diff --git "a/tldr-14b-step-832/trainer_state.json" "b/tldr-14b-step-832/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/tldr-14b-step-832/trainer_state.json"
@@ -0,0 +1,6091 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.416,
+  "eval_steps": 32,
+  "global_step": 832,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0005,
+      "grad_norm": 1.1535558177055725,
+      "learning_rate": 1e-05,
+      "loss": 0.4543,
+      "step": 1
+    },
+    {
+      "epoch": 0.001,
+      "grad_norm": 0.5039215671909039,
+      "learning_rate": 1e-05,
+      "loss": 0.3002,
+      "step": 2
+    },
+    {
+      "epoch": 0.0015,
+      "grad_norm": 0.4111085873239227,
+      "learning_rate": 1e-05,
+      "loss": 0.2138,
+      "step": 3
+    },
+    {
+      "epoch": 0.002,
+      "grad_norm": 0.3258240952011462,
+      "learning_rate": 1e-05,
+      "loss": 0.1761,
+      "step": 4
+    },
+    {
+      "epoch": 0.0025,
+      "grad_norm": 0.25417983371771585,
+      "learning_rate": 1e-05,
+      "loss": 0.1923,
+      "step": 5
+    },
+    {
+      "epoch": 0.003,
+      "grad_norm": 0.2023054568437071,
+      "learning_rate": 1e-05,
+      "loss": 0.2035,
+      "step": 6
+    },
+    {
+      "epoch": 0.0035,
+      "grad_norm": 0.29479546251817934,
+      "learning_rate": 1e-05,
+      "loss": 0.166,
+      "step": 7
+    },
+    {
+      "epoch": 0.004,
+      "grad_norm": 0.2603707531918473,
+      "learning_rate": 1e-05,
+      "loss": 0.1563,
+      "step": 8
+    },
+    {
+      "epoch": 0.0045,
+      "grad_norm": 0.27772543950112105,
+      "learning_rate": 1e-05,
+      "loss": 0.1742,
+      "step": 9
+    },
+    {
+      "epoch": 0.005,
+      "grad_norm": 0.33899668368493546,
+      "learning_rate": 1e-05,
+      "loss": 0.1623,
+      "step": 10
+    },
+    {
+      "epoch": 0.0055,
+      "grad_norm": 0.17293404759978015,
+      "learning_rate": 1e-05,
+      "loss": 0.1902,
+      "step": 11
+    },
+    {
+      "epoch": 0.006,
+      "grad_norm": 0.16812733049304054,
+      "learning_rate": 1e-05,
+      "loss": 0.1505,
+      "step": 12
+    },
+    {
+      "epoch": 0.0065,
+      "grad_norm": 0.24635507048250077,
+      "learning_rate": 1e-05,
+      "loss": 0.1496,
+      "step": 13
+    },
+    {
+      "epoch": 0.007,
+      "grad_norm": 0.23026986852648482,
+      "learning_rate": 1e-05,
+      "loss": 0.1428,
+      "step": 14
+    },
+    {
+      "epoch": 0.0075,
+      "grad_norm": 0.20312493098149043,
+      "learning_rate": 1e-05,
+      "loss": 0.1411,
+      "step": 15
+    },
+    {
+      "epoch": 0.008,
+      "grad_norm": 0.19056569111470645,
+      "learning_rate": 1e-05,
+      "loss": 0.1556,
+      "step": 16
+    },
+    {
+      "epoch": 0.0085,
+      "grad_norm": 0.19180586046592904,
+      "learning_rate": 1e-05,
+      "loss": 0.1448,
+      "step": 17
+    },
+    {
+      "epoch": 0.009,
+      "grad_norm": 0.23670452313674922,
+      "learning_rate": 1e-05,
+      "loss": 0.1272,
+      "step": 18
+    },
+    {
+      "epoch": 0.0095,
+      "grad_norm": 0.2241037709056984,
+      "learning_rate": 1e-05,
+      "loss": 0.1584,
+      "step": 19
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 0.1828368631692288,
+      "learning_rate": 1e-05,
+      "loss": 0.1347,
+      "step": 20
+    },
+    {
+      "epoch": 0.0105,
+      "grad_norm": 0.23115912688140353,
+      "learning_rate": 1e-05,
+      "loss": 0.1293,
+      "step": 21
+    },
+    {
+      "epoch": 0.011,
+      "grad_norm": 0.18886903067681482,
+      "learning_rate": 1e-05,
+      "loss": 0.1175,
+      "step": 22
+    },
+    {
+      "epoch": 0.0115,
+      "grad_norm": 0.19723236256202736,
+      "learning_rate": 1e-05,
+      "loss": 0.1665,
+      "step": 23
+    },
+    {
+      "epoch": 0.012,
+      "grad_norm": 0.2141017706399641,
+      "learning_rate": 1e-05,
+      "loss": 0.1491,
+      "step": 24
+    },
+    {
+      "epoch": 0.0125,
+      "grad_norm": 0.19090893978368054,
+      "learning_rate": 1e-05,
+      "loss": 0.1113,
+      "step": 25
+    },
+    {
+      "epoch": 0.013,
+      "grad_norm": 0.2460352826446744,
+      "learning_rate": 1e-05,
+      "loss": 0.1142,
+      "step": 26
+    },
+    {
+      "epoch": 0.0135,
+      "grad_norm": 0.21847319513439203,
+      "learning_rate": 1e-05,
+      "loss": 0.1273,
+      "step": 27
+    },
+    {
+      "epoch": 0.014,
+      "grad_norm": 0.18979686333150375,
+      "learning_rate": 1e-05,
+      "loss": 0.1441,
+      "step": 28
+    },
+    {
+      "epoch": 0.0145,
+      "grad_norm": 0.18886921913659824,
+      "learning_rate": 1e-05,
+      "loss": 0.1481,
+      "step": 29
+    },
+    {
+      "epoch": 0.015,
+      "grad_norm": 0.2115532479997522,
+      "learning_rate": 1e-05,
+      "loss": 0.0792,
+      "step": 30
+    },
+    {
+      "epoch": 0.0155,
+      "grad_norm": 0.20266301051804922,
+      "learning_rate": 1e-05,
+      "loss": 0.1379,
+      "step": 31
+    },
+    {
+      "epoch": 0.016,
+      "grad_norm": 0.1865900767381873,
+      "learning_rate": 1e-05,
+      "loss": 0.1453,
+      "step": 32
+    },
+    {
+      "epoch": 0.016,
+      "eval_dev_acc": 0.61328125,
+      "eval_dev_token": 5204.017578125,
+      "eval_runtime": 351.6008,
+      "eval_samples_per_second": 0.182,
+      "eval_steps_per_second": 0.003,
+      "step": 32
+    },
+    {
+      "epoch": 0.0165,
+      "grad_norm": 0.23402804608211233,
+      "learning_rate": 1e-05,
+      "loss": 0.1461,
+      "step": 33
+    },
+    {
+      "epoch": 0.017,
+      "grad_norm": 0.2191224488373381,
+      "learning_rate": 1e-05,
+      "loss": 0.1076,
+      "step": 34
+    },
+    {
+      "epoch": 0.0175,
+      "grad_norm": 0.18221820677824999,
+      "learning_rate": 1e-05,
+      "loss": 0.138,
+      "step": 35
+    },
+    {
+      "epoch": 0.018,
+      "grad_norm": 0.20326615377722793,
+      "learning_rate": 1e-05,
+      "loss": 0.1021,
+      "step": 36
+    },
+    {
+      "epoch": 0.0185,
+      "grad_norm": 0.2449658821374275,
+      "learning_rate": 1e-05,
+      "loss": 0.0881,
+      "step": 37
+    },
+    {
+      "epoch": 0.019,
+      "grad_norm": 0.2068478649138205,
+      "learning_rate": 1e-05,
+      "loss": 0.1249,
+      "step": 38
+    },
+    {
+      "epoch": 0.0195,
+      "grad_norm": 0.22630770783782977,
+      "learning_rate": 1e-05,
+      "loss": 0.1259,
+      "step": 39
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 0.21173641583720768,
+      "learning_rate": 1e-05,
+      "loss": 0.1331,
+      "step": 40
+    },
+    {
+      "epoch": 0.0205,
+      "grad_norm": 0.23616623193061376,
+      "learning_rate": 1e-05,
+      "loss": 0.1192,
+      "step": 41
+    },
+    {
+      "epoch": 0.021,
+      "grad_norm": 0.20789242969830385,
+      "learning_rate": 1e-05,
+      "loss": 0.159,
+      "step": 42
+    },
+    {
+      "epoch": 0.0215,
+      "grad_norm": 0.21662842275351119,
+      "learning_rate": 1e-05,
+      "loss": 0.1455,
+      "step": 43
+    },
+    {
+      "epoch": 0.022,
+      "grad_norm": 0.224914886884455,
+      "learning_rate": 1e-05,
+      "loss": 0.1565,
+      "step": 44
+    },
+    {
+      "epoch": 0.0225,
+      "grad_norm": 0.17642201019062015,
+      "learning_rate": 1e-05,
+      "loss": 0.1022,
+      "step": 45
+    },
+    {
+      "epoch": 0.023,
+      "grad_norm": 0.19476006095917964,
+      "learning_rate": 1e-05,
+      "loss": 0.1738,
+      "step": 46
+    },
+    {
+      "epoch": 0.0235,
+      "grad_norm": 0.2041987726527936,
+      "learning_rate": 1e-05,
+      "loss": 0.1545,
+      "step": 47
+    },
+    {
+      "epoch": 0.024,
+      "grad_norm": 0.15364946520211809,
+      "learning_rate": 1e-05,
+      "loss": 0.141,
+      "step": 48
+    },
+    {
+      "epoch": 0.0245,
+      "grad_norm": 0.1632596100583654,
+      "learning_rate": 1e-05,
+      "loss": 0.1493,
+      "step": 49
+    },
+    {
+      "epoch": 0.025,
+      "grad_norm": 0.23305215604624085,
+      "learning_rate": 1e-05,
+      "loss": 0.1164,
+      "step": 50
+    },
+    {
+      "epoch": 0.0255,
+      "grad_norm": 0.2697225282405861,
+      "learning_rate": 1e-05,
+      "loss": 0.0921,
+      "step": 51
+    },
+    {
+      "epoch": 0.026,
+      "grad_norm": 0.19242004892152365,
+      "learning_rate": 1e-05,
+      "loss": 0.1224,
+      "step": 52
+    },
+    {
+      "epoch": 0.0265,
+      "grad_norm": 0.21422189358330607,
+      "learning_rate": 1e-05,
+      "loss": 0.1084,
+      "step": 53
+    },
+    {
+      "epoch": 0.027,
+      "grad_norm": 0.22368520523252378,
+      "learning_rate": 1e-05,
+      "loss": 0.1018,
+      "step": 54
+    },
+    {
+      "epoch": 0.0275,
+      "grad_norm": 0.15315126193234804,
+      "learning_rate": 1e-05,
+      "loss": 0.117,
+      "step": 55
+    },
+    {
+      "epoch": 0.028,
+      "grad_norm": 0.2042745134173473,
+      "learning_rate": 1e-05,
+      "loss": 0.1355,
+      "step": 56
+    },
+    {
+      "epoch": 0.0285,
+      "grad_norm": 0.23889007522498773,
+      "learning_rate": 1e-05,
+      "loss": 0.1387,
+      "step": 57
+    },
+    {
+      "epoch": 0.029,
+      "grad_norm": 0.17240068749452392,
+      "learning_rate": 1e-05,
+      "loss": 0.1634,
+      "step": 58
+    },
+    {
+      "epoch": 0.0295,
+      "grad_norm": 0.1899912754500666,
+      "learning_rate": 1e-05,
+      "loss": 0.1027,
+      "step": 59
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 0.17118724561465037,
+      "learning_rate": 1e-05,
+      "loss": 0.1398,
+      "step": 60
+    },
+    {
+      "epoch": 0.0305,
+      "grad_norm": 0.19950059905690032,
+      "learning_rate": 1e-05,
+      "loss": 0.1118,
+      "step": 61
+    },
+    {
+      "epoch": 0.031,
+      "grad_norm": 0.19928365636144094,
+      "learning_rate": 1e-05,
+      "loss": 0.1146,
+      "step": 62
+    },
+    {
+      "epoch": 0.0315,
+      "grad_norm": 0.23146246196039105,
+      "learning_rate": 1e-05,
+      "loss": 0.0871,
+      "step": 63
+    },
+    {
+      "epoch": 0.032,
+      "grad_norm": 0.22746980065915193,
+      "learning_rate": 1e-05,
+      "loss": 0.1673,
+      "step": 64
+    },
+    {
+      "epoch": 0.032,
+      "eval_dev_acc": 0.541015625,
+      "eval_dev_token": 5677.76171875,
+      "eval_runtime": 358.7405,
+      "eval_samples_per_second": 0.178,
+      "eval_steps_per_second": 0.003,
+      "step": 64
+    },
+    {
+      "epoch": 0.0325,
+      "grad_norm": 0.1992092305273338,
+      "learning_rate": 1e-05,
+      "loss": 0.1292,
+      "step": 65
+    },
+    {
+      "epoch": 0.033,
+      "grad_norm": 0.19429880128063629,
+      "learning_rate": 1e-05,
+      "loss": 0.0911,
+      "step": 66
+    },
+    {
+      "epoch": 0.0335,
+      "grad_norm": 0.1287473705431077,
+      "learning_rate": 1e-05,
+      "loss": 0.1377,
+      "step": 67
+    },
+    {
+      "epoch": 0.034,
+      "grad_norm": 0.2228992327697556,
+      "learning_rate": 1e-05,
+      "loss": 0.0933,
+      "step": 68
+    },
+    {
+      "epoch": 0.0345,
+      "grad_norm": 0.18793882506839266,
+      "learning_rate": 1e-05,
+      "loss": 0.1097,
+      "step": 69
+    },
+    {
+      "epoch": 0.035,
+      "grad_norm": 0.1932965061071618,
+      "learning_rate": 1e-05,
+      "loss": 0.1062,
+      "step": 70
+    },
+    {
+      "epoch": 0.0355,
+      "grad_norm": 0.20585403698562318,
+      "learning_rate": 1e-05,
+      "loss": 0.1279,
+      "step": 71
+    },
+    {
+      "epoch": 0.036,
+      "grad_norm": 0.17833203451544913,
+      "learning_rate": 1e-05,
+      "loss": 0.1353,
+      "step": 72
+    },
+    {
+      "epoch": 0.0365,
+      "grad_norm": 0.21658266347490054,
+      "learning_rate": 1e-05,
+      "loss": 0.0961,
+      "step": 73
+    },
+    {
+      "epoch": 0.037,
+      "grad_norm": 0.2364173046757495,
+      "learning_rate": 1e-05,
+      "loss": 0.1038,
+      "step": 74
+    },
+    {
+      "epoch": 0.0375,
+      "grad_norm": 0.20844999333456934,
+      "learning_rate": 1e-05,
+      "loss": 0.1266,
+      "step": 75
+    },
+    {
+      "epoch": 0.038,
+      "grad_norm": 0.2288998611422715,
+      "learning_rate": 1e-05,
+      "loss": 0.0872,
+      "step": 76
+    },
+    {
+      "epoch": 0.0385,
+      "grad_norm": 0.1878179335817694,
+      "learning_rate": 1e-05,
+      "loss": 0.1038,
+      "step": 77
+    },
+    {
+      "epoch": 0.039,
+      "grad_norm": 0.19984002821227043,
+      "learning_rate": 1e-05,
+      "loss": 0.1376,
+      "step": 78
+    },
+    {
+      "epoch": 0.0395,
+      "grad_norm": 0.20050327741314344,
+      "learning_rate": 1e-05,
+      "loss": 0.135,
+      "step": 79
+    },
+    {
+      "epoch": 0.04,
+      "grad_norm": 0.19353835667751798,
+      "learning_rate": 1e-05,
+      "loss": 0.12,
+      "step": 80
+    },
+    {
+      "epoch": 0.0405,
+      "grad_norm": 0.12986340217496,
+      "learning_rate": 1e-05,
+      "loss": 0.1754,
+      "step": 81
+    },
+    {
+      "epoch": 0.041,
+      "grad_norm": 0.1889393654868388,
+      "learning_rate": 1e-05,
+      "loss": 0.1206,
+      "step": 82
+    },
+    {
+      "epoch": 0.0415,
+      "grad_norm": 0.17201063783314552,
+      "learning_rate": 1e-05,
+      "loss": 0.1401,
+      "step": 83
+    },
+    {
+      "epoch": 0.042,
+      "grad_norm": 0.19004092111820917,
+      "learning_rate": 1e-05,
+      "loss": 0.122,
+      "step": 84
+    },
+    {
+      "epoch": 0.0425,
+      "grad_norm": 0.21797692428743218,
+      "learning_rate": 1e-05,
+      "loss": 0.127,
+      "step": 85
+    },
+    {
+      "epoch": 0.043,
+      "grad_norm": 0.16779726003397347,
+      "learning_rate": 1e-05,
+      "loss": 0.1437,
+      "step": 86
+    },
+    {
+      "epoch": 0.0435,
+      "grad_norm": 0.23214262430834917,
+      "learning_rate": 1e-05,
+      "loss": 0.1187,
+      "step": 87
+    },
+    {
+      "epoch": 0.044,
+      "grad_norm": 0.19415443790822307,
+      "learning_rate": 1e-05,
+      "loss": 0.1063,
+      "step": 88
+    },
+    {
+      "epoch": 0.0445,
+      "grad_norm": 0.19254517148646239,
+      "learning_rate": 1e-05,
+      "loss": 0.1039,
+      "step": 89
+    },
+    {
+      "epoch": 0.045,
+      "grad_norm": 0.17307659461239167,
+      "learning_rate": 1e-05,
+      "loss": 0.1554,
+      "step": 90
+    },
+    {
+      "epoch": 0.0455,
+      "grad_norm": 0.24265029122082277,
+      "learning_rate": 1e-05,
+      "loss": 0.1387,
+      "step": 91
+    },
+    {
+      "epoch": 0.046,
+      "grad_norm": 0.17107218032177454,
+      "learning_rate": 1e-05,
+      "loss": 0.1301,
+      "step": 92
+    },
+    {
+      "epoch": 0.0465,
+      "grad_norm": 0.18075891622609033,
+      "learning_rate": 1e-05,
+      "loss": 0.1174,
+      "step": 93
+    },
+    {
+      "epoch": 0.047,
+      "grad_norm": 0.21595935885391185,
+      "learning_rate": 1e-05,
+      "loss": 0.1129,
+      "step": 94
+    },
+    {
+      "epoch": 0.0475,
+      "grad_norm": 0.23262812126963384,
+      "learning_rate": 1e-05,
+      "loss": 0.0983,
+      "step": 95
+    },
+    {
+      "epoch": 0.048,
+      "grad_norm": 0.20420999614021612,
+      "learning_rate": 1e-05,
+      "loss": 0.1161,
+      "step": 96
+    },
+    {
+      "epoch": 0.048,
+      "eval_dev_acc": 0.63671875,
+      "eval_dev_token": 5515.37109375,
+      "eval_runtime": 348.8656,
+      "eval_samples_per_second": 0.183,
+      "eval_steps_per_second": 0.003,
+      "step": 96
+    },
+    {
+      "epoch": 0.0485,
+      "grad_norm": 0.22980199994832454,
+      "learning_rate": 1e-05,
+      "loss": 0.1,
+      "step": 97
+    },
+    {
+      "epoch": 0.049,
+      "grad_norm": 0.19262135795217625,
+      "learning_rate": 1e-05,
+      "loss": 0.1124,
+      "step": 98
+    },
+    {
+      "epoch": 0.0495,
+      "grad_norm": 0.22869326649776367,
+      "learning_rate": 1e-05,
+      "loss": 0.1094,
+      "step": 99
+    },
+    {
+      "epoch": 0.05,
+      "grad_norm": 0.24173954720541516,
+      "learning_rate": 1e-05,
+      "loss": 0.0847,
+      "step": 100
+    },
+    {
+      "epoch": 0.0505,
+      "grad_norm": 0.2332215752101726,
+      "learning_rate": 1e-05,
+      "loss": 0.0946,
+      "step": 101
+    },
+    {
+      "epoch": 0.051,
+      "grad_norm": 0.26475494931892063,
+      "learning_rate": 1e-05,
+      "loss": 0.1094,
+      "step": 102
+    },
+    {
+      "epoch": 0.0515,
+      "grad_norm": 0.20833383536152603,
+      "learning_rate": 1e-05,
+      "loss": 0.0931,
+      "step": 103
+    },
+    {
+      "epoch": 0.052,
+      "grad_norm": 0.22676431614598225,
