diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,4588 @@
+{
+  "best_global_step": 20000,
+  "best_metric": 0.8734950423240662,
+  "best_model_checkpoint": "/scratch-shared/gwijngaard/outputs/qwen2_5-sft-nolora-nonewtoken-nofreeze_20250930_231031/checkpoint-20000",
+  "epoch": 1.0,
+  "eval_steps": 5000,
+  "global_step": 22500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.9206753075122833,
+      "epoch": 4.4444444444444447e-05,
+      "grad_norm": 8.4375,
+      "learning_rate": 0.0,
+      "loss": 1.1975,
+      "mean_token_accuracy": 0.7345979571342468,
+      "num_tokens": 29463751.0,
+      "step": 1
+    },
+    {
+      "entropy": 2.8330495211542868,
+      "epoch": 0.0022222222222222222,
+      "grad_norm": 8.5625,
+      "learning_rate": 1.451851851851852e-06,
+      "loss": 0.8656,
+      "mean_token_accuracy": 0.7418258153662389,
+      "num_tokens": 29529779.0,
+      "step": 50
+    },
+    {
+      "entropy": 2.836770453453064,
+      "epoch": 0.0044444444444444444,
+      "grad_norm": 6.625,
+      "learning_rate": 2.9333333333333338e-06,
+      "loss": 0.8785,
+      "mean_token_accuracy": 0.7382896327972412,
+      "num_tokens": 29596644.0,
+      "step": 100
+    },
+    {
+      "entropy": 2.9122318506240843,
+      "epoch": 0.006666666666666667,
+      "grad_norm": 8.3125,
+      "learning_rate": 4.4148148148148154e-06,
+      "loss": 0.8733,
+      "mean_token_accuracy": 0.7413022458553314,
+      "num_tokens": 29660760.0,
+      "step": 150
+    },
+    {
+      "entropy": 2.900973844528198,
+      "epoch": 0.008888888888888889,
+      "grad_norm": 7.65625,
+      "learning_rate": 5.896296296296296e-06,
+      "loss": 0.897,
+      "mean_token_accuracy": 0.737481083869934,
+      "num_tokens": 29728064.0,
+      "step": 200
+    },
+    {
+      "entropy": 2.8783673429489136,
+      "epoch": 0.011111111111111112,
+      "grad_norm": 7.53125,
+      "learning_rate": 7.377777777777778e-06,
+      "loss": 0.8858,
+      "mean_token_accuracy": 0.7386987507343292,
+      "num_tokens": 29793159.0,
+      "step": 250
+    },
+    {
+      "entropy": 2.855666689872742,
+      "epoch": 0.013333333333333334,
+      "grad_norm": 6.71875,
+      "learning_rate": 8.85925925925926e-06,
+      "loss": 0.8958,
+      "mean_token_accuracy": 0.736550339460373,
+      "num_tokens": 29858642.0,
+      "step": 300
+    },
+    {
+      "entropy": 2.7782880878448486,
+      "epoch": 0.015555555555555555,
+      "grad_norm": 9.125,
+      "learning_rate": 1.0340740740740743e-05,
+      "loss": 0.8976,
+      "mean_token_accuracy": 0.7359859848022461,
+      "num_tokens": 29927577.0,
+      "step": 350
+    },
+    {
+      "entropy": 2.747315707206726,
+      "epoch": 0.017777777777777778,
+      "grad_norm": 9.125,
+      "learning_rate": 1.1822222222222225e-05,
+      "loss": 0.851,
+      "mean_token_accuracy": 0.7458417963981628,
+      "num_tokens": 29993260.0,
+      "step": 400
+    },
+    {
+      "entropy": 2.777502555847168,
+      "epoch": 0.02,
+      "grad_norm": 6.28125,
+      "learning_rate": 1.3303703703703705e-05,
+      "loss": 0.8894,
+      "mean_token_accuracy": 0.7404855847358703,
+      "num_tokens": 30061485.0,
+      "step": 450
+    },
+    {
+      "entropy": 2.879872035980225,
+      "epoch": 0.022222222222222223,
+      "grad_norm": 8.4375,
+      "learning_rate": 1.4785185185185186e-05,
+      "loss": 0.9081,
+      "mean_token_accuracy": 0.7336488461494446,
+      "num_tokens": 30126648.0,
+      "step": 500
+    },
+    {
+      "entropy": 2.7876138353347777,
+      "epoch": 0.024444444444444446,
+      "grad_norm": 9.6875,
+      "learning_rate": 1.6266666666666668e-05,
+      "loss": 0.8461,
+      "mean_token_accuracy": 0.7486303246021271,
+      "num_tokens": 30192879.0,
+      "step": 550
+    },
+    {
+      "entropy": 2.9612947702407837,
+      "epoch": 0.02666666666666667,
+      "grad_norm": 8.0,
+      "learning_rate": 1.774814814814815e-05,
+      "loss": 0.8981,
+      "mean_token_accuracy": 0.7390279281139374,
+      "num_tokens": 30257537.0,
+      "step": 600
+    },
+    {
+      "entropy": 2.7916600465774537,
+      "epoch": 0.028888888888888888,
+      "grad_norm": 7.9375,
+      "learning_rate": 1.922962962962963e-05,
+      "loss": 0.8675,
+      "mean_token_accuracy": 0.7437423765659332,
+      "num_tokens": 30323879.0,
+      "step": 650
+    },
+    {
+      "entropy": 2.8081023025512697,
+      "epoch": 0.03111111111111111,
+      "grad_norm": 7.28125,
+      "learning_rate": 1.9978006872852237e-05,
+      "loss": 0.8574,
+      "mean_token_accuracy": 0.7459037184715271,
+      "num_tokens": 30393385.0,
+      "step": 700
+    },
+    {
+      "entropy": 2.9191632890701293,
+      "epoch": 0.03333333333333333,
+      "grad_norm": 8.3125,
+      "learning_rate": 1.9932187857961058e-05,
+      "loss": 0.9081,
+      "mean_token_accuracy": 0.7324011015892029,
+      "num_tokens": 30458516.0,
+      "step": 750
+    },
+    {
+      "entropy": 2.783015651702881,
+      "epoch": 0.035555555555555556,
+      "grad_norm": 6.875,
+      "learning_rate": 1.9886368843069876e-05,
+      "loss": 0.8403,
+      "mean_token_accuracy": 0.7504636573791504,
+      "num_tokens": 30522811.0,
+      "step": 800
+    },
+    {
+      "entropy": 2.782901654243469,
+      "epoch": 0.03777777777777778,
+      "grad_norm": 7.90625,
+      "learning_rate": 1.9840549828178697e-05,
+      "loss": 0.8633,
+      "mean_token_accuracy": 0.7460731649398804,
+      "num_tokens": 30588800.0,
+      "step": 850
+    },
+    {
+      "entropy": 2.852466220855713,
+      "epoch": 0.04,
+      "grad_norm": 7.875,
+      "learning_rate": 1.9794730813287515e-05,
+      "loss": 0.8939,
+      "mean_token_accuracy": 0.7356749868392944,
+      "num_tokens": 30654866.0,
+      "step": 900
+    },
+    {
+      "entropy": 2.8751544713974,
+      "epoch": 0.042222222222222223,
+      "grad_norm": 6.5625,
+      "learning_rate": 1.9748911798396336e-05,
+      "loss": 0.8526,
+      "mean_token_accuracy": 0.7440717363357544,
+      "num_tokens": 30718125.0,
+      "step": 950
+    },
+    {
+      "entropy": 2.791564016342163,
+      "epoch": 0.044444444444444446,
+      "grad_norm": 8.5625,
+      "learning_rate": 1.9703092783505157e-05,
+      "loss": 0.8528,
+      "mean_token_accuracy": 0.743047387599945,
+      "num_tokens": 30784326.0,
+      "step": 1000
+    },
+    {
+      "entropy": 2.7872579193115232,
+      "epoch": 0.04666666666666667,
+      "grad_norm": 10.3125,
+      "learning_rate": 1.9657273768613975e-05,
+      "loss": 0.8516,
+      "mean_token_accuracy": 0.747126750946045,
+      "num_tokens": 30851638.0,
+      "step": 1050
+    },
+    {
+      "entropy": 2.7552728176116945,
+      "epoch": 0.04888888888888889,
+      "grad_norm": 8.4375,
+      "learning_rate": 1.9611454753722796e-05,
+      "loss": 0.8552,
+      "mean_token_accuracy": 0.7450509345531464,
+      "num_tokens": 30917218.0,
+      "step": 1100
+    },
+    {
+      "entropy": 2.7746379947662354,
+      "epoch": 0.051111111111111114,
+      "grad_norm": 9.3125,
+      "learning_rate": 1.9565635738831617e-05,
+      "loss": 0.8364,
+      "mean_token_accuracy": 0.754111316204071,
+      "num_tokens": 30981770.0,
+      "step": 1150
+    },
+    {
+      "entropy": 2.6211288261413572,
+      "epoch": 0.05333333333333334,
+      "grad_norm": 7.03125,
+      "learning_rate": 1.9519816723940438e-05,
+      "loss": 0.8351,
+      "mean_token_accuracy": 0.7522010815143585,
+      "num_tokens": 31048479.0,
+      "step": 1200
+    },
+    {
+      "entropy": 2.6207118034362793,
+      "epoch": 0.05555555555555555,
+      "grad_norm": 8.5625,
+      "learning_rate": 1.9473997709049256e-05,
+      "loss": 0.8428,
+      "mean_token_accuracy": 0.7495422863960266,
+      "num_tokens": 31116595.0,
+      "step": 1250
+    },
+    {
+      "entropy": 2.6855200052261354,
+      "epoch": 0.057777777777777775,
+      "grad_norm": 9.1875,
+      "learning_rate": 1.9428178694158077e-05,
+      "loss": 0.8629,
+      "mean_token_accuracy": 0.7421471869945526,
+      "num_tokens": 31183980.0,
+      "step": 1300
+    },
+    {
+      "entropy": 2.528144361972809,
+      "epoch": 0.06,
+      "grad_norm": 6.0625,
+      "learning_rate": 1.93823596792669e-05,
+      "loss": 0.7948,
+      "mean_token_accuracy": 0.7594469666481019,
+      "num_tokens": 31250261.0,
+      "step": 1350
+    },
+    {
+      "entropy": 2.5050185775756835,
+      "epoch": 0.06222222222222222,
+      "grad_norm": 8.5,
+      "learning_rate": 1.933654066437572e-05,
+      "loss": 0.8032,
+      "mean_token_accuracy": 0.7555232620239258,
+      "num_tokens": 31316852.0,
+      "step": 1400
+    },
+    {
+      "entropy": 2.511464171409607,
+      "epoch": 0.06444444444444444,
+      "grad_norm": 8.4375,
+      "learning_rate": 1.9290721649484537e-05,
+      "loss": 0.8122,
+      "mean_token_accuracy": 0.7525858426094055,
+      "num_tokens": 31381166.0,
+      "step": 1450
+    },
+    {
+      "entropy": 2.47177143573761,
+      "epoch": 0.06666666666666667,
+      "grad_norm": 7.59375,
+      "learning_rate": 1.924490263459336e-05,
+      "loss": 0.8022,
+      "mean_token_accuracy": 0.7576224994659424,
+      "num_tokens": 31450722.0,
+      "step": 1500
+    },
+    {
+      "entropy": 2.451930871009827,
+      "epoch": 0.06888888888888889,
+      "grad_norm": 12.3125,
+      "learning_rate": 1.919908361970218e-05,
+      "loss": 0.8025,
+      "mean_token_accuracy": 0.7531751859188079,
+      "num_tokens": 31517729.0,
+      "step": 1550
+    },
+    {
+      "entropy": 2.46038672208786,
+      "epoch": 0.07111111111111111,
+      "grad_norm": 6.84375,
+      "learning_rate": 1.9153264604810998e-05,
+      "loss": 0.8188,
+      "mean_token_accuracy": 0.7543287444114685,
+      "num_tokens": 31584935.0,
+      "step": 1600
+    },
+    {
+      "entropy": 2.5046206855773927,
+      "epoch": 0.07333333333333333,
+      "grad_norm": 7.40625,
+      "learning_rate": 1.910744558991982e-05,
+      "loss": 0.8076,
+      "mean_token_accuracy": 0.7573912596702576,
+      "num_tokens": 31651819.0,
+      "step": 1650
+    },
+    {
+      "entropy": 2.587780613899231,
+      "epoch": 0.07555555555555556,
+      "grad_norm": 6.0625,
+      "learning_rate": 1.9061626575028637e-05,
+      "loss": 0.7887,
+      "mean_token_accuracy": 0.7603560221195221,
+      "num_tokens": 31716593.0,
+      "step": 1700
+    },
+    {
+      "entropy": 2.471900417804718,
+      "epoch": 0.07777777777777778,
+      "grad_norm": 7.0,
+      "learning_rate": 1.9015807560137458e-05,
+      "loss": 0.7814,
+      "mean_token_accuracy": 0.7607843494415283,
+      "num_tokens": 31783153.0,
+      "step": 1750
+    },
+    {
+      "entropy": 2.4517657709121705,
+      "epoch": 0.08,
+      "grad_norm": 7.1875,
+      "learning_rate": 1.896998854524628e-05,
+      "loss": 0.8497,
+      "mean_token_accuracy": 0.7433587527275085,
+      "num_tokens": 31851501.0,
+      "step": 1800
+    },
+    {
+      "entropy": 2.4117108368873597,
+      "epoch": 0.08222222222222222,
+      "grad_norm": 7.0625,
+      "learning_rate": 1.89241695303551e-05,
+      "loss": 0.799,
+      "mean_token_accuracy": 0.7592541599273681,
+      "num_tokens": 31916822.0,
+      "step": 1850
+    },
+    {
+      "entropy": 2.35961368560791,
+      "epoch": 0.08444444444444445,
+      "grad_norm": 12.4375,
+      "learning_rate": 1.8878350515463918e-05,
+      "loss": 0.8093,
+      "mean_token_accuracy": 0.7549574375152588,
+      "num_tokens": 31984311.0,
+      "step": 1900
+    },
+    {
+      "entropy": 2.444758574962616,
+      "epoch": 0.08666666666666667,
+      "grad_norm": 7.0,
+      "learning_rate": 1.883253150057274e-05,
+      "loss": 0.7778,
+      "mean_token_accuracy": 0.7624431896209717,
+      "num_tokens": 32049197.0,
+      "step": 1950
+    },
+    {
+      "entropy": 2.4109366965293884,
+      "epoch": 0.08888888888888889,
+      "grad_norm": 8.6875,
+      "learning_rate": 1.878671248568156e-05,
+      "loss": 0.8078,
+      "mean_token_accuracy": 0.7562625980377198,
+      "num_tokens": 32114585.0,
+      "step": 2000
+    },
+    {
+      "entropy": 2.338375616073608,
+      "epoch": 0.09111111111111111,
+      "grad_norm": 7.8125,
+      "learning_rate": 1.874089347079038e-05,
+      "loss": 0.8051,
+      "mean_token_accuracy": 0.7559153795242309,
+      "num_tokens": 32182189.0,
+      "step": 2050
+    },
+    {
+      "entropy": 2.3647463655471803,
+      "epoch": 0.09333333333333334,
+      "grad_norm": 8.5625,
+      "learning_rate": 1.86950744558992e-05,
+      "loss": 0.7964,
+      "mean_token_accuracy": 0.7597098231315613,
+      "num_tokens": 32246491.0,
+      "step": 2100
+    },
+    {
+      "entropy": 2.3017068338394164,
+      "epoch": 0.09555555555555556,
+      "grad_norm": 11.0625,
+      "learning_rate": 1.864925544100802e-05,
+      "loss": 0.8237,
+      "mean_token_accuracy": 0.749286116361618,
+      "num_tokens": 32313665.0,
+      "step": 2150
+    },
+    {
+      "entropy": 2.3421465373039245,
+      "epoch": 0.09777777777777778,
+      "grad_norm": 8.6875,
+      "learning_rate": 1.860343642611684e-05,
+      "loss": 0.7969,
+      "mean_token_accuracy": 0.757558046579361,
+      "num_tokens": 32377681.0,
+      "step": 2200
+    },
+    {
+      "entropy": 2.2409315156936644,
+      "epoch": 0.1,
+      "grad_norm": 8.25,
+      "learning_rate": 1.855761741122566e-05,
+      "loss": 0.782,
+      "mean_token_accuracy": 0.7633419442176819,
+      "num_tokens": 32445533.0,
+      "step": 2250
+    },
+    {
+      "entropy": 2.2817729759216308,
+      "epoch": 0.10222222222222223,
+      "grad_norm": 8.5625,
+      "learning_rate": 1.851179839633448e-05,
+      "loss": 0.803,
+      "mean_token_accuracy": 0.7561051642894745,
+      "num_tokens": 32514116.0,
+      "step": 2300
+    },
+    {
+      "entropy": 2.151771836280823,
+      "epoch": 0.10444444444444445,
+      "grad_norm": 7.125,
+      "learning_rate": 1.84659793814433e-05,
+      "loss": 0.7906,
+      "mean_token_accuracy": 0.7589080274105072,
+      "num_tokens": 32583284.0,
+      "step": 2350
+    },
+    {
+      "entropy": 2.184410240650177,
+      "epoch": 0.10666666666666667,
+      "grad_norm": 10.4375,
+      "learning_rate": 1.842016036655212e-05,
+      "loss": 0.8398,
+      "mean_token_accuracy": 0.7480292344093322,
+      "num_tokens": 32650513.0,
+      "step": 2400
+    },
+    {
+      "entropy": 2.18844452381134,
+      "epoch": 0.10888888888888888,
+      "grad_norm": 10.5,
+      "learning_rate": 1.837434135166094e-05,
+      "loss": 0.801,
+      "mean_token_accuracy": 0.7543882429599762,
+      "num_tokens": 32716407.0,
+      "step": 2450
+    },
+    {
+      "entropy": 2.172247538566589,
+      "epoch": 0.1111111111111111,
+      "grad_norm": 10.5625,
+      "learning_rate": 1.8328522336769762e-05,
+      "loss": 0.7963,
+      "mean_token_accuracy": 0.7586643397808075,
+      "num_tokens": 32781494.0,
+      "step": 2500
+    },
+    {
+      "entropy": 2.241757538318634,
+      "epoch": 0.11333333333333333,
+      "grad_norm": 9.9375,
+      "learning_rate": 1.828270332187858e-05,
+      "loss": 0.8141,
+      "mean_token_accuracy": 0.7552792716026306,
+      "num_tokens": 32846884.0,
+      "step": 2550
+    },
+    {
+      "entropy": 2.2868755722045897,
+      "epoch": 0.11555555555555555,
+      "grad_norm": 7.8125,
+      "learning_rate": 1.82368843069874e-05,
+      "loss": 0.8164,
+      "mean_token_accuracy": 0.7541338610649109,
+      "num_tokens": 32912538.0,
+      "step": 2600
+    },
+    {
+      "entropy": 2.2125061202049254,
+      "epoch": 0.11777777777777777,
+      "grad_norm": 7.1875,
+      "learning_rate": 1.8191065292096222e-05,
+      "loss": 0.7896,
+      "mean_token_accuracy": 0.7596218919754029,
+      "num_tokens": 32980319.0,
+      "step": 2650
+    },
+    {
+      "entropy": 2.155822920799255,
+      "epoch": 0.12,
+      "grad_norm": 6.78125,
+      "learning_rate": 1.8145246277205043e-05,
+      "loss": 0.8207,
+      "mean_token_accuracy": 0.754329582452774,
+      "num_tokens": 33049961.0,
+      "step": 2700
+    },
+    {
+      "entropy": 2.1824621748924256,
+      "epoch": 0.12222222222222222,
+      "grad_norm": 7.59375,
+      "learning_rate": 1.809942726231386e-05,
+      "loss": 0.7661,
+      "mean_token_accuracy": 0.7682196378707886,
+      "num_tokens": 33116366.0,
+      "step": 2750
+    },
+    {
