diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,100969 @@
+{
+  "best_metric": 0.7562862669245648,
+  "best_model_checkpoint": "/train_db_id/checkpoint-66500",
+  "epoch": 62.03357817418678,
+  "global_step": 66500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.999988339552239e-05,
+      "loss": 3.3987,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.999953358208955e-05,
+      "loss": 2.8913,
+      "step": 4
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9999067164179105e-05,
+      "loss": 1.6031,
+      "step": 8
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.999860074626866e-05,
+      "loss": 1.5674,
+      "step": 12
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.999813432835821e-05,
+      "loss": 1.075,
+      "step": 16
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.999766791044776e-05,
+      "loss": 1.0344,
+      "step": 20
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.999720149253732e-05,
+      "loss": 0.839,
+      "step": 24
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9996735074626866e-05,
+      "loss": 0.6872,
+      "step": 28
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999626865671642e-05,
+      "loss": 0.6713,
+      "step": 32
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999580223880597e-05,
+      "loss": 0.707,
+      "step": 36
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9995335820895524e-05,
+      "loss": 0.6456,
+      "step": 40
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999486940298508e-05,
+      "loss": 0.6796,
+      "step": 44
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999440298507463e-05,
+      "loss": 0.5273,
+      "step": 48
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.999393656716418e-05,
+      "loss": 0.9018,
+      "step": 52
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.999347014925374e-05,
+      "loss": 0.6358,
+      "step": 56
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9993003731343285e-05,
+      "loss": 0.462,
+      "step": 60
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9992537313432833e-05,
+      "loss": 0.5077,
+      "step": 64
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.999207089552239e-05,
+      "loss": 0.5189,
+      "step": 68
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.999160447761194e-05,
+      "loss": 0.5189,
+      "step": 72
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.999113805970149e-05,
+      "loss": 0.4512,
+      "step": 76
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9990671641791046e-05,
+      "loss": 0.4906,
+      "step": 80
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.99902052238806e-05,
+      "loss": 0.4084,
+      "step": 84
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9989738805970156e-05,
+      "loss": 0.3692,
+      "step": 88
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9989272388059704e-05,
+      "loss": 0.459,
+      "step": 92
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.998880597014925e-05,
+      "loss": 0.4347,
+      "step": 96
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9988339552238814e-05,
+      "loss": 0.39,
+      "step": 100
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.998787313432836e-05,
+      "loss": 0.4953,
+      "step": 104
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.998740671641791e-05,
+      "loss": 0.4506,
+      "step": 108
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.9986940298507465e-05,
+      "loss": 0.407,
+      "step": 112
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.998647388059702e-05,
+      "loss": 0.3425,
+      "step": 116
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.998600746268657e-05,
+      "loss": 0.4349,
+      "step": 120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.998554104477612e-05,
+      "loss": 0.3413,
+      "step": 124
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.998507462686567e-05,
+      "loss": 0.3923,
+      "step": 128
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.9984608208955226e-05,
+      "loss": 0.4086,
+      "step": 132
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.998414179104478e-05,
+      "loss": 0.3972,
+      "step": 136
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.998367537313433e-05,
+      "loss": 0.3259,
+      "step": 140
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.9983208955223884e-05,
+      "loss": 0.3852,
+      "step": 144
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.998274253731344e-05,
+      "loss": 0.3536,
+      "step": 148
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.998227611940299e-05,
+      "loss": 0.3187,
+      "step": 152
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.9981809701492535e-05,
+      "loss": 0.2841,
+      "step": 156
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.99813432835821e-05,
+      "loss": 0.4,
+      "step": 160
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.9980876865671645e-05,
+      "loss": 0.319,
+      "step": 164
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.998041044776119e-05,
+      "loss": 0.285,
+      "step": 168
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.997994402985075e-05,
+      "loss": 0.3721,
+      "step": 172
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.99794776119403e-05,
+      "loss": 0.2752,
+      "step": 176
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.997901119402985e-05,
+      "loss": 0.2895,
+      "step": 180
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.9978544776119406e-05,
+      "loss": 0.2652,
+      "step": 184
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.9978078358208954e-05,
+      "loss": 0.4065,
+      "step": 188
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.997761194029851e-05,
+      "loss": 0.2846,
+      "step": 192
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.9977145522388064e-05,
+      "loss": 0.3481,
+      "step": 196
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.997667910447761e-05,
+      "loss": 0.33,
+      "step": 200
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.997621268656717e-05,
+      "loss": 0.3122,
+      "step": 204
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.997574626865672e-05,
+      "loss": 0.2949,
+      "step": 208
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.997527985074627e-05,
+      "loss": 0.2282,
+      "step": 212
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.997481343283582e-05,
+      "loss": 0.2633,
+      "step": 216
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.997434701492538e-05,
+      "loss": 0.2847,
+      "step": 220
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.997388059701493e-05,
+      "loss": 0.252,
+      "step": 224
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.9973414179104476e-05,
+      "loss": 0.2747,
+      "step": 228
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.997294776119403e-05,
+      "loss": 0.2968,
+      "step": 232
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.9972481343283586e-05,
+      "loss": 0.3299,
+      "step": 236
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.9972014925373134e-05,
+      "loss": 0.3038,
+      "step": 240
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.997154850746269e-05,
+      "loss": 0.214,
+      "step": 244
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.997108208955224e-05,
+      "loss": 0.2138,
+      "step": 248
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.99706156716418e-05,
+      "loss": 0.2308,
+      "step": 252
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.997014925373135e-05,
+      "loss": 0.3071,
+      "step": 256
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.9969682835820895e-05,
+      "loss": 0.3456,
+      "step": 260
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.996921641791045e-05,
+      "loss": 0.2254,
+      "step": 264
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.9968750000000005e-05,
+      "loss": 0.2141,
+      "step": 268
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.996828358208955e-05,
+      "loss": 0.1988,
+      "step": 272
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.996781716417911e-05,
+      "loss": 0.2797,
+      "step": 276
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.996735074626866e-05,
+      "loss": 0.265,
+      "step": 280
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.996688432835821e-05,
+      "loss": 0.3751,
+      "step": 284
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.9966417910447766e-05,
+      "loss": 0.2993,
+      "step": 288
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.9965951492537314e-05,
+      "loss": 0.2626,
+      "step": 292
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.996548507462687e-05,
+      "loss": 0.2083,
+      "step": 296
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.9965018656716424e-05,
+      "loss": 0.1907,
+      "step": 300
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.996455223880597e-05,
+      "loss": 0.2737,
+      "step": 304
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.996408582089552e-05,
+      "loss": 0.3005,
+      "step": 308
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.996361940298508e-05,
+      "loss": 0.2399,
+      "step": 312
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.996315298507463e-05,
+      "loss": 0.2554,
+      "step": 316
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.996268656716418e-05,
+      "loss": 0.2626,
+      "step": 320