+      "learning_rate": 1e-05,
+      "loss": 0.0866,
+      "step": 104
+    },
+    {
+      "epoch": 0.0525,
+      "grad_norm": 0.22204744101927545,
+      "learning_rate": 1e-05,
+      "loss": 0.0984,
+      "step": 105
+    },
+    {
+      "epoch": 0.053,
+      "grad_norm": 0.2012349417230909,
+      "learning_rate": 1e-05,
+      "loss": 0.0899,
+      "step": 106
+    },
+    {
+      "epoch": 0.0535,
+      "grad_norm": 0.20131082411517556,
+      "learning_rate": 1e-05,
+      "loss": 0.095,
+      "step": 107
+    },
+    {
+      "epoch": 0.054,
+      "grad_norm": 0.2501268960199406,
+      "learning_rate": 1e-05,
+      "loss": 0.0872,
+      "step": 108
+    },
+    {
+      "epoch": 0.0545,
+      "grad_norm": 0.1877564425582315,
+      "learning_rate": 1e-05,
+      "loss": 0.1447,
+      "step": 109
+    },
+    {
+      "epoch": 0.055,
+      "grad_norm": 0.26373944955124323,
+      "learning_rate": 1e-05,
+      "loss": 0.1104,
+      "step": 110
+    },
+    {
+      "epoch": 0.0555,
+      "grad_norm": 0.2201267469286863,
+      "learning_rate": 1e-05,
+      "loss": 0.0864,
+      "step": 111
+    },
+    {
+      "epoch": 0.056,
+      "grad_norm": 0.2584324977531668,
+      "learning_rate": 1e-05,
+      "loss": 0.1243,
+      "step": 112
+    },
+    {
+      "epoch": 0.0565,
+      "grad_norm": 0.21251509346212935,
+      "learning_rate": 1e-05,
+      "loss": 0.1287,
+      "step": 113
+    },
+    {
+      "epoch": 0.057,
+      "grad_norm": 0.2391921924682281,
+      "learning_rate": 1e-05,
+      "loss": 0.1174,
+      "step": 114
+    },
+    {
+      "epoch": 0.0575,
+      "grad_norm": 0.2250627442441596,
+      "learning_rate": 1e-05,
+      "loss": 0.0961,
+      "step": 115
+    },
+    {
+      "epoch": 0.058,
+      "grad_norm": 0.21589217619835932,
+      "learning_rate": 1e-05,
+      "loss": 0.1041,
+      "step": 116
+    },
+    {
+      "epoch": 0.0585,
+      "grad_norm": 0.23947622053978754,
+      "learning_rate": 1e-05,
+      "loss": 0.1027,
+      "step": 117
+    },
+    {
+      "epoch": 0.059,
+      "grad_norm": 0.20944871960722214,
+      "learning_rate": 1e-05,
+      "loss": 0.1117,
+      "step": 118
+    },
+    {
+      "epoch": 0.0595,
+      "grad_norm": 0.2025724984657677,
+      "learning_rate": 1e-05,
+      "loss": 0.1102,
+      "step": 119
+    },
+    {
+      "epoch": 0.06,
+      "grad_norm": 0.1801908473330023,
+      "learning_rate": 1e-05,
+      "loss": 0.1072,
+      "step": 120
+    },
+    {
+      "epoch": 0.0605,
+      "grad_norm": 0.1940191661946107,
+      "learning_rate": 1e-05,
+      "loss": 0.0904,
+      "step": 121
+    },
+    {
+      "epoch": 0.061,
+      "grad_norm": 0.21867530545592728,
+      "learning_rate": 1e-05,
+      "loss": 0.1193,
+      "step": 122
+    },
+    {
+      "epoch": 0.0615,
+      "grad_norm": 0.22225462024793874,
+      "learning_rate": 1e-05,
+      "loss": 0.0782,
+      "step": 123
+    },
+    {
+      "epoch": 0.062,
+      "grad_norm": 0.24241920616546134,
+      "learning_rate": 1e-05,
+      "loss": 0.1099,
+      "step": 124
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 0.21360306106975577,
+      "learning_rate": 1e-05,
+      "loss": 0.091,
+      "step": 125
+    },
+    {
+      "epoch": 0.063,
+      "grad_norm": 0.2007423950283759,
+      "learning_rate": 1e-05,
+      "loss": 0.1152,
+      "step": 126
+    },
+    {
+      "epoch": 0.0635,
+      "grad_norm": 0.19343038731295426,
+      "learning_rate": 1e-05,
+      "loss": 0.1181,
+      "step": 127
+    },
+    {
+      "epoch": 0.064,
+      "grad_norm": 0.234786663580031,
+      "learning_rate": 1e-05,
+      "loss": 0.1149,
+      "step": 128
+    },
+    {
+      "epoch": 0.064,
+      "eval_dev_acc": 0.548828125,
+      "eval_dev_token": 5586.20703125,
+      "eval_runtime": 364.0015,
+      "eval_samples_per_second": 0.176,
+      "eval_steps_per_second": 0.003,
+      "step": 128
+    },
+    {
+      "epoch": 0.0645,
+      "grad_norm": 0.20468481909378916,
+      "learning_rate": 1e-05,
+      "loss": 0.0976,
+      "step": 129
+    },
+    {
+      "epoch": 0.065,
+      "grad_norm": 0.25320635754138643,
+      "learning_rate": 1e-05,
+      "loss": 0.1183,
+      "step": 130
+    },
+    {
+      "epoch": 0.0655,
+      "grad_norm": 0.21530698126365438,
+      "learning_rate": 1e-05,
+      "loss": 0.1156,
+      "step": 131
+    },
+    {
+      "epoch": 0.066,
+      "grad_norm": 0.20489101859011527,
+      "learning_rate": 1e-05,
+      "loss": 0.0991,
+      "step": 132
+    },
+    {
+      "epoch": 0.0665,
+      "grad_norm": 0.21113632835377186,
+      "learning_rate": 1e-05,
+      "loss": 0.0838,
+      "step": 133
+    },
+    {
+      "epoch": 0.067,
+      "grad_norm": 0.18341595697478763,
+      "learning_rate": 1e-05,
+      "loss": 0.1036,
+      "step": 134
+    },
+    {
+      "epoch": 0.0675,
+      "grad_norm": 0.1762785527366556,
+      "learning_rate": 1e-05,
+      "loss": 0.1674,
+      "step": 135
+    },
+    {
+      "epoch": 0.068,
+      "grad_norm": 0.2632947258005063,
+      "learning_rate": 1e-05,
+      "loss": 0.1202,
+      "step": 136
+    },
+    {
+      "epoch": 0.0685,
+      "grad_norm": 0.21085919863317307,
+      "learning_rate": 1e-05,
+      "loss": 0.1131,
+      "step": 137
+    },
+    {
+      "epoch": 0.069,
+      "grad_norm": 0.19457697084640746,
+      "learning_rate": 1e-05,
+      "loss": 0.1005,
+      "step": 138
+    },
+    {
+      "epoch": 0.0695,
+      "grad_norm": 0.17119664823512107,
+      "learning_rate": 1e-05,
+      "loss": 0.1389,
+      "step": 139
+    },
+    {
+      "epoch": 0.07,
+      "grad_norm": 0.19084161070023772,
+      "learning_rate": 1e-05,
+      "loss": 0.1527,
+      "step": 140
+    },
+    {
+      "epoch": 0.0705,
+      "grad_norm": 0.19580784724822164,
+      "learning_rate": 1e-05,
+      "loss": 0.1175,
+      "step": 141
+    },
+    {
+      "epoch": 0.071,
+      "grad_norm": 0.2110588281462844,
+      "learning_rate": 1e-05,
+      "loss": 0.1062,
+      "step": 142
+    },
+    {
+      "epoch": 0.0715,
+      "grad_norm": 0.20012341575489243,
+      "learning_rate": 1e-05,
+      "loss": 0.0848,
+      "step": 143
+    },
+    {
+      "epoch": 0.072,
+      "grad_norm": 0.1940479696118561,
+      "learning_rate": 1e-05,
+      "loss": 0.0694,
+      "step": 144
+    },
+    {
+      "epoch": 0.0725,
+      "grad_norm": 0.21593290579494073,
+      "learning_rate": 1e-05,
+      "loss": 0.0766,
+      "step": 145
+    },
+    {
+      "epoch": 0.073,
+      "grad_norm": 0.22638726501654005,
+      "learning_rate": 1e-05,
+      "loss": 0.1084,
+      "step": 146
+    },
+    {
+      "epoch": 0.0735,
+      "grad_norm": 0.20997037800742063,
+      "learning_rate": 1e-05,
+      "loss": 0.0761,
+      "step": 147
+    },
+    {
+      "epoch": 0.074,
+      "grad_norm": 0.2380179005894331,
+      "learning_rate": 1e-05,
+      "loss": 0.0927,
+      "step": 148
+    },
+    {
+      "epoch": 0.0745,
+      "grad_norm": 0.23889516090857615,
+      "learning_rate": 1e-05,
+      "loss": 0.0827,
+      "step": 149
+    },
+    {
+      "epoch": 0.075,
+      "grad_norm": 0.18767850358859223,
+      "learning_rate": 1e-05,
+      "loss": 0.0881,
+      "step": 150
+    },
+    {
+      "epoch": 0.0755,
+      "grad_norm": 0.19463069265385494,
+      "learning_rate": 1e-05,
+      "loss": 0.0917,
+      "step": 151
+    },
+    {
+      "epoch": 0.076,
+      "grad_norm": 0.212834963744102,
+      "learning_rate": 1e-05,
+      "loss": 0.0756,
+      "step": 152
+    },
+    {
+      "epoch": 0.0765,
+      "grad_norm": 0.20584172308777918,
+      "learning_rate": 1e-05,
+      "loss": 0.0762,
+      "step": 153
+    },
+    {
+      "epoch": 0.077,
+      "grad_norm": 0.2371449651260928,
+      "learning_rate": 1e-05,
+      "loss": 0.0978,
+      "step": 154
+    },
+    {
+      "epoch": 0.0775,
+      "grad_norm": 0.2049083024101962,
+      "learning_rate": 1e-05,
+      "loss": 0.0925,
+      "step": 155
+    },
+    {
+      "epoch": 0.078,
+      "grad_norm": 0.21217273061342656,
+      "learning_rate": 1e-05,
+      "loss": 0.0825,
+      "step": 156
+    },
+    {
+      "epoch": 0.0785,
+      "grad_norm": 0.20105825530151383,
+      "learning_rate": 1e-05,
+      "loss": 0.0858,
+      "step": 157
+    },
+    {
+      "epoch": 0.079,
+      "grad_norm": 0.2257052714675071,
+      "learning_rate": 1e-05,
+      "loss": 0.1105,
+      "step": 158
+    },
+    {
+      "epoch": 0.0795,
+      "grad_norm": 0.20210858652943217,
+      "learning_rate": 1e-05,
+      "loss": 0.1022,
+      "step": 159
+    },
+    {
+      "epoch": 0.08,
+      "grad_norm": 0.19482796495242663,
+      "learning_rate": 1e-05,
+      "loss": 0.1262,
+      "step": 160
+    },
+    {
+      "epoch": 0.08,
+      "eval_dev_acc": 0.6015625,
+      "eval_dev_token": 5543.318359375,
+      "eval_runtime": 353.7068,
+      "eval_samples_per_second": 0.181,
+      "eval_steps_per_second": 0.003,
+      "step": 160
+    },
+    {
+      "epoch": 0.0805,
+      "grad_norm": 0.22955090034416561,
+      "learning_rate": 1e-05,
+      "loss": 0.0988,
+      "step": 161
+    },
+    {
+      "epoch": 0.081,
+      "grad_norm": 0.23547588568917174,
+      "learning_rate": 1e-05,
+      "loss": 0.0803,
+      "step": 162
+    },
+    {
+      "epoch": 0.0815,
+      "grad_norm": 0.230658904399123,
+      "learning_rate": 1e-05,
+      "loss": 0.0881,
+      "step": 163
+    },
+    {
+      "epoch": 0.082,
+      "grad_norm": 0.2595571094242936,
+      "learning_rate": 1e-05,
+      "loss": 0.0881,
+      "step": 164
+    },
+    {
+      "epoch": 0.0825,
+      "grad_norm": 0.26763452927239884,
+      "learning_rate": 1e-05,
+      "loss": 0.114,
+      "step": 165
+    },
+    {
+      "epoch": 0.083,
+      "grad_norm": 0.257549186353109,
+      "learning_rate": 1e-05,
+      "loss": 0.1045,
+      "step": 166
+    },
+    {
+      "epoch": 0.0835,
+      "grad_norm": 0.19712751160118708,
+      "learning_rate": 1e-05,
+      "loss": 0.0899,
+      "step": 167
+    },
+    {
+      "epoch": 0.084,
+      "grad_norm": 0.17276675854807147,
+      "learning_rate": 1e-05,
+      "loss": 0.0684,
+      "step": 168
+    },
+    {
+      "epoch": 0.0845,
+      "grad_norm": 0.21103265575626073,
+      "learning_rate": 1e-05,
+      "loss": 0.0821,
+      "step": 169
+    },
+    {
+      "epoch": 0.085,
+      "grad_norm": 0.22292947141761962,
+      "learning_rate": 1e-05,
+      "loss": 0.0983,
+      "step": 170
+    },
+    {
+      "epoch": 0.0855,
+      "grad_norm": 0.21612821069411284,
+      "learning_rate": 1e-05,
+      "loss": 0.0917,
+      "step": 171
+    },
+    {
+      "epoch": 0.086,
+      "grad_norm": 0.20757781370778242,
+      "learning_rate": 1e-05,
+      "loss": 0.0919,
+      "step": 172
+    },
+    {
+      "epoch": 0.0865,
+      "grad_norm": 0.2054200483785948,
+      "learning_rate": 1e-05,
+      "loss": 0.077,
+      "step": 173
+    },
+    {
+      "epoch": 0.087,
+      "grad_norm": 0.24143931624393172,
+      "learning_rate": 1e-05,
+      "loss": 0.0919,
+      "step": 174
+    },
+    {
+      "epoch": 0.0875,
+      "grad_norm": 0.22708368965968964,
+      "learning_rate": 1e-05,
+      "loss": 0.0931,
+      "step": 175
+    },
+    {
+      "epoch": 0.088,
+      "grad_norm": 0.20044838709826737,
+      "learning_rate": 1e-05,
+      "loss": 0.0808,
+      "step": 176
+    },
+    {
+      "epoch": 0.0885,
+      "grad_norm": 0.20148090317828546,
+      "learning_rate": 1e-05,
+      "loss": 0.0905,
+      "step": 177
+    },
+    {
+      "epoch": 0.089,
+      "grad_norm": 0.2090109676571514,
+      "learning_rate": 1e-05,
+      "loss": 0.0801,
+      "step": 178
+    },
+    {
+      "epoch": 0.0895,
+      "grad_norm": 0.19147542578517765,
+      "learning_rate": 1e-05,
+      "loss": 0.0774,
+      "step": 179
+    },
+    {
+      "epoch": 0.09,
+      "grad_norm": 0.22051053694604383,
+      "learning_rate": 1e-05,
+      "loss": 0.0949,
+      "step": 180
+    },
+    {
+      "epoch": 0.0905,
+      "grad_norm": 0.20152485004966214,
+      "learning_rate": 1e-05,
+      "loss": 0.077,
+      "step": 181
+    },
+    {
+      "epoch": 0.091,
+      "grad_norm": 0.18668034234550573,
+      "learning_rate": 1e-05,
+      "loss": 0.0887,
+      "step": 182
+    },
+    {
+      "epoch": 0.0915,
+      "grad_norm": 0.21497572062563422,
+      "learning_rate": 1e-05,
+      "loss": 0.095,
+      "step": 183
+    },
+    {
+      "epoch": 0.092,
+      "grad_norm": 0.21716742542098177,
+      "learning_rate": 1e-05,
+      "loss": 0.0822,
+      "step": 184
+    },
+    {
+      "epoch": 0.0925,
+      "grad_norm": 0.20262525326126424,
+      "learning_rate": 1e-05,
+      "loss": 0.0804,
+      "step": 185
+    },
+    {
+      "epoch": 0.093,
+      "grad_norm": 0.18652482669251277,
+      "learning_rate": 1e-05,
+      "loss": 0.0959,
+      "step": 186
+    },
+    {
+      "epoch": 0.0935,
+      "grad_norm": 0.232270946947485,
+      "learning_rate": 1e-05,
+      "loss": 0.0785,
+      "step": 187
+    },
+    {
+      "epoch": 0.094,
+      "grad_norm": 0.22559277541611453,
+      "learning_rate": 1e-05,
+      "loss": 0.0868,
+      "step": 188
+    },
+    {
+      "epoch": 0.0945,
+      "grad_norm": 0.21772738934026295,
+      "learning_rate": 1e-05,
+      "loss": 0.1384,
+      "step": 189
+    },
+    {
+      "epoch": 0.095,
+      "grad_norm": 0.19366625753900965,
+      "learning_rate": 1e-05,
+      "loss": 0.0962,
+      "step": 190
+    },
+    {
+      "epoch": 0.0955,
+      "grad_norm": 0.2162137483161777,
+      "learning_rate": 1e-05,
+      "loss": 0.0753,
+      "step": 191
+    },
+    {
+      "epoch": 0.096,
+      "grad_norm": 0.2111612755929646,
+      "learning_rate": 1e-05,
+      "loss": 0.0776,
+      "step": 192
+    },
+    {
+      "epoch": 0.096,
+      "eval_dev_acc": 0.546875,
+      "eval_dev_token": 5439.14453125,
+      "eval_runtime": 358.6784,
+      "eval_samples_per_second": 0.178,
+      "eval_steps_per_second": 0.003,
+      "step": 192
+    },
+    {
+      "epoch": 0.0965,
+      "grad_norm": 0.20427587800007568,
+      "learning_rate": 1e-05,
+      "loss": 0.1026,
+      "step": 193
+    },
+    {
+      "epoch": 0.097,
+      "grad_norm": 0.1913558266102393,
+      "learning_rate": 1e-05,
+      "loss": 0.0947,
+      "step": 194
+    },
+    {
+      "epoch": 0.0975,
+      "grad_norm": 0.21823580107793827,
+      "learning_rate": 1e-05,
+      "loss": 0.0852,
+      "step": 195
+    },
+    {
+      "epoch": 0.098,
+      "grad_norm": 0.21947391378003933,
+      "learning_rate": 1e-05,
+      "loss": 0.0834,
+      "step": 196
+    },
+    {
+      "epoch": 0.0985,
+      "grad_norm": 0.2234108965736748,
+      "learning_rate": 1e-05,
+      "loss": 0.0922,
+      "step": 197
+    },
+    {
+      "epoch": 0.099,
+      "grad_norm": 0.20738196532743278,
+      "learning_rate": 1e-05,
+      "loss": 0.0814,
+      "step": 198
+    },
+    {
+      "epoch": 0.0995,
+      "grad_norm": 0.21586171895392783,
+      "learning_rate": 1e-05,
+      "loss": 0.0988,
+      "step": 199
+    },
+    {
+      "epoch": 0.1,
+      "grad_norm": 0.20785742252782521,
+      "learning_rate": 1e-05,
+      "loss": 0.0872,
+      "step": 200
+    },
+    {
+      "epoch": 0.1005,
+      "grad_norm": 0.22570317690675268,
+      "learning_rate": 1e-05,
+      "loss": 0.0937,
+      "step": 201
+    },
+    {
+      "epoch": 0.101,
+      "grad_norm": 0.19453877023547578,
+      "learning_rate": 1e-05,
+      "loss": 0.1083,
+      "step": 202
+    },
+    {
+      "epoch": 0.1015,
+      "grad_norm": 0.20591293588894416,
+      "learning_rate": 1e-05,
+      "loss": 0.0767,
+      "step": 203
+    },
+    {
+      "epoch": 0.102,
+      "grad_norm": 0.1798554117116506,
+      "learning_rate": 1e-05,
+      "loss": 0.1017,
+      "step": 204
+    },
+    {
+      "epoch": 0.1025,
+      "grad_norm": 0.1927231622743242,
+      "learning_rate": 1e-05,
+      "loss": 0.0911,
+      "step": 205
+    },
+    {
+      "epoch": 0.103,
+      "grad_norm": 0.21444122564196544,
+      "learning_rate": 1e-05,
+      "loss": 0.0856,
+      "step": 206
+    },
+    {
+      "epoch": 0.1035,
+      "grad_norm": 0.17259842125017608,
+      "learning_rate": 1e-05,
+      "loss": 0.0904,
+      "step": 207
+    },
+    {
+      "epoch": 0.104,
+      "grad_norm": 0.23979518316140722,
+      "learning_rate": 1e-05,
+      "loss": 0.084,
+      "step": 208
+    },
+    {
+      "epoch": 0.1045,