+      "entropy": 2.231255145072937,
+      "epoch": 0.12444444444444444,
+      "grad_norm": 8.875,
+      "learning_rate": 1.8053608247422682e-05,
+      "loss": 0.8048,
+      "mean_token_accuracy": 0.7537202823162079,
+      "num_tokens": 33185374.0,
+      "step": 2800
+    },
+    {
+      "entropy": 2.2291252851486205,
+      "epoch": 0.12666666666666668,
+      "grad_norm": 8.125,
+      "learning_rate": 1.8007789232531504e-05,
+      "loss": 0.7598,
+      "mean_token_accuracy": 0.766055703163147,
+      "num_tokens": 33252044.0,
+      "step": 2850
+    },
+    {
+      "entropy": 2.1579408121109007,
+      "epoch": 0.1288888888888889,
+      "grad_norm": 8.3125,
+      "learning_rate": 1.7961970217640325e-05,
+      "loss": 0.7862,
+      "mean_token_accuracy": 0.7643947994709015,
+      "num_tokens": 33319654.0,
+      "step": 2900
+    },
+    {
+      "entropy": 2.1827985763549806,
+      "epoch": 0.13111111111111112,
+      "grad_norm": 7.375,
+      "learning_rate": 1.7916151202749143e-05,
+      "loss": 0.8002,
+      "mean_token_accuracy": 0.7550258612632752,
+      "num_tokens": 33388120.0,
+      "step": 2950
+    },
+    {
+      "entropy": 2.2125383615493774,
+      "epoch": 0.13333333333333333,
+      "grad_norm": 8.0,
+      "learning_rate": 1.787033218785796e-05,
+      "loss": 0.7784,
+      "mean_token_accuracy": 0.7626236033439636,
+      "num_tokens": 33454419.0,
+      "step": 3000
+    },
+    {
+      "entropy": 2.1113911986351015,
+      "epoch": 0.13555555555555557,
+      "grad_norm": 6.6875,
+      "learning_rate": 1.782451317296678e-05,
+      "loss": 0.7674,
+      "mean_token_accuracy": 0.7665451729297638,
+      "num_tokens": 33520682.0,
+      "step": 3050
+    },
+    {
+      "entropy": 2.1803041291236878,
+      "epoch": 0.13777777777777778,
+      "grad_norm": 7.34375,
+      "learning_rate": 1.7778694158075603e-05,
+      "loss": 0.7898,
+      "mean_token_accuracy": 0.7653110408782959,
+      "num_tokens": 33588492.0,
+      "step": 3100
+    },
+    {
+      "entropy": 2.2334672379493714,
+      "epoch": 0.14,
+      "grad_norm": 7.46875,
+      "learning_rate": 1.7732875143184424e-05,
+      "loss": 0.7731,
+      "mean_token_accuracy": 0.7676819574832916,
+      "num_tokens": 33654548.0,
+      "step": 3150
+    },
+    {
+      "entropy": 2.1736029314994814,
+      "epoch": 0.14222222222222222,
+      "grad_norm": 9.3125,
+      "learning_rate": 1.768705612829324e-05,
+      "loss": 0.8031,
+      "mean_token_accuracy": 0.7566261160373687,
+      "num_tokens": 33722789.0,
+      "step": 3200
+    },
+    {
+      "entropy": 2.2151904988288877,
+      "epoch": 0.14444444444444443,
+      "grad_norm": 7.625,
+      "learning_rate": 1.7641237113402063e-05,
+      "loss": 0.7979,
+      "mean_token_accuracy": 0.7582576811313629,
+      "num_tokens": 33789866.0,
+      "step": 3250
+    },
+    {
+      "entropy": 2.213924946784973,
+      "epoch": 0.14666666666666667,
+      "grad_norm": 9.3125,
+      "learning_rate": 1.7595418098510884e-05,
+      "loss": 0.7889,
+      "mean_token_accuracy": 0.7583772194385529,
+      "num_tokens": 33857885.0,
+      "step": 3300
+    },
+    {
+      "entropy": 2.144984345436096,
+      "epoch": 0.14888888888888888,
+      "grad_norm": 7.78125,
+      "learning_rate": 1.7549599083619705e-05,
+      "loss": 0.8019,
+      "mean_token_accuracy": 0.7535349237918854,
+      "num_tokens": 33923487.0,
+      "step": 3350
+    },
+    {
+      "entropy": 2.1649972796440125,
+      "epoch": 0.1511111111111111,
+      "grad_norm": 8.125,
+      "learning_rate": 1.7503780068728523e-05,
+      "loss": 0.7989,
+      "mean_token_accuracy": 0.757453374862671,
+      "num_tokens": 33989482.0,
+      "step": 3400
+    },
+    {
+      "entropy": 2.2169123339653014,
+      "epoch": 0.15333333333333332,
+      "grad_norm": 6.5625,
+      "learning_rate": 1.7457961053837344e-05,
+      "loss": 0.8161,
+      "mean_token_accuracy": 0.7528200805187225,
+      "num_tokens": 34059798.0,
+      "step": 3450
+    },
+    {
+      "entropy": 2.2231992268562317,
+      "epoch": 0.15555555555555556,
+      "grad_norm": 7.78125,
+      "learning_rate": 1.7412142038946165e-05,
+      "loss": 0.8242,
+      "mean_token_accuracy": 0.7535739564895629,
+      "num_tokens": 34125686.0,
+      "step": 3500
+    },
+    {
+      "entropy": 2.1861121082305908,
+      "epoch": 0.15777777777777777,
+      "grad_norm": 7.125,
+      "learning_rate": 1.7366323024054987e-05,
+      "loss": 0.777,
+      "mean_token_accuracy": 0.7635022902488708,
+      "num_tokens": 34192858.0,
+      "step": 3550
+    },
+    {
+      "entropy": 2.1724994444847106,
+      "epoch": 0.16,
+      "grad_norm": 8.0625,
+      "learning_rate": 1.7320504009163804e-05,
+      "loss": 0.7834,
+      "mean_token_accuracy": 0.76077321767807,
+      "num_tokens": 34259641.0,
+      "step": 3600
+    },
+    {
+      "entropy": 2.182099552154541,
+      "epoch": 0.1622222222222222,
+      "grad_norm": 7.09375,
+      "learning_rate": 1.7274684994272622e-05,
+      "loss": 0.7983,
+      "mean_token_accuracy": 0.7584855699539185,
+      "num_tokens": 34328061.0,
+      "step": 3650
+    },
+    {
+      "entropy": 2.1422414350509644,
+      "epoch": 0.16444444444444445,
+      "grad_norm": 8.375,
+      "learning_rate": 1.7228865979381443e-05,
+      "loss": 0.7817,
+      "mean_token_accuracy": 0.7643873155117035,
+      "num_tokens": 34397172.0,
+      "step": 3700
+    },
+    {
+      "entropy": 2.1676920294761657,
+      "epoch": 0.16666666666666666,
+      "grad_norm": 7.34375,
+      "learning_rate": 1.7183046964490265e-05,
+      "loss": 0.8121,
+      "mean_token_accuracy": 0.7518685030937194,
+      "num_tokens": 34462090.0,
+      "step": 3750
+    },
+    {
+      "entropy": 2.0866295003890993,
+      "epoch": 0.1688888888888889,
+      "grad_norm": 7.5,
+      "learning_rate": 1.7137227949599086e-05,
+      "loss": 0.7967,
+      "mean_token_accuracy": 0.7572396492958069,
+      "num_tokens": 34531839.0,
+      "step": 3800
+    },
+    {
+      "entropy": 2.1308164954185487,
+      "epoch": 0.1711111111111111,
+      "grad_norm": 6.9375,
+      "learning_rate": 1.7091408934707904e-05,
+      "loss": 0.7656,
+      "mean_token_accuracy": 0.7690932631492615,
+      "num_tokens": 34597073.0,
+      "step": 3850
+    },
+    {
+      "entropy": 2.1494863390922547,
+      "epoch": 0.17333333333333334,
+      "grad_norm": 8.5625,
+      "learning_rate": 1.7045589919816725e-05,
+      "loss": 0.8129,
+      "mean_token_accuracy": 0.7566492486000062,
+      "num_tokens": 34667098.0,
+      "step": 3900
+    },
+    {
+      "entropy": 2.098915767669678,
+      "epoch": 0.17555555555555555,
+      "grad_norm": 7.9375,
+      "learning_rate": 1.6999770904925546e-05,
+      "loss": 0.7681,
+      "mean_token_accuracy": 0.764438933134079,
+      "num_tokens": 34734238.0,
+      "step": 3950
+    },
+    {
+      "entropy": 2.2060401248931885,
+      "epoch": 0.17777777777777778,
+      "grad_norm": 6.6875,
+      "learning_rate": 1.6953951890034367e-05,
+      "loss": 0.7993,
+      "mean_token_accuracy": 0.7594318461418151,
+      "num_tokens": 34801484.0,
+      "step": 4000
+    },
+    {
+      "entropy": 2.1855486106872557,
+      "epoch": 0.18,
+      "grad_norm": 7.96875,
+      "learning_rate": 1.6908132875143185e-05,
+      "loss": 0.792,
+      "mean_token_accuracy": 0.7597147989273071,
+      "num_tokens": 34869674.0,
+      "step": 4050
+    },
+    {
+      "entropy": 2.0849170541763304,
+      "epoch": 0.18222222222222223,
+      "grad_norm": 7.21875,
+      "learning_rate": 1.6862313860252006e-05,
+      "loss": 0.8095,
+      "mean_token_accuracy": 0.7556380236148834,
+      "num_tokens": 34939372.0,
+      "step": 4100
+    },
+    {
+      "entropy": 2.1427531123161314,
+      "epoch": 0.18444444444444444,
+      "grad_norm": 7.4375,
+      "learning_rate": 1.6816494845360827e-05,
+      "loss": 0.8103,
+      "mean_token_accuracy": 0.7562667608261109,
+      "num_tokens": 35007892.0,
+      "step": 4150
+    },
+    {
+      "entropy": 2.0929103469848633,
+      "epoch": 0.18666666666666668,
+      "grad_norm": 10.3125,
+      "learning_rate": 1.677067583046965e-05,
+      "loss": 0.8292,
+      "mean_token_accuracy": 0.7510438573360443,
+      "num_tokens": 35078997.0,
+      "step": 4200
+    },
+    {
+      "entropy": 2.049334568977356,
+      "epoch": 0.18888888888888888,
+      "grad_norm": 7.875,
+      "learning_rate": 1.6724856815578466e-05,
+      "loss": 0.7729,
+      "mean_token_accuracy": 0.7638446021080018,
+      "num_tokens": 35148350.0,
+      "step": 4250
+    },
+    {
+      "entropy": 2.099260985851288,
+      "epoch": 0.19111111111111112,
+      "grad_norm": 6.8125,
+      "learning_rate": 1.6679037800687284e-05,
+      "loss": 0.7974,
+      "mean_token_accuracy": 0.7584287643432617,
+      "num_tokens": 35216731.0,
+      "step": 4300
+    },
+    {
+      "entropy": 2.112381303310394,
+      "epoch": 0.19333333333333333,
+      "grad_norm": 7.0,
+      "learning_rate": 1.6633218785796105e-05,
+      "loss": 0.743,
+      "mean_token_accuracy": 0.7701405155658722,
+      "num_tokens": 35282400.0,
+      "step": 4350
+    },
+    {
+      "entropy": 2.152976577281952,
+      "epoch": 0.19555555555555557,
+      "grad_norm": 7.21875,
+      "learning_rate": 1.6587399770904926e-05,
+      "loss": 0.79,
+      "mean_token_accuracy": 0.7612162494659424,
+      "num_tokens": 35348623.0,
+      "step": 4400
+    },
+    {
+      "entropy": 2.1815468096733093,
+      "epoch": 0.19777777777777777,
+      "grad_norm": 7.125,
+      "learning_rate": 1.6541580756013748e-05,
+      "loss": 0.8279,
+      "mean_token_accuracy": 0.7511774361133575,
+      "num_tokens": 35417105.0,
+      "step": 4450
+    },
+    {
+      "entropy": 2.2018524050712585,
+      "epoch": 0.2,
+      "grad_norm": 8.8125,
+      "learning_rate": 1.6495761741122565e-05,
+      "loss": 0.7781,
+      "mean_token_accuracy": 0.763760222196579,
+      "num_tokens": 35483468.0,
+      "step": 4500
+    },
+    {
+      "entropy": 2.1301321840286254,
+      "epoch": 0.20222222222222222,
+      "grad_norm": 7.28125,
+      "learning_rate": 1.6449942726231387e-05,
+      "loss": 0.8019,
+      "mean_token_accuracy": 0.7569781160354614,
+      "num_tokens": 35550517.0,
+      "step": 4550
+    },
+    {
+      "entropy": 2.187158074378967,
+      "epoch": 0.20444444444444446,
+      "grad_norm": 8.25,
+      "learning_rate": 1.6404123711340208e-05,
+      "loss": 0.8104,
+      "mean_token_accuracy": 0.7544570553302765,
+      "num_tokens": 35617223.0,
+      "step": 4600
+    },
+    {
+      "entropy": 2.106666340827942,
+      "epoch": 0.20666666666666667,
+      "grad_norm": 7.53125,
+      "learning_rate": 1.635830469644903e-05,
+      "loss": 0.755,
+      "mean_token_accuracy": 0.7678434896469116,
+      "num_tokens": 35682799.0,
+      "step": 4650
+    },
+    {
+      "entropy": 2.109342801570892,
+      "epoch": 0.2088888888888889,
+      "grad_norm": 7.09375,
+      "learning_rate": 1.6312485681557847e-05,
+      "loss": 0.809,
+      "mean_token_accuracy": 0.7529355835914612,
+      "num_tokens": 35749281.0,
+      "step": 4700
+    },
+    {
+      "entropy": 2.0876813530921936,
+      "epoch": 0.2111111111111111,
+      "grad_norm": 8.5625,
+      "learning_rate": 1.6266666666666668e-05,
+      "loss": 0.7859,
+      "mean_token_accuracy": 0.760038149356842,
+      "num_tokens": 35817971.0,
+      "step": 4750
+    },
+    {
+      "entropy": 2.109708137512207,
+      "epoch": 0.21333333333333335,
+      "grad_norm": 8.875,
+      "learning_rate": 1.622084765177549e-05,
+      "loss": 0.8217,
+      "mean_token_accuracy": 0.7479185128211975,
+      "num_tokens": 35884686.0,
+      "step": 4800
+    },
+    {
+      "entropy": 2.181152358055115,
+      "epoch": 0.21555555555555556,
+      "grad_norm": 11.375,
+      "learning_rate": 1.617502863688431e-05,
+      "loss": 0.8565,
+      "mean_token_accuracy": 0.7458688330650329,
+      "num_tokens": 35950287.0,
+      "step": 4850
+    },
+    {
+      "entropy": 2.1443709397315978,
+      "epoch": 0.21777777777777776,
+      "grad_norm": 9.4375,
+      "learning_rate": 1.6129209621993128e-05,
+      "loss": 0.8138,
+      "mean_token_accuracy": 0.7534062623977661,
+      "num_tokens": 36015777.0,
+      "step": 4900
+    },
+    {
+      "entropy": 2.0954318118095396,
+      "epoch": 0.22,
+      "grad_norm": 7.84375,
+      "learning_rate": 1.608339060710195e-05,
+      "loss": 0.7995,
+      "mean_token_accuracy": 0.7556386387348175,
+      "num_tokens": 36084969.0,
+      "step": 4950
+    },
+    {
+      "entropy": 2.1733604526519774,
+      "epoch": 0.2222222222222222,
+      "grad_norm": 7.3125,
+      "learning_rate": 1.6037571592210767e-05,
+      "loss": 0.7857,
+      "mean_token_accuracy": 0.7591150319576263,
+      "num_tokens": 36150062.0,
+      "step": 5000
+    },
+    {
+      "epoch": 0.2222222222222222,
+      "eval_entropy": 2.060102254152298,
+      "eval_loss": 0.9120854735374451,
+      "eval_mean_token_accuracy": 0.7374376058578491,
+      "eval_num_tokens": 36150062.0,
+      "eval_runtime": 5.8568,
+      "eval_samples_per_second": 2.22,
+      "eval_steps_per_second": 0.683,
+      "step": 5000
+    },
+    {
+      "entropy": 2.12566153049469,
+      "epoch": 0.22444444444444445,
+      "grad_norm": 6.28125,
+      "learning_rate": 1.5991752577319588e-05,
+      "loss": 0.7834,
+      "mean_token_accuracy": 0.7630481898784638,
+      "num_tokens": 36218190.0,
+      "step": 5050
+    },
+    {
+      "entropy": 2.1597506642341613,
+      "epoch": 0.22666666666666666,
+      "grad_norm": 7.96875,
+      "learning_rate": 1.594593356242841e-05,
+      "loss": 0.8013,
+      "mean_token_accuracy": 0.7572294700145722,
+      "num_tokens": 36284871.0,
+      "step": 5100
+    },
+    {
+      "entropy": 2.2783447027206423,
+      "epoch": 0.2288888888888889,
+      "grad_norm": 6.90625,
+      "learning_rate": 1.5900114547537227e-05,
+      "loss": 0.8446,
+      "mean_token_accuracy": 0.7509990322589875,
+      "num_tokens": 36352201.0,
+      "step": 5150
+    },
+    {
+      "entropy": 2.0995934796333313,
+      "epoch": 0.2311111111111111,
+      "grad_norm": 7.21875,
+      "learning_rate": 1.585429553264605e-05,
+      "loss": 0.7758,
+      "mean_token_accuracy": 0.7649268102645874,
+      "num_tokens": 36422636.0,
+      "step": 5200
+    },
+    {
+      "entropy": 2.2191508650779723,
+      "epoch": 0.23333333333333334,
+      "grad_norm": 8.9375,
+      "learning_rate": 1.580847651775487e-05,
+      "loss": 0.8183,
+      "mean_token_accuracy": 0.7507635414600372,
+      "num_tokens": 36489811.0,
+      "step": 5250
+    },
+    {
+      "entropy": 2.158704442977905,
+      "epoch": 0.23555555555555555,
+      "grad_norm": 11.625,
+      "learning_rate": 1.576265750286369e-05,
+      "loss": 0.7919,
+      "mean_token_accuracy": 0.7578798067569733,
+      "num_tokens": 36558093.0,
+      "step": 5300
+    },
+    {
+      "entropy": 2.2435457158088683,
+      "epoch": 0.23777777777777778,
+      "grad_norm": 10.1875,
+      "learning_rate": 1.571683848797251e-05,
+      "loss": 0.8026,
+      "mean_token_accuracy": 0.7596748018264771,
+      "num_tokens": 36623942.0,
+      "step": 5350
+    },
+    {
+      "entropy": 2.26217898607254,
+      "epoch": 0.24,
+      "grad_norm": 6.6875,
+      "learning_rate": 1.567101947308133e-05,
+      "loss": 0.7801,
+      "mean_token_accuracy": 0.7630024456977844,
+      "num_tokens": 36693576.0,
+      "step": 5400
+    },
+    {
+      "entropy": 2.1895520281791687,
+      "epoch": 0.24222222222222223,
+      "grad_norm": 6.96875,
+      "learning_rate": 1.562520045819015e-05,
+      "loss": 0.8052,
+      "mean_token_accuracy": 0.7553786933422089,
+      "num_tokens": 36761921.0,
+      "step": 5450
+    },
+    {
+      "entropy": 2.242999081611633,
+      "epoch": 0.24444444444444444,
+      "grad_norm": 7.90625,
+      "learning_rate": 1.5579381443298972e-05,
+      "loss": 0.7577,
+      "mean_token_accuracy": 0.7669945132732391,
+      "num_tokens": 36829085.0,
+      "step": 5500
+    },
+    {
+      "entropy": 2.168731119632721,
+      "epoch": 0.24666666666666667,
+      "grad_norm": 7.71875,
+      "learning_rate": 1.553356242840779e-05,
+      "loss": 0.7417,
+      "mean_token_accuracy": 0.7682712721824646,
+      "num_tokens": 36894333.0,
+      "step": 5550
