuracy": 0.686250700801611,
+      "num_tokens": 28263002.0,
+      "step": 2890
+    },
+    {
+      "entropy": 1.3032118685543537,
+      "epoch": 2.32,
+      "grad_norm": 9.0625,
+      "learning_rate": 1.578e-05,
+      "loss": 1.2345,
+      "mean_token_accuracy": 0.6842638425529003,
+      "num_tokens": 28363594.0,
+      "step": 2900
+    },
+    {
+      "epoch": 2.32,
+      "eval_coding_entropy": 1.3769904222488403,
+      "eval_coding_loss": 2.6226658821105957,
+      "eval_coding_mean_token_accuracy": 0.4903402773141861,
+      "eval_coding_num_tokens": 28363594.0,
+      "eval_coding_runtime": 80.6873,
+      "eval_coding_samples_per_second": 6.197,
+      "eval_coding_steps_per_second": 3.098,
+      "step": 2900
+    },
+    {
+      "epoch": 2.32,
+      "eval_biology_entropy": 1.373055094242096,
+      "eval_biology_loss": 1.8568984270095825,
+      "eval_biology_mean_token_accuracy": 0.6115068056583405,
+      "eval_biology_num_tokens": 28363594.0,
+      "eval_biology_runtime": 34.5406,
+      "eval_biology_samples_per_second": 14.476,
+      "eval_biology_steps_per_second": 7.238,
+      "step": 2900
+    },
+    {
+      "entropy": 1.3078051444143057,
+      "epoch": 2.328,
+      "grad_norm": 9.0625,
+      "learning_rate": 1.575777777777778e-05,
+      "loss": 1.2442,
+      "mean_token_accuracy": 0.6830941155552864,
+      "num_tokens": 28461456.0,
+      "step": 2910
+    },
+    {
+      "entropy": 1.3277079079300165,
+      "epoch": 2.336,
+      "grad_norm": 10.1875,
+      "learning_rate": 1.5735555555555557e-05,
+      "loss": 1.2536,
+      "mean_token_accuracy": 0.6805235210806131,
+      "num_tokens": 28558104.0,
+      "step": 2920
+    },
+    {
+      "entropy": 1.3095818992704154,
+      "epoch": 2.344,
+      "grad_norm": 9.75,
+      "learning_rate": 1.5713333333333334e-05,
+      "loss": 1.2399,
+      "mean_token_accuracy": 0.6841209914535284,
+      "num_tokens": 28651333.0,
+      "step": 2930
+    },
+    {
+      "entropy": 1.3614604577422142,
+      "epoch": 2.352,
+      "grad_norm": 9.0,
+      "learning_rate": 1.5691111111111112e-05,
+      "loss": 1.2866,
+      "mean_token_accuracy": 0.674070455878973,
+      "num_tokens": 28747265.0,
+      "step": 2940
+    },
+    {
+      "entropy": 1.3213990192860365,
+      "epoch": 2.36,
+      "grad_norm": 9.5,
+      "learning_rate": 1.5668888888888893e-05,
+      "loss": 1.2554,
+      "mean_token_accuracy": 0.6805708345025778,
+      "num_tokens": 28845078.0,
+      "step": 2950
+    },
+    {
+      "entropy": 1.313858350366354,
+      "epoch": 2.368,
+      "grad_norm": 10.4375,
+      "learning_rate": 1.5646666666666667e-05,
+      "loss": 1.2317,
+      "mean_token_accuracy": 0.6821391180157661,
+      "num_tokens": 28944694.0,
+      "step": 2960
+    },
+    {
+      "entropy": 1.3028883308172226,
+      "epoch": 2.376,
+      "grad_norm": 10.3125,
+      "learning_rate": 1.5624444444444448e-05,
+      "loss": 1.2547,
+      "mean_token_accuracy": 0.6808353718370199,
+      "num_tokens": 29043273.0,
+      "step": 2970
+    },
+    {
+      "entropy": 1.357869551703334,
+      "epoch": 2.384,
+      "grad_norm": 9.625,
+      "learning_rate": 1.5602222222222222e-05,
+      "loss": 1.2662,
+      "mean_token_accuracy": 0.6764214668422938,
+      "num_tokens": 29136617.0,
+      "step": 2980
+    },
+    {
+      "entropy": 1.3357534304261207,
+      "epoch": 2.392,
+      "grad_norm": 8.0,
+      "learning_rate": 1.5580000000000003e-05,
+      "loss": 1.2801,
+      "mean_token_accuracy": 0.676220154389739,
+      "num_tokens": 29233852.0,
+      "step": 2990
+    },
+    {
+      "entropy": 1.2693344183266162,
+      "epoch": 2.4,
+      "grad_norm": 9.4375,
+      "learning_rate": 1.5557777777777778e-05,
+      "loss": 1.1995,
+      "mean_token_accuracy": 0.6897810094058514,
+      "num_tokens": 29333526.0,
+      "step": 3000
+    },
+    {
+      "epoch": 2.4,
+      "eval_coding_entropy": 1.3863299145698547,
+      "eval_coding_loss": 2.6043550968170166,
+      "eval_coding_mean_token_accuracy": 0.48935880374908447,
+      "eval_coding_num_tokens": 29333526.0,
+      "eval_coding_runtime": 81.0232,
+      "eval_coding_samples_per_second": 6.171,
+      "eval_coding_steps_per_second": 3.086,
+      "step": 3000
+    },
+    {
+      "epoch": 2.4,
+      "eval_biology_entropy": 1.3638584072589874,
+      "eval_biology_loss": 1.8524683713912964,
+      "eval_biology_mean_token_accuracy": 0.6125637955665588,
+      "eval_biology_num_tokens": 29333526.0,
+      "eval_biology_runtime": 34.2837,
+      "eval_biology_samples_per_second": 14.584,
+      "eval_biology_steps_per_second": 7.292,
+      "step": 3000
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 10000,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 8,
+  "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": 1.951075940066513e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}