+      "grad_norm": 0.22442151254111703,
+      "learning_rate": 1e-05,
+      "loss": 0.0949,
+      "step": 209
+    },
+    {
+      "epoch": 0.105,
+      "grad_norm": 0.19615294291751353,
+      "learning_rate": 1e-05,
+      "loss": 0.0864,
+      "step": 210
+    },
+    {
+      "epoch": 0.1055,
+      "grad_norm": 0.18344154651920094,
+      "learning_rate": 1e-05,
+      "loss": 0.0843,
+      "step": 211
+    },
+    {
+      "epoch": 0.106,
+      "grad_norm": 0.21335203803361255,
+      "learning_rate": 1e-05,
+      "loss": 0.077,
+      "step": 212
+    },
+    {
+      "epoch": 0.1065,
+      "grad_norm": 0.252518444172673,
+      "learning_rate": 1e-05,
+      "loss": 0.0888,
+      "step": 213
+    },
+    {
+      "epoch": 0.107,
+      "grad_norm": 0.2116629068211744,
+      "learning_rate": 1e-05,
+      "loss": 0.098,
+      "step": 214
+    },
+    {
+      "epoch": 0.1075,
+      "grad_norm": 0.21920052784491295,
+      "learning_rate": 1e-05,
+      "loss": 0.0823,
+      "step": 215
+    },
+    {
+      "epoch": 0.108,
+      "grad_norm": 0.19779399311082105,
+      "learning_rate": 1e-05,
+      "loss": 0.0752,
+      "step": 216
+    },
+    {
+      "epoch": 0.1085,
+      "grad_norm": 0.19973286372655655,
+      "learning_rate": 1e-05,
+      "loss": 0.074,
+      "step": 217
+    },
+    {
+      "epoch": 0.109,
+      "grad_norm": 0.22343594632856933,
+      "learning_rate": 1e-05,
+      "loss": 0.0907,
+      "step": 218
+    },
+    {
+      "epoch": 0.1095,
+      "grad_norm": 0.25396426928555105,
+      "learning_rate": 1e-05,
+      "loss": 0.1075,
+      "step": 219
+    },
+    {
+      "epoch": 0.11,
+      "grad_norm": 0.1945715284952783,
+      "learning_rate": 1e-05,
+      "loss": 0.1053,
+      "step": 220
+    },
+    {
+      "epoch": 0.1105,
+      "grad_norm": 0.21559958220587308,
+      "learning_rate": 1e-05,
+      "loss": 0.0865,
+      "step": 221
+    },
+    {
+      "epoch": 0.111,
+      "grad_norm": 0.22369755043282374,
+      "learning_rate": 1e-05,
+      "loss": 0.0841,
+      "step": 222
+    },
+    {
+      "epoch": 0.1115,
+      "grad_norm": 0.2097379198995065,
+      "learning_rate": 1e-05,
+      "loss": 0.0753,
+      "step": 223
+    },
+    {
+      "epoch": 0.112,
+      "grad_norm": 0.17224880947033328,
+      "learning_rate": 1e-05,
+      "loss": 0.0699,
+      "step": 224
+    },
+    {
+      "epoch": 0.112,
+      "eval_dev_acc": 0.560546875,
+      "eval_dev_token": 5619.009765625,
+      "eval_runtime": 357.3428,
+      "eval_samples_per_second": 0.179,
+      "eval_steps_per_second": 0.003,
+      "step": 224
+    },
+    {
+      "epoch": 0.1125,
+      "grad_norm": 0.19442868536212735,
+      "learning_rate": 1e-05,
+      "loss": 0.0846,
+      "step": 225
+    },
+    {
+      "epoch": 0.113,
+      "grad_norm": 0.1573515275602218,
+      "learning_rate": 1e-05,
+      "loss": 0.0634,
+      "step": 226
+    },
+    {
+      "epoch": 0.1135,
+      "grad_norm": 0.17680777096637068,
+      "learning_rate": 1e-05,
+      "loss": 0.0761,
+      "step": 227
+    },
+    {
+      "epoch": 0.114,
+      "grad_norm": 0.20935704393341403,
+      "learning_rate": 1e-05,
+      "loss": 0.0549,
+      "step": 228
+    },
+    {
+      "epoch": 0.1145,
+      "grad_norm": 0.19829321187204563,
+      "learning_rate": 1e-05,
+      "loss": 0.05,
+      "step": 229
+    },
+    {
+      "epoch": 0.115,
+      "grad_norm": 0.18833561824346334,
+      "learning_rate": 1e-05,
+      "loss": 0.0656,
+      "step": 230
+    },
+    {
+      "epoch": 0.1155,
+      "grad_norm": 0.17277292328026173,
+      "learning_rate": 1e-05,
+      "loss": 0.08,
+      "step": 231
+    },
+    {
+      "epoch": 0.116,
+      "grad_norm": 0.2083709354078263,
+      "learning_rate": 1e-05,
+      "loss": 0.0628,
+      "step": 232
+    },
+    {
+      "epoch": 0.1165,
+      "grad_norm": 0.18113171413223286,
+      "learning_rate": 1e-05,
+      "loss": 0.0633,
+      "step": 233
+    },
+    {
+      "epoch": 0.117,
+      "grad_norm": 0.19985236934976783,
+      "learning_rate": 1e-05,
+      "loss": 0.0676,
+      "step": 234
+    },
+    {
+      "epoch": 0.1175,
+      "grad_norm": 0.2023196182410912,
+      "learning_rate": 1e-05,
+      "loss": 0.062,
+      "step": 235
+    },
+    {
+      "epoch": 0.118,
+      "grad_norm": 0.19446408540605106,
+      "learning_rate": 1e-05,
+      "loss": 0.0609,
+      "step": 236
+    },
+    {
+      "epoch": 0.1185,
+      "grad_norm": 0.1879635955015942,
+      "learning_rate": 1e-05,
+      "loss": 0.0631,
+      "step": 237
+    },
+    {
+      "epoch": 0.119,
+      "grad_norm": 0.18000098291861588,
+      "learning_rate": 1e-05,
+      "loss": 0.0628,
+      "step": 238
+    },
+    {
+      "epoch": 0.1195,
+      "grad_norm": 0.2091766063587954,
+      "learning_rate": 1e-05,
+      "loss": 0.0642,
+      "step": 239
+    },
+    {
+      "epoch": 0.12,
+      "grad_norm": 0.19781458462773657,
+      "learning_rate": 1e-05,
+      "loss": 0.0551,
+      "step": 240
+    },
+    {
+      "epoch": 0.1205,
+      "grad_norm": 0.20570535681429145,
+      "learning_rate": 1e-05,
+      "loss": 0.062,
+      "step": 241
+    },
+    {
+      "epoch": 0.121,
+      "grad_norm": 0.17479813291913535,
+      "learning_rate": 1e-05,
+      "loss": 0.0657,
+      "step": 242
+    },
+    {
+      "epoch": 0.1215,
+      "grad_norm": 0.20023576215167263,
+      "learning_rate": 1e-05,
+      "loss": 0.0545,
+      "step": 243
+    },
+    {
+      "epoch": 0.122,
+      "grad_norm": 0.21569894284947272,
+      "learning_rate": 1e-05,
+      "loss": 0.0641,
+      "step": 244
+    },
+    {
+      "epoch": 0.1225,
+      "grad_norm": 0.16426082027771785,
+      "learning_rate": 1e-05,
+      "loss": 0.0488,
+      "step": 245
+    },
+    {
+      "epoch": 0.123,
+      "grad_norm": 0.23142807048539513,
+      "learning_rate": 1e-05,
+      "loss": 0.0617,
+      "step": 246
+    },
+    {
+      "epoch": 0.1235,
+      "grad_norm": 0.21054969399806525,
+      "learning_rate": 1e-05,
+      "loss": 0.0566,
+      "step": 247
+    },
+    {
+      "epoch": 0.124,
+      "grad_norm": 0.1533567582820314,
+      "learning_rate": 1e-05,
+      "loss": 0.0559,
+      "step": 248
+    },
+    {
+      "epoch": 0.1245,
+      "grad_norm": 0.19025683614022437,
+      "learning_rate": 1e-05,
+      "loss": 0.051,
+      "step": 249
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 0.16411772241541067,
+      "learning_rate": 1e-05,
+      "loss": 0.068,
+      "step": 250
+    },
+    {
+      "epoch": 0.1255,
+      "grad_norm": 0.23699209914417435,
+      "learning_rate": 1e-05,
+      "loss": 0.0845,
+      "step": 251
+    },
+    {
+      "epoch": 0.126,
+      "grad_norm": 0.2047456890167515,
+      "learning_rate": 1e-05,
+      "loss": 0.0588,
+      "step": 252
+    },
+    {
+      "epoch": 0.1265,
+      "grad_norm": 0.20625503133501016,
+      "learning_rate": 1e-05,
+      "loss": 0.0543,
+      "step": 253
+    },
+    {
+      "epoch": 0.127,
+      "grad_norm": 0.16081505489801892,
+      "learning_rate": 1e-05,
+      "loss": 0.0413,
+      "step": 254
+    },
+    {
+      "epoch": 0.1275,
+      "grad_norm": 0.20221086748641462,
+      "learning_rate": 1e-05,
+      "loss": 0.0704,
+      "step": 255
+    },
+    {
+      "epoch": 0.128,
+      "grad_norm": 0.20711399972324054,
+      "learning_rate": 1e-05,
+      "loss": 0.0563,
+      "step": 256
+    },
+    {
+      "epoch": 0.128,
+      "eval_dev_acc": 0.587890625,
+      "eval_dev_token": 5704.3125,
+      "eval_runtime": 356.7726,
+      "eval_samples_per_second": 0.179,
+      "eval_steps_per_second": 0.003,
+      "step": 256
+    },
+    {
+      "epoch": 0.1285,
+      "grad_norm": 0.18942704583355485,
+      "learning_rate": 1e-05,
+      "loss": 0.0692,
+      "step": 257
+    },
+    {
+      "epoch": 0.129,
+      "grad_norm": 0.19171984906136896,
+      "learning_rate": 1e-05,
+      "loss": 0.05,
+      "step": 258
+    },
+    {
+      "epoch": 0.1295,
+      "grad_norm": 0.2125382567332232,
+      "learning_rate": 1e-05,
+      "loss": 0.0615,
+      "step": 259
+    },
+    {
+      "epoch": 0.13,
+      "grad_norm": 0.17877261040661208,
+      "learning_rate": 1e-05,
+      "loss": 0.0597,
+      "step": 260
+    },
+    {
+      "epoch": 0.1305,
+      "grad_norm": 0.1708210973205684,
+      "learning_rate": 1e-05,
+      "loss": 0.067,
+      "step": 261
+    },
+    {
+      "epoch": 0.131,
+      "grad_norm": 0.1850887764718648,
+      "learning_rate": 1e-05,
+      "loss": 0.0544,
+      "step": 262
+    },
+    {
+      "epoch": 0.1315,
+      "grad_norm": 0.2010060822058417,
+      "learning_rate": 1e-05,
+      "loss": 0.0696,
+      "step": 263
+    },
+    {
+      "epoch": 0.132,
+      "grad_norm": 0.18460835555899294,
+      "learning_rate": 1e-05,
+      "loss": 0.0607,
+      "step": 264
+    },
+    {
+      "epoch": 0.1325,
+      "grad_norm": 0.2264686856441524,
+      "learning_rate": 1e-05,
+      "loss": 0.0638,
+      "step": 265
+    },
+    {
+      "epoch": 0.133,
+      "grad_norm": 0.17252712599660533,
+      "learning_rate": 1e-05,
+      "loss": 0.055,
+      "step": 266
+    },
+    {
+      "epoch": 0.1335,
+      "grad_norm": 0.1668268871760919,
+      "learning_rate": 1e-05,
+      "loss": 0.0467,
+      "step": 267
+    },
+    {
+      "epoch": 0.134,
+      "grad_norm": 0.17711472127782535,
+      "learning_rate": 1e-05,
+      "loss": 0.0462,
+      "step": 268
+    },
+    {
+      "epoch": 0.1345,
+      "grad_norm": 0.16354851889499628,
+      "learning_rate": 1e-05,
+      "loss": 0.0687,
+      "step": 269
+    },
+    {
+      "epoch": 0.135,
+      "grad_norm": 0.17844367901102645,
+      "learning_rate": 1e-05,
+      "loss": 0.0455,
+      "step": 270
+    },
+    {
+      "epoch": 0.1355,
+      "grad_norm": 0.19303024902618696,
+      "learning_rate": 1e-05,
+      "loss": 0.0565,
+      "step": 271
+    },
+    {
+      "epoch": 0.136,
+      "grad_norm": 0.19049740233006035,
+      "learning_rate": 1e-05,
+      "loss": 0.0575,
+      "step": 272
+    },
+    {
+      "epoch": 0.1365,
+      "grad_norm": 0.18444934835307936,
+      "learning_rate": 1e-05,
+      "loss": 0.0495,
+      "step": 273
+    },
+    {
+      "epoch": 0.137,
+      "grad_norm": 0.2029153556589725,
+      "learning_rate": 1e-05,
+      "loss": 0.0632,
+      "step": 274
+    },
+    {
+      "epoch": 0.1375,
+      "grad_norm": 0.17742276348080663,
+      "learning_rate": 1e-05,
+      "loss": 0.0591,
+      "step": 275
+    },
+    {
+      "epoch": 0.138,
+      "grad_norm": 0.2086941032177491,
+      "learning_rate": 1e-05,
+      "loss": 0.0453,
+      "step": 276
+    },
+    {
+      "epoch": 0.1385,
+      "grad_norm": 0.1599025673655474,
+      "learning_rate": 1e-05,
+      "loss": 0.0346,
+      "step": 277
+    },
+    {
+      "epoch": 0.139,
+      "grad_norm": 0.20223917188200294,
+      "learning_rate": 1e-05,
+      "loss": 0.0516,
+      "step": 278
+    },
+    {
+      "epoch": 0.1395,
+      "grad_norm": 0.168572629555483,
+      "learning_rate": 1e-05,
+      "loss": 0.0502,
+      "step": 279
+    },
+    {
+      "epoch": 0.14,
+      "grad_norm": 0.21316857087336016,
+      "learning_rate": 1e-05,
+      "loss": 0.0585,
+      "step": 280
+    },
+    {
+      "epoch": 0.1405,
+      "grad_norm": 0.1533009935920478,
+      "learning_rate": 1e-05,
+      "loss": 0.0732,
+      "step": 281
+    },
+    {
+      "epoch": 0.141,
+      "grad_norm": 0.2226592493652288,
+      "learning_rate": 1e-05,
+      "loss": 0.0683,
+      "step": 282
+    },
+    {
+      "epoch": 0.1415,
+      "grad_norm": 0.2005052634299014,
+      "learning_rate": 1e-05,
+      "loss": 0.064,
+      "step": 283
+    },
+    {
+      "epoch": 0.142,
+      "grad_norm": 0.16977898059181232,
+      "learning_rate": 1e-05,
+      "loss": 0.0759,
+      "step": 284
+    },
+    {
+      "epoch": 0.1425,
+      "grad_norm": 0.17622048198257903,
+      "learning_rate": 1e-05,
+      "loss": 0.0546,
+      "step": 285
+    },
+    {
+      "epoch": 0.143,
+      "grad_norm": 0.15734345344681852,
+      "learning_rate": 1e-05,
+      "loss": 0.0371,
+      "step": 286
+    },
+    {
+      "epoch": 0.1435,
+      "grad_norm": 0.16097527322211574,
+      "learning_rate": 1e-05,
+      "loss": 0.0523,
+      "step": 287
+    },
+    {
+      "epoch": 0.144,
+      "grad_norm": 0.18490744056020517,
+      "learning_rate": 1e-05,
+      "loss": 0.0518,
+      "step": 288
+    },
+    {
+      "epoch": 0.144,
+      "eval_dev_acc": 0.5234375,
+      "eval_dev_token": 5945.451171875,
+      "eval_runtime": 402.5197,
+      "eval_samples_per_second": 0.159,
+      "eval_steps_per_second": 0.002,
+      "step": 288
+    },
+    {
+      "epoch": 0.1445,
+      "grad_norm": 0.16803932674344507,
+      "learning_rate": 1e-05,
+      "loss": 0.0564,
+      "step": 289
+    },
+    {
+      "epoch": 0.145,
+      "grad_norm": 0.18279683357896828,
+      "learning_rate": 1e-05,
+      "loss": 0.0604,
+      "step": 290
+    },
+    {
+      "epoch": 0.1455,
+      "grad_norm": 0.1773550481655194,
+      "learning_rate": 1e-05,
+      "loss": 0.0465,
+      "step": 291
+    },
+    {
+      "epoch": 0.146,
+      "grad_norm": 0.37855180864427374,
+      "learning_rate": 1e-05,
+      "loss": 0.0687,
+      "step": 292
+    },
+    {
+      "epoch": 0.1465,
+      "grad_norm": 0.2256173969538139,
+      "learning_rate": 1e-05,
+      "loss": 0.0597,
+      "step": 293
+    },
+    {
+      "epoch": 0.147,
+      "grad_norm": 0.19318315047474646,
+      "learning_rate": 1e-05,
+      "loss": 0.05,
+      "step": 294
+    },
+    {
+      "epoch": 0.1475,
+      "grad_norm": 0.1949573691249157,
+      "learning_rate": 1e-05,
+      "loss": 0.054,
+      "step": 295
+    },
+    {
+      "epoch": 0.148,
+      "grad_norm": 0.1784893290182381,
+      "learning_rate": 1e-05,
+      "loss": 0.0551,
+      "step": 296
+    },
+    {
+      "epoch": 0.1485,
+      "grad_norm": 0.1893518286787237,
+      "learning_rate": 1e-05,
+      "loss": 0.0582,
+      "step": 297
+    },
+    {
+      "epoch": 0.149,
+      "grad_norm": 0.16491423015511872,
+      "learning_rate": 1e-05,
+      "loss": 0.0393,
+      "step": 298
+    },
+    {
+      "epoch": 0.1495,
+      "grad_norm": 0.1762274160177828,
+      "learning_rate": 1e-05,
+      "loss": 0.0485,
+      "step": 299
+    },
+    {
+      "epoch": 0.15,
+      "grad_norm": 0.18398727773770782,
+      "learning_rate": 1e-05,
+      "loss": 0.0573,
+      "step": 300
+    },
+    {
+      "epoch": 0.1505,
+      "grad_norm": 0.18217281656040227,
+      "learning_rate": 1e-05,
+      "loss": 0.0485,
+      "step": 301
+    },
+    {
+      "epoch": 0.151,
+      "grad_norm": 0.17276763991718358,
+      "learning_rate": 1e-05,
+      "loss": 0.0571,
+      "step": 302
+    },
+    {
+      "epoch": 0.1515,
+      "grad_norm": 0.15668609882813492,
+      "learning_rate": 1e-05,
+      "loss": 0.0741,
+      "step": 303
+    },
+    {
+      "epoch": 0.152,
+      "grad_norm": 0.18965888700513778,
+      "learning_rate": 1e-05,
+      "loss": 0.0548,
+      "step": 304
+    },
+    {
+      "epoch": 0.1525,
+      "grad_norm": 0.1514861962476675,
+      "learning_rate": 1e-05,
+      "loss": 0.0408,
+      "step": 305
+    },
+    {
+      "epoch": 0.153,
+      "grad_norm": 0.21027930287961952,
+      "learning_rate": 1e-05,
+      "loss": 0.0473,
+      "step": 306
+    },
+    {
+      "epoch": 0.1535,
+      "grad_norm": 0.20086505297048218,
+      "learning_rate": 1e-05,
+      "loss": 0.0457,
+      "step": 307
+    },
+    {
+      "epoch": 0.154,
+      "grad_norm": 0.1834058060370301,
+      "learning_rate": 1e-05,
+      "loss": 0.0391,
+      "step": 308
+    },
+    {
+      "epoch": 0.1545,
+      "grad_norm": 0.1675035648173745,
+      "learning_rate": 1e-05,
+      "loss": 0.0346,
+      "step": 309
+    },
+    {
+      "epoch": 0.155,
+      "grad_norm": 0.19041217604042332,
+      "learning_rate": 1e-05,
+      "loss": 0.0447,
+      "step": 310
+    },
+    {
+      "epoch": 0.1555,
+      "grad_norm": 0.2063641120441124,
+      "learning_rate": 1e-05,
+      "loss": 0.0454,
+      "step": 311
+    },
+    {
+      "epoch": 0.156,
+      "grad_norm": 0.18205494058640856,
+      "learning_rate": 1e-05,
+      "loss": 0.0402,
+      "step": 312
+    },
+    {
+      "epoch": 0.1565,
+      "grad_norm": 0.18642039675473027,
+      "learning_rate": 1e-05,
+      "loss": 0.0397,
+      "step": 313
+    },
+    {
+      "epoch": 0.157,