+    },
+    {
+      "entropy": 2.209793448448181,
+      "epoch": 0.24888888888888888,
+      "grad_norm": 10.875,
+      "learning_rate": 1.548774341351661e-05,
+      "loss": 0.787,
+      "mean_token_accuracy": 0.7617891025543213,
+      "num_tokens": 36964522.0,
+      "step": 5600
+    },
+    {
+      "entropy": 2.2578547739982606,
+      "epoch": 0.2511111111111111,
+      "grad_norm": 7.59375,
+      "learning_rate": 1.5441924398625432e-05,
+      "loss": 0.8494,
+      "mean_token_accuracy": 0.7462784695625305,
+      "num_tokens": 37031831.0,
+      "step": 5650
+    },
+    {
+      "entropy": 2.2453298950195313,
+      "epoch": 0.25333333333333335,
+      "grad_norm": 7.09375,
+      "learning_rate": 1.539610538373425e-05,
+      "loss": 0.7898,
+      "mean_token_accuracy": 0.7597798085212708,
+      "num_tokens": 37096360.0,
+      "step": 5700
+    },
+    {
+      "entropy": 2.1777192115783692,
+      "epoch": 0.25555555555555554,
+      "grad_norm": 10.0,
+      "learning_rate": 1.535028636884307e-05,
+      "loss": 0.8147,
+      "mean_token_accuracy": 0.7537575364112854,
+      "num_tokens": 37161633.0,
+      "step": 5750
+    },
+    {
+      "entropy": 2.171867892742157,
+      "epoch": 0.2577777777777778,
+      "grad_norm": 7.71875,
+      "learning_rate": 1.530446735395189e-05,
+      "loss": 0.8001,
+      "mean_token_accuracy": 0.7575827407836914,
+      "num_tokens": 37228323.0,
+      "step": 5800
+    },
+    {
+      "entropy": 2.1830092740058897,
+      "epoch": 0.26,
+      "grad_norm": 8.8125,
+      "learning_rate": 1.525864833906071e-05,
+      "loss": 0.7722,
+      "mean_token_accuracy": 0.7659056377410889,
+      "num_tokens": 37295036.0,
+      "step": 5850
+    },
+    {
+      "entropy": 2.2247348022460938,
+      "epoch": 0.26222222222222225,
+      "grad_norm": 7.75,
+      "learning_rate": 1.5212829324169531e-05,
+      "loss": 0.786,
+      "mean_token_accuracy": 0.7639349389076233,
+      "num_tokens": 37363275.0,
+      "step": 5900
+    },
+    {
+      "entropy": 2.1976384329795837,
+      "epoch": 0.2644444444444444,
+      "grad_norm": 7.21875,
+      "learning_rate": 1.5167010309278351e-05,
+      "loss": 0.7832,
+      "mean_token_accuracy": 0.7637232232093811,
+      "num_tokens": 37429362.0,
+      "step": 5950
+    },
+    {
+      "entropy": 2.1915416312217713,
+      "epoch": 0.26666666666666666,
+      "grad_norm": 7.53125,
+      "learning_rate": 1.5121191294387172e-05,
+      "loss": 0.7911,
+      "mean_token_accuracy": 0.7595293188095092,
+      "num_tokens": 37496221.0,
+      "step": 6000
+    },
+    {
+      "entropy": 2.1550126123428344,
+      "epoch": 0.2688888888888889,
+      "grad_norm": 7.90625,
+      "learning_rate": 1.5075372279495992e-05,
+      "loss": 0.7866,
+      "mean_token_accuracy": 0.7628938972949981,
+      "num_tokens": 37563946.0,
+      "step": 6050
+    },
+    {
+      "entropy": 2.1578181266784666,
+      "epoch": 0.27111111111111114,
+      "grad_norm": 11.625,
+      "learning_rate": 1.5029553264604813e-05,
+      "loss": 0.8002,
+      "mean_token_accuracy": 0.7602377796173095,
+      "num_tokens": 37632154.0,
+      "step": 6100
+    },
+    {
+      "entropy": 2.1704964351654055,
+      "epoch": 0.2733333333333333,
+      "grad_norm": 7.09375,
+      "learning_rate": 1.4983734249713632e-05,
+      "loss": 0.7968,
+      "mean_token_accuracy": 0.7589374840259552,
+      "num_tokens": 37700403.0,
+      "step": 6150
+    },
+    {
+      "entropy": 2.1958404278755186,
+      "epoch": 0.27555555555555555,
+      "grad_norm": 7.71875,
+      "learning_rate": 1.4937915234822453e-05,
+      "loss": 0.8021,
+      "mean_token_accuracy": 0.7585455322265625,
+      "num_tokens": 37763144.0,
+      "step": 6200
+    },
+    {
+      "entropy": 2.2267961406707766,
+      "epoch": 0.2777777777777778,
+      "grad_norm": 9.375,
+      "learning_rate": 1.4892096219931273e-05,
+      "loss": 0.8,
+      "mean_token_accuracy": 0.756108273267746,
+      "num_tokens": 37828258.0,
+      "step": 6250
+    },
+    {
+      "entropy": 2.2654725027084353,
+      "epoch": 0.28,
+      "grad_norm": 8.875,
+      "learning_rate": 1.4846277205040094e-05,
+      "loss": 0.8398,
+      "mean_token_accuracy": 0.7474746882915497,
+      "num_tokens": 37894712.0,
+      "step": 6300
+    },
+    {
+      "entropy": 2.2119923210144044,
+      "epoch": 0.2822222222222222,
+      "grad_norm": 7.125,
+      "learning_rate": 1.4800458190148912e-05,
+      "loss": 0.8204,
+      "mean_token_accuracy": 0.7523449230194091,
+      "num_tokens": 37961686.0,
+      "step": 6350
+    },
+    {
+      "entropy": 2.176643466949463,
+      "epoch": 0.28444444444444444,
+      "grad_norm": 8.5625,
+      "learning_rate": 1.4754639175257731e-05,
+      "loss": 0.7743,
+      "mean_token_accuracy": 0.7683572423458099,
+      "num_tokens": 38028734.0,
+      "step": 6400
+    },
+    {
+      "entropy": 2.175390179157257,
+      "epoch": 0.2866666666666667,
+      "grad_norm": 7.03125,
+      "learning_rate": 1.4708820160366553e-05,
+      "loss": 0.8181,
+      "mean_token_accuracy": 0.7544300401210785,
+      "num_tokens": 38093844.0,
+      "step": 6450
+    },
+    {
+      "entropy": 2.2051888942718505,
+      "epoch": 0.28888888888888886,
+      "grad_norm": 6.5,
+      "learning_rate": 1.4663001145475372e-05,
+      "loss": 0.8126,
+      "mean_token_accuracy": 0.7564982330799103,
+      "num_tokens": 38159450.0,
+      "step": 6500
+    },
+    {
+      "entropy": 2.217521970272064,
+      "epoch": 0.2911111111111111,
+      "grad_norm": 7.90625,
+      "learning_rate": 1.4617182130584193e-05,
+      "loss": 0.816,
+      "mean_token_accuracy": 0.7553452157974243,
+      "num_tokens": 38224878.0,
+      "step": 6550
+    },
+    {
+      "entropy": 2.190767750740051,
+      "epoch": 0.29333333333333333,
+      "grad_norm": 8.875,
+      "learning_rate": 1.4571363115693013e-05,
+      "loss": 0.8144,
+      "mean_token_accuracy": 0.7556489539146424,
+      "num_tokens": 38291865.0,
+      "step": 6600
+    },
+    {
+      "entropy": 2.2169429779052736,
+      "epoch": 0.29555555555555557,
+      "grad_norm": 7.0,
+      "learning_rate": 1.4525544100801834e-05,
+      "loss": 0.8167,
+      "mean_token_accuracy": 0.7550825190544128,
+      "num_tokens": 38360809.0,
+      "step": 6650
+    },
+    {
+      "entropy": 2.2201628613471986,
+      "epoch": 0.29777777777777775,
+      "grad_norm": 8.375,
+      "learning_rate": 1.4479725085910653e-05,
+      "loss": 0.8098,
+      "mean_token_accuracy": 0.7543143463134766,
+      "num_tokens": 38428799.0,
+      "step": 6700
+    },
+    {
+      "entropy": 2.198468723297119,
+      "epoch": 0.3,
+      "grad_norm": 7.59375,
+      "learning_rate": 1.4433906071019475e-05,
+      "loss": 0.841,
+      "mean_token_accuracy": 0.7505165827274323,
+      "num_tokens": 38496550.0,
+      "step": 6750
+    },
+    {
+      "entropy": 2.1353413486480712,
+      "epoch": 0.3022222222222222,
+      "grad_norm": 7.625,
+      "learning_rate": 1.4388087056128294e-05,
+      "loss": 0.794,
+      "mean_token_accuracy": 0.7610986518859864,
+      "num_tokens": 38564172.0,
+      "step": 6800
+    },
+    {
+      "entropy": 2.249865219593048,
+      "epoch": 0.30444444444444446,
+      "grad_norm": 7.5,
+      "learning_rate": 1.4342268041237115e-05,
+      "loss": 0.8153,
+      "mean_token_accuracy": 0.7510960531234742,
+      "num_tokens": 38630875.0,
+      "step": 6850
+    },
+    {
+      "entropy": 2.1702062487602234,
+      "epoch": 0.30666666666666664,
+      "grad_norm": 10.6875,
+      "learning_rate": 1.4296449026345935e-05,
+      "loss": 0.7737,
+      "mean_token_accuracy": 0.763483716249466,
+      "num_tokens": 38694435.0,
+      "step": 6900
+    },
+    {
+      "entropy": 2.2601536536216735,
+      "epoch": 0.3088888888888889,
+      "grad_norm": 10.375,
+      "learning_rate": 1.4250630011454756e-05,
+      "loss": 0.8578,
+      "mean_token_accuracy": 0.7417992842197418,
+      "num_tokens": 38759789.0,
+      "step": 6950
+    },
+    {
+      "entropy": 2.276434664726257,
+      "epoch": 0.3111111111111111,
+      "grad_norm": 6.8125,
+      "learning_rate": 1.4204810996563575e-05,
+      "loss": 0.7678,
+      "mean_token_accuracy": 0.7661231434345246,
+      "num_tokens": 38826171.0,
+      "step": 7000
+    },
+    {
+      "entropy": 2.2015094637870787,
+      "epoch": 0.31333333333333335,
+      "grad_norm": 7.9375,
+      "learning_rate": 1.4158991981672395e-05,
+      "loss": 0.8136,
+      "mean_token_accuracy": 0.7545084583759308,
+      "num_tokens": 38890825.0,
+      "step": 7050
+    },
+    {
+      "entropy": 2.1954299902915952,
+      "epoch": 0.31555555555555553,
+      "grad_norm": 7.875,
+      "learning_rate": 1.4113172966781214e-05,
+      "loss": 0.8102,
+      "mean_token_accuracy": 0.7536708974838257,
+      "num_tokens": 38957597.0,
+      "step": 7100
+    },
+    {
+      "entropy": 2.1553618788719175,
+      "epoch": 0.31777777777777777,
+      "grad_norm": 10.625,
+      "learning_rate": 1.4067353951890036e-05,
+      "loss": 0.806,
+      "mean_token_accuracy": 0.756889488697052,
+      "num_tokens": 39024094.0,
+      "step": 7150
+    },
+    {
+      "entropy": 2.1930714321136473,
+      "epoch": 0.32,
+      "grad_norm": 7.34375,
+      "learning_rate": 1.4021534936998855e-05,
+      "loss": 0.8191,
+      "mean_token_accuracy": 0.7526458752155304,
+      "num_tokens": 39088926.0,
+      "step": 7200
+    },
+    {
+      "entropy": 2.1648390030860902,
+      "epoch": 0.32222222222222224,
+      "grad_norm": 6.40625,
+      "learning_rate": 1.3975715922107675e-05,
+      "loss": 0.8018,
+      "mean_token_accuracy": 0.7559799265861511,
+      "num_tokens": 39155958.0,
+      "step": 7250
+    },
+    {
+      "entropy": 2.163375446796417,
+      "epoch": 0.3244444444444444,
+      "grad_norm": 7.8125,
+      "learning_rate": 1.3929896907216496e-05,
+      "loss": 0.8065,
+      "mean_token_accuracy": 0.7592847204208374,
+      "num_tokens": 39224450.0,
+      "step": 7300
+    },
+    {
+      "entropy": 2.1846578884124757,
+      "epoch": 0.32666666666666666,
+      "grad_norm": 8.0625,
+      "learning_rate": 1.3884077892325315e-05,
+      "loss": 0.7933,
+      "mean_token_accuracy": 0.761990624666214,
+      "num_tokens": 39288896.0,
+      "step": 7350
+    },
+    {
+      "entropy": 2.2050635480880736,
+      "epoch": 0.3288888888888889,
+      "grad_norm": 6.84375,
+      "learning_rate": 1.3838258877434137e-05,
+      "loss": 0.8317,
+      "mean_token_accuracy": 0.7510676419734955,
+      "num_tokens": 39354824.0,
+      "step": 7400
+    },
+    {
+      "entropy": 2.1733724021911622,
+      "epoch": 0.33111111111111113,
+      "grad_norm": 10.5,
+      "learning_rate": 1.3792439862542956e-05,
+      "loss": 0.8164,
+      "mean_token_accuracy": 0.755161405801773,
+      "num_tokens": 39423233.0,
+      "step": 7450
+    },
+    {
+      "entropy": 2.1624359107017517,
+      "epoch": 0.3333333333333333,
+      "grad_norm": 8.875,
+      "learning_rate": 1.3746620847651777e-05,
+      "loss": 0.8105,
+      "mean_token_accuracy": 0.7563036870956421,
+      "num_tokens": 39488844.0,
+      "step": 7500
+    },
+    {
+      "entropy": 2.2275446701049804,
+      "epoch": 0.33555555555555555,
+      "grad_norm": 8.625,
+      "learning_rate": 1.3700801832760597e-05,
+      "loss": 0.8569,
+      "mean_token_accuracy": 0.7444409322738648,
+      "num_tokens": 39557151.0,
+      "step": 7550
+    },
+    {
+      "entropy": 2.160377633571625,
+      "epoch": 0.3377777777777778,
+      "grad_norm": 8.0,
+      "learning_rate": 1.3654982817869418e-05,
+      "loss": 0.7943,
+      "mean_token_accuracy": 0.7591172540187836,
+      "num_tokens": 39622145.0,
+      "step": 7600
+    },
+    {
+      "entropy": 2.1659918451309204,
+      "epoch": 0.34,
+      "grad_norm": 8.8125,
+      "learning_rate": 1.3609163802978237e-05,
+      "loss": 0.8312,
+      "mean_token_accuracy": 0.7525255751609802,
+      "num_tokens": 39692633.0,
+      "step": 7650
+    },
+    {
+      "entropy": 2.15194406747818,
+      "epoch": 0.3422222222222222,
+      "grad_norm": 9.8125,
+      "learning_rate": 1.3563344788087059e-05,
+      "loss": 0.7778,
+      "mean_token_accuracy": 0.7636078727245331,
+      "num_tokens": 39759159.0,
+      "step": 7700
+    },
+    {
+      "entropy": 2.217858202457428,
+      "epoch": 0.34444444444444444,
+      "grad_norm": 7.09375,
+      "learning_rate": 1.3517525773195876e-05,
+      "loss": 0.8241,
+      "mean_token_accuracy": 0.7522653472423554,
+      "num_tokens": 39826283.0,
+      "step": 7750
+    },
+    {
+      "entropy": 2.1696537256240847,
+      "epoch": 0.3466666666666667,
+      "grad_norm": 9.0625,
+      "learning_rate": 1.3471706758304698e-05,
+      "loss": 0.815,
+      "mean_token_accuracy": 0.7559839642047882,
+      "num_tokens": 39895721.0,
+      "step": 7800
+    },
+    {
+      "entropy": 2.1727401876449584,
+      "epoch": 0.3488888888888889,
+      "grad_norm": 8.375,
+      "learning_rate": 1.3425887743413517e-05,
+      "loss": 0.8366,
+      "mean_token_accuracy": 0.7508484077453613,
+      "num_tokens": 39963266.0,
+      "step": 7850
+    },
+    {
+      "entropy": 2.1629985857009886,
+      "epoch": 0.3511111111111111,
+      "grad_norm": 7.25,
+      "learning_rate": 1.3380068728522338e-05,
+      "loss": 0.782,
+      "mean_token_accuracy": 0.761427184343338,
+      "num_tokens": 40031913.0,
+      "step": 7900
+    },
+    {
+      "entropy": 2.157064917087555,
+      "epoch": 0.35333333333333333,
+      "grad_norm": 14.125,
+      "learning_rate": 1.3334249713631158e-05,
+      "loss": 0.7747,
+      "mean_token_accuracy": 0.7661045718193055,
+      "num_tokens": 40098930.0,
+      "step": 7950
+    },
+    {
+      "entropy": 2.0971979546546935,
+      "epoch": 0.35555555555555557,
+      "grad_norm": 8.75,
+      "learning_rate": 1.3288430698739979e-05,
+      "loss": 0.8186,
+      "mean_token_accuracy": 0.7531020665168762,
+      "num_tokens": 40165756.0,
+      "step": 8000
+    },
+    {
+      "entropy": 2.1666273212432863,
+      "epoch": 0.35777777777777775,
+      "grad_norm": 7.03125,
+      "learning_rate": 1.3242611683848798e-05,
+      "loss": 0.842,
+      "mean_token_accuracy": 0.7474488174915314,
+      "num_tokens": 40233071.0,
+      "step": 8050
+    },
+    {
+      "entropy": 2.2938923120498655,
+      "epoch": 0.36,
+      "grad_norm": 11.875,
+      "learning_rate": 1.3196792668957618e-05,
+      "loss": 0.8804,
+      "mean_token_accuracy": 0.7407302105426788,
+      "num_tokens": 40297749.0,
+      "step": 8100
+    },
+    {
+      "entropy": 2.2303052496910096,
+      "epoch": 0.3622222222222222,
+      "grad_norm": 8.125,
+      "learning_rate": 1.3150973654066439e-05,
+      "loss": 0.8033,
+      "mean_token_accuracy": 0.7578864073753357,
+      "num_tokens": 40366213.0,
+      "step": 8150
+    },
+    {
+      "entropy": 2.2161930847167968,
+      "epoch": 0.36444444444444446,
+      "grad_norm": 9.6875,
+      "learning_rate": 1.3105154639175259e-05,
+      "loss": 0.7947,
+      "mean_token_accuracy": 0.7593982243537902,
+      "num_tokens": 40433288.0,
+      "step": 8200
+    },
+    {
+      "entropy": 2.1040697479248047,
+      "epoch": 0.36666666666666664,
+      "grad_norm": 7.21875,
+      "learning_rate": 1.305933562428408e-05,
+      "loss": 0.7959,
+      "mean_token_accuracy": 0.7588922154903411,
+      "num_tokens": 40499720.0,
+      "step": 8250
+    },
+    {
+      "entropy": 2.1772580099105836,
+      "epoch": 0.3688888888888889,
+      "grad_norm": 7.25,
+      "learning_rate": 1.30135166093929e-05,
+      "loss": 0.8284,
+      "mean_token_accuracy": 0.7538638985157013,
+      "num_tokens": 40565376.0,
+      "step": 8300
+    },
+    {
+      "entropy": 2.128792498111725,
+      "epoch": 0.3711111111111111,
+      "grad_norm": 8.6875,
+      "learning_rate": 1.296769759450172e-05,
+      "loss": 0.7672,
+      "mean_token_accuracy": 0.7641964781284333,
+      "num_tokens": 40631675.0,
+      "step": 8350
+    },
+    {
+      "entropy": 2.1719995403289794,
+      "epoch": 0.37333333333333335,
+      "grad_norm": 9.25,
+      "learning_rate": 1.2921878579610538e-05,
+      "loss": 0.801,
+      "mean_token_accuracy": 0.76089714884758,
+      "num_tokens": 40697335.0,
+      "step": 8400
+    },
+    {
+      "entropy": 2.156131479740143,