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.996222014925373e-05,
+      "loss": 0.2391,
+      "step": 324
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.996175373134329e-05,
+      "loss": 0.2105,
+      "step": 328
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.9961287313432836e-05,
+      "loss": 0.2333,
+      "step": 332
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.996082089552239e-05,
+      "loss": 0.2182,
+      "step": 336
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.9960354477611946e-05,
+      "loss": 0.2743,
+      "step": 340
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.9959888059701494e-05,
+      "loss": 0.1997,
+      "step": 344
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.995942164179105e-05,
+      "loss": 0.3057,
+      "step": 348
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.99589552238806e-05,
+      "loss": 0.1665,
+      "step": 352
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.995848880597015e-05,
+      "loss": 0.2077,
+      "step": 356
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.995802238805971e-05,
+      "loss": 0.229,
+      "step": 360
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.9957555970149255e-05,
+      "loss": 0.1626,
+      "step": 364
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.99570895522388e-05,
+      "loss": 0.2124,
+      "step": 368
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.9956623134328365e-05,
+      "loss": 0.1981,
+      "step": 372
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.995615671641791e-05,
+      "loss": 0.1941,
+      "step": 376
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.995569029850746e-05,
+      "loss": 0.1985,
+      "step": 380
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.9955223880597016e-05,
+      "loss": 0.2346,
+      "step": 384
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.995475746268657e-05,
+      "loss": 0.1867,
+      "step": 388
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.995429104477612e-05,
+      "loss": 0.1955,
+      "step": 392
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.9953824626865674e-05,
+      "loss": 0.2611,
+      "step": 396
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.995335820895522e-05,
+      "loss": 0.2636,
+      "step": 400
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.995289179104478e-05,
+      "loss": 0.2358,
+      "step": 404
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.995242537313433e-05,
+      "loss": 0.2215,
+      "step": 408
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.995195895522388e-05,
+      "loss": 0.1965,
+      "step": 412
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.9951492537313435e-05,
+      "loss": 0.2427,
+      "step": 416
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.995102611940299e-05,
+      "loss": 0.1671,
+      "step": 420
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.995055970149254e-05,
+      "loss": 0.1296,
+      "step": 424
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.995009328358209e-05,
+      "loss": 0.2114,
+      "step": 428
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.994962686567165e-05,
+      "loss": 0.1675,
+      "step": 432
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.9949160447761196e-05,
+      "loss": 0.162,
+      "step": 436
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.994869402985075e-05,
+      "loss": 0.1415,
+      "step": 440
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.99482276119403e-05,
+      "loss": 0.148,
+      "step": 444
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.9947761194029854e-05,
+      "loss": 0.1717,
+      "step": 448
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.994729477611941e-05,
+      "loss": 0.157,
+      "step": 452
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.994682835820896e-05,
+      "loss": 0.226,
+      "step": 456
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.9946361940298505e-05,
+      "loss": 0.2185,
+      "step": 460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.994589552238807e-05,
+      "loss": 0.2423,
+      "step": 464
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.9945429104477615e-05,
+      "loss": 0.2652,
+      "step": 468
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.994496268656716e-05,
+      "loss": 0.1864,
+      "step": 472
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.994449626865672e-05,
+      "loss": 0.1418,
+      "step": 476
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.994402985074627e-05,
+      "loss": 0.1907,
+      "step": 480
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.994356343283582e-05,
+      "loss": 0.1826,
+      "step": 484
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.9943097014925376e-05,
+      "loss": 0.1699,
+      "step": 488
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.994263059701493e-05,
+      "loss": 0.1956,
+      "step": 492
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.994216417910448e-05,
+      "loss": 0.1386,
+      "step": 496
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.9941697761194034e-05,
+      "loss": 0.1573,
+      "step": 500
+    },
+    {
+      "epoch": 0.47,
+      "eval_exact_match": 0.4410058027079304,
+      "eval_exec": 0.534816247582205,
+      "eval_loss": 0.1741107553243637,
+      "eval_runtime": 975.177,
+      "eval_samples_per_second": 1.06,
+      "step": 500
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.994123134328358e-05,
+      "loss": 0.165,
+      "step": 504
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.994076492537314e-05,
+      "loss": 0.1845,
+      "step": 508
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.994029850746269e-05,
+      "loss": 0.1452,
+      "step": 512
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.993983208955224e-05,
+      "loss": 0.1927,
+      "step": 516
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.993936567164179e-05,
+      "loss": 0.1153,
+      "step": 520
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.993889925373135e-05,
+      "loss": 0.1446,
+      "step": 524
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.99384328358209e-05,
+      "loss": 0.1632,
+      "step": 528
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.9937966417910446e-05,
+      "loss": 0.1256,
+      "step": 532
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.99375e-05,
+      "loss": 0.2019,
+      "step": 536
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.9937033582089556e-05,
+      "loss": 0.1862,
+      "step": 540
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.9936567164179104e-05,
+      "loss": 0.1931,
+      "step": 544
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.993610074626866e-05,
+      "loss": 0.16,
+      "step": 548
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.9935634328358214e-05,
+      "loss": 0.2193,
+      "step": 552
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.993516791044776e-05,
+      "loss": 0.1208,
+      "step": 556
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.993470149253732e-05,
+      "loss": 0.1524,
+      "step": 560
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.9934235074626865e-05,
+      "loss": 0.1348,
+      "step": 564
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.993376865671642e-05,
+      "loss": 0.1796,
+      "step": 568
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.9933302238805975e-05,
+      "loss": 0.1142,
+      "step": 572
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.993283582089552e-05,
+      "loss": 0.1417,
+      "step": 576
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.993236940298508e-05,
+      "loss": 0.1696,
+      "step": 580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.993190298507463e-05,
+      "loss": 0.1412,
+      "step": 584
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.993143656716418e-05,
+      "loss": 0.1188,
+      "step": 588
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.9930970149253736e-05,
+      "loss": 0.1944,
+      "step": 592
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.9930503731343284e-05,
+      "loss": 0.0786,
+      "step": 596
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.993003731343284e-05,
+      "loss": 0.1258,
+      "step": 600
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.9929570895522394e-05,
+      "loss": 0.1359,
+      "step": 604
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.992910447761194e-05,
+      "loss": 0.1297,
+      "step": 608
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.99286380597015e-05,
+      "loss": 0.2223,
+      "step": 612
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.992817164179105e-05,
+      "loss": 0.1363,
+      "step": 616
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.99277052238806e-05,
+      "loss": 0.1362,
+      "step": 620
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.992723880597015e-05,
+      "loss": 0.1977,
+      "step": 624
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.99267723880597e-05,
+      "loss": 0.1176,
+      "step": 628
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.992630597014926e-05,
+      "loss": 0.1254,
+      "step": 632
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.9925839552238806e-05,
+      "loss": 0.1825,
+      "step": 636
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.992537313432836e-05,
+      "loss": 0.1363,
+      "step": 640
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.9924906716417916e-05,