+      "grad_norm": 0.16971154384699963,
+      "learning_rate": 1e-05,
+      "loss": 0.0426,
+      "step": 314
+    },
+    {
+      "epoch": 0.1575,
+      "grad_norm": 0.19779499600374534,
+      "learning_rate": 1e-05,
+      "loss": 0.042,
+      "step": 315
+    },
+    {
+      "epoch": 0.158,
+      "grad_norm": 0.1597506922805199,
+      "learning_rate": 1e-05,
+      "loss": 0.0561,
+      "step": 316
+    },
+    {
+      "epoch": 0.1585,
+      "grad_norm": 0.22455898943377722,
+      "learning_rate": 1e-05,
+      "loss": 0.0508,
+      "step": 317
+    },
+    {
+      "epoch": 0.159,
+      "grad_norm": 0.22816888070811367,
+      "learning_rate": 1e-05,
+      "loss": 0.0684,
+      "step": 318
+    },
+    {
+      "epoch": 0.1595,
+      "grad_norm": 0.20829314487846406,
+      "learning_rate": 1e-05,
+      "loss": 0.0558,
+      "step": 319
+    },
+    {
+      "epoch": 0.16,
+      "grad_norm": 0.1726376642892394,
+      "learning_rate": 1e-05,
+      "loss": 0.0382,
+      "step": 320
+    },
+    {
+      "epoch": 0.16,
+      "eval_dev_acc": 0.62109375,
+      "eval_dev_token": 5348.43359375,
+      "eval_runtime": 348.0694,
+      "eval_samples_per_second": 0.184,
+      "eval_steps_per_second": 0.003,
+      "step": 320
+    },
+    {
+      "epoch": 0.1605,
+      "grad_norm": 0.21533911997871472,
+      "learning_rate": 1e-05,
+      "loss": 0.0573,
+      "step": 321
+    },
+    {
+      "epoch": 0.161,
+      "grad_norm": 0.1945926926705701,
+      "learning_rate": 1e-05,
+      "loss": 0.0482,
+      "step": 322
+    },
+    {
+      "epoch": 0.1615,
+      "grad_norm": 0.16282045186801866,
+      "learning_rate": 1e-05,
+      "loss": 0.0395,
+      "step": 323
+    },
+    {
+      "epoch": 0.162,
+      "grad_norm": 0.21192465805738625,
+      "learning_rate": 1e-05,
+      "loss": 0.0534,
+      "step": 324
+    },
+    {
+      "epoch": 0.1625,
+      "grad_norm": 0.18412473332196624,
+      "learning_rate": 1e-05,
+      "loss": 0.0457,
+      "step": 325
+    },
+    {
+      "epoch": 0.163,
+      "grad_norm": 0.1760683716395308,
+      "learning_rate": 1e-05,
+      "loss": 0.0621,
+      "step": 326
+    },
+    {
+      "epoch": 0.1635,
+      "grad_norm": 0.1977120327808734,
+      "learning_rate": 1e-05,
+      "loss": 0.0443,
+      "step": 327
+    },
+    {
+      "epoch": 0.164,
+      "grad_norm": 0.17834928432327893,
+      "learning_rate": 1e-05,
+      "loss": 0.0474,
+      "step": 328
+    },
+    {
+      "epoch": 0.1645,
+      "grad_norm": 0.19810362807929732,
+      "learning_rate": 1e-05,
+      "loss": 0.0472,
+      "step": 329
+    },
+    {
+      "epoch": 0.165,
+      "grad_norm": 0.17306554655229037,
+      "learning_rate": 1e-05,
+      "loss": 0.0514,
+      "step": 330
+    },
+    {
+      "epoch": 0.1655,
+      "grad_norm": 0.1732660830513622,
+      "learning_rate": 1e-05,
+      "loss": 0.0487,
+      "step": 331
+    },
+    {
+      "epoch": 0.166,
+      "grad_norm": 0.15153749458139032,
+      "learning_rate": 1e-05,
+      "loss": 0.044,
+      "step": 332
+    },
+    {
+      "epoch": 0.1665,
+      "grad_norm": 0.20187085045913772,
+      "learning_rate": 1e-05,
+      "loss": 0.0551,
+      "step": 333
+    },
+    {
+      "epoch": 0.167,
+      "grad_norm": 0.16579582791974742,
+      "learning_rate": 1e-05,
+      "loss": 0.0497,
+      "step": 334
+    },
+    {
+      "epoch": 0.1675,
+      "grad_norm": 0.19316064563692958,
+      "learning_rate": 1e-05,
+      "loss": 0.0549,
+      "step": 335
+    },
+    {
+      "epoch": 0.168,
+      "grad_norm": 0.16491714800111232,
+      "learning_rate": 1e-05,
+      "loss": 0.0371,
+      "step": 336
+    },
+    {
+      "epoch": 0.1685,
+      "grad_norm": 0.17551519178449135,
+      "learning_rate": 1e-05,
+      "loss": 0.041,
+      "step": 337
+    },
+    {
+      "epoch": 0.169,
+      "grad_norm": 0.1734781492111704,
+      "learning_rate": 1e-05,
+      "loss": 0.0441,
+      "step": 338
+    },
+    {
+      "epoch": 0.1695,
+      "grad_norm": 0.18684315556716974,
+      "learning_rate": 1e-05,
+      "loss": 0.0488,
+      "step": 339
+    },
+    {
+      "epoch": 0.17,
+      "grad_norm": 0.16917376679576818,
+      "learning_rate": 1e-05,
+      "loss": 0.0437,
+      "step": 340
+    },
+    {
+      "epoch": 0.1705,
+      "grad_norm": 0.1501957333162884,
+      "learning_rate": 1e-05,
+      "loss": 0.0393,
+      "step": 341
+    },
+    {
+      "epoch": 0.171,
+      "grad_norm": 0.2205121105977978,
+      "learning_rate": 1e-05,
+      "loss": 0.0579,
+      "step": 342
+    },
+    {
+      "epoch": 0.1715,
+      "grad_norm": 0.18041963938373073,
+      "learning_rate": 1e-05,
+      "loss": 0.0492,
+      "step": 343
+    },
+    {
+      "epoch": 0.172,
+      "grad_norm": 0.17449103579952038,
+      "learning_rate": 1e-05,
+      "loss": 0.0401,
+      "step": 344
+    },
+    {
+      "epoch": 0.1725,
+      "grad_norm": 0.15528177663303303,
+      "learning_rate": 1e-05,
+      "loss": 0.0379,
+      "step": 345
+    },
+    {
+      "epoch": 0.173,
+      "grad_norm": 0.20473861699417553,
+      "learning_rate": 1e-05,
+      "loss": 0.0499,
+      "step": 346
+    },
+    {
+      "epoch": 0.1735,
+      "grad_norm": 0.22954222855804748,
+      "learning_rate": 1e-05,
+      "loss": 0.066,
+      "step": 347
+    },
+    {
+      "epoch": 0.174,
+      "grad_norm": 0.18144976711639327,
+      "learning_rate": 1e-05,
+      "loss": 0.0417,
+      "step": 348
+    },
+    {
+      "epoch": 0.1745,
+      "grad_norm": 0.1647885822020398,
+      "learning_rate": 1e-05,
+      "loss": 0.0406,
+      "step": 349
+    },
+    {
+      "epoch": 0.175,
+      "grad_norm": 0.13381930823784724,
+      "learning_rate": 1e-05,
+      "loss": 0.039,
+      "step": 350
+    },
+    {
+      "epoch": 0.1755,
+      "grad_norm": 0.17233840695200286,
+      "learning_rate": 1e-05,
+      "loss": 0.0504,
+      "step": 351
+    },
+    {
+      "epoch": 0.176,
+      "grad_norm": 0.18220085493590332,
+      "learning_rate": 1e-05,
+      "loss": 0.0411,
+      "step": 352
+    },
+    {
+      "epoch": 0.176,
+      "eval_dev_acc": 0.53515625,
+      "eval_dev_token": 5787.59375,
+      "eval_runtime": 364.0085,
+      "eval_samples_per_second": 0.176,
+      "eval_steps_per_second": 0.003,
+      "step": 352
+    },
+    {
+      "epoch": 0.1765,
+      "grad_norm": 0.20744185724682074,
+      "learning_rate": 1e-05,
+      "loss": 0.0606,
+      "step": 353
+    },
+    {
+      "epoch": 0.177,
+      "grad_norm": 0.18600570998480834,
+      "learning_rate": 1e-05,
+      "loss": 0.0416,
+      "step": 354
+    },
+    {
+      "epoch": 0.1775,
+      "grad_norm": 0.1776469292641811,
+      "learning_rate": 1e-05,
+      "loss": 0.042,
+      "step": 355
+    },
+    {
+      "epoch": 0.178,
+      "grad_norm": 0.1783568782453835,
+      "learning_rate": 1e-05,
+      "loss": 0.0457,
+      "step": 356
+    },
+    {
+      "epoch": 0.1785,
+      "grad_norm": 0.1981729227656145,
+      "learning_rate": 1e-05,
+      "loss": 0.0578,
+      "step": 357
+    },
+    {
+      "epoch": 0.179,
+      "grad_norm": 0.18984703197303243,
+      "learning_rate": 1e-05,
+      "loss": 0.0427,
+      "step": 358
+    },
+    {
+      "epoch": 0.1795,
+      "grad_norm": 0.21799846739281004,
+      "learning_rate": 1e-05,
+      "loss": 0.0492,
+      "step": 359
+    },
+    {
+      "epoch": 0.18,
+      "grad_norm": 0.2263278306065525,
+      "learning_rate": 1e-05,
+      "loss": 0.0708,
+      "step": 360
+    },
+    {
+      "epoch": 0.1805,
+      "grad_norm": 0.16612125798758726,
+      "learning_rate": 1e-05,
+      "loss": 0.0447,
+      "step": 361
+    },
+    {
+      "epoch": 0.181,
+      "grad_norm": 0.15311766872023147,
+      "learning_rate": 1e-05,
+      "loss": 0.0369,
+      "step": 362
+    },
+    {
+      "epoch": 0.1815,
+      "grad_norm": 0.18614757917185834,
+      "learning_rate": 1e-05,
+      "loss": 0.0484,
+      "step": 363
+    },
+    {
+      "epoch": 0.182,
+      "grad_norm": 0.18253431030668557,
+      "learning_rate": 1e-05,
+      "loss": 0.044,
+      "step": 364
+    },
+    {
+      "epoch": 0.1825,
+      "grad_norm": 0.19238661256236803,
+      "learning_rate": 1e-05,
+      "loss": 0.0615,
+      "step": 365
+    },
+    {
+      "epoch": 0.183,
+      "grad_norm": 0.1808155264273284,
+      "learning_rate": 1e-05,
+      "loss": 0.0444,
+      "step": 366
+    },
+    {
+      "epoch": 0.1835,
+      "grad_norm": 0.1743762662626829,
+      "learning_rate": 1e-05,
+      "loss": 0.0501,
+      "step": 367
+    },
+    {
+      "epoch": 0.184,
+      "grad_norm": 0.16508774246157967,
+      "learning_rate": 1e-05,
+      "loss": 0.0448,
+      "step": 368
+    },
+    {
+      "epoch": 0.1845,
+      "grad_norm": 0.1546243168773746,
+      "learning_rate": 1e-05,
+      "loss": 0.0366,
+      "step": 369
+    },
+    {
+      "epoch": 0.185,
+      "grad_norm": 0.1746189186464954,
+      "learning_rate": 1e-05,
+      "loss": 0.0471,
+      "step": 370
+    },
+    {
+      "epoch": 0.1855,
+      "grad_norm": 0.17995461422580256,
+      "learning_rate": 1e-05,
+      "loss": 0.0405,
+      "step": 371
+    },
+    {
+      "epoch": 0.186,
+      "grad_norm": 0.16745033647841967,
+      "learning_rate": 1e-05,
+      "loss": 0.0371,
+      "step": 372
+    },
+    {
+      "epoch": 0.1865,
+      "grad_norm": 0.14177227347565124,
+      "learning_rate": 1e-05,
+      "loss": 0.0336,
+      "step": 373
+    },
+    {
+      "epoch": 0.187,
+      "grad_norm": 0.19568633642105135,
+      "learning_rate": 1e-05,
+      "loss": 0.0419,
+      "step": 374
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 0.1694809590901385,
+      "learning_rate": 1e-05,
+      "loss": 0.0365,
+      "step": 375
+    },
+    {
+      "epoch": 0.188,
+      "grad_norm": 0.16086017791775223,
+      "learning_rate": 1e-05,
+      "loss": 0.0382,
+      "step": 376
+    },
+    {
+      "epoch": 0.1885,
+      "grad_norm": 0.14863922525565496,
+      "learning_rate": 1e-05,
+      "loss": 0.039,
+      "step": 377
+    },
+    {
+      "epoch": 0.189,
+      "grad_norm": 0.16084357735487792,
+      "learning_rate": 1e-05,
+      "loss": 0.0312,
+      "step": 378
+    },
+    {
+      "epoch": 0.1895,
+      "grad_norm": 0.19070881724879324,
+      "learning_rate": 1e-05,
+      "loss": 0.0491,
+      "step": 379
+    },
+    {
+      "epoch": 0.19,
+      "grad_norm": 0.17240390839318184,
+      "learning_rate": 1e-05,
+      "loss": 0.0455,
+      "step": 380
+    },
+    {
+      "epoch": 0.1905,
+      "grad_norm": 0.13331277326103189,
+      "learning_rate": 1e-05,
+      "loss": 0.0289,
+      "step": 381
+    },
+    {
+      "epoch": 0.191,
+      "grad_norm": 0.18756729894366522,
+      "learning_rate": 1e-05,
+      "loss": 0.0468,
+      "step": 382
+    },
+    {
+      "epoch": 0.1915,
+      "grad_norm": 0.1660248717735821,
+      "learning_rate": 1e-05,
+      "loss": 0.0424,
+      "step": 383
+    },
+    {
+      "epoch": 0.192,
+      "grad_norm": 0.16346974130070938,
+      "learning_rate": 1e-05,
+      "loss": 0.0311,
+      "step": 384
+    },
+    {
+      "epoch": 0.192,
+      "eval_dev_acc": 0.599609375,
+      "eval_dev_token": 5596.130859375,
+      "eval_runtime": 352.4793,
+      "eval_samples_per_second": 0.182,
+      "eval_steps_per_second": 0.003,
+      "step": 384
+    },
+    {
+      "epoch": 0.1925,
+      "grad_norm": 0.16858862798497806,
+      "learning_rate": 1e-05,
+      "loss": 0.0389,
+      "step": 385
+    },
+    {
+      "epoch": 0.193,
+      "grad_norm": 0.1484958580298565,
+      "learning_rate": 1e-05,
+      "loss": 0.0397,
+      "step": 386
+    },
+    {
+      "epoch": 0.1935,
+      "grad_norm": 0.17660261356555002,
+      "learning_rate": 1e-05,
+      "loss": 0.0515,
+      "step": 387
+    },
+    {
+      "epoch": 0.194,
+      "grad_norm": 0.1783517215939047,
+      "learning_rate": 1e-05,
+      "loss": 0.0431,
+      "step": 388
+    },
+    {
+      "epoch": 0.1945,
+      "grad_norm": 0.14136150090913457,
+      "learning_rate": 1e-05,
+      "loss": 0.0323,
+      "step": 389
+    },
+    {
+      "epoch": 0.195,
+      "grad_norm": 0.16595913921658337,
+      "learning_rate": 1e-05,
+      "loss": 0.0394,
+      "step": 390
+    },
+    {
+      "epoch": 0.1955,
+      "grad_norm": 0.17788297569443248,
+      "learning_rate": 1e-05,
+      "loss": 0.0698,
+      "step": 391
+    },
+    {
+      "epoch": 0.196,
+      "grad_norm": 0.14755167079389797,
+      "learning_rate": 1e-05,
+      "loss": 0.0308,
+      "step": 392
+    },
+    {
+      "epoch": 0.1965,
+      "grad_norm": 0.20681855290430337,
+      "learning_rate": 1e-05,
+      "loss": 0.0494,
+      "step": 393
+    },
+    {
+      "epoch": 0.197,
+      "grad_norm": 0.19060439020439998,
+      "learning_rate": 1e-05,
+      "loss": 0.0445,
+      "step": 394
+    },
+    {
+      "epoch": 0.1975,
+      "grad_norm": 0.17199443698076167,
+      "learning_rate": 1e-05,
+      "loss": 0.0414,
+      "step": 395
+    },
+    {
+      "epoch": 0.198,
+      "grad_norm": 0.15210077373082737,
+      "learning_rate": 1e-05,
+      "loss": 0.0296,
+      "step": 396
+    },
+    {
+      "epoch": 0.1985,
+      "grad_norm": 0.17482591540638856,
+      "learning_rate": 1e-05,
+      "loss": 0.044,
+      "step": 397
+    },
+    {
+      "epoch": 0.199,
+      "grad_norm": 0.15501601608099658,
+      "learning_rate": 1e-05,
+      "loss": 0.0376,
+      "step": 398
+    },
+    {
+      "epoch": 0.1995,
+      "grad_norm": 0.17142493205422682,
+      "learning_rate": 1e-05,
+      "loss": 0.0386,
+      "step": 399
+    },
+    {
+      "epoch": 0.2,
+      "grad_norm": 0.1921162644413309,
+      "learning_rate": 1e-05,
+      "loss": 0.0469,
+      "step": 400
+    },
+    {
+      "epoch": 0.2005,
+      "grad_norm": 0.15938080403417312,
+      "learning_rate": 1e-05,
+      "loss": 0.0496,
+      "step": 401
+    },
+    {
+      "epoch": 0.201,
+      "grad_norm": 0.14786848292294155,
+      "learning_rate": 1e-05,
+      "loss": 0.0426,
+      "step": 402
+    },
+    {
+      "epoch": 0.2015,
+      "grad_norm": 0.18628997533329272,
+      "learning_rate": 1e-05,
+      "loss": 0.0581,
+      "step": 403
+    },
+    {
+      "epoch": 0.202,
+      "grad_norm": 0.16058096254934043,
+      "learning_rate": 1e-05,
+      "loss": 0.0336,
+      "step": 404
+    },
+    {
+      "epoch": 0.2025,
+      "grad_norm": 0.19319024386507233,
+      "learning_rate": 1e-05,
+      "loss": 0.047,
+      "step": 405
+    },
+    {
+      "epoch": 0.203,
+      "grad_norm": 0.17328115011013,
+      "learning_rate": 1e-05,
+      "loss": 0.049,
+      "step": 406
+    },
+    {
+      "epoch": 0.2035,
+      "grad_norm": 0.13258378170371796,
+      "learning_rate": 1e-05,
+      "loss": 0.0286,
+      "step": 407
+    },
+    {
+      "epoch": 0.204,
+      "grad_norm": 0.17945245697241183,
+      "learning_rate": 1e-05,
+      "loss": 0.0518,
+      "step": 408
+    },
+    {
+      "epoch": 0.2045,
+      "grad_norm": 0.16689764407399071,
+      "learning_rate": 1e-05,
+      "loss": 0.0458,
+      "step": 409
+    },
+    {
+      "epoch": 0.205,
+      "grad_norm": 0.18446815699746041,
+      "learning_rate": 1e-05,
+      "loss": 0.0408,
+      "step": 410
+    },
+    {
+      "epoch": 0.2055,
+      "grad_norm": 0.1489326060726689,
+      "learning_rate": 1e-05,
+      "loss": 0.0656,
+      "step": 411
+    },
+    {
+      "epoch": 0.206,
+      "grad_norm": 0.14974593012017515,
+      "learning_rate": 1e-05,
+      "loss": 0.0297,
+      "step": 412
+    },
+    {
+      "epoch": 0.2065,
+      "grad_norm": 0.1918114395748189,
+      "learning_rate": 1e-05,
+      "loss": 0.0439,
+      "step": 413
+    },
+    {
+      "epoch": 0.207,
+      "grad_norm": 0.1689953495305046,
+      "learning_rate": 1e-05,
+      "loss": 0.04,
+      "step": 414
+    },
+    {
+      "epoch": 0.2075,
+      "grad_norm": 0.1403733317703667,
+      "learning_rate": 1e-05,
+      "loss": 0.0472,
+      "step": 415
+    },
+    {
+      "epoch": 0.208,
+      "grad_norm": 0.173982074128614,
+      "learning_rate": 1e-05,
+      "loss": 0.0395,
+      "step": 416
+    },
+    {
+      "epoch": 0.208,
+      "eval_dev_acc": 0.49609375,
+      "eval_dev_token": 5482.95703125,
+      "eval_runtime": 374.469,
+      "eval_samples_per_second": 0.171,
+      "eval_steps_per_second": 0.003,
+      "step": 416
+    },
+    {
+      "epoch": 0.2085,