+      "epoch": 0.37555555555555553,
+      "grad_norm": 8.75,
+      "learning_rate": 1.287605956471936e-05,
+      "loss": 0.8384,
+      "mean_token_accuracy": 0.7503017449378967,
+      "num_tokens": 40764922.0,
+      "step": 8450
+    },
+    {
+      "entropy": 2.284397015571594,
+      "epoch": 0.37777777777777777,
+      "grad_norm": 9.25,
+      "learning_rate": 1.2830240549828179e-05,
+      "loss": 0.8418,
+      "mean_token_accuracy": 0.7493084251880646,
+      "num_tokens": 40828443.0,
+      "step": 8500
+    },
+    {
+      "entropy": 2.147591190338135,
+      "epoch": 0.38,
+      "grad_norm": 8.625,
+      "learning_rate": 1.2784421534937e-05,
+      "loss": 0.8173,
+      "mean_token_accuracy": 0.756924353837967,
+      "num_tokens": 40894414.0,
+      "step": 8550
+    },
+    {
+      "entropy": 2.142817313671112,
+      "epoch": 0.38222222222222224,
+      "grad_norm": 8.5625,
+      "learning_rate": 1.273860252004582e-05,
+      "loss": 0.7962,
+      "mean_token_accuracy": 0.7598996949195862,
+      "num_tokens": 40962274.0,
+      "step": 8600
+    },
+    {
+      "entropy": 2.1330949759483335,
+      "epoch": 0.3844444444444444,
+      "grad_norm": 7.0625,
+      "learning_rate": 1.269278350515464e-05,
+      "loss": 0.7939,
+      "mean_token_accuracy": 0.7612218356132507,
+      "num_tokens": 41030193.0,
+      "step": 8650
+    },
+    {
+      "entropy": 2.166394736766815,
+      "epoch": 0.38666666666666666,
+      "grad_norm": 7.40625,
+      "learning_rate": 1.264696449026346e-05,
+      "loss": 0.8336,
+      "mean_token_accuracy": 0.7500250363349914,
+      "num_tokens": 41099766.0,
+      "step": 8700
+    },
+    {
+      "entropy": 2.2078594040870665,
+      "epoch": 0.3888888888888889,
+      "grad_norm": 8.625,
+      "learning_rate": 1.2601145475372281e-05,
+      "loss": 0.803,
+      "mean_token_accuracy": 0.7575648665428162,
+      "num_tokens": 41164744.0,
+      "step": 8750
+    },
+    {
+      "entropy": 2.2065504598617554,
+      "epoch": 0.39111111111111113,
+      "grad_norm": 6.9375,
+      "learning_rate": 1.2555326460481101e-05,
+      "loss": 0.816,
+      "mean_token_accuracy": 0.7564821767807007,
+      "num_tokens": 41232194.0,
+      "step": 8800
+    },
+    {
+      "entropy": 2.1662226915359497,
+      "epoch": 0.3933333333333333,
+      "grad_norm": 8.0625,
+      "learning_rate": 1.2509507445589922e-05,
+      "loss": 0.8317,
+      "mean_token_accuracy": 0.7503245520591736,
+      "num_tokens": 41297353.0,
+      "step": 8850
+    },
+    {
+      "entropy": 2.236915967464447,
+      "epoch": 0.39555555555555555,
+      "grad_norm": 9.875,
+      "learning_rate": 1.2463688430698742e-05,
+      "loss": 0.7977,
+      "mean_token_accuracy": 0.759481954574585,
+      "num_tokens": 41361134.0,
+      "step": 8900
+    },
+    {
+      "entropy": 2.1871555137634275,
+      "epoch": 0.3977777777777778,
+      "grad_norm": 7.90625,
+      "learning_rate": 1.2417869415807561e-05,
+      "loss": 0.8034,
+      "mean_token_accuracy": 0.757278825044632,
+      "num_tokens": 41427995.0,
+      "step": 8950
+    },
+    {
+      "entropy": 2.2592362785339355,
+      "epoch": 0.4,
+      "grad_norm": 9.875,
+      "learning_rate": 1.2372050400916382e-05,
+      "loss": 0.8186,
+      "mean_token_accuracy": 0.7548012447357177,
+      "num_tokens": 41494201.0,
+      "step": 9000
+    },
+    {
+      "entropy": 2.196383099555969,
+      "epoch": 0.4022222222222222,
+      "grad_norm": 7.625,
+      "learning_rate": 1.2326231386025202e-05,
+      "loss": 0.8138,
+      "mean_token_accuracy": 0.7561362779140473,
+      "num_tokens": 41559989.0,
+      "step": 9050
+    },
+    {
+      "entropy": 2.1516851663589476,
+      "epoch": 0.40444444444444444,
+      "grad_norm": 13.75,
+      "learning_rate": 1.2280412371134021e-05,
+      "loss": 0.7951,
+      "mean_token_accuracy": 0.7594835031032562,
+      "num_tokens": 41626379.0,
+      "step": 9100
+    },
+    {
+      "entropy": 2.1745046091079714,
+      "epoch": 0.4066666666666667,
+      "grad_norm": 8.0,
+      "learning_rate": 1.223459335624284e-05,
+      "loss": 0.8048,
+      "mean_token_accuracy": 0.7599712920188904,
+      "num_tokens": 41692143.0,
+      "step": 9150
+    },
+    {
+      "entropy": 2.154158115386963,
+      "epoch": 0.4088888888888889,
+      "grad_norm": 8.25,
+      "learning_rate": 1.2188774341351662e-05,
+      "loss": 0.7877,
+      "mean_token_accuracy": 0.7622714912891388,
+      "num_tokens": 41759805.0,
+      "step": 9200
+    },
+    {
+      "entropy": 2.2480538749694823,
+      "epoch": 0.4111111111111111,
+      "grad_norm": 8.4375,
+      "learning_rate": 1.2142955326460481e-05,
+      "loss": 0.8052,
+      "mean_token_accuracy": 0.7560270345211029,
+      "num_tokens": 41823593.0,
+      "step": 9250
+    },
+    {
+      "entropy": 2.2491649985313416,
+      "epoch": 0.41333333333333333,
+      "grad_norm": 9.125,
+      "learning_rate": 1.2097136311569303e-05,
+      "loss": 0.8696,
+      "mean_token_accuracy": 0.7419601881504059,
+      "num_tokens": 41888809.0,
+      "step": 9300
+    },
+    {
+      "entropy": 2.258921926021576,
+      "epoch": 0.41555555555555557,
+      "grad_norm": 10.0625,
+      "learning_rate": 1.2051317296678122e-05,
+      "loss": 0.8289,
+      "mean_token_accuracy": 0.75118199467659,
+      "num_tokens": 41952730.0,
+      "step": 9350
+    },
+    {
+      "entropy": 2.163720915317535,
+      "epoch": 0.4177777777777778,
+      "grad_norm": 9.25,
+      "learning_rate": 1.2005498281786943e-05,
+      "loss": 0.8323,
+      "mean_token_accuracy": 0.7530957090854645,
+      "num_tokens": 42019573.0,
+      "step": 9400
+    },
+    {
+      "entropy": 2.2212659883499146,
+      "epoch": 0.42,
+      "grad_norm": 7.40625,
+      "learning_rate": 1.1959679266895763e-05,
+      "loss": 0.8361,
+      "mean_token_accuracy": 0.7518267011642457,
+      "num_tokens": 42088291.0,
+      "step": 9450
+    },
+    {
+      "entropy": 2.2050712847709657,
+      "epoch": 0.4222222222222222,
+      "grad_norm": 9.75,
+      "learning_rate": 1.1913860252004584e-05,
+      "loss": 0.8456,
+      "mean_token_accuracy": 0.7448002827167511,
+      "num_tokens": 42158120.0,
+      "step": 9500
+    },
+    {
+      "entropy": 2.216916351318359,
+      "epoch": 0.42444444444444446,
+      "grad_norm": 6.8125,
+      "learning_rate": 1.1868041237113403e-05,
+      "loss": 0.8415,
+      "mean_token_accuracy": 0.7452654683589935,
+      "num_tokens": 42226742.0,
+      "step": 9550
+    },
+    {
+      "entropy": 2.1955419325828553,
+      "epoch": 0.4266666666666667,
+      "grad_norm": 8.1875,
+      "learning_rate": 1.1822222222222225e-05,
+      "loss": 0.8166,
+      "mean_token_accuracy": 0.7562698805332184,
+      "num_tokens": 42292822.0,
+      "step": 9600
+    },
+    {
+      "entropy": 2.2368349695205687,
+      "epoch": 0.4288888888888889,
+      "grad_norm": 7.34375,
+      "learning_rate": 1.1776403207331044e-05,
+      "loss": 0.8089,
+      "mean_token_accuracy": 0.757049810886383,
+      "num_tokens": 42356424.0,
+      "step": 9650
+    },
+    {
+      "entropy": 2.178975234031677,
+      "epoch": 0.4311111111111111,
+      "grad_norm": 8.375,
+      "learning_rate": 1.1730584192439865e-05,
+      "loss": 0.8438,
+      "mean_token_accuracy": 0.7456352376937866,
+      "num_tokens": 42422979.0,
+      "step": 9700
+    },
+    {
+      "entropy": 2.0868096995353698,
+      "epoch": 0.43333333333333335,
+      "grad_norm": 7.53125,
+      "learning_rate": 1.1684765177548683e-05,
+      "loss": 0.8235,
+      "mean_token_accuracy": 0.7529381263256073,
+      "num_tokens": 42490375.0,
+      "step": 9750
+    },
+    {
+      "entropy": 2.147622332572937,
+      "epoch": 0.43555555555555553,
+      "grad_norm": 9.75,
+      "learning_rate": 1.1638946162657503e-05,
+      "loss": 0.7728,
+      "mean_token_accuracy": 0.7673224699497223,
+      "num_tokens": 42557143.0,
+      "step": 9800
+    },
+    {
+      "entropy": 2.114510886669159,
+      "epoch": 0.43777777777777777,
+      "grad_norm": 9.375,
+      "learning_rate": 1.1593127147766324e-05,
+      "loss": 0.8238,
+      "mean_token_accuracy": 0.7501462066173553,
+      "num_tokens": 42626026.0,
+      "step": 9850
+    },
+    {
+      "entropy": 2.133998863697052,
+      "epoch": 0.44,
+      "grad_norm": 7.4375,
+      "learning_rate": 1.1547308132875143e-05,
+      "loss": 0.8005,
+      "mean_token_accuracy": 0.7565454721450806,
+      "num_tokens": 42688911.0,
+      "step": 9900
+    },
+    {
+      "entropy": 2.146713092327118,
+      "epoch": 0.44222222222222224,
+      "grad_norm": 8.625,
+      "learning_rate": 1.1501489117983964e-05,
+      "loss": 0.801,
+      "mean_token_accuracy": 0.7562527394294739,
+      "num_tokens": 42756274.0,
+      "step": 9950
+    },
+    {
+      "entropy": 2.205234091281891,
+      "epoch": 0.4444444444444444,
+      "grad_norm": 6.875,
+      "learning_rate": 1.1455670103092784e-05,
+      "loss": 0.8242,
+      "mean_token_accuracy": 0.7537857472896576,
+      "num_tokens": 42825342.0,
+      "step": 10000
+    },
+    {
+      "epoch": 0.4444444444444444,
+      "eval_entropy": 2.0749987959861755,
+      "eval_loss": 0.8938310146331787,
+      "eval_mean_token_accuracy": 0.7446183115243912,
+      "eval_num_tokens": 42825342.0,
+      "eval_runtime": 7.8855,
+      "eval_samples_per_second": 1.649,
+      "eval_steps_per_second": 0.507,
+      "step": 10000
+    },
+    {
+      "entropy": 2.232636868953705,
+      "epoch": 0.44666666666666666,
+      "grad_norm": 6.3125,
+      "learning_rate": 1.1409851088201605e-05,
+      "loss": 0.8088,
+      "mean_token_accuracy": 0.7571802771091461,
+      "num_tokens": 42892615.0,
+      "step": 10050
+    },
+    {
+      "entropy": 2.1724118828773498,
+      "epoch": 0.4488888888888889,
+      "grad_norm": 7.4375,
+      "learning_rate": 1.1364032073310425e-05,
+      "loss": 0.7938,
+      "mean_token_accuracy": 0.7575406277179718,
+      "num_tokens": 42961116.0,
+      "step": 10100
+    },
+    {
+      "entropy": 2.2688467741012572,
+      "epoch": 0.45111111111111113,
+      "grad_norm": 8.25,
+      "learning_rate": 1.1318213058419246e-05,
+      "loss": 0.8109,
+      "mean_token_accuracy": 0.7592162156105041,
+      "num_tokens": 43026002.0,
+      "step": 10150
+    },
+    {
+      "entropy": 2.189846224784851,
+      "epoch": 0.4533333333333333,
+      "grad_norm": 7.4375,
+      "learning_rate": 1.1272394043528065e-05,
+      "loss": 0.8393,
+      "mean_token_accuracy": 0.750606085062027,
+      "num_tokens": 43096450.0,
+      "step": 10200
+    },
+    {
+      "entropy": 2.1485534167289733,
+      "epoch": 0.45555555555555555,
+      "grad_norm": 6.96875,
+      "learning_rate": 1.1226575028636886e-05,
+      "loss": 0.8005,
+      "mean_token_accuracy": 0.7586869549751282,
+      "num_tokens": 43164974.0,
+      "step": 10250
+    },
+    {
+      "entropy": 2.114973647594452,
+      "epoch": 0.4577777777777778,
+      "grad_norm": 7.71875,
+      "learning_rate": 1.1180756013745706e-05,
+      "loss": 0.81,
+      "mean_token_accuracy": 0.7530654954910279,
+      "num_tokens": 43230676.0,
+      "step": 10300
+    },
+    {
+      "entropy": 2.210615482330322,
+      "epoch": 0.46,
+      "grad_norm": 11.6875,
+      "learning_rate": 1.1134936998854527e-05,
+      "loss": 0.8133,
+      "mean_token_accuracy": 0.7568694865703582,
+      "num_tokens": 43295393.0,
+      "step": 10350
+    },
+    {
+      "entropy": 2.1622642707824706,
+      "epoch": 0.4622222222222222,
+      "grad_norm": 7.03125,
+      "learning_rate": 1.1089117983963347e-05,
+      "loss": 0.7942,
+      "mean_token_accuracy": 0.7580956876277923,
+      "num_tokens": 43361602.0,
+      "step": 10400
+    },
+    {
+      "entropy": 2.164878299236298,
+      "epoch": 0.46444444444444444,
+      "grad_norm": 6.6875,
+      "learning_rate": 1.1043298969072164e-05,
+      "loss": 0.7759,
+      "mean_token_accuracy": 0.7644217216968536,
+      "num_tokens": 43427335.0,
+      "step": 10450
+    },
+    {
+      "entropy": 2.1177299284934996,
+      "epoch": 0.4666666666666667,
+      "grad_norm": 10.0,
+      "learning_rate": 1.0997479954180986e-05,
+      "loss": 0.85,
+      "mean_token_accuracy": 0.7469760966300965,
+      "num_tokens": 43494368.0,
+      "step": 10500
+    },
+    {
+      "entropy": 2.151867859363556,
+      "epoch": 0.4688888888888889,
+      "grad_norm": 9.25,
+      "learning_rate": 1.0951660939289805e-05,
+      "loss": 0.8053,
+      "mean_token_accuracy": 0.7571006786823272,
+      "num_tokens": 43560267.0,
+      "step": 10550
+    },
+    {
+      "entropy": 2.1284614849090575,
+      "epoch": 0.4711111111111111,
+      "grad_norm": 6.6875,
+      "learning_rate": 1.0905841924398626e-05,
+      "loss": 0.8309,
+      "mean_token_accuracy": 0.752720388174057,
+      "num_tokens": 43626812.0,
+      "step": 10600
+    },
+    {
+      "entropy": 2.1522640681266783,
+      "epoch": 0.47333333333333333,
+      "grad_norm": 8.4375,
+      "learning_rate": 1.0860022909507446e-05,
+      "loss": 0.8633,
+      "mean_token_accuracy": 0.7454240775108337,
+      "num_tokens": 43693459.0,
+      "step": 10650
+    },
+    {
+      "entropy": 2.1508279252052307,
+      "epoch": 0.47555555555555556,
+      "grad_norm": 7.78125,
+      "learning_rate": 1.0814203894616267e-05,
+      "loss": 0.7701,
+      "mean_token_accuracy": 0.767644681930542,
+      "num_tokens": 43758441.0,
+      "step": 10700
+    },
+    {
+      "entropy": 2.1406164813041686,
+      "epoch": 0.4777777777777778,
+      "grad_norm": 8.3125,
+      "learning_rate": 1.0768384879725086e-05,
+      "loss": 0.8299,
+      "mean_token_accuracy": 0.7507808887958527,
+      "num_tokens": 43827587.0,
+      "step": 10750
+    },
+    {
+      "entropy": 2.0773234295845033,
+      "epoch": 0.48,
+      "grad_norm": 6.65625,
+      "learning_rate": 1.0722565864833908e-05,
+      "loss": 0.8059,
+      "mean_token_accuracy": 0.7575164914131165,
+      "num_tokens": 43893454.0,
+      "step": 10800
+    },
+    {
+      "entropy": 2.0964699006080627,
+      "epoch": 0.4822222222222222,
+      "grad_norm": 7.09375,
+      "learning_rate": 1.0676746849942727e-05,
+      "loss": 0.8538,
+      "mean_token_accuracy": 0.7474709522724151,
+      "num_tokens": 43961221.0,
+      "step": 10850
+    },
+    {
+      "entropy": 2.127323613166809,
+      "epoch": 0.48444444444444446,
+      "grad_norm": 8.25,
+      "learning_rate": 1.0630927835051548e-05,
+      "loss": 0.823,
+      "mean_token_accuracy": 0.7566704392433167,
+      "num_tokens": 44030179.0,
+      "step": 10900
+    },
+    {
+      "entropy": 2.1277793073654174,
+      "epoch": 0.4866666666666667,
+      "grad_norm": 8.0,
+      "learning_rate": 1.0585108820160368e-05,
+      "loss": 0.7951,
+      "mean_token_accuracy": 0.7619584739208222,
+      "num_tokens": 44096190.0,
+      "step": 10950
+    },
+    {
+      "entropy": 2.100730609893799,
+      "epoch": 0.4888888888888889,
+      "grad_norm": 7.5625,
+      "learning_rate": 1.0539289805269189e-05,
+      "loss": 0.8355,
+      "mean_token_accuracy": 0.7507687473297119,
+      "num_tokens": 44164149.0,
+      "step": 11000
+    },
+    {
+      "entropy": 2.1630342602729797,
+      "epoch": 0.4911111111111111,
+      "grad_norm": 9.0,
+      "learning_rate": 1.0493470790378008e-05,
+      "loss": 0.8394,
+      "mean_token_accuracy": 0.7504077112674713,
+      "num_tokens": 44232826.0,
+      "step": 11050
+    },
+    {
+      "entropy": 2.135989320278168,
+      "epoch": 0.49333333333333335,
+      "grad_norm": 7.25,
+      "learning_rate": 1.044765177548683e-05,
+      "loss": 0.826,
+      "mean_token_accuracy": 0.7503847754001618,
+      "num_tokens": 44300785.0,
+      "step": 11100
+    },
+    {
+      "entropy": 2.1245554232597352,
+      "epoch": 0.4955555555555556,
+      "grad_norm": 13.125,
+      "learning_rate": 1.0401832760595647e-05,
+      "loss": 0.8276,
+      "mean_token_accuracy": 0.7498899948596954,
+      "num_tokens": 44366272.0,
+      "step": 11150
+    },
+    {
+      "entropy": 2.0881606268882753,
+      "epoch": 0.49777777777777776,
+      "grad_norm": 8.5625,
+      "learning_rate": 1.0356013745704467e-05,
+      "loss": 0.8272,
+      "mean_token_accuracy": 0.7556144452095032,
+      "num_tokens": 44432287.0,
+      "step": 11200
+    },
+    {
+      "entropy": 2.1587840700149536,