+      "loss": 0.1388,
+      "step": 644
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.9924440298507464e-05,
+      "loss": 0.1409,
+      "step": 648
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.992397388059702e-05,
+      "loss": 0.1481,
+      "step": 652
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.992350746268657e-05,
+      "loss": 0.183,
+      "step": 656
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.992304104477612e-05,
+      "loss": 0.1344,
+      "step": 660
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.9922574626865677e-05,
+      "loss": 0.1149,
+      "step": 664
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.9922108208955225e-05,
+      "loss": 0.1584,
+      "step": 668
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.992164179104478e-05,
+      "loss": 0.1784,
+      "step": 672
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.9921175373134335e-05,
+      "loss": 0.066,
+      "step": 676
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.992070895522388e-05,
+      "loss": 0.1293,
+      "step": 680
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.992024253731343e-05,
+      "loss": 0.1601,
+      "step": 684
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.9919776119402986e-05,
+      "loss": 0.1549,
+      "step": 688
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.991930970149254e-05,
+      "loss": 0.102,
+      "step": 692
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.991884328358209e-05,
+      "loss": 0.1538,
+      "step": 696
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.9918376865671644e-05,
+      "loss": 0.1375,
+      "step": 700
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.99179104477612e-05,
+      "loss": 0.1162,
+      "step": 704
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.991744402985075e-05,
+      "loss": 0.1154,
+      "step": 708
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.99169776119403e-05,
+      "loss": 0.2463,
+      "step": 712
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.991651119402985e-05,
+      "loss": 0.1519,
+      "step": 716
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.9916044776119405e-05,
+      "loss": 0.0805,
+      "step": 720
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.991557835820896e-05,
+      "loss": 0.1146,
+      "step": 724
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.991511194029851e-05,
+      "loss": 0.1365,
+      "step": 728
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.991464552238806e-05,
+      "loss": 0.1944,
+      "step": 732
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.991417910447762e-05,
+      "loss": 0.1529,
+      "step": 736
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.9913712686567166e-05,
+      "loss": 0.1301,
+      "step": 740
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.991324626865672e-05,
+      "loss": 0.1387,
+      "step": 744
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.991277985074627e-05,
+      "loss": 0.1241,
+      "step": 748
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.9912313432835824e-05,
+      "loss": 0.1165,
+      "step": 752
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.991184701492538e-05,
+      "loss": 0.1206,
+      "step": 756
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.991138059701493e-05,
+      "loss": 0.1771,
+      "step": 760
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.991091417910448e-05,
+      "loss": 0.1229,
+      "step": 764
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.9910447761194036e-05,
+      "loss": 0.1462,
+      "step": 768
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.9909981343283585e-05,
+      "loss": 0.119,
+      "step": 772
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.990951492537313e-05,
+      "loss": 0.1318,
+      "step": 776
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.9909048507462694e-05,
+      "loss": 0.1514,
+      "step": 780
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.990858208955224e-05,
+      "loss": 0.1043,
+      "step": 784
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.990811567164179e-05,
+      "loss": 0.1466,
+      "step": 788
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.9907649253731346e-05,
+      "loss": 0.1052,
+      "step": 792
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.99071828358209e-05,
+      "loss": 0.1098,
+      "step": 796
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.990671641791045e-05,
+      "loss": 0.2,
+      "step": 800
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.9906250000000004e-05,
+      "loss": 0.1416,
+      "step": 804
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.990578358208955e-05,
+      "loss": 0.1137,
+      "step": 808
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.9905317164179107e-05,
+      "loss": 0.14,
+      "step": 812
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.990485074626866e-05,
+      "loss": 0.1023,
+      "step": 816
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.990438432835821e-05,
+      "loss": 0.1276,
+      "step": 820
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.9903917910447764e-05,
+      "loss": 0.1445,
+      "step": 824
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.990345149253732e-05,
+      "loss": 0.1165,
+      "step": 828
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.990298507462687e-05,
+      "loss": 0.1369,
+      "step": 832
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.9902518656716416e-05,
+      "loss": 0.1656,
+      "step": 836
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.990205223880598e-05,
+      "loss": 0.1383,
+      "step": 840
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.9901585820895525e-05,
+      "loss": 0.1065,
+      "step": 844
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.9901119402985074e-05,
+      "loss": 0.1279,
+      "step": 848
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.990065298507463e-05,
+      "loss": 0.111,
+      "step": 852
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.9900186567164183e-05,
+      "loss": 0.1729,
+      "step": 856
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.989972014925373e-05,
+      "loss": 0.1522,
+      "step": 860
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.9899253731343286e-05,
+      "loss": 0.1157,
+      "step": 864
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.9898787313432835e-05,
+      "loss": 0.0954,
+      "step": 868
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.989832089552239e-05,
+      "loss": 0.1079,
+      "step": 872
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.9897854477611944e-05,
+      "loss": 0.094,
+      "step": 876
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.989738805970149e-05,
+      "loss": 0.1084,
+      "step": 880
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.989692164179105e-05,
+      "loss": 0.1579,
+      "step": 884
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.98964552238806e-05,
+      "loss": 0.1245,
+      "step": 888
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.989598880597015e-05,
+      "loss": 0.1674,
+      "step": 892
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.98955223880597e-05,
+      "loss": 0.0896,
+      "step": 896
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.989505597014926e-05,
+      "loss": 0.1532,
+      "step": 900
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.989458955223881e-05,
+      "loss": 0.1221,
+      "step": 904
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 4.989412313432836e-05,
+      "loss": 0.1003,
+      "step": 908
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 4.989365671641791e-05,
+      "loss": 0.149,
+      "step": 912
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 4.9893190298507466e-05,
+      "loss": 0.1192,
+      "step": 916
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.989272388059702e-05,
+      "loss": 0.122,
+      "step": 920
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.989225746268657e-05,
+      "loss": 0.0945,
+      "step": 924
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.989179104477612e-05,
+      "loss": 0.1248,
+      "step": 928
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.989132462686568e-05,
+      "loss": 0.0974,
+      "step": 932
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.989085820895523e-05,
+      "loss": 0.0956,
+      "step": 936
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.9890391791044775e-05,
+      "loss": 0.1214,
+      "step": 940
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.988992537313433e-05,
+      "loss": 0.1591,
+      "step": 944
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.9889458955223885e-05,
+      "loss": 0.1303,
+      "step": 948
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 4.9888992537313433e-05,
+      "loss": 0.1071,
+      "step": 952
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 4.988852611940299e-05,
+      "loss": 0.1266,
+      "step": 956
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.988805970149254e-05,
+      "loss": 0.1142,
+      "step": 960
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.988759328358209e-05,
+      "loss": 0.0919,
+      "step": 964
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.9887126865671646e-05,
+      "loss": 0.1704,
+      "step": 968
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.9886660447761194e-05,
+      "loss": 0.1074,