+      "grad_norm": 0.16717306009352031,
+      "learning_rate": 1e-05,
+      "loss": 0.0369,
+      "step": 417
+    },
+    {
+      "epoch": 0.209,
+      "grad_norm": 0.17317803700896214,
+      "learning_rate": 1e-05,
+      "loss": 0.0581,
+      "step": 418
+    },
+    {
+      "epoch": 0.2095,
+      "grad_norm": 0.14729050118039705,
+      "learning_rate": 1e-05,
+      "loss": 0.0469,
+      "step": 419
+    },
+    {
+      "epoch": 0.21,
+      "grad_norm": 0.14599122830811173,
+      "learning_rate": 1e-05,
+      "loss": 0.0571,
+      "step": 420
+    },
+    {
+      "epoch": 0.2105,
+      "grad_norm": 0.16285142688584706,
+      "learning_rate": 1e-05,
+      "loss": 0.0291,
+      "step": 421
+    },
+    {
+      "epoch": 0.211,
+      "grad_norm": 0.19044973230329837,
+      "learning_rate": 1e-05,
+      "loss": 0.0599,
+      "step": 422
+    },
+    {
+      "epoch": 0.2115,
+      "grad_norm": 0.188861960333507,
+      "learning_rate": 1e-05,
+      "loss": 0.0471,
+      "step": 423
+    },
+    {
+      "epoch": 0.212,
+      "grad_norm": 0.19188548951756218,
+      "learning_rate": 1e-05,
+      "loss": 0.0529,
+      "step": 424
+    },
+    {
+      "epoch": 0.2125,
+      "grad_norm": 0.16267402517673002,
+      "learning_rate": 1e-05,
+      "loss": 0.0305,
+      "step": 425
+    },
+    {
+      "epoch": 0.213,
+      "grad_norm": 0.1447850696130614,
+      "learning_rate": 1e-05,
+      "loss": 0.0324,
+      "step": 426
+    },
+    {
+      "epoch": 0.2135,
+      "grad_norm": 0.15248164794588065,
+      "learning_rate": 1e-05,
+      "loss": 0.0388,
+      "step": 427
+    },
+    {
+      "epoch": 0.214,
+      "grad_norm": 0.1661241871100943,
+      "learning_rate": 1e-05,
+      "loss": 0.0328,
+      "step": 428
+    },
+    {
+      "epoch": 0.2145,
+      "grad_norm": 0.16566625624023265,
+      "learning_rate": 1e-05,
+      "loss": 0.031,
+      "step": 429
+    },
+    {
+      "epoch": 0.215,
+      "grad_norm": 0.15249287061514458,
+      "learning_rate": 1e-05,
+      "loss": 0.0319,
+      "step": 430
+    },
+    {
+      "epoch": 0.2155,
+      "grad_norm": 0.12995530917181783,
+      "learning_rate": 1e-05,
+      "loss": 0.0233,
+      "step": 431
+    },
+    {
+      "epoch": 0.216,
+      "grad_norm": 0.1704276552962093,
+      "learning_rate": 1e-05,
+      "loss": 0.0405,
+      "step": 432
+    },
+    {
+      "epoch": 0.2165,
+      "grad_norm": 0.17386329346754434,
+      "learning_rate": 1e-05,
+      "loss": 0.0336,
+      "step": 433
+    },
+    {
+      "epoch": 0.217,
+      "grad_norm": 0.15704760833763615,
+      "learning_rate": 1e-05,
+      "loss": 0.0325,
+      "step": 434
+    },
+    {
+      "epoch": 0.2175,
+      "grad_norm": 0.1495524799308763,
+      "learning_rate": 1e-05,
+      "loss": 0.0341,
+      "step": 435
+    },
+    {
+      "epoch": 0.218,
+      "grad_norm": 0.1686891909288217,
+      "learning_rate": 1e-05,
+      "loss": 0.0343,
+      "step": 436
+    },
+    {
+      "epoch": 0.2185,
+      "grad_norm": 0.13995459985426573,
+      "learning_rate": 1e-05,
+      "loss": 0.0398,
+      "step": 437
+    },
+    {
+      "epoch": 0.219,
+      "grad_norm": 0.15473569116081692,
+      "learning_rate": 1e-05,
+      "loss": 0.0412,
+      "step": 438
+    },
+    {
+      "epoch": 0.2195,
+      "grad_norm": 0.1801609077983992,
+      "learning_rate": 1e-05,
+      "loss": 0.0534,
+      "step": 439
+    },
+    {
+      "epoch": 0.22,
+      "grad_norm": 0.17809364795872226,
+      "learning_rate": 1e-05,
+      "loss": 0.0548,
+      "step": 440
+    },
+    {
+      "epoch": 0.2205,
+      "grad_norm": 0.1535032114151188,
+      "learning_rate": 1e-05,
+      "loss": 0.0593,
+      "step": 441
+    },
+    {
+      "epoch": 0.221,
+      "grad_norm": 0.20882248884544774,
+      "learning_rate": 1e-05,
+      "loss": 0.0402,
+      "step": 442
+    },
+    {
+      "epoch": 0.2215,
+      "grad_norm": 0.14517381058327564,
+      "learning_rate": 1e-05,
+      "loss": 0.0436,
+      "step": 443
+    },
+    {
+      "epoch": 0.222,
+      "grad_norm": 0.17014179155102424,
+      "learning_rate": 1e-05,
+      "loss": 0.0333,
+      "step": 444
+    },
+    {
+      "epoch": 0.2225,
+      "grad_norm": 0.1729306341614305,
+      "learning_rate": 1e-05,
+      "loss": 0.0301,
+      "step": 445
+    },
+    {
+      "epoch": 0.223,
+      "grad_norm": 0.1686712423851483,
+      "learning_rate": 1e-05,
+      "loss": 0.0406,
+      "step": 446
+    },
+    {
+      "epoch": 0.2235,
+      "grad_norm": 0.1535287640721648,
+      "learning_rate": 1e-05,
+      "loss": 0.0282,
+      "step": 447
+    },
+    {
+      "epoch": 0.224,
+      "grad_norm": 0.1406783148617548,
+      "learning_rate": 1e-05,
+      "loss": 0.0264,
+      "step": 448
+    },
+    {
+      "epoch": 0.224,
+      "eval_dev_acc": 0.576171875,
+      "eval_dev_token": 5738.91015625,
+      "eval_runtime": 360.4892,
+      "eval_samples_per_second": 0.178,
+      "eval_steps_per_second": 0.003,
+      "step": 448
+    },
+    {
+      "epoch": 0.2245,
+      "grad_norm": 0.15218760519443988,
+      "learning_rate": 1e-05,
+      "loss": 0.0227,
+      "step": 449
+    },
+    {
+      "epoch": 0.225,
+      "grad_norm": 0.13165646820927943,
+      "learning_rate": 1e-05,
+      "loss": 0.0244,
+      "step": 450
+    },
+    {
+      "epoch": 0.2255,
+      "grad_norm": 0.13658776890249372,
+      "learning_rate": 1e-05,
+      "loss": 0.0235,
+      "step": 451
+    },
+    {
+      "epoch": 0.226,
+      "grad_norm": 0.1595102880182028,
+      "learning_rate": 1e-05,
+      "loss": 0.0361,
+      "step": 452
+    },
+    {
+      "epoch": 0.2265,
+      "grad_norm": 0.18272272767076744,
+      "learning_rate": 1e-05,
+      "loss": 0.0341,
+      "step": 453
+    },
+    {
+      "epoch": 0.227,
+      "grad_norm": 0.15970250529787045,
+      "learning_rate": 1e-05,
+      "loss": 0.0277,
+      "step": 454
+    },
+    {
+      "epoch": 0.2275,
+      "grad_norm": 0.1641447638351716,
+      "learning_rate": 1e-05,
+      "loss": 0.0284,
+      "step": 455
+    },
+    {
+      "epoch": 0.228,
+      "grad_norm": 0.1294308434226962,
+      "learning_rate": 1e-05,
+      "loss": 0.022,
+      "step": 456
+    },
+    {
+      "epoch": 0.2285,
+      "grad_norm": 0.11954195360401737,
+      "learning_rate": 1e-05,
+      "loss": 0.0202,
+      "step": 457
+    },
+    {
+      "epoch": 0.229,
+      "grad_norm": 0.16068276912989043,
+      "learning_rate": 1e-05,
+      "loss": 0.0251,
+      "step": 458
+    },
+    {
+      "epoch": 0.2295,
+      "grad_norm": 0.15307414897001792,
+      "learning_rate": 1e-05,
+      "loss": 0.027,
+      "step": 459
+    },
+    {
+      "epoch": 0.23,
+      "grad_norm": 0.15979117725555442,
+      "learning_rate": 1e-05,
+      "loss": 0.0375,
+      "step": 460
+    },
+    {
+      "epoch": 0.2305,
+      "grad_norm": 0.2020059964338148,
+      "learning_rate": 1e-05,
+      "loss": 0.043,
+      "step": 461
+    },
+    {
+      "epoch": 0.231,
+      "grad_norm": 0.15074817260440432,
+      "learning_rate": 1e-05,
+      "loss": 0.0241,
+      "step": 462
+    },
+    {
+      "epoch": 0.2315,
+      "grad_norm": 0.16521962645814686,
+      "learning_rate": 1e-05,
+      "loss": 0.0447,
+      "step": 463
+    },
+    {
+      "epoch": 0.232,
+      "grad_norm": 0.14710027259702427,
+      "learning_rate": 1e-05,
+      "loss": 0.0322,
+      "step": 464
+    },
+    {
+      "epoch": 0.2325,
+      "grad_norm": 0.1879875645942318,
+      "learning_rate": 1e-05,
+      "loss": 0.0324,
+      "step": 465
+    },
+    {
+      "epoch": 0.233,
+      "grad_norm": 0.16624112738581265,
+      "learning_rate": 1e-05,
+      "loss": 0.0308,
+      "step": 466
+    },
+    {
+      "epoch": 0.2335,
+      "grad_norm": 0.19050875236463863,
+      "learning_rate": 1e-05,
+      "loss": 0.0342,
+      "step": 467
+    },
+    {
+      "epoch": 0.234,
+      "grad_norm": 0.11998525830480301,
+      "learning_rate": 1e-05,
+      "loss": 0.021,
+      "step": 468
+    },
+    {
+      "epoch": 0.2345,
+      "grad_norm": 0.1656701890014892,
+      "learning_rate": 1e-05,
+      "loss": 0.027,
+      "step": 469
+    },
+    {
+      "epoch": 0.235,
+      "grad_norm": 0.15307753326429366,
+      "learning_rate": 1e-05,
+      "loss": 0.0452,
+      "step": 470
+    },
+    {
+      "epoch": 0.2355,
+      "grad_norm": 0.1897260442883158,
+      "learning_rate": 1e-05,
+      "loss": 0.0245,
+      "step": 471
+    },
+    {
+      "epoch": 0.236,
+      "grad_norm": 0.11931107646228578,
+      "learning_rate": 1e-05,
+      "loss": 0.017,
+      "step": 472
+    },
+    {
+      "epoch": 0.2365,
+      "grad_norm": 0.18498950301005707,
+      "learning_rate": 1e-05,
+      "loss": 0.0342,
+      "step": 473
+    },
+    {
+      "epoch": 0.237,
+      "grad_norm": 0.14355239628818517,
+      "learning_rate": 1e-05,
+      "loss": 0.0221,
+      "step": 474
+    },
+    {
+      "epoch": 0.2375,
+      "grad_norm": 0.16525653932908532,
+      "learning_rate": 1e-05,
+      "loss": 0.029,
+      "step": 475
+    },
+    {
+      "epoch": 0.238,
+      "grad_norm": 0.1518153688638394,
+      "learning_rate": 1e-05,
+      "loss": 0.0267,
+      "step": 476
+    },
+    {
+      "epoch": 0.2385,
+      "grad_norm": 0.15987321641272437,
+      "learning_rate": 1e-05,
+      "loss": 0.0279,
+      "step": 477
+    },
+    {
+      "epoch": 0.239,
+      "grad_norm": 0.1442274823944727,
+      "learning_rate": 1e-05,
+      "loss": 0.0349,
+      "step": 478
+    },
+    {
+      "epoch": 0.2395,
+      "grad_norm": 0.11710766704672448,
+      "learning_rate": 1e-05,
+      "loss": 0.0179,
+      "step": 479
+    },
+    {
+      "epoch": 0.24,
+      "grad_norm": 0.15497604683020938,
+      "learning_rate": 1e-05,
+      "loss": 0.023,
+      "step": 480
+    },
+    {
+      "epoch": 0.24,
+      "eval_dev_acc": 0.544921875,
+      "eval_dev_token": 5840.61328125,
+      "eval_runtime": 373.4708,
+      "eval_samples_per_second": 0.171,
+      "eval_steps_per_second": 0.003,
+      "step": 480
+    },
+    {
+      "epoch": 0.2405,
+      "grad_norm": 0.13492229768745556,
+      "learning_rate": 1e-05,
+      "loss": 0.0205,
+      "step": 481
+    },
+    {
+      "epoch": 0.241,
+      "grad_norm": 0.1704648731314998,
+      "learning_rate": 1e-05,
+      "loss": 0.0353,
+      "step": 482
+    },
+    {
+      "epoch": 0.2415,
+      "grad_norm": 0.1491861836462168,
+      "learning_rate": 1e-05,
+      "loss": 0.0365,
+      "step": 483
+    },
+    {
+      "epoch": 0.242,
+      "grad_norm": 0.17050828891525746,
+      "learning_rate": 1e-05,
+      "loss": 0.0277,
+      "step": 484
+    },
+    {
+      "epoch": 0.2425,
+      "grad_norm": 0.17980691606220936,
+      "learning_rate": 1e-05,
+      "loss": 0.0335,
+      "step": 485
+    },
+    {
+      "epoch": 0.243,
+      "grad_norm": 0.16998825524724584,
+      "learning_rate": 1e-05,
+      "loss": 0.0362,
+      "step": 486
+    },
+    {
+      "epoch": 0.2435,
+      "grad_norm": 0.11641133365996917,
+      "learning_rate": 1e-05,
+      "loss": 0.0181,
+      "step": 487
+    },
+    {
+      "epoch": 0.244,
+      "grad_norm": 0.14362674831456992,
+      "learning_rate": 1e-05,
+      "loss": 0.0365,
+      "step": 488
+    },
+    {
+      "epoch": 0.2445,
+      "grad_norm": 0.14488123923452778,
+      "learning_rate": 1e-05,
+      "loss": 0.024,
+      "step": 489
+    },
+    {
+      "epoch": 0.245,
+      "grad_norm": 0.1517003378019991,
+      "learning_rate": 1e-05,
+      "loss": 0.0271,
+      "step": 490
+    },
+    {
+      "epoch": 0.2455,
+      "grad_norm": 0.14967074987714707,
+      "learning_rate": 1e-05,
+      "loss": 0.0294,
+      "step": 491
+    },
+    {
+      "epoch": 0.246,
+      "grad_norm": 0.15791993394836015,
+      "learning_rate": 1e-05,
+      "loss": 0.0283,
+      "step": 492
+    },
+    {
+      "epoch": 0.2465,
+      "grad_norm": 0.13495006239387555,
+      "learning_rate": 1e-05,
+      "loss": 0.0251,
+      "step": 493
+    },
+    {
+      "epoch": 0.247,
+      "grad_norm": 0.18930054102351096,
+      "learning_rate": 1e-05,
+      "loss": 0.0373,
+      "step": 494
+    },
+    {
+      "epoch": 0.2475,
+      "grad_norm": 0.13152234060084034,
+      "learning_rate": 1e-05,
+      "loss": 0.0233,
+      "step": 495
+    },
+    {
+      "epoch": 0.248,
+      "grad_norm": 0.1341531691510106,
+      "learning_rate": 1e-05,
+      "loss": 0.0269,
+      "step": 496
+    },
+    {
+      "epoch": 0.2485,
+      "grad_norm": 0.13741586371551992,
+      "learning_rate": 1e-05,
+      "loss": 0.0277,
+      "step": 497
+    },
+    {
+      "epoch": 0.249,
+      "grad_norm": 0.1554051684617337,
+      "learning_rate": 1e-05,
+      "loss": 0.0276,
+      "step": 498
+    },
+    {
+      "epoch": 0.2495,
+      "grad_norm": 0.14814577647609775,
+      "learning_rate": 1e-05,
+      "loss": 0.0235,
+      "step": 499
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.14930720560226657,
+      "learning_rate": 1e-05,
+      "loss": 0.039,
+      "step": 500
+    },
+    {
+      "epoch": 0.2505,
+      "grad_norm": 0.1244942117603243,
+      "learning_rate": 1e-05,
+      "loss": 0.0202,
+      "step": 501
+    },
+    {
+      "epoch": 0.251,
+      "grad_norm": 0.14244145658079232,
+      "learning_rate": 1e-05,
+      "loss": 0.0209,
+      "step": 502
+    },
+    {
+      "epoch": 0.2515,
+      "grad_norm": 0.146145096145696,
+      "learning_rate": 1e-05,
+      "loss": 0.024,
+      "step": 503
+    },
+    {
+      "epoch": 0.252,
+      "grad_norm": 0.13594585715406687,
+      "learning_rate": 1e-05,
+      "loss": 0.0269,
+      "step": 504
+    },
+    {
+      "epoch": 0.2525,
+      "grad_norm": 0.1490412459954878,
+      "learning_rate": 1e-05,
+      "loss": 0.0345,
+      "step": 505
+    },
+    {
+      "epoch": 0.253,
+      "grad_norm": 0.11950170266380834,
+      "learning_rate": 1e-05,
+      "loss": 0.0181,
+      "step": 506
+    },
+    {
+      "epoch": 0.2535,
+      "grad_norm": 0.18548215823845707,
+      "learning_rate": 1e-05,
+      "loss": 0.0275,
+      "step": 507
+    },
+    {
+      "epoch": 0.254,
+      "grad_norm": 0.15108980653404058,
+      "learning_rate": 1e-05,
+      "loss": 0.0197,
+      "step": 508
+    },
+    {
+      "epoch": 0.2545,
+      "grad_norm": 0.16504836098536718,
+      "learning_rate": 1e-05,
+      "loss": 0.0289,
+      "step": 509
+    },
+    {
+      "epoch": 0.255,
+      "grad_norm": 0.15746261920489785,
+      "learning_rate": 1e-05,
+      "loss": 0.0253,
+      "step": 510
+    },
+    {
+      "epoch": 0.2555,
+      "grad_norm": 0.14071771991438595,
+      "learning_rate": 1e-05,
+      "loss": 0.0219,
+      "step": 511
+    },
+    {
+      "epoch": 0.256,
+      "grad_norm": 0.16079872072377113,
+      "learning_rate": 1e-05,
+      "loss": 0.0204,
+      "step": 512
+    },
+    {
+      "epoch": 0.256,
+      "eval_dev_acc": 0.56640625,
+      "eval_dev_token": 5634.1015625,
+      "eval_runtime": 361.9891,
+      "eval_samples_per_second": 0.177,
+      "eval_steps_per_second": 0.003,
+      "step": 512
+    },
+    {
+      "epoch": 0.2565,
+      "grad_norm": 0.13549471484008802,
+      "learning_rate": 1e-05,
+      "loss": 0.0164,
+      "step": 513
+    },
+    {
+      "epoch": 0.257,
+      "grad_norm": 0.12072963489745359,
+      "learning_rate": 1e-05,
+      "loss": 0.0302,
+      "step": 514
+    },
+    {
+      "epoch": 0.2575,
+      "grad_norm": 0.14026647684897994,
+      "learning_rate": 1e-05,
+      "loss": 0.0194,
+      "step": 515
+    },
+    {
+      "epoch": 0.258,
+      "grad_norm": 0.1634484411344168,
+      "learning_rate": 1e-05,
+      "loss": 0.0277,
+      "step": 516
+    },
+    {
+      "epoch": 0.2585,
+      "grad_norm": 0.15844211231505426,
+      "learning_rate": 1e-05,
+      "loss": 0.0214,
+      "step": 517
+    },
+    {
+      "epoch": 0.259,
+      "grad_norm": 0.1567910605652928,
+      "learning_rate": 1e-05,
+      "loss": 0.024,
+      "step": 518
+    },
+    {
+      "epoch": 0.2595,
+      "grad_norm": 0.17902606156745304,
+      "learning_rate": 1e-05,
+      "loss": 0.033,
+      "step": 519
+    },
+    {
+      "epoch": 0.26,
+      "grad_norm": 0.12339744538286439,
+      "learning_rate": 1e-05,
+      "loss": 0.0172,
+      "step": 520
+    },
+    {
+      "epoch": 0.2605,
+      "grad_norm": 0.13532209502494125,
+      "learning_rate": 1e-05,
+      "loss": 0.0206,
+      "step": 521
+    },
+    {
+      "epoch": 0.261,