+      "epoch": 0.5,
+      "grad_norm": 22.625,
+      "learning_rate": 1.0310194730813288e-05,
+      "loss": 0.8169,
+      "mean_token_accuracy": 0.7587198996543885,
+      "num_tokens": 44497099.0,
+      "step": 11250
+    },
+    {
+      "entropy": 2.059075405597687,
+      "epoch": 0.5022222222222222,
+      "grad_norm": 6.15625,
+      "learning_rate": 1.0264375715922108e-05,
+      "loss": 0.8128,
+      "mean_token_accuracy": 0.7540943920612335,
+      "num_tokens": 44568074.0,
+      "step": 11300
+    },
+    {
+      "entropy": 2.132368493080139,
+      "epoch": 0.5044444444444445,
+      "grad_norm": 5.65625,
+      "learning_rate": 1.0218556701030929e-05,
+      "loss": 0.7915,
+      "mean_token_accuracy": 0.7632870030403137,
+      "num_tokens": 44634710.0,
+      "step": 11350
+    },
+    {
+      "entropy": 2.075038847923279,
+      "epoch": 0.5066666666666667,
+      "grad_norm": 11.3125,
+      "learning_rate": 1.0172737686139748e-05,
+      "loss": 0.8615,
+      "mean_token_accuracy": 0.7465237331390381,
+      "num_tokens": 44705095.0,
+      "step": 11400
+    },
+    {
+      "entropy": 2.11560346364975,
+      "epoch": 0.5088888888888888,
+      "grad_norm": 7.3125,
+      "learning_rate": 1.012691867124857e-05,
+      "loss": 0.8016,
+      "mean_token_accuracy": 0.758952580690384,
+      "num_tokens": 44772217.0,
+      "step": 11450
+    },
+    {
+      "entropy": 2.1024736380577087,
+      "epoch": 0.5111111111111111,
+      "grad_norm": 13.5625,
+      "learning_rate": 1.0081099656357389e-05,
+      "loss": 0.7976,
+      "mean_token_accuracy": 0.7605179595947266,
+      "num_tokens": 44840575.0,
+      "step": 11500
+    },
+    {
+      "entropy": 2.145506045818329,
+      "epoch": 0.5133333333333333,
+      "grad_norm": 7.03125,
+      "learning_rate": 1.003528064146621e-05,
+      "loss": 0.8572,
+      "mean_token_accuracy": 0.7433122813701629,
+      "num_tokens": 44906721.0,
+      "step": 11550
+    },
+    {
+      "entropy": 2.1501906871795655,
+      "epoch": 0.5155555555555555,
+      "grad_norm": 7.9375,
+      "learning_rate": 9.98946162657503e-06,
+      "loss": 0.8199,
+      "mean_token_accuracy": 0.7566520750522614,
+      "num_tokens": 44973401.0,
+      "step": 11600
+    },
+    {
+      "entropy": 2.180145356655121,
+      "epoch": 0.5177777777777778,
+      "grad_norm": 7.96875,
+      "learning_rate": 9.943642611683849e-06,
+      "loss": 0.8497,
+      "mean_token_accuracy": 0.747704974412918,
+      "num_tokens": 45038114.0,
+      "step": 11650
+    },
+    {
+      "entropy": 2.1444980192184446,
+      "epoch": 0.52,
+      "grad_norm": 7.3125,
+      "learning_rate": 9.897823596792669e-06,
+      "loss": 0.8054,
+      "mean_token_accuracy": 0.7571915662288666,
+      "num_tokens": 45105740.0,
+      "step": 11700
+    },
+    {
+      "entropy": 2.1226493763923644,
+      "epoch": 0.5222222222222223,
+      "grad_norm": 8.6875,
+      "learning_rate": 9.85200458190149e-06,
+      "loss": 0.8543,
+      "mean_token_accuracy": 0.74732057929039,
+      "num_tokens": 45173497.0,
+      "step": 11750
+    },
+    {
+      "entropy": 2.105964336395264,
+      "epoch": 0.5244444444444445,
+      "grad_norm": 6.90625,
+      "learning_rate": 9.80618556701031e-06,
+      "loss": 0.8664,
+      "mean_token_accuracy": 0.7433609414100647,
+      "num_tokens": 45240881.0,
+      "step": 11800
+    },
+    {
+      "entropy": 2.1655562567710875,
+      "epoch": 0.5266666666666666,
+      "grad_norm": 7.78125,
+      "learning_rate": 9.76036655211913e-06,
+      "loss": 0.8134,
+      "mean_token_accuracy": 0.7605512762069702,
+      "num_tokens": 45306709.0,
+      "step": 11850
+    },
+    {
+      "entropy": 2.115200798511505,
+      "epoch": 0.5288888888888889,
+      "grad_norm": 8.125,
+      "learning_rate": 9.71454753722795e-06,
+      "loss": 0.7942,
+      "mean_token_accuracy": 0.76005859375,
+      "num_tokens": 45373791.0,
+      "step": 11900
+    },
+    {
+      "entropy": 2.1086077737808226,
+      "epoch": 0.5311111111111111,
+      "grad_norm": 8.375,
+      "learning_rate": 9.668728522336771e-06,
+      "loss": 0.832,
+      "mean_token_accuracy": 0.7541840577125549,
+      "num_tokens": 45442329.0,
+      "step": 11950
+    },
+    {
+      "entropy": 2.1453971552848814,
+      "epoch": 0.5333333333333333,
+      "grad_norm": 9.625,
+      "learning_rate": 9.62290950744559e-06,
+      "loss": 0.8618,
+      "mean_token_accuracy": 0.74840935587883,
+      "num_tokens": 45509133.0,
+      "step": 12000
+    },
+    {
+      "entropy": 2.185004472732544,
+      "epoch": 0.5355555555555556,
+      "grad_norm": 6.46875,
+      "learning_rate": 9.57709049255441e-06,
+      "loss": 0.8269,
+      "mean_token_accuracy": 0.75218052983284,
+      "num_tokens": 45574414.0,
+      "step": 12050
+    },
+    {
+      "entropy": 2.0952596497535705,
+      "epoch": 0.5377777777777778,
+      "grad_norm": 7.59375,
+      "learning_rate": 9.531271477663231e-06,
+      "loss": 0.8312,
+      "mean_token_accuracy": 0.7515332329273224,
+      "num_tokens": 45645759.0,
+      "step": 12100
+    },
+    {
+      "entropy": 2.1197858357429507,
+      "epoch": 0.54,
+      "grad_norm": 8.0,
+      "learning_rate": 9.48545246277205e-06,
+      "loss": 0.8165,
+      "mean_token_accuracy": 0.7548327159881592,
+      "num_tokens": 45711167.0,
+      "step": 12150
+    },
+    {
+      "entropy": 2.0839271712303162,
+      "epoch": 0.5422222222222223,
+      "grad_norm": 7.34375,
+      "learning_rate": 9.439633447880872e-06,
+      "loss": 0.8007,
+      "mean_token_accuracy": 0.7602275168895721,
+      "num_tokens": 45781410.0,
+      "step": 12200
+    },
+    {
+      "entropy": 2.1347982597351076,
+      "epoch": 0.5444444444444444,
+      "grad_norm": 7.09375,
+      "learning_rate": 9.393814432989692e-06,
+      "loss": 0.8163,
+      "mean_token_accuracy": 0.752629029750824,
+      "num_tokens": 45845816.0,
+      "step": 12250
+    },
+    {
+      "entropy": 2.124971535205841,
+      "epoch": 0.5466666666666666,
+      "grad_norm": 8.1875,
+      "learning_rate": 9.347995418098513e-06,
+      "loss": 0.8166,
+      "mean_token_accuracy": 0.7571777141094208,
+      "num_tokens": 45911093.0,
+      "step": 12300
+    },
+    {
+      "entropy": 2.133582751750946,
+      "epoch": 0.5488888888888889,
+      "grad_norm": 7.8125,
+      "learning_rate": 9.302176403207332e-06,
+      "loss": 0.8724,
+      "mean_token_accuracy": 0.7431828391551971,
+      "num_tokens": 45980429.0,
+      "step": 12350
+    },
+    {
+      "entropy": 2.102145965099335,
+      "epoch": 0.5511111111111111,
+      "grad_norm": 6.8125,
+      "learning_rate": 9.256357388316152e-06,
+      "loss": 0.799,
+      "mean_token_accuracy": 0.7613983142375946,
+      "num_tokens": 46047747.0,
+      "step": 12400
+    },
+    {
+      "entropy": 2.19835401058197,
+      "epoch": 0.5533333333333333,
+      "grad_norm": 8.25,
+      "learning_rate": 9.210538373424973e-06,
+      "loss": 0.8433,
+      "mean_token_accuracy": 0.750613443851471,
+      "num_tokens": 46109196.0,
+      "step": 12450
+    },
+    {
+      "entropy": 2.192390067577362,
+      "epoch": 0.5555555555555556,
+      "grad_norm": 6.6875,
+      "learning_rate": 9.164719358533792e-06,
+      "loss": 0.8233,
+      "mean_token_accuracy": 0.7548783671855926,
+      "num_tokens": 46175945.0,
+      "step": 12500
+    },
+    {
+      "entropy": 2.1229864621162413,
+      "epoch": 0.5577777777777778,
+      "grad_norm": 7.0625,
+      "learning_rate": 9.118900343642612e-06,
+      "loss": 0.8412,
+      "mean_token_accuracy": 0.7494572842121124,
+      "num_tokens": 46241375.0,
+      "step": 12550
+    },
+    {
+      "entropy": 2.1183197784423826,
+      "epoch": 0.56,
+      "grad_norm": 6.875,
+      "learning_rate": 9.073081328751433e-06,
+      "loss": 0.7941,
+      "mean_token_accuracy": 0.7617311882972717,
+      "num_tokens": 46309203.0,
+      "step": 12600
+    },
+    {
+      "entropy": 2.1563652181625366,
+      "epoch": 0.5622222222222222,
+      "grad_norm": 7.34375,
+      "learning_rate": 9.027262313860253e-06,
+      "loss": 0.8333,
+      "mean_token_accuracy": 0.7494297361373902,
+      "num_tokens": 46377337.0,
+      "step": 12650
+    },
+    {
+      "entropy": 2.149436049461365,
+      "epoch": 0.5644444444444444,
+      "grad_norm": 9.0,
+      "learning_rate": 8.981443298969072e-06,
+      "loss": 0.8053,
+      "mean_token_accuracy": 0.7584555840492249,
+      "num_tokens": 46444416.0,
+      "step": 12700
+    },
+    {
+      "entropy": 2.1461888194084167,
+      "epoch": 0.5666666666666667,
+      "grad_norm": 6.6875,
+      "learning_rate": 8.935624284077893e-06,
+      "loss": 0.8286,
+      "mean_token_accuracy": 0.7490329504013061,
+      "num_tokens": 46513786.0,
+      "step": 12750
+    },
+    {
+      "entropy": 2.149775302410126,
+      "epoch": 0.5688888888888889,
+      "grad_norm": 7.5625,
+      "learning_rate": 8.889805269186713e-06,
+      "loss": 0.8296,
+      "mean_token_accuracy": 0.7498440980911255,
+      "num_tokens": 46580058.0,
+      "step": 12800
+    },
+    {
+      "entropy": 2.1283646035194397,
+      "epoch": 0.5711111111111111,
+      "grad_norm": 6.96875,
+      "learning_rate": 8.843986254295534e-06,
+      "loss": 0.8538,
+      "mean_token_accuracy": 0.7479376924037934,
+      "num_tokens": 46645685.0,
+      "step": 12850
+    },
+    {
+      "entropy": 2.136440978050232,
+      "epoch": 0.5733333333333334,
+      "grad_norm": 10.1875,
+      "learning_rate": 8.798167239404353e-06,
+      "loss": 0.8558,
+      "mean_token_accuracy": 0.7462117183208465,
+      "num_tokens": 46713622.0,
+      "step": 12900
+    },
+    {
+      "entropy": 2.044099681377411,
+      "epoch": 0.5755555555555556,
+      "grad_norm": 7.65625,
+      "learning_rate": 8.752348224513175e-06,
+      "loss": 0.8078,
+      "mean_token_accuracy": 0.755551530122757,
+      "num_tokens": 46784415.0,
+      "step": 12950
+    },
+    {
+      "entropy": 2.1400314664840696,
+      "epoch": 0.5777777777777777,
+      "grad_norm": 9.1875,
+      "learning_rate": 8.706529209621994e-06,
+      "loss": 0.7706,
+      "mean_token_accuracy": 0.7651760971546173,
+      "num_tokens": 46849509.0,
+      "step": 13000
+    },
+    {
+      "entropy": 2.1189517760276795,
+      "epoch": 0.58,
+      "grad_norm": 8.0,
+      "learning_rate": 8.660710194730814e-06,
+      "loss": 0.8118,
+      "mean_token_accuracy": 0.7566381883621216,
+      "num_tokens": 46916191.0,
+      "step": 13050
+    },
+    {
+      "entropy": 2.107829716205597,
+      "epoch": 0.5822222222222222,
+      "grad_norm": 6.5625,
+      "learning_rate": 8.614891179839635e-06,
+      "loss": 0.7543,
+      "mean_token_accuracy": 0.7710553121566772,
+      "num_tokens": 46984133.0,
+      "step": 13100
+    },
+    {
+      "entropy": 2.2209577679634096,
+      "epoch": 0.5844444444444444,
+      "grad_norm": 10.4375,
+      "learning_rate": 8.569072164948454e-06,
+      "loss": 0.8554,
+      "mean_token_accuracy": 0.7479790794849396,
+      "num_tokens": 47046937.0,
+      "step": 13150
+    },
+    {
+      "entropy": 2.1490628361701964,
+      "epoch": 0.5866666666666667,
+      "grad_norm": 7.40625,
+      "learning_rate": 8.523253150057275e-06,
+      "loss": 0.8488,
+      "mean_token_accuracy": 0.7476764714717865,
+      "num_tokens": 47115124.0,
+      "step": 13200
+    },
+    {
+      "entropy": 2.1350867557525635,
+      "epoch": 0.5888888888888889,
+      "grad_norm": 7.4375,
+      "learning_rate": 8.477434135166095e-06,
+      "loss": 0.8266,
+      "mean_token_accuracy": 0.7515107154846191,
+      "num_tokens": 47183436.0,
+      "step": 13250
+    },
+    {
+      "entropy": 2.203930015563965,
+      "epoch": 0.5911111111111111,
+      "grad_norm": 7.375,
+      "learning_rate": 8.431615120274916e-06,
+      "loss": 0.8206,
+      "mean_token_accuracy": 0.758274735212326,
+      "num_tokens": 47250973.0,
+      "step": 13300
+    },
+    {
+      "entropy": 2.17663028717041,
+      "epoch": 0.5933333333333334,
+      "grad_norm": 7.125,
+      "learning_rate": 8.385796105383734e-06,
+      "loss": 0.8627,
+      "mean_token_accuracy": 0.7431224703788757,
+      "num_tokens": 47319629.0,
+      "step": 13350
+    },
+    {
+      "entropy": 2.1186418867111207,
+      "epoch": 0.5955555555555555,
+      "grad_norm": 8.0625,
+      "learning_rate": 8.339977090492555e-06,
+      "loss": 0.8079,
+      "mean_token_accuracy": 0.7619412040710449,
+      "num_tokens": 47387168.0,
+      "step": 13400
+    },
+    {
+      "entropy": 2.1040466976165773,
+      "epoch": 0.5977777777777777,
+      "grad_norm": 8.0625,
+      "learning_rate": 8.294158075601375e-06,
+      "loss": 0.8375,
+      "mean_token_accuracy": 0.7491306090354919,
+      "num_tokens": 47458980.0,
+      "step": 13450
+    },
+    {
+      "entropy": 2.1847296571731567,
+      "epoch": 0.6,
+      "grad_norm": 7.75,
+      "learning_rate": 8.248339060710196e-06,
+      "loss": 0.8656,
+      "mean_token_accuracy": 0.7442897510528564,
+      "num_tokens": 47527776.0,
+      "step": 13500
+    },
+    {
+      "entropy": 2.1727059721946715,
+      "epoch": 0.6022222222222222,
+      "grad_norm": 8.875,
+      "learning_rate": 8.202520045819015e-06,
+      "loss": 0.8416,
+      "mean_token_accuracy": 0.7498923110961914,
+      "num_tokens": 47592067.0,
+      "step": 13550
+    },
+    {
+      "entropy": 2.146069450378418,
+      "epoch": 0.6044444444444445,
+      "grad_norm": 8.1875,
+      "learning_rate": 8.156701030927836e-06,
+      "loss": 0.8304,
+      "mean_token_accuracy": 0.7521639728546142,
+      "num_tokens": 47660520.0,
+      "step": 13600
+    },
+    {
+      "entropy": 2.1906869626045227,
+      "epoch": 0.6066666666666667,
+      "grad_norm": 8.0625,
+      "learning_rate": 8.110882016036656e-06,
+      "loss": 0.8126,
+      "mean_token_accuracy": 0.7562307071685791,
+      "num_tokens": 47726291.0,
+      "step": 13650
+    },
+    {
+      "entropy": 2.256641490459442,
+      "epoch": 0.6088888888888889,
+      "grad_norm": 8.6875,
+      "learning_rate": 8.065063001145475e-06,
+      "loss": 0.8229,
+      "mean_token_accuracy": 0.7553273499011993,
+      "num_tokens": 47789796.0,
+      "step": 13700
+    },
+    {
+      "entropy": 2.15241749048233,
+      "epoch": 0.6111111111111112,
+      "grad_norm": 7.625,
+      "learning_rate": 8.019243986254297e-06,
+      "loss": 0.7972,
+      "mean_token_accuracy": 0.7611013460159302,
+      "num_tokens": 47854791.0,
+      "step": 13750
+    },
+    {
+      "entropy": 2.1587498426437377,
+      "epoch": 0.6133333333333333,
+      "grad_norm": 10.1875,
+      "learning_rate": 7.973424971363116e-06,
+      "loss": 0.8219,
+      "mean_token_accuracy": 0.7551308906078339,
+      "num_tokens": 47922964.0,
+      "step": 13800
+    },
+    {
+      "entropy": 2.141474585533142,
+      "epoch": 0.6155555555555555,
+      "grad_norm": 8.875,
+      "learning_rate": 7.927605956471937e-06,
+      "loss": 0.8584,
+      "mean_token_accuracy": 0.7452862620353699,
+      "num_tokens": 47989985.0,
+      "step": 13850
+    },
+    {
+      "entropy": 2.116554036140442,
+      "epoch": 0.6177777777777778,
+      "grad_norm": 7.0625,
+      "learning_rate": 7.881786941580757e-06,
+      "loss": 0.7575,
+      "mean_token_accuracy": 0.7697462677955628,
+      "num_tokens": 48055485.0,
+      "step": 13900
+    },
+    {
+      "entropy": 2.145723969936371,
+      "epoch": 0.62,
+      "grad_norm": 7.21875,
+      "learning_rate": 7.835967926689578e-06,
+      "loss": 0.8424,
+      "mean_token_accuracy": 0.7527009093761444,
+      "num_tokens": 48123791.0,
+      "step": 13950
+    },
+    {
+      "entropy": 2.228086581230164,
+      "epoch": 0.6222222222222222,
+      "grad_norm": 16.25,
+      "learning_rate": 7.790148911798397e-06,
+      "loss": 0.8609,
+      "mean_token_accuracy": 0.7481016480922699,
+      "num_tokens": 48188322.0,
+      "step": 14000
+    },
+    {
+      "entropy": 2.1463445258140563,
+      "epoch": 0.6244444444444445,
+      "grad_norm": 8.0,
+      "learning_rate": 7.744329896907217e-06,
+      "loss": 0.7909,
+      "mean_token_accuracy": 0.7623829674720765,
+      "num_tokens": 48256913.0,
+      "step": 14050
+    },
+    {
+      "entropy": 2.1033449459075926,
+      "epoch": 0.6266666666666667,
+      "grad_norm": 6.5,
+      "learning_rate": 7.698510882016036e-06,