+      "step": 972
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.988619402985075e-05,
+      "loss": 0.1104,
+      "step": 976
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.9885727611940304e-05,
+      "loss": 0.1298,
+      "step": 980
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.988526119402985e-05,
+      "loss": 0.1069,
+      "step": 984
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.98847947761194e-05,
+      "loss": 0.0986,
+      "step": 988
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.988432835820896e-05,
+      "loss": 0.1057,
+      "step": 992
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.988386194029851e-05,
+      "loss": 0.1513,
+      "step": 996
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.988339552238806e-05,
+      "loss": 0.16,
+      "step": 1000
+    },
+    {
+      "epoch": 0.93,
+      "eval_exact_match": 0.597678916827853,
+      "eval_exec": 0.6266924564796905,
+      "eval_loss": 0.14075149595737457,
+      "eval_runtime": 1596.4093,
+      "eval_samples_per_second": 0.648,
+      "step": 1000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.988292910447761e-05,
+      "loss": 0.1095,
+      "step": 1004
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.988246268656717e-05,
+      "loss": 0.0819,
+      "step": 1008
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.9881996268656716e-05,
+      "loss": 0.1231,
+      "step": 1012
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 4.988152985074627e-05,
+      "loss": 0.0878,
+      "step": 1016
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 4.9881063432835826e-05,
+      "loss": 0.1158,
+      "step": 1020
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.9880597014925374e-05,
+      "loss": 0.1312,
+      "step": 1024
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.988013059701493e-05,
+      "loss": 0.1238,
+      "step": 1028
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.987966417910448e-05,
+      "loss": 0.1167,
+      "step": 1032
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.987919776119403e-05,
+      "loss": 0.0814,
+      "step": 1036
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.987873134328359e-05,
+      "loss": 0.0817,
+      "step": 1040
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.9878264925373135e-05,
+      "loss": 0.1221,
+      "step": 1044
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.9877798507462683e-05,
+      "loss": 0.131,
+      "step": 1048
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.9877332089552245e-05,
+      "loss": 0.1129,
+      "step": 1052
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.987686567164179e-05,
+      "loss": 0.1051,
+      "step": 1056
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.987639925373134e-05,
+      "loss": 0.0943,
+      "step": 1060
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.9875932835820896e-05,
+      "loss": 0.0652,
+      "step": 1064
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.987546641791045e-05,
+      "loss": 0.109,
+      "step": 1068
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.9875000000000006e-05,
+      "loss": 0.1217,
+      "step": 1072
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.9874533582089554e-05,
+      "loss": 0.0951,
+      "step": 1076
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.98740671641791e-05,
+      "loss": 0.0788,
+      "step": 1080
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.9873600746268664e-05,
+      "loss": 0.0627,
+      "step": 1084
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.987313432835821e-05,
+      "loss": 0.1035,
+      "step": 1088
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.987266791044776e-05,
+      "loss": 0.0611,
+      "step": 1092
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.9872201492537315e-05,
+      "loss": 0.0593,
+      "step": 1096
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.987173507462687e-05,
+      "loss": 0.0997,
+      "step": 1100
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.987126865671642e-05,
+      "loss": 0.0991,
+      "step": 1104
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.987080223880597e-05,
+      "loss": 0.0917,
+      "step": 1108
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.987033582089553e-05,
+      "loss": 0.0519,
+      "step": 1112
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.9869869402985076e-05,
+      "loss": 0.0923,
+      "step": 1116
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.986940298507463e-05,
+      "loss": 0.071,
+      "step": 1120
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.986893656716418e-05,
+      "loss": 0.0654,
+      "step": 1124
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.9868470149253734e-05,
+      "loss": 0.0823,
+      "step": 1128
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.986800373134329e-05,
+      "loss": 0.0674,
+      "step": 1132
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.986753731343284e-05,
+      "loss": 0.0838,
+      "step": 1136
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.9867070895522385e-05,
+      "loss": 0.115,
+      "step": 1140
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.986660447761195e-05,
+      "loss": 0.0886,
+      "step": 1144
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.9866138059701495e-05,
+      "loss": 0.086,
+      "step": 1148
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.986567164179104e-05,
+      "loss": 0.0902,
+      "step": 1152
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.98652052238806e-05,
+      "loss": 0.0724,
+      "step": 1156
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.986473880597015e-05,
+      "loss": 0.0832,
+      "step": 1160
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.98642723880597e-05,
+      "loss": 0.0507,
+      "step": 1164
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.9863805970149256e-05,
+      "loss": 0.0754,
+      "step": 1168
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.986333955223881e-05,
+      "loss": 0.0738,
+      "step": 1172
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.986287313432836e-05,
+      "loss": 0.0834,
+      "step": 1176
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.9862406716417914e-05,
+      "loss": 0.0867,
+      "step": 1180
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.986194029850746e-05,
+      "loss": 0.0967,
+      "step": 1184
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.986147388059702e-05,
+      "loss": 0.0759,
+      "step": 1188
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.986100746268657e-05,
+      "loss": 0.0858,
+      "step": 1192
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.986054104477612e-05,
+      "loss": 0.143,
+      "step": 1196
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.986007462686567e-05,
+      "loss": 0.0648,
+      "step": 1200
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.985960820895523e-05,
+      "loss": 0.0706,
+      "step": 1204
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.985914179104478e-05,
+      "loss": 0.0687,
+      "step": 1208
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.9858675373134326e-05,
+      "loss": 0.098,
+      "step": 1212
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.985820895522388e-05,
+      "loss": 0.0809,
+      "step": 1216
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.9857742537313436e-05,
+      "loss": 0.0634,
+      "step": 1220
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.9857276119402984e-05,
+      "loss": 0.0899,
+      "step": 1224
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.985680970149254e-05,
+      "loss": 0.0905,
+      "step": 1228
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.9856343283582094e-05,
+      "loss": 0.0669,
+      "step": 1232
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.985587686567165e-05,
+      "loss": 0.0898,
+      "step": 1236
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.98554104477612e-05,
+      "loss": 0.1037,
+      "step": 1240
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.9854944029850745e-05,
+      "loss": 0.1286,
+      "step": 1244
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.98544776119403e-05,
+      "loss": 0.0882,
+      "step": 1248
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.9854011194029855e-05,
+      "loss": 0.0847,
+      "step": 1252
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.98535447761194e-05,
+      "loss": 0.0579,
+      "step": 1256
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.985307835820896e-05,
+      "loss": 0.0864,
+      "step": 1260
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.985261194029851e-05,
+      "loss": 0.0831,
+      "step": 1264
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.985214552238806e-05,
+      "loss": 0.1051,
+      "step": 1268
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.9851679104477616e-05,
+      "loss": 0.0406,
+      "step": 1272
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.9851212686567164e-05,
+      "loss": 0.1047,
+      "step": 1276
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.985074626865672e-05,
+      "loss": 0.0855,
+      "step": 1280
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.9850279850746274e-05,
+      "loss": 0.1054,
+      "step": 1284
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.984981343283582e-05,
+      "loss": 0.102,
+      "step": 1288
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 4.984934701492538e-05,