+      "grad_norm": 0.15623082886780087,
+      "learning_rate": 1e-05,
+      "loss": 0.0213,
+      "step": 522
+    },
+    {
+      "epoch": 0.2615,
+      "grad_norm": 0.14428427308597647,
+      "learning_rate": 1e-05,
+      "loss": 0.0201,
+      "step": 523
+    },
+    {
+      "epoch": 0.262,
+      "grad_norm": 0.14835567545470982,
+      "learning_rate": 1e-05,
+      "loss": 0.0244,
+      "step": 524
+    },
+    {
+      "epoch": 0.2625,
+      "grad_norm": 0.14068070672711747,
+      "learning_rate": 1e-05,
+      "loss": 0.0239,
+      "step": 525
+    },
+    {
+      "epoch": 0.263,
+      "grad_norm": 0.1460843289248216,
+      "learning_rate": 1e-05,
+      "loss": 0.0223,
+      "step": 526
+    },
+    {
+      "epoch": 0.2635,
+      "grad_norm": 0.13777430449621855,
+      "learning_rate": 1e-05,
+      "loss": 0.0229,
+      "step": 527
+    },
+    {
+      "epoch": 0.264,
+      "grad_norm": 0.15161607294549337,
+      "learning_rate": 1e-05,
+      "loss": 0.0272,
+      "step": 528
+    },
+    {
+      "epoch": 0.2645,
+      "grad_norm": 0.13410519048089503,
+      "learning_rate": 1e-05,
+      "loss": 0.0181,
+      "step": 529
+    },
+    {
+      "epoch": 0.265,
+      "grad_norm": 0.15931617673254456,
+      "learning_rate": 1e-05,
+      "loss": 0.0244,
+      "step": 530
+    },
+    {
+      "epoch": 0.2655,
+      "grad_norm": 0.1410700523457689,
+      "learning_rate": 1e-05,
+      "loss": 0.0251,
+      "step": 531
+    },
+    {
+      "epoch": 0.266,
+      "grad_norm": 0.11388951846034073,
+      "learning_rate": 1e-05,
+      "loss": 0.0144,
+      "step": 532
+    },
+    {
+      "epoch": 0.2665,
+      "grad_norm": 0.12253780956369799,
+      "learning_rate": 1e-05,
+      "loss": 0.0177,
+      "step": 533
+    },
+    {
+      "epoch": 0.267,
+      "grad_norm": 0.15575473599510573,
+      "learning_rate": 1e-05,
+      "loss": 0.0192,
+      "step": 534
+    },
+    {
+      "epoch": 0.2675,
+      "grad_norm": 0.14690747155640696,
+      "learning_rate": 1e-05,
+      "loss": 0.0222,
+      "step": 535
+    },
+    {
+      "epoch": 0.268,
+      "grad_norm": 0.13584546405544728,
+      "learning_rate": 1e-05,
+      "loss": 0.0237,
+      "step": 536
+    },
+    {
+      "epoch": 0.2685,
+      "grad_norm": 0.13430763220790742,
+      "learning_rate": 1e-05,
+      "loss": 0.0291,
+      "step": 537
+    },
+    {
+      "epoch": 0.269,
+      "grad_norm": 0.14208572873353734,
+      "learning_rate": 1e-05,
+      "loss": 0.0187,
+      "step": 538
+    },
+    {
+      "epoch": 0.2695,
+      "grad_norm": 0.14058928149963162,
+      "learning_rate": 1e-05,
+      "loss": 0.0199,
+      "step": 539
+    },
+    {
+      "epoch": 0.27,
+      "grad_norm": 0.15100703501541832,
+      "learning_rate": 1e-05,
+      "loss": 0.0348,
+      "step": 540
+    },
+    {
+      "epoch": 0.2705,
+      "grad_norm": 0.12269452397268416,
+      "learning_rate": 1e-05,
+      "loss": 0.0247,
+      "step": 541
+    },
+    {
+      "epoch": 0.271,
+      "grad_norm": 0.1364796501674048,
+      "learning_rate": 1e-05,
+      "loss": 0.0227,
+      "step": 542
+    },
+    {
+      "epoch": 0.2715,
+      "grad_norm": 0.13163932605554884,
+      "learning_rate": 1e-05,
+      "loss": 0.0262,
+      "step": 543
+    },
+    {
+      "epoch": 0.272,
+      "grad_norm": 0.13497428740182482,
+      "learning_rate": 1e-05,
+      "loss": 0.0206,
+      "step": 544
+    },
+    {
+      "epoch": 0.272,
+      "eval_dev_acc": 0.59765625,
+      "eval_dev_token": 5639.98828125,
+      "eval_runtime": 359.2369,
+      "eval_samples_per_second": 0.178,
+      "eval_steps_per_second": 0.003,
+      "step": 544
+    },
+    {
+      "epoch": 0.2725,
+      "grad_norm": 0.13930735859181714,
+      "learning_rate": 1e-05,
+      "loss": 0.0234,
+      "step": 545
+    },
+    {
+      "epoch": 0.273,
+      "grad_norm": 0.11985280096835198,
+      "learning_rate": 1e-05,
+      "loss": 0.0269,
+      "step": 546
+    },
+    {
+      "epoch": 0.2735,
+      "grad_norm": 0.17031723198491708,
+      "learning_rate": 1e-05,
+      "loss": 0.028,
+      "step": 547
+    },
+    {
+      "epoch": 0.274,
+      "grad_norm": 0.17166197772315975,
+      "learning_rate": 1e-05,
+      "loss": 0.0281,
+      "step": 548
+    },
+    {
+      "epoch": 0.2745,
+      "grad_norm": 0.1167335581681914,
+      "learning_rate": 1e-05,
+      "loss": 0.022,
+      "step": 549
+    },
+    {
+      "epoch": 0.275,
+      "grad_norm": 0.1443441971157384,
+      "learning_rate": 1e-05,
+      "loss": 0.0198,
+      "step": 550
+    },
+    {
+      "epoch": 0.2755,
+      "grad_norm": 0.1268787923602722,
+      "learning_rate": 1e-05,
+      "loss": 0.017,
+      "step": 551
+    },
+    {
+      "epoch": 0.276,
+      "grad_norm": 0.11065296478824395,
+      "learning_rate": 1e-05,
+      "loss": 0.0157,
+      "step": 552
+    },
+    {
+      "epoch": 0.2765,
+      "grad_norm": 0.12047907824944362,
+      "learning_rate": 1e-05,
+      "loss": 0.016,
+      "step": 553
+    },
+    {
+      "epoch": 0.277,
+      "grad_norm": 0.13956303855472266,
+      "learning_rate": 1e-05,
+      "loss": 0.0216,
+      "step": 554
+    },
+    {
+      "epoch": 0.2775,
+      "grad_norm": 0.10533407777378404,
+      "learning_rate": 1e-05,
+      "loss": 0.0137,
+      "step": 555
+    },
+    {
+      "epoch": 0.278,
+      "grad_norm": 0.11532013491755984,
+      "learning_rate": 1e-05,
+      "loss": 0.0282,
+      "step": 556
+    },
+    {
+      "epoch": 0.2785,
+      "grad_norm": 0.11921463919727264,
+      "learning_rate": 1e-05,
+      "loss": 0.0163,
+      "step": 557
+    },
+    {
+      "epoch": 0.279,
+      "grad_norm": 0.15645731769207732,
+      "learning_rate": 1e-05,
+      "loss": 0.0241,
+      "step": 558
+    },
+    {
+      "epoch": 0.2795,
+      "grad_norm": 0.12096274696840706,
+      "learning_rate": 1e-05,
+      "loss": 0.0167,
+      "step": 559
+    },
+    {
+      "epoch": 0.28,
+      "grad_norm": 0.149157783124579,
+      "learning_rate": 1e-05,
+      "loss": 0.0193,
+      "step": 560
+    },
+    {
+      "epoch": 0.2805,
+      "grad_norm": 0.16982490839988412,
+      "learning_rate": 1e-05,
+      "loss": 0.0283,
+      "step": 561
+    },
+    {
+      "epoch": 0.281,
+      "grad_norm": 0.12038107977310454,
+      "learning_rate": 1e-05,
+      "loss": 0.0154,
+      "step": 562
+    },
+    {
+      "epoch": 0.2815,
+      "grad_norm": 0.16469919524412158,
+      "learning_rate": 1e-05,
+      "loss": 0.0214,
+      "step": 563
+    },
+    {
+      "epoch": 0.282,
+      "grad_norm": 0.15827423056846177,
+      "learning_rate": 1e-05,
+      "loss": 0.0216,
+      "step": 564
+    },
+    {
+      "epoch": 0.2825,
+      "grad_norm": 0.12058245559465251,
+      "learning_rate": 1e-05,
+      "loss": 0.0141,
+      "step": 565
+    },
+    {
+      "epoch": 0.283,
+      "grad_norm": 0.163789727088167,
+      "learning_rate": 1e-05,
+      "loss": 0.0241,
+      "step": 566
+    },
+    {
+      "epoch": 0.2835,
+      "grad_norm": 0.1390884369932456,
+      "learning_rate": 1e-05,
+      "loss": 0.0221,
+      "step": 567
+    },
+    {
+      "epoch": 0.284,
+      "grad_norm": 0.14472941005878595,
+      "learning_rate": 1e-05,
+      "loss": 0.0164,
+      "step": 568
+    },
+    {
+      "epoch": 0.2845,
+      "grad_norm": 0.15437454042645973,
+      "learning_rate": 1e-05,
+      "loss": 0.024,
+      "step": 569
+    },
+    {
+      "epoch": 0.285,
+      "grad_norm": 0.1207487307624573,
+      "learning_rate": 1e-05,
+      "loss": 0.0172,
+      "step": 570
+    },
+    {
+      "epoch": 0.2855,
+      "grad_norm": 0.1502409849611173,
+      "learning_rate": 1e-05,
+      "loss": 0.0298,
+      "step": 571
+    },
+    {
+      "epoch": 0.286,
+      "grad_norm": 0.16401355690597133,
+      "learning_rate": 1e-05,
+      "loss": 0.0225,
+      "step": 572
+    },
+    {
+      "epoch": 0.2865,
+      "grad_norm": 0.15181464752177645,
+      "learning_rate": 1e-05,
+      "loss": 0.0189,
+      "step": 573
+    },
+    {
+      "epoch": 0.287,
+      "grad_norm": 0.14560432645081878,
+      "learning_rate": 1e-05,
+      "loss": 0.0186,
+      "step": 574
+    },
+    {
+      "epoch": 0.2875,
+      "grad_norm": 0.12603042660981642,
+      "learning_rate": 1e-05,
+      "loss": 0.0155,
+      "step": 575
+    },
+    {
+      "epoch": 0.288,
+      "grad_norm": 0.11638577111126014,
+      "learning_rate": 1e-05,
+      "loss": 0.0164,
+      "step": 576
+    },
+    {
+      "epoch": 0.288,
+      "eval_dev_acc": 0.54296875,
+      "eval_dev_token": 5801.126953125,
+      "eval_runtime": 373.1703,
+      "eval_samples_per_second": 0.172,
+      "eval_steps_per_second": 0.003,
+      "step": 576
+    },
+    {
+      "epoch": 0.2885,
+      "grad_norm": 0.13420942083968396,
+      "learning_rate": 1e-05,
+      "loss": 0.0181,
+      "step": 577
+    },
+    {
+      "epoch": 0.289,
+      "grad_norm": 0.12122809121871923,
+      "learning_rate": 1e-05,
+      "loss": 0.0134,
+      "step": 578
+    },
+    {
+      "epoch": 0.2895,
+      "grad_norm": 0.13114866603642533,
+      "learning_rate": 1e-05,
+      "loss": 0.0174,
+      "step": 579
+    },
+    {
+      "epoch": 0.29,
+      "grad_norm": 0.1498609312158644,
+      "learning_rate": 1e-05,
+      "loss": 0.0214,
+      "step": 580
+    },
+    {
+      "epoch": 0.2905,
+      "grad_norm": 0.1527812218308566,
+      "learning_rate": 1e-05,
+      "loss": 0.0194,
+      "step": 581
+    },
+    {
+      "epoch": 0.291,
+      "grad_norm": 0.14711876695343454,
+      "learning_rate": 1e-05,
+      "loss": 0.018,
+      "step": 582
+    },
+    {
+      "epoch": 0.2915,
+      "grad_norm": 0.11529220604038168,
+      "learning_rate": 1e-05,
+      "loss": 0.0144,
+      "step": 583
+    },
+    {
+      "epoch": 0.292,
+      "grad_norm": 0.16180671831014115,
+      "learning_rate": 1e-05,
+      "loss": 0.0165,
+      "step": 584
+    },
+    {
+      "epoch": 0.2925,
+      "grad_norm": 0.13621545825638848,
+      "learning_rate": 1e-05,
+      "loss": 0.015,
+      "step": 585
+    },
+    {
+      "epoch": 0.293,
+      "grad_norm": 0.15473239935591382,
+      "learning_rate": 1e-05,
+      "loss": 0.0164,
+      "step": 586
+    },
+    {
+      "epoch": 0.2935,
+      "grad_norm": 0.15716799171541335,
+      "learning_rate": 1e-05,
+      "loss": 0.0194,
+      "step": 587
+    },
+    {
+      "epoch": 0.294,
+      "grad_norm": 0.1684941322847538,
+      "learning_rate": 1e-05,
+      "loss": 0.022,
+      "step": 588
+    },
+    {
+      "epoch": 0.2945,
+      "grad_norm": 0.15453918821249785,
+      "learning_rate": 1e-05,
+      "loss": 0.0188,
+      "step": 589
+    },
+    {
+      "epoch": 0.295,
+      "grad_norm": 0.140163345657633,
+      "learning_rate": 1e-05,
+      "loss": 0.0208,
+      "step": 590
+    },
+    {
+      "epoch": 0.2955,
+      "grad_norm": 0.15010258665645038,
+      "learning_rate": 1e-05,
+      "loss": 0.0215,
+      "step": 591
+    },
+    {
+      "epoch": 0.296,
+      "grad_norm": 0.14661643221841641,
+      "learning_rate": 1e-05,
+      "loss": 0.0191,
+      "step": 592
+    },
+    {
+      "epoch": 0.2965,
+      "grad_norm": 0.15435066476462508,
+      "learning_rate": 1e-05,
+      "loss": 0.0231,
+      "step": 593
+    },
+    {
+      "epoch": 0.297,
+      "grad_norm": 0.17094702806791945,
+      "learning_rate": 1e-05,
+      "loss": 0.0251,
+      "step": 594
+    },
+    {
+      "epoch": 0.2975,
+      "grad_norm": 0.1371139566901347,
+      "learning_rate": 1e-05,
+      "loss": 0.0186,
+      "step": 595
+    },
+    {
+      "epoch": 0.298,
+      "grad_norm": 0.11779673830033237,
+      "learning_rate": 1e-05,
+      "loss": 0.0269,
+      "step": 596
+    },
+    {
+      "epoch": 0.2985,
+      "grad_norm": 0.11853976704548681,
+      "learning_rate": 1e-05,
+      "loss": 0.0154,
+      "step": 597
+    },
+    {
+      "epoch": 0.299,
+      "grad_norm": 0.14881574569113099,
+      "learning_rate": 1e-05,
+      "loss": 0.0246,
+      "step": 598
+    },
+    {
+      "epoch": 0.2995,
+      "grad_norm": 0.11792409287393274,
+      "learning_rate": 1e-05,
+      "loss": 0.0164,
+      "step": 599
+    },
+    {
+      "epoch": 0.3,
+      "grad_norm": 0.13831559531762572,
+      "learning_rate": 1e-05,
+      "loss": 0.0257,
+      "step": 600
+    },
+    {
+      "epoch": 0.3005,
+      "grad_norm": 0.13756632720301187,
+      "learning_rate": 1e-05,
+      "loss": 0.0214,
+      "step": 601
+    },
+    {
+      "epoch": 0.301,
+      "grad_norm": 0.10998907458045305,
+      "learning_rate": 1e-05,
+      "loss": 0.015,
+      "step": 602
+    },
+    {
+      "epoch": 0.3015,
+      "grad_norm": 0.135955562101373,
+      "learning_rate": 1e-05,
+      "loss": 0.0211,
+      "step": 603
+    },
+    {
+      "epoch": 0.302,
+      "grad_norm": 0.1214956422000124,
+      "learning_rate": 1e-05,
+      "loss": 0.0219,
+      "step": 604
+    },
+    {
+      "epoch": 0.3025,
+      "grad_norm": 0.15757702522309963,
+      "learning_rate": 1e-05,
+      "loss": 0.018,
+      "step": 605
+    },
+    {
+      "epoch": 0.303,
+      "grad_norm": 0.1350858708023801,
+      "learning_rate": 1e-05,
+      "loss": 0.0211,
+      "step": 606
+    },
+    {
+      "epoch": 0.3035,
+      "grad_norm": 0.10610433140412452,
+      "learning_rate": 1e-05,
+      "loss": 0.0147,
+      "step": 607
+    },
+    {
+      "epoch": 0.304,
+      "grad_norm": 0.11514647079357257,
+      "learning_rate": 1e-05,
+      "loss": 0.0192,
+      "step": 608
+    },
+    {
+      "epoch": 0.304,
+      "eval_dev_acc": 0.62109375,
+      "eval_dev_token": 5367.916015625,
+      "eval_runtime": 354.8095,
+      "eval_samples_per_second": 0.18,
+      "eval_steps_per_second": 0.003,
+      "step": 608
+    },
+    {
+      "epoch": 0.3045,
+      "grad_norm": 0.12603567923188372,
+      "learning_rate": 1e-05,
+      "loss": 0.0188,
+      "step": 609
+    },
+    {
+      "epoch": 0.305,
+      "grad_norm": 0.14277125754270012,
+      "learning_rate": 1e-05,
+      "loss": 0.022,
+      "step": 610
+    },
+    {
+      "epoch": 0.3055,
+      "grad_norm": 0.12862855181841676,
+      "learning_rate": 1e-05,
+      "loss": 0.021,
+      "step": 611
+    },
+    {
+      "epoch": 0.306,
+      "grad_norm": 0.1227953424403543,
+      "learning_rate": 1e-05,
+      "loss": 0.018,
+      "step": 612
+    },
+    {
+      "epoch": 0.3065,
+      "grad_norm": 0.11646820367498804,
+      "learning_rate": 1e-05,
+      "loss": 0.0131,
+      "step": 613
+    },
+    {
+      "epoch": 0.307,
+      "grad_norm": 0.14701145754992329,
+      "learning_rate": 1e-05,
+      "loss": 0.0186,
+      "step": 614
+    },
+    {
+      "epoch": 0.3075,
+      "grad_norm": 0.1493073818813876,
+      "learning_rate": 1e-05,
+      "loss": 0.0254,
+      "step": 615
+    },
+    {
+      "epoch": 0.308,
+      "grad_norm": 0.1352952895732537,
+      "learning_rate": 1e-05,
+      "loss": 0.0181,
+      "step": 616
+    },
+    {
+      "epoch": 0.3085,
+      "grad_norm": 0.13007743097982305,
+      "learning_rate": 1e-05,
+      "loss": 0.0171,
+      "step": 617
+    },
+    {
+      "epoch": 0.309,
+      "grad_norm": 0.1665432351262121,
+      "learning_rate": 1e-05,
+      "loss": 0.0228,
+      "step": 618
+    },
+    {
+      "epoch": 0.3095,
+      "grad_norm": 0.16442931730443322,
+      "learning_rate": 1e-05,
+      "loss": 0.0238,
+      "step": 619
+    },
+    {
+      "epoch": 0.31,
+      "grad_norm": 0.16320986192220768,
+      "learning_rate": 1e-05,
+      "loss": 0.0238,
+      "step": 620
+    },
+    {
+      "epoch": 0.3105,
+      "grad_norm": 0.13880254871235365,
+      "learning_rate": 1e-05,
+      "loss": 0.0179,
+      "step": 621
+    },
+    {
+      "epoch": 0.311,
+      "grad_norm": 0.13609379700738453,
+      "learning_rate": 1e-05,
+      "loss": 0.0195,
+      "step": 622
+    },
+    {
+      "epoch": 0.3115,
+      "grad_norm": 0.1368415516519621,
+      "learning_rate": 1e-05,
+      "loss": 0.024,
+      "step": 623
+    },
+    {
+      "epoch": 0.312,
+      "grad_norm": 0.12821586481120512,
+      "learning_rate": 1e-05,
+      "loss": 0.0191,
+      "step": 624
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.13644927854222083,
+      "learning_rate": 1e-05,
+      "loss": 0.0154,
+      "step": 625
+    },
+    {
+      "epoch": 0.313,
+      "grad_norm": 0.15158164143556496,
+      "learning_rate": 1e-05,
+      "loss": 0.0318,
+      "step": 626