+      "loss": 0.8107,
+      "mean_token_accuracy": 0.7558232533931732,
+      "num_tokens": 48322383.0,
+      "step": 14100
+    },
+    {
+      "entropy": 2.1505328583717347,
+      "epoch": 0.6288888888888889,
+      "grad_norm": 8.0,
+      "learning_rate": 7.652691867124858e-06,
+      "loss": 0.8591,
+      "mean_token_accuracy": 0.7444049680233001,
+      "num_tokens": 48390728.0,
+      "step": 14150
+    },
+    {
+      "entropy": 2.1623160433769226,
+      "epoch": 0.6311111111111111,
+      "grad_norm": 7.875,
+      "learning_rate": 7.606872852233678e-06,
+      "loss": 0.783,
+      "mean_token_accuracy": 0.7643996751308442,
+      "num_tokens": 48457254.0,
+      "step": 14200
+    },
+    {
+      "entropy": 2.1291307163238526,
+      "epoch": 0.6333333333333333,
+      "grad_norm": 9.125,
+      "learning_rate": 7.561053837342498e-06,
+      "loss": 0.8297,
+      "mean_token_accuracy": 0.755406551361084,
+      "num_tokens": 48526165.0,
+      "step": 14250
+    },
+    {
+      "entropy": 2.0992329573631285,
+      "epoch": 0.6355555555555555,
+      "grad_norm": 9.5625,
+      "learning_rate": 7.515234822451319e-06,
+      "loss": 0.8357,
+      "mean_token_accuracy": 0.7475246119499207,
+      "num_tokens": 48594409.0,
+      "step": 14300
+    },
+    {
+      "entropy": 2.096374764442444,
+      "epoch": 0.6377777777777778,
+      "grad_norm": 7.5,
+      "learning_rate": 7.469415807560139e-06,
+      "loss": 0.7929,
+      "mean_token_accuracy": 0.7609642744064331,
+      "num_tokens": 48662016.0,
+      "step": 14350
+    },
+    {
+      "entropy": 2.1556865262985228,
+      "epoch": 0.64,
+      "grad_norm": 7.96875,
+      "learning_rate": 7.4235967926689576e-06,
+      "loss": 0.8432,
+      "mean_token_accuracy": 0.7489691793918609,
+      "num_tokens": 48727783.0,
+      "step": 14400
+    },
+    {
+      "entropy": 2.2297734808921814,
+      "epoch": 0.6422222222222222,
+      "grad_norm": 7.25,
+      "learning_rate": 7.377777777777778e-06,
+      "loss": 0.845,
+      "mean_token_accuracy": 0.7471547079086304,
+      "num_tokens": 48795872.0,
+      "step": 14450
+    },
+    {
+      "entropy": 2.093806471824646,
+      "epoch": 0.6444444444444445,
+      "grad_norm": 9.875,
+      "learning_rate": 7.331958762886598e-06,
+      "loss": 0.7843,
+      "mean_token_accuracy": 0.764157919883728,
+      "num_tokens": 48864940.0,
+      "step": 14500
+    },
+    {
+      "entropy": 2.1970014452934263,
+      "epoch": 0.6466666666666666,
+      "grad_norm": 6.96875,
+      "learning_rate": 7.286139747995419e-06,
+      "loss": 0.8726,
+      "mean_token_accuracy": 0.7447559666633606,
+      "num_tokens": 48932155.0,
+      "step": 14550
+    },
+    {
+      "entropy": 2.173217701911926,
+      "epoch": 0.6488888888888888,
+      "grad_norm": 7.0,
+      "learning_rate": 7.240320733104239e-06,
+      "loss": 0.8209,
+      "mean_token_accuracy": 0.7533725011348724,
+      "num_tokens": 48996865.0,
+      "step": 14600
+    },
+    {
+      "entropy": 2.1182384943962096,
+      "epoch": 0.6511111111111111,
+      "grad_norm": 7.25,
+      "learning_rate": 7.194501718213059e-06,
+      "loss": 0.8433,
+      "mean_token_accuracy": 0.743918879032135,
+      "num_tokens": 49062177.0,
+      "step": 14650
+    },
+    {
+      "entropy": 2.182834794521332,
+      "epoch": 0.6533333333333333,
+      "grad_norm": 15.5,
+      "learning_rate": 7.148682703321879e-06,
+      "loss": 0.8163,
+      "mean_token_accuracy": 0.7540112996101379,
+      "num_tokens": 49126666.0,
+      "step": 14700
+    },
+    {
+      "entropy": 2.1658547282218934,
+      "epoch": 0.6555555555555556,
+      "grad_norm": 7.0625,
+      "learning_rate": 7.102863688430699e-06,
+      "loss": 0.8318,
+      "mean_token_accuracy": 0.7548858106136322,
+      "num_tokens": 49193802.0,
+      "step": 14750
+    },
+    {
+      "entropy": 2.22178261756897,
+      "epoch": 0.6577777777777778,
+      "grad_norm": 7.34375,
+      "learning_rate": 7.0570446735395194e-06,
+      "loss": 0.8171,
+      "mean_token_accuracy": 0.755473929643631,
+      "num_tokens": 49256927.0,
+      "step": 14800
+    },
+    {
+      "entropy": 2.1972863483428955,
+      "epoch": 0.66,
+      "grad_norm": 17.25,
+      "learning_rate": 7.01122565864834e-06,
+      "loss": 0.868,
+      "mean_token_accuracy": 0.7451527345180512,
+      "num_tokens": 49324296.0,
+      "step": 14850
+    },
+    {
+      "entropy": 2.13280175447464,
+      "epoch": 0.6622222222222223,
+      "grad_norm": 7.28125,
+      "learning_rate": 6.96540664375716e-06,
+      "loss": 0.8565,
+      "mean_token_accuracy": 0.7481217682361603,
+      "num_tokens": 49393359.0,
+      "step": 14900
+    },
+    {
+      "entropy": 2.223068025112152,
+      "epoch": 0.6644444444444444,
+      "grad_norm": 9.9375,
+      "learning_rate": 6.9195876288659804e-06,
+      "loss": 0.8344,
+      "mean_token_accuracy": 0.7528722262382508,
+      "num_tokens": 49455849.0,
+      "step": 14950
+    },
+    {
+      "entropy": 2.1984812426567077,
+      "epoch": 0.6666666666666666,
+      "grad_norm": 7.875,
+      "learning_rate": 6.873768613974801e-06,
+      "loss": 0.8534,
+      "mean_token_accuracy": 0.7472201347351074,
+      "num_tokens": 49521825.0,
+      "step": 15000
+    },
+    {
+      "epoch": 0.6666666666666666,
+      "eval_entropy": 2.0471889674663544,
+      "eval_loss": 0.8776370286941528,
+      "eval_mean_token_accuracy": 0.7463443726301193,
+      "eval_num_tokens": 49521825.0,
+      "eval_runtime": 5.8695,
+      "eval_samples_per_second": 2.215,
+      "eval_steps_per_second": 0.681,
+      "step": 15000
+    },
+    {
+      "entropy": 2.1753278255462645,
+      "epoch": 0.6688888888888889,
+      "grad_norm": 8.6875,
+      "learning_rate": 6.8279495990836194e-06,
+      "loss": 0.8559,
+      "mean_token_accuracy": 0.7446865785121918,
+      "num_tokens": 49586773.0,
+      "step": 15050
+    },
+    {
+      "entropy": 2.137886700630188,
+      "epoch": 0.6711111111111111,
+      "grad_norm": 8.6875,
+      "learning_rate": 6.78213058419244e-06,
+      "loss": 0.8128,
+      "mean_token_accuracy": 0.757747368812561,
+      "num_tokens": 49654785.0,
+      "step": 15100
+    },
+    {
+      "entropy": 2.1542887043952943,
+      "epoch": 0.6733333333333333,
+      "grad_norm": 9.5625,
+      "learning_rate": 6.73631156930126e-06,
+      "loss": 0.8498,
+      "mean_token_accuracy": 0.7489660286903381,
+      "num_tokens": 49720666.0,
+      "step": 15150
+    },
+    {
+      "entropy": 2.1714869332313538,
+      "epoch": 0.6755555555555556,
+      "grad_norm": 8.25,
+      "learning_rate": 6.6904925544100804e-06,
+      "loss": 0.8183,
+      "mean_token_accuracy": 0.7538860237598419,
+      "num_tokens": 49790606.0,
+      "step": 15200
+    },
+    {
+      "entropy": 2.1753248810768127,
+      "epoch": 0.6777777777777778,
+      "grad_norm": 7.96875,
+      "learning_rate": 6.644673539518901e-06,
+      "loss": 0.8017,
+      "mean_token_accuracy": 0.7590576016902923,
+      "num_tokens": 49855120.0,
+      "step": 15250
+    },
+    {
+      "entropy": 2.1427893018722535,
+      "epoch": 0.68,
+      "grad_norm": 10.25,
+      "learning_rate": 6.598854524627721e-06,
+      "loss": 0.8308,
+      "mean_token_accuracy": 0.7554869890213013,
+      "num_tokens": 49923395.0,
+      "step": 15300
+    },
+    {
+      "entropy": 2.0930906391143798,
+      "epoch": 0.6822222222222222,
+      "grad_norm": 8.1875,
+      "learning_rate": 6.5530355097365415e-06,
+      "loss": 0.8226,
+      "mean_token_accuracy": 0.7525417697429657,
+      "num_tokens": 49988632.0,
+      "step": 15350
+    },
+    {
+      "entropy": 2.183457748889923,
+      "epoch": 0.6844444444444444,
+      "grad_norm": 7.34375,
+      "learning_rate": 6.507216494845361e-06,
+      "loss": 0.8382,
+      "mean_token_accuracy": 0.7489526784420013,
+      "num_tokens": 50055473.0,
+      "step": 15400
+    },
+    {
+      "entropy": 2.1905418968200685,
+      "epoch": 0.6866666666666666,
+      "grad_norm": 8.25,
+      "learning_rate": 6.461397479954181e-06,
+      "loss": 0.8143,
+      "mean_token_accuracy": 0.7581993770599366,
+      "num_tokens": 50122557.0,
+      "step": 15450
+    },
+    {
+      "entropy": 2.1051335525512695,
+      "epoch": 0.6888888888888889,
+      "grad_norm": 8.25,
+      "learning_rate": 6.415578465063002e-06,
+      "loss": 0.8459,
+      "mean_token_accuracy": 0.7460039758682251,
+      "num_tokens": 50190091.0,
+      "step": 15500
+    },
+    {
+      "entropy": 2.191064128875732,
+      "epoch": 0.6911111111111111,
+      "grad_norm": 6.96875,
+      "learning_rate": 6.369759450171822e-06,
+      "loss": 0.8458,
+      "mean_token_accuracy": 0.7485956978797913,
+      "num_tokens": 50255412.0,
+      "step": 15550
+    },
+    {
+      "entropy": 2.1825055122375487,
+      "epoch": 0.6933333333333334,
+      "grad_norm": 8.25,
+      "learning_rate": 6.323940435280642e-06,
+      "loss": 0.8686,
+      "mean_token_accuracy": 0.7439832353591919,
+      "num_tokens": 50323487.0,
+      "step": 15600
+    },
+    {
+      "entropy": 2.2044737410545348,
+      "epoch": 0.6955555555555556,
+      "grad_norm": 6.71875,
+      "learning_rate": 6.278121420389463e-06,
+      "loss": 0.8422,
+      "mean_token_accuracy": 0.7480260360240937,
+      "num_tokens": 50389804.0,
+      "step": 15650
+    },
+    {
+      "entropy": 2.194586501121521,
+      "epoch": 0.6977777777777778,
+      "grad_norm": 7.65625,
+      "learning_rate": 6.232302405498283e-06,
+      "loss": 0.8351,
+      "mean_token_accuracy": 0.7537850439548492,
+      "num_tokens": 50452603.0,
+      "step": 15700
+    },
+    {
+      "entropy": 2.1760614275932313,
+      "epoch": 0.7,
+      "grad_norm": 7.90625,
+      "learning_rate": 6.186483390607102e-06,
+      "loss": 0.8462,
+      "mean_token_accuracy": 0.7481569695472717,
+      "num_tokens": 50522865.0,
+      "step": 15750
+    },
+    {
+      "entropy": 2.1524909257888796,
+      "epoch": 0.7022222222222222,
+      "grad_norm": 15.375,
+      "learning_rate": 6.140664375715922e-06,
+      "loss": 0.8255,
+      "mean_token_accuracy": 0.7523048520088196,
+      "num_tokens": 50592440.0,
+      "step": 15800
+    },
+    {
+      "entropy": 2.160615997314453,
+      "epoch": 0.7044444444444444,
+      "grad_norm": 8.1875,
+      "learning_rate": 6.094845360824742e-06,
+      "loss": 0.8195,
+      "mean_token_accuracy": 0.7542084169387817,
+      "num_tokens": 50661169.0,
+      "step": 15850
+    },
+    {
+      "entropy": 2.1588108134269715,
+      "epoch": 0.7066666666666667,
+      "grad_norm": 7.28125,
+      "learning_rate": 6.049026345933563e-06,
+      "loss": 0.8343,
+      "mean_token_accuracy": 0.7516414785385132,
+      "num_tokens": 50730313.0,
+      "step": 15900
+    },
+    {
+      "entropy": 2.2916067910194395,
+      "epoch": 0.7088888888888889,
+      "grad_norm": 8.9375,
+      "learning_rate": 6.003207331042383e-06,
+      "loss": 0.8779,
+      "mean_token_accuracy": 0.7434639298915863,
+      "num_tokens": 50793896.0,
+      "step": 15950
+    },
+    {
+      "entropy": 2.165027015209198,
+      "epoch": 0.7111111111111111,
+      "grad_norm": 9.75,
+      "learning_rate": 5.957388316151203e-06,
+      "loss": 0.8546,
+      "mean_token_accuracy": 0.7454061865806579,
+      "num_tokens": 50861288.0,
+      "step": 16000
+    },
+    {
+      "entropy": 2.0912844824790953,
+      "epoch": 0.7133333333333334,
+      "grad_norm": 8.375,
+      "learning_rate": 5.911569301260024e-06,
+      "loss": 0.7921,
+      "mean_token_accuracy": 0.7610292685031891,
+      "num_tokens": 50928175.0,
+      "step": 16050
+    },
+    {
+      "entropy": 2.193835806846619,
+      "epoch": 0.7155555555555555,
+      "grad_norm": 9.375,
+      "learning_rate": 5.865750286368843e-06,
+      "loss": 0.8429,
+      "mean_token_accuracy": 0.748592312335968,
+      "num_tokens": 50991843.0,
+      "step": 16100
+    },
+    {
+      "entropy": 2.150286679267883,
+      "epoch": 0.7177777777777777,
+      "grad_norm": 8.0,
+      "learning_rate": 5.8199312714776635e-06,
+      "loss": 0.8731,
+      "mean_token_accuracy": 0.7444507765769959,
+      "num_tokens": 51057738.0,
+      "step": 16150
+    },
+    {
+      "entropy": 2.2080190443992613,
+      "epoch": 0.72,
+      "grad_norm": 6.59375,
+      "learning_rate": 5.774112256586484e-06,
+      "loss": 0.8394,
+      "mean_token_accuracy": 0.7506636822223663,
+      "num_tokens": 51122791.0,
+      "step": 16200
+    },
+    {
+      "entropy": 2.1831669211387634,
+      "epoch": 0.7222222222222222,
+      "grad_norm": 6.78125,
+      "learning_rate": 5.728293241695304e-06,
+      "loss": 0.8655,
+      "mean_token_accuracy": 0.7454936730861664,
+      "num_tokens": 51189529.0,
+      "step": 16250
+    },
+    {
+      "entropy": 2.187736349105835,
+      "epoch": 0.7244444444444444,
+      "grad_norm": 7.15625,
+      "learning_rate": 5.6824742268041245e-06,
+      "loss": 0.8503,
+      "mean_token_accuracy": 0.7467346620559693,
+      "num_tokens": 51255571.0,
+      "step": 16300
+    },
+    {
+      "entropy": 2.228825159072876,
+      "epoch": 0.7266666666666667,
+      "grad_norm": 7.09375,
+      "learning_rate": 5.636655211912945e-06,
+      "loss": 0.8646,
+      "mean_token_accuracy": 0.7439834475517273,
+      "num_tokens": 51321902.0,
+      "step": 16350
+    },
+    {
+      "entropy": 2.1766190052032472,
+      "epoch": 0.7288888888888889,
+      "grad_norm": 8.0,
+      "learning_rate": 5.590836197021764e-06,
+      "loss": 0.8582,
+      "mean_token_accuracy": 0.7422134637832641,
+      "num_tokens": 51387236.0,
+      "step": 16400
+    },
+    {
+      "entropy": 2.1550874876976014,
+      "epoch": 0.7311111111111112,
+      "grad_norm": 6.84375,
+      "learning_rate": 5.545017182130585e-06,
+      "loss": 0.803,
+      "mean_token_accuracy": 0.7592871415615082,
+      "num_tokens": 51454340.0,
+      "step": 16450
+    },
+    {
+      "entropy": 2.1886205792427065,
+      "epoch": 0.7333333333333333,
+      "grad_norm": 8.5625,
+      "learning_rate": 5.499198167239405e-06,
+      "loss": 0.8312,
+      "mean_token_accuracy": 0.7522400307655335,
+      "num_tokens": 51521289.0,
+      "step": 16500
+    },
+    {
+      "entropy": 2.1630620193481445,
+      "epoch": 0.7355555555555555,
+      "grad_norm": 11.125,
+      "learning_rate": 5.4533791523482245e-06,
+      "loss": 0.8429,
+      "mean_token_accuracy": 0.7494550979137421,
+      "num_tokens": 51590453.0,
+      "step": 16550
+    },
+    {
+      "entropy": 2.2003714513778685,
+      "epoch": 0.7377777777777778,
+      "grad_norm": 7.90625,
+      "learning_rate": 5.407560137457045e-06,
+      "loss": 0.8594,
+      "mean_token_accuracy": 0.7431814324855804,
+      "num_tokens": 51659077.0,
+      "step": 16600
+    },
+    {
+      "entropy": 2.159868106842041,
+      "epoch": 0.74,
+      "grad_norm": 7.09375,
+      "learning_rate": 5.361741122565865e-06,
+      "loss": 0.8511,
+      "mean_token_accuracy": 0.7491097986698151,
+      "num_tokens": 51724327.0,
+      "step": 16650
+    },
+    {
+      "entropy": 2.1559503626823426,
+      "epoch": 0.7422222222222222,
+      "grad_norm": 7.03125,
+      "learning_rate": 5.3159221076746855e-06,
+      "loss": 0.866,
+      "mean_token_accuracy": 0.7423543095588684,
+      "num_tokens": 51793571.0,
+      "step": 16700
+    },
+    {
+      "entropy": 2.1431314539909363,
+      "epoch": 0.7444444444444445,
+      "grad_norm": 6.96875,
+      "learning_rate": 5.270103092783505e-06,
+      "loss": 0.8175,
+      "mean_token_accuracy": 0.7550452971458435,
+      "num_tokens": 51862535.0,
+      "step": 16750
+    },
+    {
+      "entropy": 2.1364932513237,
+      "epoch": 0.7466666666666667,
+      "grad_norm": 9.5,
+      "learning_rate": 5.224284077892325e-06,
+      "loss": 0.8622,
+      "mean_token_accuracy": 0.7470258843898773,
+      "num_tokens": 51930936.0,
+      "step": 16800
+    },
+    {
+      "entropy": 2.213224956989288,
+      "epoch": 0.7488888888888889,
+      "grad_norm": 7.0,
+      "learning_rate": 5.178465063001146e-06,
+      "loss": 0.8664,
+      "mean_token_accuracy": 0.7476742577552795,
+      "num_tokens": 51996312.0,
+      "step": 16850
+    },
+    {
+      "entropy": 2.3157504415512085,
+      "epoch": 0.7511111111111111,
+      "grad_norm": 7.53125,
+      "learning_rate": 5.132646048109966e-06,
+      "loss": 0.821,