+      "loss": 0.0599,
+      "step": 1292
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 4.984888059701493e-05,
+      "loss": 0.054,
+      "step": 1296
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 4.984841417910448e-05,
+      "loss": 0.0712,
+      "step": 1300
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 4.984794776119403e-05,
+      "loss": 0.0768,
+      "step": 1304
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 4.984748134328358e-05,
+      "loss": 0.0646,
+      "step": 1308
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 4.984701492537314e-05,
+      "loss": 0.0775,
+      "step": 1312
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 4.9846548507462686e-05,
+      "loss": 0.0643,
+      "step": 1316
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 4.984608208955224e-05,
+      "loss": 0.0972,
+      "step": 1320
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 4.9845615671641796e-05,
+      "loss": 0.0801,
+      "step": 1324
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 4.9845149253731344e-05,
+      "loss": 0.114,
+      "step": 1328
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 4.98446828358209e-05,
+      "loss": 0.0623,
+      "step": 1332
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 4.984421641791045e-05,
+      "loss": 0.07,
+      "step": 1336
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 4.984375e-05,
+      "loss": 0.0768,
+      "step": 1340
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 4.984328358208956e-05,
+      "loss": 0.0684,
+      "step": 1344
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 4.9842817164179105e-05,
+      "loss": 0.08,
+      "step": 1348
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 4.984235074626866e-05,
+      "loss": 0.062,
+      "step": 1352
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 4.9841884328358215e-05,
+      "loss": 0.1145,
+      "step": 1356
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 4.984141791044776e-05,
+      "loss": 0.088,
+      "step": 1360
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 4.984095149253731e-05,
+      "loss": 0.0713,
+      "step": 1364
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 4.9840485074626866e-05,
+      "loss": 0.0372,
+      "step": 1368
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 4.984001865671642e-05,
+      "loss": 0.0618,
+      "step": 1372
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 4.983955223880597e-05,
+      "loss": 0.119,
+      "step": 1376
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 4.9839085820895524e-05,
+      "loss": 0.0979,
+      "step": 1380
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 4.983861940298508e-05,
+      "loss": 0.056,
+      "step": 1384
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 4.983815298507463e-05,
+      "loss": 0.0805,
+      "step": 1388
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 4.983768656716418e-05,
+      "loss": 0.0768,
+      "step": 1392
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 4.983722014925373e-05,
+      "loss": 0.0488,
+      "step": 1396
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 4.983675373134329e-05,
+      "loss": 0.0914,
+      "step": 1400
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 4.983628731343284e-05,
+      "loss": 0.0765,
+      "step": 1404
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 4.983582089552239e-05,
+      "loss": 0.0862,
+      "step": 1408
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 4.983535447761194e-05,
+      "loss": 0.0692,
+      "step": 1412
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 4.98348880597015e-05,
+      "loss": 0.0611,
+      "step": 1416
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 4.9834421641791046e-05,
+      "loss": 0.0947,
+      "step": 1420
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 4.98339552238806e-05,
+      "loss": 0.0615,
+      "step": 1424
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 4.983348880597015e-05,
+      "loss": 0.0774,
+      "step": 1428
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 4.9833022388059704e-05,
+      "loss": 0.0725,
+      "step": 1432
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 4.983255597014926e-05,
+      "loss": 0.0784,
+      "step": 1436
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 4.983208955223881e-05,
+      "loss": 0.061,
+      "step": 1440
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.983162313432836e-05,
+      "loss": 0.1046,
+      "step": 1444
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.983115671641792e-05,
+      "loss": 0.0768,
+      "step": 1448
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.9830690298507465e-05,
+      "loss": 0.0608,
+      "step": 1452
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.983022388059701e-05,
+      "loss": 0.0959,
+      "step": 1456
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9829757462686575e-05,
+      "loss": 0.0602,
+      "step": 1460
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.982929104477612e-05,
+      "loss": 0.0655,
+      "step": 1464
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.982882462686567e-05,
+      "loss": 0.1009,
+      "step": 1468
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.9828358208955226e-05,
+      "loss": 0.0563,
+      "step": 1472
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.982789179104478e-05,
+      "loss": 0.0763,
+      "step": 1476
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.982742537313433e-05,
+      "loss": 0.0651,
+      "step": 1480
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.9826958955223884e-05,
+      "loss": 0.0346,
+      "step": 1484
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.982649253731343e-05,
+      "loss": 0.0437,
+      "step": 1488
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.982602611940299e-05,
+      "loss": 0.0944,
+      "step": 1492
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.982555970149254e-05,
+      "loss": 0.0718,
+      "step": 1496
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.982509328358209e-05,
+      "loss": 0.098,
+      "step": 1500
+    },
+    {
+      "epoch": 1.4,
+      "eval_exact_match": 0.660541586073501,
+      "eval_exec": 0.6818181818181818,
+      "eval_loss": 0.1437511444091797,
+      "eval_runtime": 1123.3017,
+      "eval_samples_per_second": 0.921,
+      "step": 1500
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.9824626865671645e-05,
+      "loss": 0.087,
+      "step": 1504
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.98241604477612e-05,
+      "loss": 0.0929,
+      "step": 1508
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.982369402985075e-05,
+      "loss": 0.1199,
+      "step": 1512
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.9823227611940296e-05,
+      "loss": 0.0896,
+      "step": 1516
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.982276119402986e-05,
+      "loss": 0.0901,
+      "step": 1520
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.9822294776119406e-05,
+      "loss": 0.0724,
+      "step": 1524
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.9821828358208954e-05,
+      "loss": 0.067,
+      "step": 1528
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.982136194029851e-05,
+      "loss": 0.0706,
+      "step": 1532
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.9820895522388064e-05,
+      "loss": 0.0507,
+      "step": 1536
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 4.982042910447761e-05,
+      "loss": 0.0632,
+      "step": 1540
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 4.981996268656717e-05,
+      "loss": 0.0833,
+      "step": 1544
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 4.9819496268656715e-05,
+      "loss": 0.1091,
+      "step": 1548
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 4.981902985074627e-05,
+      "loss": 0.0629,
+      "step": 1552
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 4.9818563432835825e-05,
+      "loss": 0.1016,
+      "step": 1556
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 4.981809701492537e-05,
+      "loss": 0.0879,
+      "step": 1560
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 4.981763059701493e-05,
+      "loss": 0.0938,
+      "step": 1564
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 4.981716417910448e-05,
+      "loss": 0.0596,
+      "step": 1568
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 4.981669776119403e-05,
+      "loss": 0.1145,
+      "step": 1572
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 4.9816231343283586e-05,
+      "loss": 0.0723,
+      "step": 1576
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 4.981576492537314e-05,
+      "loss": 0.0335,
+      "step": 1580
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 4.981529850746269e-05,
+      "loss": 0.0567,
+      "step": 1584
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 4.9814832089552244e-05,
+      "loss": 0.0653,
+      "step": 1588
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.981436567164179e-05,
+      "loss": 0.0643,
+      "step": 1592
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.981389925373135e-05,
+      "loss": 0.0719,
+      "step": 1596
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.98134328358209e-05,
+      "loss": 0.0477,
+      "step": 1600
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 4.981296641791045e-05,
+      "loss": 0.0702,
+      "step": 1604
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 4.98125e-05,