+    },
+    {
+      "epoch": 0.3135,
+      "grad_norm": 0.12404550422721679,
+      "learning_rate": 1e-05,
+      "loss": 0.0202,
+      "step": 627
+    },
+    {
+      "epoch": 0.314,
+      "grad_norm": 0.1235074023832298,
+      "learning_rate": 1e-05,
+      "loss": 0.0167,
+      "step": 628
+    },
+    {
+      "epoch": 0.3145,
+      "grad_norm": 0.16094487436899907,
+      "learning_rate": 1e-05,
+      "loss": 0.0227,
+      "step": 629
+    },
+    {
+      "epoch": 0.315,
+      "grad_norm": 0.11086598912590964,
+      "learning_rate": 1e-05,
+      "loss": 0.0158,
+      "step": 630
+    },
+    {
+      "epoch": 0.3155,
+      "grad_norm": 0.1147741974179167,
+      "learning_rate": 1e-05,
+      "loss": 0.0136,
+      "step": 631
+    },
+    {
+      "epoch": 0.316,
+      "grad_norm": 0.12346095617438974,
+      "learning_rate": 1e-05,
+      "loss": 0.017,
+      "step": 632
+    },
+    {
+      "epoch": 0.3165,
+      "grad_norm": 0.1235267138232638,
+      "learning_rate": 1e-05,
+      "loss": 0.0131,
+      "step": 633
+    },
+    {
+      "epoch": 0.317,
+      "grad_norm": 0.11979162262432065,
+      "learning_rate": 1e-05,
+      "loss": 0.0194,
+      "step": 634
+    },
+    {
+      "epoch": 0.3175,
+      "grad_norm": 0.12253729986288973,
+      "learning_rate": 1e-05,
+      "loss": 0.0205,
+      "step": 635
+    },
+    {
+      "epoch": 0.318,
+      "grad_norm": 0.1374736081434109,
+      "learning_rate": 1e-05,
+      "loss": 0.0207,
+      "step": 636
+    },
+    {
+      "epoch": 0.3185,
+      "grad_norm": 0.11667911740285354,
+      "learning_rate": 1e-05,
+      "loss": 0.0164,
+      "step": 637
+    },
+    {
+      "epoch": 0.319,
+      "grad_norm": 0.13725799823509804,
+      "learning_rate": 1e-05,
+      "loss": 0.0177,
+      "step": 638
+    },
+    {
+      "epoch": 0.3195,
+      "grad_norm": 0.1461325036101512,
+      "learning_rate": 1e-05,
+      "loss": 0.0233,
+      "step": 639
+    },
+    {
+      "epoch": 0.32,
+      "grad_norm": 0.1486586288765987,
+      "learning_rate": 1e-05,
+      "loss": 0.022,
+      "step": 640
+    },
+    {
+      "epoch": 0.32,
+      "eval_dev_acc": 0.52734375,
+      "eval_dev_token": 5585.80859375,
+      "eval_runtime": 360.5581,
+      "eval_samples_per_second": 0.178,
+      "eval_steps_per_second": 0.003,
+      "step": 640
+    },
+    {
+      "epoch": 0.3205,
+      "grad_norm": 0.13037027619165104,
+      "learning_rate": 1e-05,
+      "loss": 0.0181,
+      "step": 641
+    },
+    {
+      "epoch": 0.321,
+      "grad_norm": 0.1384777662387777,
+      "learning_rate": 1e-05,
+      "loss": 0.0284,
+      "step": 642
+    },
+    {
+      "epoch": 0.3215,
+      "grad_norm": 0.11731142200376247,
+      "learning_rate": 1e-05,
+      "loss": 0.0136,
+      "step": 643
+    },
+    {
+      "epoch": 0.322,
+      "grad_norm": 0.13199000719975476,
+      "learning_rate": 1e-05,
+      "loss": 0.0147,
+      "step": 644
+    },
+    {
+      "epoch": 0.3225,
+      "grad_norm": 0.145679314202878,
+      "learning_rate": 1e-05,
+      "loss": 0.0227,
+      "step": 645
+    },
+    {
+      "epoch": 0.323,
+      "grad_norm": 0.13813521110883425,
+      "learning_rate": 1e-05,
+      "loss": 0.0173,
+      "step": 646
+    },
+    {
+      "epoch": 0.3235,
+      "grad_norm": 0.11216370610734963,
+      "learning_rate": 1e-05,
+      "loss": 0.0125,
+      "step": 647
+    },
+    {
+      "epoch": 0.324,
+      "grad_norm": 0.09898218700430327,
+      "learning_rate": 1e-05,
+      "loss": 0.0165,
+      "step": 648
+    },
+    {
+      "epoch": 0.3245,
+      "grad_norm": 0.13555813085878698,
+      "learning_rate": 1e-05,
+      "loss": 0.0163,
+      "step": 649
+    },
+    {
+      "epoch": 0.325,
+      "grad_norm": 0.11552480540546263,
+      "learning_rate": 1e-05,
+      "loss": 0.0161,
+      "step": 650
+    },
+    {
+      "epoch": 0.3255,
+      "grad_norm": 0.12028253181729011,
+      "learning_rate": 1e-05,
+      "loss": 0.014,
+      "step": 651
+    },
+    {
+      "epoch": 0.326,
+      "grad_norm": 0.11773491790468957,
+      "learning_rate": 1e-05,
+      "loss": 0.0143,
+      "step": 652
+    },
+    {
+      "epoch": 0.3265,
+      "grad_norm": 0.11052002031571376,
+      "learning_rate": 1e-05,
+      "loss": 0.0177,
+      "step": 653
+    },
+    {
+      "epoch": 0.327,
+      "grad_norm": 0.1251016102612686,
+      "learning_rate": 1e-05,
+      "loss": 0.0134,
+      "step": 654
+    },
+    {
+      "epoch": 0.3275,
+      "grad_norm": 0.13329112874236815,
+      "learning_rate": 1e-05,
+      "loss": 0.0148,
+      "step": 655
+    },
+    {
+      "epoch": 0.328,
+      "grad_norm": 0.11382501473889628,
+      "learning_rate": 1e-05,
+      "loss": 0.0145,
+      "step": 656
+    },
+    {
+      "epoch": 0.3285,
+      "grad_norm": 0.1319495587226548,
+      "learning_rate": 1e-05,
+      "loss": 0.0139,
+      "step": 657
+    },
+    {
+      "epoch": 0.329,
+      "grad_norm": 0.12070785233460224,
+      "learning_rate": 1e-05,
+      "loss": 0.014,
+      "step": 658
+    },
+    {
+      "epoch": 0.3295,
+      "grad_norm": 0.12142680030465443,
+      "learning_rate": 1e-05,
+      "loss": 0.0159,
+      "step": 659
+    },
+    {
+      "epoch": 0.33,
+      "grad_norm": 0.12182666413282645,
+      "learning_rate": 1e-05,
+      "loss": 0.0114,
+      "step": 660
+    },
+    {
+      "epoch": 0.3305,
+      "grad_norm": 0.13190056217824564,
+      "learning_rate": 1e-05,
+      "loss": 0.0163,
+      "step": 661
+    },
+    {
+      "epoch": 0.331,
+      "grad_norm": 0.12804397999806036,
+      "learning_rate": 1e-05,
+      "loss": 0.016,
+      "step": 662
+    },
+    {
+      "epoch": 0.3315,
+      "grad_norm": 0.1461414693983946,
+      "learning_rate": 1e-05,
+      "loss": 0.0179,
+      "step": 663
+    },
+    {
+      "epoch": 0.332,
+      "grad_norm": 0.1467248720157808,
+      "learning_rate": 1e-05,
+      "loss": 0.0171,
+      "step": 664
+    },
+    {
+      "epoch": 0.3325,
+      "grad_norm": 0.1393703645756879,
+      "learning_rate": 1e-05,
+      "loss": 0.017,
+      "step": 665
+    },
+    {
+      "epoch": 0.333,
+      "grad_norm": 0.1677851073130961,
+      "learning_rate": 1e-05,
+      "loss": 0.033,
+      "step": 666
+    },
+    {
+      "epoch": 0.3335,
+      "grad_norm": 0.13245384270982133,
+      "learning_rate": 1e-05,
+      "loss": 0.0163,
+      "step": 667
+    },
+    {
+      "epoch": 0.334,
+      "grad_norm": 0.12289205645457782,
+      "learning_rate": 1e-05,
+      "loss": 0.015,
+      "step": 668
+    },
+    {
+      "epoch": 0.3345,
+      "grad_norm": 0.13290806824539741,
+      "learning_rate": 1e-05,
+      "loss": 0.0131,
+      "step": 669
+    },
+    {
+      "epoch": 0.335,
+      "grad_norm": 0.12256221521384754,
+      "learning_rate": 1e-05,
+      "loss": 0.0138,
+      "step": 670
+    },
+    {
+      "epoch": 0.3355,
+      "grad_norm": 0.11871572464334247,
+      "learning_rate": 1e-05,
+      "loss": 0.0156,
+      "step": 671
+    },
+    {
+      "epoch": 0.336,
+      "grad_norm": 0.1425273432040885,
+      "learning_rate": 1e-05,
+      "loss": 0.0132,
+      "step": 672
+    },
+    {
+      "epoch": 0.336,
+      "eval_dev_acc": 0.5625,
+      "eval_dev_token": 5492.04296875,
+      "eval_runtime": 348.043,
+      "eval_samples_per_second": 0.184,
+      "eval_steps_per_second": 0.003,
+      "step": 672
+    },
+    {
+      "epoch": 0.3365,
+      "grad_norm": 0.12612838576017849,
+      "learning_rate": 1e-05,
+      "loss": 0.0127,
+      "step": 673
+    },
+    {
+      "epoch": 0.337,
+      "grad_norm": 0.1294118778329942,
+      "learning_rate": 1e-05,
+      "loss": 0.0161,
+      "step": 674
+    },
+    {
+      "epoch": 0.3375,
+      "grad_norm": 0.11771619437889824,
+      "learning_rate": 1e-05,
+      "loss": 0.0142,
+      "step": 675
+    },
+    {
+      "epoch": 0.338,
+      "grad_norm": 0.13434389137713848,
+      "learning_rate": 1e-05,
+      "loss": 0.0157,
+      "step": 676
+    },
+    {
+      "epoch": 0.3385,
+      "grad_norm": 0.2579148520419769,
+      "learning_rate": 1e-05,
+      "loss": 0.0322,
+      "step": 677
+    },
+    {
+      "epoch": 0.339,
+      "grad_norm": 0.12876645046050272,
+      "learning_rate": 1e-05,
+      "loss": 0.0127,
+      "step": 678
+    },
+    {
+      "epoch": 0.3395,
+      "grad_norm": 0.12336988858308351,
+      "learning_rate": 1e-05,
+      "loss": 0.0114,
+      "step": 679
+    },
+    {
+      "epoch": 0.34,
+      "grad_norm": 0.11038801232074134,
+      "learning_rate": 1e-05,
+      "loss": 0.0112,
+      "step": 680
+    },
+    {
+      "epoch": 0.3405,
+      "grad_norm": 0.13782079916676085,
+      "learning_rate": 1e-05,
+      "loss": 0.0156,
+      "step": 681
+    },
+    {
+      "epoch": 0.341,
+      "grad_norm": 0.12593807949317973,
+      "learning_rate": 1e-05,
+      "loss": 0.0179,
+      "step": 682
+    },
+    {
+      "epoch": 0.3415,
+      "grad_norm": 0.13416291611922937,
+      "learning_rate": 1e-05,
+      "loss": 0.0167,
+      "step": 683
+    },
+    {
+      "epoch": 0.342,
+      "grad_norm": 0.14107509427243767,
+      "learning_rate": 1e-05,
+      "loss": 0.0182,
+      "step": 684
+    },
+    {
+      "epoch": 0.3425,
+      "grad_norm": 0.13829290958101634,
+      "learning_rate": 1e-05,
+      "loss": 0.0133,
+      "step": 685
+    },
+    {
+      "epoch": 0.343,
+      "grad_norm": 0.12072602797225344,
+      "learning_rate": 1e-05,
+      "loss": 0.0135,
+      "step": 686
+    },
+    {
+      "epoch": 0.3435,
+      "grad_norm": 0.09808985286323638,
+      "learning_rate": 1e-05,
+      "loss": 0.0109,
+      "step": 687
+    },
+    {
+      "epoch": 0.344,
+      "grad_norm": 0.1237155662701831,
+      "learning_rate": 1e-05,
+      "loss": 0.0137,
+      "step": 688
+    },
+    {
+      "epoch": 0.3445,
+      "grad_norm": 0.11512509686864711,
+      "learning_rate": 1e-05,
+      "loss": 0.0131,
+      "step": 689
+    },
+    {
+      "epoch": 0.345,
+      "grad_norm": 0.10310189822258317,
+      "learning_rate": 1e-05,
+      "loss": 0.011,
+      "step": 690
+    },
+    {
+      "epoch": 0.3455,
+      "grad_norm": 0.17008589258309467,
+      "learning_rate": 1e-05,
+      "loss": 0.0219,
+      "step": 691
+    },
+    {
+      "epoch": 0.346,
+      "grad_norm": 0.12175425158539896,
+      "learning_rate": 1e-05,
+      "loss": 0.014,
+      "step": 692
+    },
+    {
+      "epoch": 0.3465,
+      "grad_norm": 0.11242731302801981,
+      "learning_rate": 1e-05,
+      "loss": 0.0142,
+      "step": 693
+    },
+    {
+      "epoch": 0.347,
+      "grad_norm": 0.13975508336225442,
+      "learning_rate": 1e-05,
+      "loss": 0.0149,
+      "step": 694
+    },
+    {
+      "epoch": 0.3475,
+      "grad_norm": 0.08820258618918772,
+      "learning_rate": 1e-05,
+      "loss": 0.008,
+      "step": 695
+    },
+    {
+      "epoch": 0.348,
+      "grad_norm": 0.1343553646574964,
+      "learning_rate": 1e-05,
+      "loss": 0.014,
+      "step": 696
+    },
+    {
+      "epoch": 0.3485,
+      "grad_norm": 0.13274844121268298,
+      "learning_rate": 1e-05,
+      "loss": 0.0143,
+      "step": 697
+    },
+    {
+      "epoch": 0.349,
+      "grad_norm": 0.1412865550899799,
+      "learning_rate": 1e-05,
+      "loss": 0.0164,
+      "step": 698
+    },
+    {
+      "epoch": 0.3495,
+      "grad_norm": 0.1471608978417448,
+      "learning_rate": 1e-05,
+      "loss": 0.0144,
+      "step": 699
+    },
+    {
+      "epoch": 0.35,
+      "grad_norm": 0.13861936123681107,
+      "learning_rate": 1e-05,
+      "loss": 0.0151,
+      "step": 700
+    },
+    {
+      "epoch": 0.3505,
+      "grad_norm": 0.11790173787959472,
+      "learning_rate": 1e-05,
+      "loss": 0.0252,
+      "step": 701
+    },
+    {
+      "epoch": 0.351,
+      "grad_norm": 0.12166501219045824,
+      "learning_rate": 1e-05,
+      "loss": 0.0145,
+      "step": 702
+    },
+    {
+      "epoch": 0.3515,
+      "grad_norm": 0.1082871078184047,
+      "learning_rate": 1e-05,
+      "loss": 0.0095,
+      "step": 703
+    },
+    {
+      "epoch": 0.352,
+      "grad_norm": 0.12226846848831563,
+      "learning_rate": 1e-05,
+      "loss": 0.0141,
+      "step": 704
+    },
+    {
+      "epoch": 0.352,
+      "eval_dev_acc": 0.52734375,
+      "eval_dev_token": 5760.470703125,
+      "eval_runtime": 408.3746,
+      "eval_samples_per_second": 0.157,
+      "eval_steps_per_second": 0.002,
+      "step": 704
+    },
+    {
+      "epoch": 0.3525,
+      "grad_norm": 0.09666366634628189,
+      "learning_rate": 1e-05,
+      "loss": 0.0097,
+      "step": 705
+    },
+    {
+      "epoch": 0.353,
+      "grad_norm": 0.11283217042776503,
+      "learning_rate": 1e-05,
+      "loss": 0.0143,
+      "step": 706
+    },
+    {
+      "epoch": 0.3535,
+      "grad_norm": 0.12289011751838193,
+      "learning_rate": 1e-05,
+      "loss": 0.0173,
+      "step": 707
+    },
+    {
+      "epoch": 0.354,
+      "grad_norm": 0.11335657340497375,
+      "learning_rate": 1e-05,
+      "loss": 0.0193,
+      "step": 708
+    },
+    {
+      "epoch": 0.3545,
+      "grad_norm": 0.12878402333619293,
+      "learning_rate": 1e-05,
+      "loss": 0.0198,
+      "step": 709
+    },
+    {
+      "epoch": 0.355,
+      "grad_norm": 0.09768170830344952,
+      "learning_rate": 1e-05,
+      "loss": 0.0123,
+      "step": 710
+    },
+    {
+      "epoch": 0.3555,
+      "grad_norm": 0.11142711749521704,
+      "learning_rate": 1e-05,
+      "loss": 0.0142,
+      "step": 711
+    },
+    {
+      "epoch": 0.356,
+      "grad_norm": 0.12263582497303413,
+      "learning_rate": 1e-05,
+      "loss": 0.0125,
+      "step": 712
+    },
+    {
+      "epoch": 0.3565,
+      "grad_norm": 0.10602173709981424,
+      "learning_rate": 1e-05,
+      "loss": 0.013,
+      "step": 713
+    },
+    {
+      "epoch": 0.357,
+      "grad_norm": 0.09797092044857854,
+      "learning_rate": 1e-05,
+      "loss": 0.0097,
+      "step": 714
+    },
+    {
+      "epoch": 0.3575,
+      "grad_norm": 0.10267028455310084,
+      "learning_rate": 1e-05,
+      "loss": 0.0153,
+      "step": 715
+    },
+    {
+      "epoch": 0.358,
+      "grad_norm": 0.11247269902696218,
+      "learning_rate": 1e-05,
+      "loss": 0.0124,
+      "step": 716
+    },
+    {
+      "epoch": 0.3585,
+      "grad_norm": 0.10192090643527008,
+      "learning_rate": 1e-05,
+      "loss": 0.01,
+      "step": 717
+    },
+    {
+      "epoch": 0.359,
+      "grad_norm": 0.15447030723754146,
+      "learning_rate": 1e-05,
+      "loss": 0.0191,
+      "step": 718
+    },
+    {
+      "epoch": 0.3595,
+      "grad_norm": 0.12653173296866044,
+      "learning_rate": 1e-05,
+      "loss": 0.0144,
+      "step": 719
+    },
+    {
+      "epoch": 0.36,
+      "grad_norm": 0.09798115770372441,
+      "learning_rate": 1e-05,
+      "loss": 0.0113,
+      "step": 720
+    },
+    {
+      "epoch": 0.3605,
+      "grad_norm": 0.13673705245890774,
+      "learning_rate": 1e-05,
+      "loss": 0.0176,
+      "step": 721
+    },
+    {
+      "epoch": 0.361,
+      "grad_norm": 0.10484924187213118,
+      "learning_rate": 1e-05,
+      "loss": 0.011,
+      "step": 722
+    },
+    {
+      "epoch": 0.3615,
+      "grad_norm": 0.12026390551992476,
+      "learning_rate": 1e-05,
+      "loss": 0.0161,
+      "step": 723
+    },
+    {
+      "epoch": 0.362,
+      "grad_norm": 0.12248845158519388,
+      "learning_rate": 1e-05,
+      "loss": 0.0135,
+      "step": 724
+    },
+    {
+      "epoch": 0.3625,
+      "grad_norm": 0.08242254382606763,
+      "learning_rate": 1e-05,
+      "loss": 0.0103,
+      "step": 725
+    },
+    {
+      "epoch": 0.363,
+      "grad_norm": 0.1310711546429888,
+      "learning_rate": 1e-05,
+      "loss": 0.0213,
+      "step": 726
+    },
+    {
+      "epoch": 0.3635,
+      "grad_norm": 0.11504216894290854,
+      "learning_rate": 1e-05,
+      "loss": 0.0122,
+      "step": 727
+    },
+    {
+      "epoch": 0.364,
+      "grad_norm": 0.12368075631500317,
+      "learning_rate": 1e-05,
+      "loss": 0.0184,
+      "step": 728
+    },
+    {
+      "epoch": 0.3645,
+      "grad_norm": 0.12437061544598803,
+      "learning_rate": 1e-05,
+      "loss": 0.0152,
+      "step": 729
+    },
+    {
+      "epoch": 0.365,
+      "grad_norm": 0.13397286445240938,
+      "learning_rate": 1e-05,
+      "loss": 0.0158,
+      "step": 730
+    },
+    {
+      "epoch": 0.3655,
+      "grad_norm": 0.0982171426550068,
+      "learning_rate": 1e-05,