+      "mean_token_accuracy": 0.7552366864681244,
+      "num_tokens": 52061146.0,
+      "step": 16900
+    },
+    {
+      "entropy": 2.258919379711151,
+      "epoch": 0.7533333333333333,
+      "grad_norm": 7.1875,
+      "learning_rate": 5.086827033218786e-06,
+      "loss": 0.8122,
+      "mean_token_accuracy": 0.7566414999961854,
+      "num_tokens": 52126822.0,
+      "step": 16950
+    },
+    {
+      "entropy": 2.2003828430175782,
+      "epoch": 0.7555555555555555,
+      "grad_norm": 8.6875,
+      "learning_rate": 5.041008018327607e-06,
+      "loss": 0.8321,
+      "mean_token_accuracy": 0.7529778003692627,
+      "num_tokens": 52192915.0,
+      "step": 17000
+    },
+    {
+      "entropy": 2.2184296226501465,
+      "epoch": 0.7577777777777778,
+      "grad_norm": 9.1875,
+      "learning_rate": 4.995189003436426e-06,
+      "loss": 0.8577,
+      "mean_token_accuracy": 0.7449932956695556,
+      "num_tokens": 52261692.0,
+      "step": 17050
+    },
+    {
+      "entropy": 2.171595447063446,
+      "epoch": 0.76,
+      "grad_norm": 9.125,
+      "learning_rate": 4.9493699885452465e-06,
+      "loss": 0.8545,
+      "mean_token_accuracy": 0.7480556511878967,
+      "num_tokens": 52329194.0,
+      "step": 17100
+    },
+    {
+      "entropy": 2.2223637771606444,
+      "epoch": 0.7622222222222222,
+      "grad_norm": 7.375,
+      "learning_rate": 4.903550973654067e-06,
+      "loss": 0.8367,
+      "mean_token_accuracy": 0.7503479218482971,
+      "num_tokens": 52395383.0,
+      "step": 17150
+    },
+    {
+      "entropy": 2.2046579623222353,
+      "epoch": 0.7644444444444445,
+      "grad_norm": 7.53125,
+      "learning_rate": 4.857731958762887e-06,
+      "loss": 0.8635,
+      "mean_token_accuracy": 0.7457800447940827,
+      "num_tokens": 52463232.0,
+      "step": 17200
+    },
+    {
+      "entropy": 2.173711452484131,
+      "epoch": 0.7666666666666667,
+      "grad_norm": 6.625,
+      "learning_rate": 4.8119129438717075e-06,
+      "loss": 0.8209,
+      "mean_token_accuracy": 0.7542583394050598,
+      "num_tokens": 52531656.0,
+      "step": 17250
+    },
+    {
+      "entropy": 2.2189766001701354,
+      "epoch": 0.7688888888888888,
+      "grad_norm": 10.6875,
+      "learning_rate": 4.766093928980528e-06,
+      "loss": 0.8685,
+      "mean_token_accuracy": 0.7441653323173523,
+      "num_tokens": 52595389.0,
+      "step": 17300
+    },
+    {
+      "entropy": 2.1452325272560118,
+      "epoch": 0.7711111111111111,
+      "grad_norm": 9.625,
+      "learning_rate": 4.720274914089347e-06,
+      "loss": 0.8348,
+      "mean_token_accuracy": 0.7506163120269775,
+      "num_tokens": 52663028.0,
+      "step": 17350
+    },
+    {
+      "entropy": 2.2454160952568056,
+      "epoch": 0.7733333333333333,
+      "grad_norm": 7.75,
+      "learning_rate": 4.674455899198168e-06,
+      "loss": 0.8597,
+      "mean_token_accuracy": 0.7458804631233216,
+      "num_tokens": 52728656.0,
+      "step": 17400
+    },
+    {
+      "entropy": 2.292165369987488,
+      "epoch": 0.7755555555555556,
+      "grad_norm": 6.65625,
+      "learning_rate": 4.628636884306988e-06,
+      "loss": 0.8405,
+      "mean_token_accuracy": 0.7525258159637451,
+      "num_tokens": 52794230.0,
+      "step": 17450
+    },
+    {
+      "entropy": 2.1589057970047,
+      "epoch": 0.7777777777777778,
+      "grad_norm": 7.9375,
+      "learning_rate": 4.582817869415808e-06,
+      "loss": 0.8481,
+      "mean_token_accuracy": 0.7463726258277893,
+      "num_tokens": 52862165.0,
+      "step": 17500
+    },
+    {
+      "entropy": 2.169076681137085,
+      "epoch": 0.78,
+      "grad_norm": 7.78125,
+      "learning_rate": 4.536998854524628e-06,
+      "loss": 0.8224,
+      "mean_token_accuracy": 0.7554120934009552,
+      "num_tokens": 52930052.0,
+      "step": 17550
+    },
+    {
+      "entropy": 2.21035982131958,
+      "epoch": 0.7822222222222223,
+      "grad_norm": 8.625,
+      "learning_rate": 4.491179839633448e-06,
+      "loss": 0.7947,
+      "mean_token_accuracy": 0.7617706823348999,
+      "num_tokens": 52997539.0,
+      "step": 17600
+    },
+    {
+      "entropy": 2.285029878616333,
+      "epoch": 0.7844444444444445,
+      "grad_norm": 6.90625,
+      "learning_rate": 4.4453608247422685e-06,
+      "loss": 0.8251,
+      "mean_token_accuracy": 0.7533777153491974,
+      "num_tokens": 53062000.0,
+      "step": 17650
+    },
+    {
+      "entropy": 2.169939410686493,
+      "epoch": 0.7866666666666666,
+      "grad_norm": 6.875,
+      "learning_rate": 4.399541809851088e-06,
+      "loss": 0.8256,
+      "mean_token_accuracy": 0.7538401031494141,
+      "num_tokens": 53128912.0,
+      "step": 17700
+    },
+    {
+      "entropy": 2.2041222214698792,
+      "epoch": 0.7888888888888889,
+      "grad_norm": 10.375,
+      "learning_rate": 4.353722794959908e-06,
+      "loss": 0.7965,
+      "mean_token_accuracy": 0.7599800097942352,
+      "num_tokens": 53195055.0,
+      "step": 17750
+    },
+    {
+      "entropy": 2.2029996418952944,
+      "epoch": 0.7911111111111111,
+      "grad_norm": 8.25,
+      "learning_rate": 4.307903780068729e-06,
+      "loss": 0.8447,
+      "mean_token_accuracy": 0.748561098575592,
+      "num_tokens": 53259948.0,
+      "step": 17800
+    },
+    {
+      "entropy": 2.139230773448944,
+      "epoch": 0.7933333333333333,
+      "grad_norm": 6.15625,
+      "learning_rate": 4.262084765177549e-06,
+      "loss": 0.8634,
+      "mean_token_accuracy": 0.7443192017078399,
+      "num_tokens": 53330653.0,
+      "step": 17850
+    },
+    {
+      "entropy": 2.1335135221481325,
+      "epoch": 0.7955555555555556,
+      "grad_norm": 7.09375,
+      "learning_rate": 4.216265750286369e-06,
+      "loss": 0.8332,
+      "mean_token_accuracy": 0.7517373490333558,
+      "num_tokens": 53398724.0,
+      "step": 17900
+    },
+    {
+      "entropy": 2.164382312297821,
+      "epoch": 0.7977777777777778,
+      "grad_norm": 8.375,
+      "learning_rate": 4.17044673539519e-06,
+      "loss": 0.8511,
+      "mean_token_accuracy": 0.7499736166000366,
+      "num_tokens": 53466863.0,
+      "step": 17950
+    },
+    {
+      "entropy": 2.190858428478241,
+      "epoch": 0.8,
+      "grad_norm": 6.59375,
+      "learning_rate": 4.12462772050401e-06,
+      "loss": 0.8523,
+      "mean_token_accuracy": 0.7486334586143494,
+      "num_tokens": 53532021.0,
+      "step": 18000
+    },
+    {
+      "entropy": 2.22938738822937,
+      "epoch": 0.8022222222222222,
+      "grad_norm": 8.3125,
+      "learning_rate": 4.0788087056128295e-06,
+      "loss": 0.8667,
+      "mean_token_accuracy": 0.7432076168060303,
+      "num_tokens": 53600120.0,
+      "step": 18050
+    },
+    {
+      "entropy": 2.233106544017792,
+      "epoch": 0.8044444444444444,
+      "grad_norm": 9.375,
+      "learning_rate": 4.03298969072165e-06,
+      "loss": 0.8267,
+      "mean_token_accuracy": 0.7550076794624329,
+      "num_tokens": 53665356.0,
+      "step": 18100
+    },
+    {
+      "entropy": 2.214742834568024,
+      "epoch": 0.8066666666666666,
+      "grad_norm": 10.375,
+      "learning_rate": 3.98717067583047e-06,
+      "loss": 0.8315,
+      "mean_token_accuracy": 0.752135591506958,
+      "num_tokens": 53732051.0,
+      "step": 18150
+    },
+    {
+      "entropy": 2.1795053052902222,
+      "epoch": 0.8088888888888889,
+      "grad_norm": 8.0625,
+      "learning_rate": 3.94135166093929e-06,
+      "loss": 0.8391,
+      "mean_token_accuracy": 0.7515090310573578,
+      "num_tokens": 53798222.0,
+      "step": 18200
+    },
+    {
+      "entropy": 2.1936265587806703,
+      "epoch": 0.8111111111111111,
+      "grad_norm": 7.875,
+      "learning_rate": 3.89553264604811e-06,
+      "loss": 0.8642,
+      "mean_token_accuracy": 0.7441390895843506,
+      "num_tokens": 53861370.0,
+      "step": 18250
+    },
+    {
+      "entropy": 2.261586802005768,
+      "epoch": 0.8133333333333334,
+      "grad_norm": 8.0625,
+      "learning_rate": 3.84971363115693e-06,
+      "loss": 0.8219,
+      "mean_token_accuracy": 0.7518996393680573,
+      "num_tokens": 53927168.0,
+      "step": 18300
+    },
+    {
+      "entropy": 2.1711619758605956,
+      "epoch": 0.8155555555555556,
+      "grad_norm": 7.125,
+      "learning_rate": 3.8038946162657507e-06,
+      "loss": 0.854,
+      "mean_token_accuracy": 0.7472757256031036,
+      "num_tokens": 53995523.0,
+      "step": 18350
+    },
+    {
+      "entropy": 2.1286602544784547,
+      "epoch": 0.8177777777777778,
+      "grad_norm": 10.0625,
+      "learning_rate": 3.7580756013745706e-06,
+      "loss": 0.8391,
+      "mean_token_accuracy": 0.7513591229915619,
+      "num_tokens": 54064144.0,
+      "step": 18400
+    },
+    {
+      "entropy": 2.227732105255127,
+      "epoch": 0.82,
+      "grad_norm": 8.5,
+      "learning_rate": 3.712256586483391e-06,
+      "loss": 0.8569,
+      "mean_token_accuracy": 0.7463644111156463,
+      "num_tokens": 54130199.0,
+      "step": 18450
+    },
+    {
+      "entropy": 2.242250292301178,
+      "epoch": 0.8222222222222222,
+      "grad_norm": 7.4375,
+      "learning_rate": 3.6664375715922113e-06,
+      "loss": 0.8458,
+      "mean_token_accuracy": 0.7489724898338318,
+      "num_tokens": 54193967.0,
+      "step": 18500
+    },
+    {
+      "entropy": 2.2216247129440307,
+      "epoch": 0.8244444444444444,
+      "grad_norm": 9.5,
+      "learning_rate": 3.620618556701031e-06,
+      "loss": 0.8631,
+      "mean_token_accuracy": 0.7474298918247223,
+      "num_tokens": 54258735.0,
+      "step": 18550
+    },
+    {
+      "entropy": 2.19566166639328,
+      "epoch": 0.8266666666666667,
+      "grad_norm": 8.875,
+      "learning_rate": 3.574799541809851e-06,
+      "loss": 0.8383,
+      "mean_token_accuracy": 0.7505941188335419,
+      "num_tokens": 54325869.0,
+      "step": 18600
+    },
+    {
+      "entropy": 2.2029261493682863,
+      "epoch": 0.8288888888888889,
+      "grad_norm": 6.71875,
+      "learning_rate": 3.5289805269186715e-06,
+      "loss": 0.8685,
+      "mean_token_accuracy": 0.7453620088100433,
+      "num_tokens": 54391495.0,
+      "step": 18650
+    },
+    {
+      "entropy": 2.321768162250519,
+      "epoch": 0.8311111111111111,
+      "grad_norm": 8.1875,
+      "learning_rate": 3.4831615120274914e-06,
+      "loss": 0.8268,
+      "mean_token_accuracy": 0.7540508365631103,
+      "num_tokens": 54457662.0,
+      "step": 18700
+    },
+    {
+      "entropy": 2.204611828327179,
+      "epoch": 0.8333333333333334,
+      "grad_norm": 7.03125,
+      "learning_rate": 3.4373424971363117e-06,
+      "loss": 0.8534,
+      "mean_token_accuracy": 0.7491276848316193,
+      "num_tokens": 54522465.0,
+      "step": 18750
+    },
+    {
+      "entropy": 2.180729539394379,
+      "epoch": 0.8355555555555556,
+      "grad_norm": 7.96875,
+      "learning_rate": 3.391523482245132e-06,
+      "loss": 0.8335,
+      "mean_token_accuracy": 0.7529357576370239,
+      "num_tokens": 54590053.0,
+      "step": 18800
+    },
+    {
+      "entropy": 2.205647897720337,
+      "epoch": 0.8377777777777777,
+      "grad_norm": 7.53125,
+      "learning_rate": 3.3457044673539524e-06,
+      "loss": 0.8733,
+      "mean_token_accuracy": 0.7421969878673553,
+      "num_tokens": 54656959.0,
+      "step": 18850
+    },
+    {
+      "entropy": 2.22350492477417,
+      "epoch": 0.84,
+      "grad_norm": 7.96875,
+      "learning_rate": 3.2998854524627723e-06,
+      "loss": 0.8657,
+      "mean_token_accuracy": 0.7474096655845642,
+      "num_tokens": 54722156.0,
+      "step": 18900
+    },
+    {
+      "entropy": 2.1877153444290163,
+      "epoch": 0.8422222222222222,
+      "grad_norm": 6.15625,
+      "learning_rate": 3.2540664375715927e-06,
+      "loss": 0.8894,
+      "mean_token_accuracy": 0.7431522738933564,
+      "num_tokens": 54789152.0,
+      "step": 18950
+    },
+    {
+      "entropy": 2.225923342704773,
+      "epoch": 0.8444444444444444,
+      "grad_norm": 11.125,
+      "learning_rate": 3.2082474226804126e-06,
+      "loss": 0.8555,
+      "mean_token_accuracy": 0.7505220258235932,
+      "num_tokens": 54858257.0,
+      "step": 19000
+    },
+    {
+      "entropy": 2.1716776871681214,
+      "epoch": 0.8466666666666667,
+      "grad_norm": 7.75,
+      "learning_rate": 3.1624284077892325e-06,
+      "loss": 0.8272,
+      "mean_token_accuracy": 0.7512453496456146,
+      "num_tokens": 54927221.0,
+      "step": 19050
+    },
+    {
+      "entropy": 2.2240468645095826,
+      "epoch": 0.8488888888888889,
+      "grad_norm": 12.5,
+      "learning_rate": 3.116609392898053e-06,
+      "loss": 0.8363,
+      "mean_token_accuracy": 0.7510754656791687,
+      "num_tokens": 54995938.0,
+      "step": 19100
+    },
+    {
+      "entropy": 2.1904780864715576,
+      "epoch": 0.8511111111111112,
+      "grad_norm": 9.75,
+      "learning_rate": 3.070790378006873e-06,
+      "loss": 0.834,
+      "mean_token_accuracy": 0.75088552236557,
+      "num_tokens": 55058163.0,
+      "step": 19150
+    },
+    {
+      "entropy": 2.215365264415741,
+      "epoch": 0.8533333333333334,
+      "grad_norm": 8.25,
+      "learning_rate": 3.0249713631156935e-06,
+      "loss": 0.8597,
+      "mean_token_accuracy": 0.7446792745590209,
+      "num_tokens": 55124026.0,
+      "step": 19200
+    },
+    {
+      "entropy": 2.202007007598877,
+      "epoch": 0.8555555555555555,
+      "grad_norm": 9.0625,
+      "learning_rate": 2.9791523482245134e-06,
+      "loss": 0.8144,
+      "mean_token_accuracy": 0.7558258211612702,
+      "num_tokens": 55192976.0,
+      "step": 19250
+    },
+    {
+      "entropy": 2.2235484766960143,
+      "epoch": 0.8577777777777778,
+      "grad_norm": 8.3125,
+      "learning_rate": 2.9333333333333338e-06,
+      "loss": 0.8662,
+      "mean_token_accuracy": 0.7427694439888001,
+      "num_tokens": 55260374.0,
+      "step": 19300
+    },
+    {
+      "entropy": 2.256188449859619,
+      "epoch": 0.86,
+      "grad_norm": 8.5625,
+      "learning_rate": 2.887514318442154e-06,
+      "loss": 0.8496,
+      "mean_token_accuracy": 0.7467195224761963,
+      "num_tokens": 55326060.0,
+      "step": 19350
+    },
+    {
+      "entropy": 2.2128705596923828,
+      "epoch": 0.8622222222222222,
+      "grad_norm": 11.75,
+      "learning_rate": 2.8416953035509736e-06,
+      "loss": 0.8645,
+      "mean_token_accuracy": 0.7435001564025879,
+      "num_tokens": 55391552.0,
+      "step": 19400
+    },
+    {
+      "entropy": 2.1930198001861574,
+      "epoch": 0.8644444444444445,
+      "grad_norm": 8.125,
+      "learning_rate": 2.795876288659794e-06,
+      "loss": 0.8781,
+      "mean_token_accuracy": 0.7410417079925538,
+      "num_tokens": 55461963.0,
+      "step": 19450
+    },
+    {
+      "entropy": 2.1576480197906496,
+      "epoch": 0.8666666666666667,
+      "grad_norm": 7.75,
+      "learning_rate": 2.7500572737686143e-06,
+      "loss": 0.832,
+      "mean_token_accuracy": 0.7539917016029358,
+      "num_tokens": 55529650.0,
+      "step": 19500
+    },
+    {
+      "entropy": 2.1880473136901855,
+      "epoch": 0.8688888888888889,
+      "grad_norm": 7.75,
+      "learning_rate": 2.7042382588774346e-06,
+      "loss": 0.868,
+      "mean_token_accuracy": 0.7416273355484009,
+      "num_tokens": 55596036.0,
+      "step": 19550
+    },
+    {
+      "entropy": 2.2236318159103394,
+      "epoch": 0.8711111111111111,
+      "grad_norm": 8.1875,
+      "learning_rate": 2.6584192439862545e-06,
+      "loss": 0.8535,
+      "mean_token_accuracy": 0.7471865510940552,
+      "num_tokens": 55662244.0,
+      "step": 19600
+    },
+    {
+      "entropy": 2.2074507117271422,
+      "epoch": 0.8733333333333333,
+      "grad_norm": 8.125,
+      "learning_rate": 2.612600229095075e-06,
+      "loss": 0.857,
+      "mean_token_accuracy": 0.746492406129837,
+      "num_tokens": 55730056.0,
+      "step": 19650
+    },
+    {
+      "entropy": 2.175764639377594,
+      "epoch": 0.8755555555555555,
+      "grad_norm": 7.875,
+      "learning_rate": 2.566781214203895e-06,
+      "loss": 0.8402,
+      "mean_token_accuracy": 0.7526404368877411,
+      "num_tokens": 55796368.0,
+      "step": 19700
+    },
+    {
+      "entropy": 2.2133652138710023,
+      "epoch": 0.8777777777777778,
+      "grad_norm": 7.40625,
+      "learning_rate": 2.5209621993127147e-06,
+      "loss": 0.8401,
+      "mean_token_accuracy": 0.7528079390525818,
+      "num_tokens": 55862412.0,