+      "loss": 0.0685,
+      "step": 1608
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 4.981203358208956e-05,
+      "loss": 0.0908,
+      "step": 1612
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 4.981156716417911e-05,
+      "loss": 0.1234,
+      "step": 1616
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 4.9811100746268656e-05,
+      "loss": 0.0749,
+      "step": 1620
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 4.981063432835821e-05,
+      "loss": 0.0627,
+      "step": 1624
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9810167910447766e-05,
+      "loss": 0.0642,
+      "step": 1628
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9809701492537314e-05,
+      "loss": 0.0632,
+      "step": 1632
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.980923507462687e-05,
+      "loss": 0.0874,
+      "step": 1636
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.9808768656716424e-05,
+      "loss": 0.1122,
+      "step": 1640
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.980830223880597e-05,
+      "loss": 0.0583,
+      "step": 1644
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.9807835820895527e-05,
+      "loss": 0.0584,
+      "step": 1648
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.9807369402985075e-05,
+      "loss": 0.1076,
+      "step": 1652
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.980690298507463e-05,
+      "loss": 0.0854,
+      "step": 1656
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.9806436567164185e-05,
+      "loss": 0.0862,
+      "step": 1660
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.980597014925373e-05,
+      "loss": 0.0705,
+      "step": 1664
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.980550373134328e-05,
+      "loss": 0.0803,
+      "step": 1668
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.980503731343284e-05,
+      "loss": 0.0387,
+      "step": 1672
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.980457089552239e-05,
+      "loss": 0.0556,
+      "step": 1676
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.980410447761194e-05,
+      "loss": 0.0559,
+      "step": 1680
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.9803638059701494e-05,
+      "loss": 0.0766,
+      "step": 1684
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.980317164179105e-05,
+      "loss": 0.0778,
+      "step": 1688
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.98027052238806e-05,
+      "loss": 0.0976,
+      "step": 1692
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.980223880597015e-05,
+      "loss": 0.0818,
+      "step": 1696
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.9801772388059707e-05,
+      "loss": 0.0597,
+      "step": 1700
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.9801305970149255e-05,
+      "loss": 0.0626,
+      "step": 1704
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.980083955223881e-05,
+      "loss": 0.117,
+      "step": 1708
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.980037313432836e-05,
+      "loss": 0.0725,
+      "step": 1712
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.979990671641791e-05,
+      "loss": 0.0788,
+      "step": 1716
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.979944029850747e-05,
+      "loss": 0.0874,
+      "step": 1720
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.9798973880597016e-05,
+      "loss": 0.0781,
+      "step": 1724
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.979850746268657e-05,
+      "loss": 0.0971,
+      "step": 1728
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 4.9798041044776125e-05,
+      "loss": 0.1099,
+      "step": 1732
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 4.9797574626865674e-05,
+      "loss": 0.0748,
+      "step": 1736
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 4.979710820895523e-05,
+      "loss": 0.073,
+      "step": 1740
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 4.9796641791044777e-05,
+      "loss": 0.0584,
+      "step": 1744
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 4.979617537313433e-05,
+      "loss": 0.0749,
+      "step": 1748
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 4.9795708955223886e-05,
+      "loss": 0.0712,
+      "step": 1752
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 4.9795242537313435e-05,
+      "loss": 0.082,
+      "step": 1756
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 4.979477611940298e-05,
+      "loss": 0.0701,
+      "step": 1760
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 4.9794309701492544e-05,
+      "loss": 0.0424,
+      "step": 1764
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 4.979384328358209e-05,
+      "loss": 0.1098,
+      "step": 1768
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 4.979337686567164e-05,
+      "loss": 0.0737,
+      "step": 1772
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 4.9792910447761196e-05,
+      "loss": 0.0571,
+      "step": 1776
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 4.979244402985075e-05,
+      "loss": 0.0605,
+      "step": 1780
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 4.97919776119403e-05,
+      "loss": 0.0686,
+      "step": 1784
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.9791511194029853e-05,
+      "loss": 0.0497,
+      "step": 1788
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.979104477611941e-05,
+      "loss": 0.0528,
+      "step": 1792
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.9790578358208957e-05,
+      "loss": 0.0618,
+      "step": 1796
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.979011194029851e-05,
+      "loss": 0.0575,
+      "step": 1800
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.978964552238806e-05,
+      "loss": 0.0448,
+      "step": 1804
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.9789179104477614e-05,
+      "loss": 0.0505,
+      "step": 1808
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.978871268656717e-05,
+      "loss": 0.0813,
+      "step": 1812
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.978824626865672e-05,
+      "loss": 0.0487,
+      "step": 1816
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.9787779850746266e-05,
+      "loss": 0.0912,
+      "step": 1820
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.978731343283583e-05,
+      "loss": 0.0839,
+      "step": 1824
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.9786847014925375e-05,
+      "loss": 0.0734,
+      "step": 1828
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.9786380597014924e-05,
+      "loss": 0.0664,
+      "step": 1832
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.978591417910448e-05,
+      "loss": 0.0435,
+      "step": 1836
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.9785447761194033e-05,
+      "loss": 0.0893,
+      "step": 1840
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.978498134328358e-05,
+      "loss": 0.0807,
+      "step": 1844
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.9784514925373136e-05,
+      "loss": 0.0587,
+      "step": 1848
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.978404850746269e-05,
+      "loss": 0.072,
+      "step": 1852
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.978358208955224e-05,
+      "loss": 0.0707,
+      "step": 1856
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.9783115671641794e-05,
+      "loss": 0.0658,
+      "step": 1860
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 4.978264925373134e-05,
+      "loss": 0.0782,
+      "step": 1864
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 4.97821828358209e-05,
+      "loss": 0.124,
+      "step": 1868
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 4.978171641791045e-05,
+      "loss": 0.0764,
+      "step": 1872
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 4.978125e-05,
+      "loss": 0.0901,
+      "step": 1876
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 4.9780783582089555e-05,
+      "loss": 0.0531,
+      "step": 1880
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 4.978031716417911e-05,
+      "loss": 0.0964,
+      "step": 1884
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 4.977985074626866e-05,
+      "loss": 0.0847,
+      "step": 1888
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 4.977938432835821e-05,
+      "loss": 0.0643,
+      "step": 1892
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 4.977891791044776e-05,
+      "loss": 0.0743,
+      "step": 1896
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 4.9778451492537316e-05,
+      "loss": 0.0533,
+      "step": 1900
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 4.977798507462687e-05,
+      "loss": 0.0736,
+      "step": 1904
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 4.977751865671642e-05,
+      "loss": 0.0535,
+      "step": 1908
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 4.9777052238805974e-05,
+      "loss": 0.0636,
+      "step": 1912
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 4.977658582089553e-05,
+      "loss": 0.0381,
+      "step": 1916
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 4.977611940298508e-05,
+      "loss": 0.0298,
+      "step": 1920
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 4.9775652985074625e-05,
+      "loss": 0.0838,
+      "step": 1924
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.977518656716418e-05,
+      "loss": 0.1048,
+      "step": 1928
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.9774720149253735e-05,