+      "loss": 0.0096,
+      "step": 731
+    },
+    {
+      "epoch": 0.366,
+      "grad_norm": 0.10415731829601732,
+      "learning_rate": 1e-05,
+      "loss": 0.0105,
+      "step": 732
+    },
+    {
+      "epoch": 0.3665,
+      "grad_norm": 0.13524982542339864,
+      "learning_rate": 1e-05,
+      "loss": 0.0183,
+      "step": 733
+    },
+    {
+      "epoch": 0.367,
+      "grad_norm": 0.09664339538491498,
+      "learning_rate": 1e-05,
+      "loss": 0.0115,
+      "step": 734
+    },
+    {
+      "epoch": 0.3675,
+      "grad_norm": 0.0851015567710096,
+      "learning_rate": 1e-05,
+      "loss": 0.0112,
+      "step": 735
+    },
+    {
+      "epoch": 0.368,
+      "grad_norm": 0.10062979284816816,
+      "learning_rate": 1e-05,
+      "loss": 0.0113,
+      "step": 736
+    },
+    {
+      "epoch": 0.368,
+      "eval_dev_acc": 0.53515625,
+      "eval_dev_token": 5898.23046875,
+      "eval_runtime": 365.57,
+      "eval_samples_per_second": 0.175,
+      "eval_steps_per_second": 0.003,
+      "step": 736
+    },
+    {
+      "epoch": 0.3685,
+      "grad_norm": 0.11264161967062039,
+      "learning_rate": 1e-05,
+      "loss": 0.0108,
+      "step": 737
+    },
+    {
+      "epoch": 0.369,
+      "grad_norm": 0.11222433927286389,
+      "learning_rate": 1e-05,
+      "loss": 0.0104,
+      "step": 738
+    },
+    {
+      "epoch": 0.3695,
+      "grad_norm": 0.20055167671089585,
+      "learning_rate": 1e-05,
+      "loss": 0.013,
+      "step": 739
+    },
+    {
+      "epoch": 0.37,
+      "grad_norm": 0.10491538007846005,
+      "learning_rate": 1e-05,
+      "loss": 0.0131,
+      "step": 740
+    },
+    {
+      "epoch": 0.3705,
+      "grad_norm": 0.09174967069041229,
+      "learning_rate": 1e-05,
+      "loss": 0.0095,
+      "step": 741
+    },
+    {
+      "epoch": 0.371,
+      "grad_norm": 0.10582756451393364,
+      "learning_rate": 1e-05,
+      "loss": 0.012,
+      "step": 742
+    },
+    {
+      "epoch": 0.3715,
+      "grad_norm": 0.09194993895839906,
+      "learning_rate": 1e-05,
+      "loss": 0.0097,
+      "step": 743
+    },
+    {
+      "epoch": 0.372,
+      "grad_norm": 0.08858743590625234,
+      "learning_rate": 1e-05,
+      "loss": 0.01,
+      "step": 744
+    },
+    {
+      "epoch": 0.3725,
+      "grad_norm": 0.10826764845042156,
+      "learning_rate": 1e-05,
+      "loss": 0.0145,
+      "step": 745
+    },
+    {
+      "epoch": 0.373,
+      "grad_norm": 0.11033032167695773,
+      "learning_rate": 1e-05,
+      "loss": 0.0113,
+      "step": 746
+    },
+    {
+      "epoch": 0.3735,
+      "grad_norm": 0.15107498232603372,
+      "learning_rate": 1e-05,
+      "loss": 0.0157,
+      "step": 747
+    },
+    {
+      "epoch": 0.374,
+      "grad_norm": 0.11411131491498813,
+      "learning_rate": 1e-05,
+      "loss": 0.0114,
+      "step": 748
+    },
+    {
+      "epoch": 0.3745,
+      "grad_norm": 0.12425138196940645,
+      "learning_rate": 1e-05,
+      "loss": 0.0138,
+      "step": 749
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.10647093971928946,
+      "learning_rate": 1e-05,
+      "loss": 0.0107,
+      "step": 750
+    },
+    {
+      "epoch": 0.3755,
+      "grad_norm": 0.12413448325714936,
+      "learning_rate": 1e-05,
+      "loss": 0.014,
+      "step": 751
+    },
+    {
+      "epoch": 0.376,
+      "grad_norm": 0.105340917878943,
+      "learning_rate": 1e-05,
+      "loss": 0.0103,
+      "step": 752
+    },
+    {
+      "epoch": 0.3765,
+      "grad_norm": 0.11802541191692037,
+      "learning_rate": 1e-05,
+      "loss": 0.0127,
+      "step": 753
+    },
+    {
+      "epoch": 0.377,
+      "grad_norm": 0.09864810460683521,
+      "learning_rate": 1e-05,
+      "loss": 0.0128,
+      "step": 754
+    },
+    {
+      "epoch": 0.3775,
+      "grad_norm": 0.12491513055109611,
+      "learning_rate": 1e-05,
+      "loss": 0.0102,
+      "step": 755
+    },
+    {
+      "epoch": 0.378,
+      "grad_norm": 0.11778790720208179,
+      "learning_rate": 1e-05,
+      "loss": 0.0093,
+      "step": 756
+    },
+    {
+      "epoch": 0.3785,
+      "grad_norm": 0.11902239371415295,
+      "learning_rate": 1e-05,
+      "loss": 0.0115,
+      "step": 757
+    },
+    {
+      "epoch": 0.379,
+      "grad_norm": 0.1025498591924567,
+      "learning_rate": 1e-05,
+      "loss": 0.0099,
+      "step": 758
+    },
+    {
+      "epoch": 0.3795,
+      "grad_norm": 0.14120840112868438,
+      "learning_rate": 1e-05,
+      "loss": 0.0123,
+      "step": 759
+    },
+    {
+      "epoch": 0.38,
+      "grad_norm": 0.08964665828244849,
+      "learning_rate": 1e-05,
+      "loss": 0.0116,
+      "step": 760
+    },
+    {
+      "epoch": 0.3805,
+      "grad_norm": 0.13533754381134197,
+      "learning_rate": 1e-05,
+      "loss": 0.0142,
+      "step": 761
+    },
+    {
+      "epoch": 0.381,
+      "grad_norm": 0.11151852649444051,
+      "learning_rate": 1e-05,
+      "loss": 0.0122,
+      "step": 762
+    },
+    {
+      "epoch": 0.3815,
+      "grad_norm": 0.12448616858204287,
+      "learning_rate": 1e-05,
+      "loss": 0.0148,
+      "step": 763
+    },
+    {
+      "epoch": 0.382,
+      "grad_norm": 0.10160526390201502,
+      "learning_rate": 1e-05,
+      "loss": 0.0102,
+      "step": 764
+    },
+    {
+      "epoch": 0.3825,
+      "grad_norm": 0.12329039775788013,
+      "learning_rate": 1e-05,
+      "loss": 0.0226,
+      "step": 765
+    },
+    {
+      "epoch": 0.383,
+      "grad_norm": 0.1706851101549876,
+      "learning_rate": 1e-05,
+      "loss": 0.013,
+      "step": 766
+    },
+    {
+      "epoch": 0.3835,
+      "grad_norm": 0.11518698920716465,
+      "learning_rate": 1e-05,
+      "loss": 0.0136,
+      "step": 767
+    },
+    {
+      "epoch": 0.384,
+      "grad_norm": 0.1130084278865893,
+      "learning_rate": 1e-05,
+      "loss": 0.0151,
+      "step": 768
+    },
+    {
+      "epoch": 0.384,
+      "eval_dev_acc": 0.505859375,
+      "eval_dev_token": 5660.353515625,
+      "eval_runtime": 362.2613,
+      "eval_samples_per_second": 0.177,
+      "eval_steps_per_second": 0.003,
+      "step": 768
+    },
+    {
+      "epoch": 0.3845,
+      "grad_norm": 0.12092852490034717,
+      "learning_rate": 1e-05,
+      "loss": 0.0111,
+      "step": 769
+    },
+    {
+      "epoch": 0.385,
+      "grad_norm": 0.12041159291779986,
+      "learning_rate": 1e-05,
+      "loss": 0.0157,
+      "step": 770
+    },
+    {
+      "epoch": 0.3855,
+      "grad_norm": 0.13135896730332378,
+      "learning_rate": 1e-05,
+      "loss": 0.0149,
+      "step": 771
+    },
+    {
+      "epoch": 0.386,
+      "grad_norm": 0.09838622926936438,
+      "learning_rate": 1e-05,
+      "loss": 0.0113,
+      "step": 772
+    },
+    {
+      "epoch": 0.3865,
+      "grad_norm": 0.1118823935585986,
+      "learning_rate": 1e-05,
+      "loss": 0.016,
+      "step": 773
+    },
+    {
+      "epoch": 0.387,
+      "grad_norm": 0.10549287704509497,
+      "learning_rate": 1e-05,
+      "loss": 0.0097,
+      "step": 774
+    },
+    {
+      "epoch": 0.3875,
+      "grad_norm": 0.1232106501038667,
+      "learning_rate": 1e-05,
+      "loss": 0.013,
+      "step": 775
+    },
+    {
+      "epoch": 0.388,
+      "grad_norm": 0.12975165037101072,
+      "learning_rate": 1e-05,
+      "loss": 0.0121,
+      "step": 776
+    },
+    {
+      "epoch": 0.3885,
+      "grad_norm": 0.13595262265606586,
+      "learning_rate": 1e-05,
+      "loss": 0.0126,
+      "step": 777
+    },
+    {
+      "epoch": 0.389,
+      "grad_norm": 0.1371181387315353,
+      "learning_rate": 1e-05,
+      "loss": 0.0123,
+      "step": 778
+    },
+    {
+      "epoch": 0.3895,
+      "grad_norm": 0.13359565264106144,
+      "learning_rate": 1e-05,
+      "loss": 0.0153,
+      "step": 779
+    },
+    {
+      "epoch": 0.39,
+      "grad_norm": 0.11480062436980976,
+      "learning_rate": 1e-05,
+      "loss": 0.0107,
+      "step": 780
+    },
+    {
+      "epoch": 0.3905,
+      "grad_norm": 0.09947922362782227,
+      "learning_rate": 1e-05,
+      "loss": 0.0086,
+      "step": 781
+    },
+    {
+      "epoch": 0.391,
+      "grad_norm": 0.08734624126570777,
+      "learning_rate": 1e-05,
+      "loss": 0.0076,
+      "step": 782
+    },
+    {
+      "epoch": 0.3915,
+      "grad_norm": 0.10267346744796824,
+      "learning_rate": 1e-05,
+      "loss": 0.0097,
+      "step": 783
+    },
+    {
+      "epoch": 0.392,
+      "grad_norm": 0.09576192367034056,
+      "learning_rate": 1e-05,
+      "loss": 0.0083,
+      "step": 784
+    },
+    {
+      "epoch": 0.3925,
+      "grad_norm": 0.14141610877259214,
+      "learning_rate": 1e-05,
+      "loss": 0.0139,
+      "step": 785
+    },
+    {
+      "epoch": 0.393,
+      "grad_norm": 0.12071117391984963,
+      "learning_rate": 1e-05,
+      "loss": 0.0115,
+      "step": 786
+    },
+    {
+      "epoch": 0.3935,
+      "grad_norm": 0.13172358788988653,
+      "learning_rate": 1e-05,
+      "loss": 0.0129,
+      "step": 787
+    },
+    {
+      "epoch": 0.394,
+      "grad_norm": 0.1345424657278948,
+      "learning_rate": 1e-05,
+      "loss": 0.014,
+      "step": 788
+    },
+    {
+      "epoch": 0.3945,
+      "grad_norm": 0.1166885723717584,
+      "learning_rate": 1e-05,
+      "loss": 0.0125,
+      "step": 789
+    },
+    {
+      "epoch": 0.395,
+      "grad_norm": 0.1513194581227132,
+      "learning_rate": 1e-05,
+      "loss": 0.0145,
+      "step": 790
+    },
+    {
+      "epoch": 0.3955,
+      "grad_norm": 0.11105132556458801,
+      "learning_rate": 1e-05,
+      "loss": 0.011,
+      "step": 791
+    },
+    {
+      "epoch": 0.396,
+      "grad_norm": 0.0989302250949553,
+      "learning_rate": 1e-05,
+      "loss": 0.0086,
+      "step": 792
+    },
+    {
+      "epoch": 0.3965,
+      "grad_norm": 0.1384321782028591,
+      "learning_rate": 1e-05,
+      "loss": 0.0122,
+      "step": 793
+    },
+    {
+      "epoch": 0.397,
+      "grad_norm": 0.1439582094147518,
+      "learning_rate": 1e-05,
+      "loss": 0.0142,
+      "step": 794
+    },
+    {
+      "epoch": 0.3975,
+      "grad_norm": 0.11905534302900607,
+      "learning_rate": 1e-05,
+      "loss": 0.0102,
+      "step": 795
+    },
+    {
+      "epoch": 0.398,
+      "grad_norm": 0.1353943654211476,
+      "learning_rate": 1e-05,
+      "loss": 0.015,
+      "step": 796
+    },
+    {
+      "epoch": 0.3985,
+      "grad_norm": 0.1227008050574365,
+      "learning_rate": 1e-05,
+      "loss": 0.0159,
+      "step": 797
+    },
+    {
+      "epoch": 0.399,
+      "grad_norm": 0.10084479670557596,
+      "learning_rate": 1e-05,
+      "loss": 0.0103,
+      "step": 798
+    },
+    {
+      "epoch": 0.3995,
+      "grad_norm": 0.11430101131806415,
+      "learning_rate": 1e-05,
+      "loss": 0.0128,
+      "step": 799
+    },
+    {
+      "epoch": 0.4,
+      "grad_norm": 0.12720193634725216,
+      "learning_rate": 1e-05,
+      "loss": 0.0166,
+      "step": 800
+    },
+    {
+      "epoch": 0.4,
+      "eval_dev_acc": 0.49609375,
+      "eval_dev_token": 5933.87109375,
+      "eval_runtime": 371.4205,
+      "eval_samples_per_second": 0.172,
+      "eval_steps_per_second": 0.003,
+      "step": 800
+    },
+    {
+      "epoch": 0.4005,
+      "grad_norm": 0.10747534966369286,
+      "learning_rate": 1e-05,
+      "loss": 0.0113,
+      "step": 801
+    },
+    {
+      "epoch": 0.401,
+      "grad_norm": 0.09755157150747987,
+      "learning_rate": 1e-05,
+      "loss": 0.0094,
+      "step": 802
+    },
+    {
+      "epoch": 0.4015,
+      "grad_norm": 0.1093754763997552,
+      "learning_rate": 1e-05,
+      "loss": 0.0136,
+      "step": 803
+    },
+    {
+      "epoch": 0.402,
+      "grad_norm": 0.12186788180512477,
+      "learning_rate": 1e-05,
+      "loss": 0.0112,
+      "step": 804
+    },
+    {
+      "epoch": 0.4025,
+      "grad_norm": 0.12341274893925132,
+      "learning_rate": 1e-05,
+      "loss": 0.0126,
+      "step": 805
+    },
+    {
+      "epoch": 0.403,
+      "grad_norm": 0.1000051489528446,
+      "learning_rate": 1e-05,
+      "loss": 0.0171,
+      "step": 806
+    },
+    {
+      "epoch": 0.4035,
+      "grad_norm": 0.09210699513663904,
+      "learning_rate": 1e-05,
+      "loss": 0.0097,
+      "step": 807
+    },
+    {
+      "epoch": 0.404,
+      "grad_norm": 0.12629053026155362,
+      "learning_rate": 1e-05,
+      "loss": 0.0126,
+      "step": 808
+    },
+    {
+      "epoch": 0.4045,
+      "grad_norm": 0.11979707428750866,
+      "learning_rate": 1e-05,
+      "loss": 0.0129,
+      "step": 809
+    },
+    {
+      "epoch": 0.405,
+      "grad_norm": 0.13240620090939892,
+      "learning_rate": 1e-05,
+      "loss": 0.0173,
+      "step": 810
+    },
+    {
+      "epoch": 0.4055,
+      "grad_norm": 0.11509826968311627,
+      "learning_rate": 1e-05,
+      "loss": 0.0133,
+      "step": 811
+    },
+    {
+      "epoch": 0.406,
+      "grad_norm": 0.09930674216692448,
+      "learning_rate": 1e-05,
+      "loss": 0.0103,
+      "step": 812
+    },
+    {
+      "epoch": 0.4065,
+      "grad_norm": 0.12486374889735856,
+      "learning_rate": 1e-05,
+      "loss": 0.0152,
+      "step": 813
+    },
+    {
+      "epoch": 0.407,
+      "grad_norm": 0.09439307719243419,
+      "learning_rate": 1e-05,
+      "loss": 0.0113,
+      "step": 814
+    },
+    {
+      "epoch": 0.4075,
+      "grad_norm": 0.09010069214916468,
+      "learning_rate": 1e-05,
+      "loss": 0.0101,
+      "step": 815
+    },
+    {
+      "epoch": 0.408,
+      "grad_norm": 0.10807091465236611,
+      "learning_rate": 1e-05,
+      "loss": 0.0136,
+      "step": 816
+    },
+    {
+      "epoch": 0.4085,
+      "grad_norm": 0.09320554728801374,
+      "learning_rate": 1e-05,
+      "loss": 0.0102,
+      "step": 817
+    },
+    {
+      "epoch": 0.409,
+      "grad_norm": 0.09578310039513883,
+      "learning_rate": 1e-05,
+      "loss": 0.0105,
+      "step": 818
+    },
+    {
+      "epoch": 0.4095,
+      "grad_norm": 0.10795646602867415,
+      "learning_rate": 1e-05,
+      "loss": 0.0141,
+      "step": 819
+    },
+    {
+      "epoch": 0.41,
+      "grad_norm": 0.11260765818079684,
+      "learning_rate": 1e-05,
+      "loss": 0.0123,
+      "step": 820
+    },
+    {
+      "epoch": 0.4105,
+      "grad_norm": 0.13162780397133028,
+      "learning_rate": 1e-05,
+      "loss": 0.0141,
+      "step": 821
+    },
+    {
+      "epoch": 0.411,
+      "grad_norm": 0.10884500486087925,
+      "learning_rate": 1e-05,
+      "loss": 0.0122,
+      "step": 822
+    },
+    {
+      "epoch": 0.4115,
+      "grad_norm": 0.12472839162847292,
+      "learning_rate": 1e-05,
+      "loss": 0.0135,
+      "step": 823
+    },
+    {
+      "epoch": 0.412,
+      "grad_norm": 0.13289735991638021,
+      "learning_rate": 1e-05,
+      "loss": 0.0112,
+      "step": 824
+    },
+    {
+      "epoch": 0.4125,
+      "grad_norm": 0.12509974441211302,
+      "learning_rate": 1e-05,
+      "loss": 0.0256,
+      "step": 825
+    },
+    {
+      "epoch": 0.413,
+      "grad_norm": 0.12014632147622897,
+      "learning_rate": 1e-05,
+      "loss": 0.0113,
+      "step": 826
+    },
+    {
+      "epoch": 0.4135,
+      "grad_norm": 0.09172916013688245,
+      "learning_rate": 1e-05,
+      "loss": 0.0082,
+      "step": 827
+    },
+    {
+      "epoch": 0.414,
+      "grad_norm": 0.09305774811224422,
+      "learning_rate": 1e-05,
+      "loss": 0.0074,
+      "step": 828
+    },
+    {
+      "epoch": 0.4145,
+      "grad_norm": 0.12720506111352092,
+      "learning_rate": 1e-05,
+      "loss": 0.0091,
+      "step": 829
+    },
+    {
+      "epoch": 0.415,
+      "grad_norm": 0.0815065287803298,
+      "learning_rate": 1e-05,
+      "loss": 0.0061,
+      "step": 830
+    },
+    {
+      "epoch": 0.4155,
+      "grad_norm": 0.09730425117259746,
+      "learning_rate": 1e-05,
+      "loss": 0.0079,
+      "step": 831
+    },
+    {
+      "epoch": 0.416,
+      "grad_norm": 0.09579694307116488,
+      "learning_rate": 1e-05,
+      "loss": 0.0098,
+      "step": 832
+    },
+    {
+      "epoch": 0.416,
+      "eval_dev_acc": 0.513671875,
+      "eval_dev_token": 4446.53125,
+      "eval_runtime": 329.208,
+      "eval_samples_per_second": 0.194,
+      "eval_steps_per_second": 0.003,
+      "step": 832
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 2000,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 9223372036854775807,
+  "save_steps": 32,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 224786117410816.0,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}