+      "step": 19750
+    },
+    {
+      "entropy": 2.16952513217926,
+      "epoch": 0.88,
+      "grad_norm": 11.125,
+      "learning_rate": 2.475143184421535e-06,
+      "loss": 0.8162,
+      "mean_token_accuracy": 0.7567255461215973,
+      "num_tokens": 55929780.0,
+      "step": 19800
+    },
+    {
+      "entropy": 2.2304827785491943,
+      "epoch": 0.8822222222222222,
+      "grad_norm": 9.125,
+      "learning_rate": 2.4293241695303554e-06,
+      "loss": 0.8747,
+      "mean_token_accuracy": 0.7438322114944458,
+      "num_tokens": 55997356.0,
+      "step": 19850
+    },
+    {
+      "entropy": 2.1662241196632386,
+      "epoch": 0.8844444444444445,
+      "grad_norm": 7.71875,
+      "learning_rate": 2.3835051546391753e-06,
+      "loss": 0.8738,
+      "mean_token_accuracy": 0.7397397947311402,
+      "num_tokens": 56066573.0,
+      "step": 19900
+    },
+    {
+      "entropy": 2.1665696811676027,
+      "epoch": 0.8866666666666667,
+      "grad_norm": 7.15625,
+      "learning_rate": 2.3376861397479956e-06,
+      "loss": 0.8558,
+      "mean_token_accuracy": 0.7482295572757721,
+      "num_tokens": 56136956.0,
+      "step": 19950
+    },
+    {
+      "entropy": 2.2371082639694215,
+      "epoch": 0.8888888888888888,
+      "grad_norm": 8.75,
+      "learning_rate": 2.291867124856816e-06,
+      "loss": 0.8136,
+      "mean_token_accuracy": 0.7589144480228424,
+      "num_tokens": 56202618.0,
+      "step": 20000
+    },
+    {
+      "epoch": 0.8888888888888888,
+      "eval_entropy": 2.108439266681671,
+      "eval_loss": 0.8734950423240662,
+      "eval_mean_token_accuracy": 0.7479927390813828,
+      "eval_num_tokens": 56202618.0,
+      "eval_runtime": 5.499,
+      "eval_samples_per_second": 2.364,
+      "eval_steps_per_second": 0.727,
+      "step": 20000
+    },
+    {
+      "entropy": 2.2510220575332642,
+      "epoch": 0.8911111111111111,
+      "grad_norm": 9.5,
+      "learning_rate": 2.246048109965636e-06,
+      "loss": 0.8871,
+      "mean_token_accuracy": 0.7370022451877594,
+      "num_tokens": 56266792.0,
+      "step": 20050
+    },
+    {
+      "entropy": 2.1528950190544127,
+      "epoch": 0.8933333333333333,
+      "grad_norm": 9.75,
+      "learning_rate": 2.200229095074456e-06,
+      "loss": 0.8586,
+      "mean_token_accuracy": 0.7451235044002533,
+      "num_tokens": 56334124.0,
+      "step": 20100
+    },
+    {
+      "entropy": 2.1923975205421447,
+      "epoch": 0.8955555555555555,
+      "grad_norm": 6.71875,
+      "learning_rate": 2.154410080183276e-06,
+      "loss": 0.874,
+      "mean_token_accuracy": 0.7444364356994629,
+      "num_tokens": 56401545.0,
+      "step": 20150
+    },
+    {
+      "entropy": 2.222883083820343,
+      "epoch": 0.8977777777777778,
+      "grad_norm": 9.1875,
+      "learning_rate": 2.1085910652920965e-06,
+      "loss": 0.8376,
+      "mean_token_accuracy": 0.7498513388633729,
+      "num_tokens": 56471261.0,
+      "step": 20200
+    },
+    {
+      "entropy": 2.228672001361847,
+      "epoch": 0.9,
+      "grad_norm": 7.21875,
+      "learning_rate": 2.062772050400917e-06,
+      "loss": 0.8231,
+      "mean_token_accuracy": 0.755576502084732,
+      "num_tokens": 56535034.0,
+      "step": 20250
+    },
+    {
+      "entropy": 2.196291310787201,
+      "epoch": 0.9022222222222223,
+      "grad_norm": 7.21875,
+      "learning_rate": 2.0169530355097367e-06,
+      "loss": 0.8711,
+      "mean_token_accuracy": 0.7452339708805085,
+      "num_tokens": 56605038.0,
+      "step": 20300
+    },
+    {
+      "entropy": 2.206197905540466,
+      "epoch": 0.9044444444444445,
+      "grad_norm": 9.0,
+      "learning_rate": 1.971134020618557e-06,
+      "loss": 0.8707,
+      "mean_token_accuracy": 0.7420957183837891,
+      "num_tokens": 56670283.0,
+      "step": 20350
+    },
+    {
+      "entropy": 2.1762799286842345,
+      "epoch": 0.9066666666666666,
+      "grad_norm": 11.0,
+      "learning_rate": 1.925315005727377e-06,
+      "loss": 0.8375,
+      "mean_token_accuracy": 0.751966392993927,
+      "num_tokens": 56735920.0,
+      "step": 20400
+    },
+    {
+      "entropy": 2.218523108959198,
+      "epoch": 0.9088888888888889,
+      "grad_norm": 9.25,
+      "learning_rate": 1.879495990836197e-06,
+      "loss": 0.8468,
+      "mean_token_accuracy": 0.7500053834915161,
+      "num_tokens": 56801907.0,
+      "step": 20450
+    },
+    {
+      "entropy": 2.289003756046295,
+      "epoch": 0.9111111111111111,
+      "grad_norm": 11.75,
+      "learning_rate": 1.8336769759450174e-06,
+      "loss": 0.8643,
+      "mean_token_accuracy": 0.7460965967178345,
+      "num_tokens": 56867853.0,
+      "step": 20500
+    },
+    {
+      "entropy": 2.1768972492218017,
+      "epoch": 0.9133333333333333,
+      "grad_norm": 9.1875,
+      "learning_rate": 1.7878579610538373e-06,
+      "loss": 0.8613,
+      "mean_token_accuracy": 0.7483590936660767,
+      "num_tokens": 56936042.0,
+      "step": 20550
+    },
+    {
+      "entropy": 2.211531710624695,
+      "epoch": 0.9155555555555556,
+      "grad_norm": 12.125,
+      "learning_rate": 1.7420389461626577e-06,
+      "loss": 0.8518,
+      "mean_token_accuracy": 0.7475169038772583,
+      "num_tokens": 57003712.0,
+      "step": 20600
+    },
+    {
+      "entropy": 2.2285032725334166,
+      "epoch": 0.9177777777777778,
+      "grad_norm": 9.0,
+      "learning_rate": 1.6962199312714778e-06,
+      "loss": 0.8805,
+      "mean_token_accuracy": 0.7423696160316468,
+      "num_tokens": 57071218.0,
+      "step": 20650
+    },
+    {
+      "entropy": 2.2252227783203127,
+      "epoch": 0.92,
+      "grad_norm": 13.1875,
+      "learning_rate": 1.6504009163802981e-06,
+      "loss": 0.8401,
+      "mean_token_accuracy": 0.751958976984024,
+      "num_tokens": 57136636.0,
+      "step": 20700
+    },
+    {
+      "entropy": 2.220773038864136,
+      "epoch": 0.9222222222222223,
+      "grad_norm": 6.9375,
+      "learning_rate": 1.604581901489118e-06,
+      "loss": 0.8293,
+      "mean_token_accuracy": 0.7526602661609649,
+      "num_tokens": 57201653.0,
+      "step": 20750
+    },
+    {
+      "entropy": 2.2865464878082276,
+      "epoch": 0.9244444444444444,
+      "grad_norm": 7.3125,
+      "learning_rate": 1.5587628865979382e-06,
+      "loss": 0.8497,
+      "mean_token_accuracy": 0.7494020068645477,
+      "num_tokens": 57266565.0,
+      "step": 20800
+    },
+    {
+      "entropy": 2.2102810382843017,
+      "epoch": 0.9266666666666666,
+      "grad_norm": 6.8125,
+      "learning_rate": 1.5129438717067585e-06,
+      "loss": 0.8608,
+      "mean_token_accuracy": 0.748868852853775,
+      "num_tokens": 57334744.0,
+      "step": 20850
+    },
+    {
+      "entropy": 2.1809313702583313,
+      "epoch": 0.9288888888888889,
+      "grad_norm": 15.625,
+      "learning_rate": 1.4671248568155784e-06,
+      "loss": 0.8679,
+      "mean_token_accuracy": 0.7432131195068359,
+      "num_tokens": 57403405.0,
+      "step": 20900
+    },
+    {
+      "entropy": 2.2084882354736326,
+      "epoch": 0.9311111111111111,
+      "grad_norm": 6.90625,
+      "learning_rate": 1.4213058419243988e-06,
+      "loss": 0.8544,
+      "mean_token_accuracy": 0.7461783969402314,
+      "num_tokens": 57470110.0,
+      "step": 20950
+    },
+    {
+      "entropy": 2.143686933517456,
+      "epoch": 0.9333333333333333,
+      "grad_norm": 7.375,
+      "learning_rate": 1.375486827033219e-06,
+      "loss": 0.8695,
+      "mean_token_accuracy": 0.7416430413722992,
+      "num_tokens": 57537171.0,
+      "step": 21000
+    },
+    {
+      "entropy": 2.210377869606018,
+      "epoch": 0.9355555555555556,
+      "grad_norm": 6.75,
+      "learning_rate": 1.329667812142039e-06,
+      "loss": 0.8413,
+      "mean_token_accuracy": 0.7495397543907165,
+      "num_tokens": 57602216.0,
+      "step": 21050
+    },
+    {
+      "entropy": 2.186688220500946,
+      "epoch": 0.9377777777777778,
+      "grad_norm": 6.96875,
+      "learning_rate": 1.2838487972508592e-06,
+      "loss": 0.8411,
+      "mean_token_accuracy": 0.7513405895233154,
+      "num_tokens": 57669645.0,
+      "step": 21100
+    },
+    {
+      "entropy": 2.1919109773635865,
+      "epoch": 0.94,
+      "grad_norm": 6.96875,
+      "learning_rate": 1.2380297823596793e-06,
+      "loss": 0.8498,
+      "mean_token_accuracy": 0.7503270518779754,
+      "num_tokens": 57738922.0,
+      "step": 21150
+    },
+    {
+      "entropy": 2.2736938905715944,
+      "epoch": 0.9422222222222222,
+      "grad_norm": 9.0625,
+      "learning_rate": 1.1922107674684994e-06,
+      "loss": 0.8467,
+      "mean_token_accuracy": 0.7496572816371918,
+      "num_tokens": 57805121.0,
+      "step": 21200
+    },
+    {
+      "entropy": 2.2604011583328245,
+      "epoch": 0.9444444444444444,
+      "grad_norm": 7.28125,
+      "learning_rate": 1.1463917525773197e-06,
+      "loss": 0.8553,
+      "mean_token_accuracy": 0.7471484684944153,
+      "num_tokens": 57869045.0,
+      "step": 21250
+    },
+    {
+      "entropy": 2.236326413154602,
+      "epoch": 0.9466666666666667,
+      "grad_norm": 8.625,
+      "learning_rate": 1.1005727376861399e-06,
+      "loss": 0.8684,
+      "mean_token_accuracy": 0.7434519910812378,
+      "num_tokens": 57937773.0,
+      "step": 21300
+    },
+    {
+      "entropy": 2.2119676208496095,
+      "epoch": 0.9488888888888889,
+      "grad_norm": 7.5,
+      "learning_rate": 1.05475372279496e-06,
+      "loss": 0.8414,
+      "mean_token_accuracy": 0.7505972480773926,
+      "num_tokens": 57999509.0,
+      "step": 21350
+    },
+    {
+      "entropy": 2.19934800863266,
+      "epoch": 0.9511111111111111,
+      "grad_norm": 10.0,
+      "learning_rate": 1.0089347079037801e-06,
+      "loss": 0.8283,
+      "mean_token_accuracy": 0.7531795799732208,
+      "num_tokens": 58066324.0,
+      "step": 21400
+    },
+    {
+      "entropy": 2.189887263774872,
+      "epoch": 0.9533333333333334,
+      "grad_norm": 8.6875,
+      "learning_rate": 9.631156930126003e-07,
+      "loss": 0.8438,
+      "mean_token_accuracy": 0.7486509644985199,
+      "num_tokens": 58133522.0,
+      "step": 21450
+    },
+    {
+      "entropy": 2.2022621488571166,
+      "epoch": 0.9555555555555556,
+      "grad_norm": 12.0,
+      "learning_rate": 9.172966781214204e-07,
+      "loss": 0.8816,
+      "mean_token_accuracy": 0.7386648190021515,
+      "num_tokens": 58201032.0,
+      "step": 21500
+    },
+    {
+      "entropy": 2.2088757705688478,
+      "epoch": 0.9577777777777777,
+      "grad_norm": 7.65625,
+      "learning_rate": 8.714776632302406e-07,
+      "loss": 0.8685,
+      "mean_token_accuracy": 0.745578328371048,
+      "num_tokens": 58266051.0,
+      "step": 21550
+    },
+    {
+      "entropy": 2.2617397117614746,
+      "epoch": 0.96,
+      "grad_norm": 7.5625,
+      "learning_rate": 8.256586483390607e-07,
+      "loss": 0.857,
+      "mean_token_accuracy": 0.7448407518863678,
+      "num_tokens": 58332021.0,
+      "step": 21600
+    },
+    {
+      "entropy": 2.2229527854919433,
+      "epoch": 0.9622222222222222,
+      "grad_norm": 6.625,
+      "learning_rate": 7.79839633447881e-07,
+      "loss": 0.8636,
+      "mean_token_accuracy": 0.7455604159832001,
+      "num_tokens": 58396998.0,
+      "step": 21650
+    },
+    {
+      "entropy": 2.179794452190399,
+      "epoch": 0.9644444444444444,
+      "grad_norm": 7.71875,
+      "learning_rate": 7.340206185567011e-07,
+      "loss": 0.8293,
+      "mean_token_accuracy": 0.754916387796402,
+      "num_tokens": 58464526.0,
+      "step": 21700
+    },
+    {
+      "entropy": 2.2393770956993104,
+      "epoch": 0.9666666666666667,
+      "grad_norm": 12.4375,
+      "learning_rate": 6.882016036655212e-07,
+      "loss": 0.8434,
+      "mean_token_accuracy": 0.7503564131259918,
+      "num_tokens": 58530361.0,
+      "step": 21750
+    },
+    {
+      "entropy": 2.263950316905975,
+      "epoch": 0.9688888888888889,
+      "grad_norm": 6.40625,
+      "learning_rate": 6.423825887743414e-07,
+      "loss": 0.8466,
+      "mean_token_accuracy": 0.748320734500885,
+      "num_tokens": 58595415.0,
+      "step": 21800
+    },
+    {
+      "entropy": 2.3069614815711974,
+      "epoch": 0.9711111111111111,
+      "grad_norm": 8.4375,
+      "learning_rate": 5.965635738831616e-07,
+      "loss": 0.8578,
+      "mean_token_accuracy": 0.7452424871921539,
+      "num_tokens": 58658566.0,
+      "step": 21850
+    },
+    {
+      "entropy": 2.2174816036224367,
+      "epoch": 0.9733333333333334,
+      "grad_norm": 8.875,
+      "learning_rate": 5.507445589919817e-07,
+      "loss": 0.8814,
+      "mean_token_accuracy": 0.743362922668457,
+      "num_tokens": 58728364.0,
+      "step": 21900
+    },
+    {
+      "entropy": 2.2033449959754945,
+      "epoch": 0.9755555555555555,
+      "grad_norm": 7.15625,
+      "learning_rate": 5.049255441008018e-07,
+      "loss": 0.8709,
+      "mean_token_accuracy": 0.7446471619606018,
+      "num_tokens": 58796528.0,
+      "step": 21950
+    },
+    {
+      "entropy": 2.262139241695404,
+      "epoch": 0.9777777777777777,
+      "grad_norm": 12.75,
+      "learning_rate": 4.59106529209622e-07,
+      "loss": 0.8688,
+      "mean_token_accuracy": 0.7436162507534028,
+      "num_tokens": 58862439.0,
+      "step": 22000
+    },
+    {
+      "entropy": 2.2621299147605898,
+      "epoch": 0.98,
+      "grad_norm": 6.25,
+      "learning_rate": 4.132875143184422e-07,
+      "loss": 0.866,
+      "mean_token_accuracy": 0.7452659630775451,
+      "num_tokens": 58930357.0,
+      "step": 22050
+    },
+    {
+      "entropy": 2.2046650099754332,
+      "epoch": 0.9822222222222222,
+      "grad_norm": 8.0,
+      "learning_rate": 3.674684994272623e-07,
+      "loss": 0.8945,
+      "mean_token_accuracy": 0.7410675585269928,
+      "num_tokens": 58997717.0,
+      "step": 22100
+    },
+    {
+      "entropy": 2.210694715976715,
+      "epoch": 0.9844444444444445,
+      "grad_norm": 7.71875,
+      "learning_rate": 3.216494845360825e-07,
+      "loss": 0.8986,
+      "mean_token_accuracy": 0.7420401775836944,
+      "num_tokens": 59064018.0,
+      "step": 22150
+    },
+    {
+      "entropy": 2.196241397857666,
+      "epoch": 0.9866666666666667,
+      "grad_norm": 7.34375,
+      "learning_rate": 2.758304696449026e-07,
+      "loss": 0.9046,
+      "mean_token_accuracy": 0.7343410396575928,
+      "num_tokens": 59131839.0,
+      "step": 22200
+    },
+    {
+      "entropy": 2.1624761509895323,
+      "epoch": 0.9888888888888889,
+      "grad_norm": 8.6875,
+      "learning_rate": 2.3001145475372283e-07,
+      "loss": 0.8625,
+      "mean_token_accuracy": 0.749553587436676,
+      "num_tokens": 59199340.0,
+      "step": 22250
+    },
+    {
+      "entropy": 2.214530596733093,
+      "epoch": 0.9911111111111112,
+      "grad_norm": 7.21875,
+      "learning_rate": 1.8419243986254296e-07,
+      "loss": 0.9018,
+      "mean_token_accuracy": 0.7367948520183564,
+      "num_tokens": 59265961.0,
+      "step": 22300
+    },
+    {
+      "entropy": 2.189036545753479,
+      "epoch": 0.9933333333333333,
+      "grad_norm": 7.0625,
+      "learning_rate": 1.3837342497136314e-07,
+      "loss": 0.9118,
+      "mean_token_accuracy": 0.7363495242595672,
+      "num_tokens": 59335350.0,
+      "step": 22350
+    },
+    {
+      "entropy": 2.2046823930740356,
+      "epoch": 0.9955555555555555,
+      "grad_norm": 7.65625,
+      "learning_rate": 9.255441008018328e-08,
+      "loss": 0.8634,
+      "mean_token_accuracy": 0.7454369437694549,
+      "num_tokens": 59401509.0,
+      "step": 22400
+    },
+    {
+      "entropy": 2.1603779411315918,
+      "epoch": 0.9977777777777778,
+      "grad_norm": 8.25,
+      "learning_rate": 4.673539518900344e-08,
+      "loss": 0.8957,
+      "mean_token_accuracy": 0.741243121623993,
+      "num_tokens": 59470524.0,
+      "step": 22450
+    },
+    {
+      "entropy": 2.1624831557273865,
+      "epoch": 1.0,
+      "grad_norm": 9.0625,
+      "learning_rate": 9.163802978235968e-10,
+      "loss": 0.8587,
+      "mean_token_accuracy": 0.7448894453048706,
+      "num_tokens": 59536445.0,
+      "step": 22500
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 22500,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 9223372036854775807,
+  "save_steps": 5000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.515437740795392e+18,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}