+      "loss": 0.031,
+      "step": 1932
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.9774253731343283e-05,
+      "loss": 0.0617,
+      "step": 1936
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.977378731343284e-05,
+      "loss": 0.1009,
+      "step": 1940
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.977332089552239e-05,
+      "loss": 0.057,
+      "step": 1944
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.977285447761194e-05,
+      "loss": 0.0519,
+      "step": 1948
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.9772388059701496e-05,
+      "loss": 0.061,
+      "step": 1952
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.9771921641791044e-05,
+      "loss": 0.1017,
+      "step": 1956
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 4.97714552238806e-05,
+      "loss": 0.0552,
+      "step": 1960
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 4.9770988805970154e-05,
+      "loss": 0.0943,
+      "step": 1964
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 4.97705223880597e-05,
+      "loss": 0.0545,
+      "step": 1968
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 4.977005597014926e-05,
+      "loss": 0.0931,
+      "step": 1972
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 4.976958955223881e-05,
+      "loss": 0.0426,
+      "step": 1976
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 4.976912313432836e-05,
+      "loss": 0.0939,
+      "step": 1980
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 4.976865671641791e-05,
+      "loss": 0.086,
+      "step": 1984
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 4.976819029850746e-05,
+      "loss": 0.0391,
+      "step": 1988
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 4.976772388059702e-05,
+      "loss": 0.0389,
+      "step": 1992
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 4.9767257462686566e-05,
+      "loss": 0.0725,
+      "step": 1996
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 4.976679104477612e-05,
+      "loss": 0.074,
+      "step": 2000
+    },
+    {
+      "epoch": 1.87,
+      "eval_exact_match": 0.6721470019342359,
+      "eval_exec": 0.7147001934235977,
+      "eval_loss": 0.14614403247833252,
+      "eval_runtime": 1069.1464,
+      "eval_samples_per_second": 0.967,
+      "step": 2000
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 4.9766324626865676e-05,
+      "loss": 0.0657,
+      "step": 2004
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 4.9765858208955224e-05,
+      "loss": 0.0452,
+      "step": 2008
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 4.976539179104478e-05,
+      "loss": 0.0343,
+      "step": 2012
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 4.976492537313433e-05,
+      "loss": 0.0896,
+      "step": 2016
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 4.976445895522388e-05,
+      "loss": 0.043,
+      "step": 2020
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 4.976399253731344e-05,
+      "loss": 0.0543,
+      "step": 2024
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 4.9763526119402985e-05,
+      "loss": 0.0976,
+      "step": 2028
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.976305970149254e-05,
+      "loss": 0.0623,
+      "step": 2032
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.9762593283582095e-05,
+      "loss": 0.0817,
+      "step": 2036
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.976212686567164e-05,
+      "loss": 0.0442,
+      "step": 2040
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.97616604477612e-05,
+      "loss": 0.0688,
+      "step": 2044
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.9761194029850746e-05,
+      "loss": 0.0729,
+      "step": 2048
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.97607276119403e-05,
+      "loss": 0.055,
+      "step": 2052
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.9760261194029856e-05,
+      "loss": 0.055,
+      "step": 2056
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.9759794776119404e-05,
+      "loss": 0.0533,
+      "step": 2060
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 4.975932835820896e-05,
+      "loss": 0.0656,
+      "step": 2064
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 4.9758861940298514e-05,
+      "loss": 0.0534,
+      "step": 2068
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 4.975839552238806e-05,
+      "loss": 0.0565,
+      "step": 2072
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.975792910447761e-05,
+      "loss": 0.0608,
+      "step": 2076
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.975746268656717e-05,
+      "loss": 0.0383,
+      "step": 2080
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.975699626865672e-05,
+      "loss": 0.1284,
+      "step": 2084
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.975652985074627e-05,
+      "loss": 0.0556,
+      "step": 2088
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.975606343283582e-05,
+      "loss": 0.0666,
+      "step": 2092
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.975559701492538e-05,
+      "loss": 0.0705,
+      "step": 2096
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.9755130597014926e-05,
+      "loss": 0.0421,
+      "step": 2100
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.975466417910448e-05,
+      "loss": 0.0666,
+      "step": 2104
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.975419776119403e-05,
+      "loss": 0.0333,
+      "step": 2108
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.9753731343283584e-05,
+      "loss": 0.05,
+      "step": 2112
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.975326492537314e-05,
+      "loss": 0.059,
+      "step": 2116
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.975279850746269e-05,
+      "loss": 0.0567,
+      "step": 2120
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.975233208955224e-05,
+      "loss": 0.0447,
+      "step": 2124
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.97518656716418e-05,
+      "loss": 0.0653,
+      "step": 2128
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.9751399253731345e-05,
+      "loss": 0.0653,
+      "step": 2132
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.975093283582089e-05,
+      "loss": 0.0908,
+      "step": 2136
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.9750466417910455e-05,
+      "loss": 0.086,
+      "step": 2140
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.975e-05,
+      "loss": 0.0909,
+      "step": 2144
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.974953358208955e-05,
+      "loss": 0.0282,
+      "step": 2148
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.9749067164179106e-05,
+      "loss": 0.0412,
+      "step": 2152
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.974860074626866e-05,
+      "loss": 0.0324,
+      "step": 2156
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.974813432835821e-05,
+      "loss": 0.0381,
+      "step": 2160
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.9747667910447764e-05,
+      "loss": 0.0609,
+      "step": 2164
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.974720149253731e-05,
+      "loss": 0.0252,
+      "step": 2168
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.974673507462687e-05,
+      "loss": 0.0282,
+      "step": 2172
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.974626865671642e-05,
+      "loss": 0.0329,
+      "step": 2176
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.974580223880597e-05,
+      "loss": 0.0479,
+      "step": 2180
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.9745335820895525e-05,
+      "loss": 0.0323,
+      "step": 2184
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.974486940298508e-05,
+      "loss": 0.0309,
+      "step": 2188
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.974440298507463e-05,
+      "loss": 0.0482,
+      "step": 2192
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.9743936567164176e-05,
+      "loss": 0.0378,
+      "step": 2196
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.974347014925374e-05,
+      "loss": 0.057,
+      "step": 2200
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.9743003731343286e-05,
+      "loss": 0.0485,
+      "step": 2204
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.974253731343284e-05,
+      "loss": 0.0409,
+      "step": 2208
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.974207089552239e-05,
+      "loss": 0.0497,
+      "step": 2212
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.9741604477611944e-05,
+      "loss": 0.0286,
+      "step": 2216
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.97411380597015e-05,
+      "loss": 0.0323,
+      "step": 2220
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.974067164179105e-05,
+      "loss": 0.0414,
+      "step": 2224
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.9740205223880595e-05,
+      "loss": 0.0469,
+      "step": 2228
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.973973880597016e-05,
+      "loss": 0.0267,
+      "step": 2232
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.9739272388059705e-05,
+      "loss": 0.0438,
+      "step": 2236
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.973880597014925e-05,
+      "loss": 0.0423,
+      "step": 2240
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.973833955223881e-05,
+      "loss": 0.0424,
+      "step": 2244
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.973787313432836e-05,
+      "loss": 0.0562,
+      "step": 2248
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.973740671641791e-05,
+      "loss": 0.0538,
+      "step": 2252
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.9736940298507466e-05,
+      "loss": 0.0554,
+      "step": 2256
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.973647388059702e-05,
+      "loss": 0.0595,
+      "step": 2260
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.973600746268657e-05,
+      "loss": 0.0652,
+      "step": 2264
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.9735541044776124e-05,
+      "loss": 0.05,
+      "step": 2268
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.973507462686567e-05,
+      "loss": 0.0699,
+      "step": 2272
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.973460820895523e-05,
+      "loss": 0.0484,
+      "step": 2276
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.973414179104478e-05,
+      "loss": 0.0502,
+      "step": 2280
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.973367537313433e-05,
+      "loss": 0.0437,
+      "step": 2284
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.973320895522388e-05,
+      "loss": 0.072,
+      "step": 2288
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.973274253731344e-05,
+      "loss": 0.0548,
+      "step": 2292
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.973227611940299e-05,
+      "loss": 0.0734,
+      "step": 2296
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.9731809701492536e-05,
+      "loss": 0.0451,
+      "step": 2300
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.973134328358209e-05,
+      "loss": 0.0202,
+      "step": 2304
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.9730876865671646e-05,
+      "loss": 0.0281,
+      "step": 2308
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 4.9730410447761194e-05,
+      "loss": 0.0335,
+      "step": 2312
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 4.972994402985075e-05,
+      "loss": 0.042,
+      "step": 2316
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 4.9729477611940304e-05,
+      "loss": 0.0429,
+      "step": 2320
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 4.972901119402985e-05,
+      "loss": 0.0548,
+      "step": 2324
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 4.972854477611941e-05,
+      "loss": 0.0355,
+      "step": 2328
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 4.9728078358208955e-05,
+      "loss": 0.0399,
+      "step": 2332
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 4.972761194029851e-05,
+      "loss": 0.0273,
+      "step": 2336
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 4.9727145522388065e-05,
+      "loss": 0.0439,
+      "step": 2340
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 4.972667910447761e-05,
+      "loss": 0.0303,
+      "step": 2344
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 4.972621268656716e-05,
+      "loss": 0.0439,
+      "step": 2348
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 4.972574626865672e-05,
+      "loss": 0.0417,
+      "step": 2352
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 4.972527985074627e-05,
+      "loss": 0.0286,
+      "step": 2356
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 4.972481343283582e-05,
+      "loss": 0.0292,
+      "step": 2360
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 4.9724347014925374e-05,
+      "loss": 0.0515,
+      "step": 2364
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 4.972388059701493e-05,
+      "loss": 0.0473,
+      "step": 2368
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 4.9723414179104484e-05,
+      "loss": 0.0406,
+      "step": 2372
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 4.972294776119403e-05,
+      "loss": 0.0629,
+      "step": 2376
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 4.972248134328359e-05,
+      "loss": 0.0504,
+      "step": 2380
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 4.972201492537314e-05,
+      "loss": 0.0267,
+      "step": 2384
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 4.972154850746269e-05,
+      "loss": 0.0302,
+      "step": 2388
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 4.972108208955224e-05,
+      "loss": 0.0223,
+      "step": 2392
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 4.972061567164179e-05,
+      "loss": 0.0636,
+      "step": 2396
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 4.972014925373135e-05,
+      "loss": 0.0347,
+      "step": 2400
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 4.9719682835820896e-05,
+      "loss": 0.0412,
+      "step": 2404
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.971921641791045e-05,
+      "loss": 0.0416,
+      "step": 2408
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.9718750000000006e-05,
+      "loss": 0.0442,
+      "step": 2412
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.9718283582089554e-05,
+      "loss": 0.0268,
+      "step": 2416
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.971781716417911e-05,
+      "loss": 0.043,
+      "step": 2420
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.971735074626866e-05,
+      "loss": 0.0315,
+      "step": 2424
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.971688432835821e-05,
+      "loss": 0.0238,
+      "step": 2428
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.971641791044777e-05,
+      "loss": 0.0405,
+      "step": 2432
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.9715951492537315e-05,
+      "loss": 0.0445,
+      "step": 2436
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.971548507462686e-05,
+      "loss": 0.0537,
+      "step": 2440
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.9715018656716425e-05,
+      "loss": 0.0656,
+      "step": 2444
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.971455223880597e-05,
+      "loss": 0.0381,
+      "step": 2448
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.971408582089552e-05,
+      "loss": 0.0455,
+      "step": 2452
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.9713619402985076e-05,
+      "loss": 0.0489,
+      "step": 2456
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.971315298507463e-05,
+      "loss": 0.0505,
+      "step": 2460
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.971268656716418e-05,
+      "loss": 0.0454,
+      "step": 2464
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.9712220149253734e-05,
+      "loss": 0.0409,
+      "step": 2468
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.971175373134329e-05,
+      "loss": 0.0429,
+      "step": 2472
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.971128731343284e-05,
+      "loss": 0.0383,
+      "step": 2476
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.971082089552239e-05,
+      "loss": 0.0349,
+      "step": 2480
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.971035447761194e-05,
+      "loss": 0.0369,
+      "step": 2484
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.9709888059701495e-05,
+      "loss": 0.052,
+      "step": 2488
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.970942164179105e-05,
+      "loss": 0.0564,
+      "step": 2492
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.97089552238806e-05,
+      "loss": 0.0585,
+      "step": 2496
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.9708488805970146e-05,
+      "loss": 0.0382,
+      "step": 2500
+    },
+    {
+      "epoch": 2.33,
+      "eval_exact_match": 0.6847195357833655,
+      "eval_exec": 0.7408123791102514,
+      "eval_loss": 0.16533711552619934,
+      "eval_runtime": 1046.3454,
+      "eval_samples_per_second": 0.988,
+      "step": 2500
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.970802238805971e-05,
+      "loss": 0.0608,
+      "step": 2504
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.9707555970149256e-05,
+      "loss": 0.0526,
+      "step": 2508
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.9707089552238804e-05,
+      "loss": 0.0666,
+      "step": 2512
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.970662313432836e-05,
+      "loss": 0.019,
+      "step": 2516
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.9706156716417914e-05,
+      "loss": 0.0384,
+      "step": 2520
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.970569029850746e-05,
+      "loss": 0.0553,
+      "step": 2524
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.970522388059702e-05,
+      "loss": 0.0517,
+      "step": 2528
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.970475746268657e-05,
+      "loss": 0.086,
+      "step": 2532
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.9704291044776127e-05,
+      "loss": 0.0685,
+      "step": 2536
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.9703824626865675e-05,
+      "loss": 0.0479,
+      "step": 2540
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.970335820895522e-05,
+      "loss": 0.0378,
+      "step": 2544
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.9702891791044784e-05,
+      "loss": 0.046,
+      "step": 2548
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.970242537313433e-05,
+      "loss": 0.056,
+      "step": 2552
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.970195895522388e-05,
+      "loss": 0.0294,
+      "step": 2556
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.9701492537313436e-05,
+      "loss": 0.0497,
+      "step": 2560
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.970102611940299e-05,
+      "loss": 0.039,
+      "step": 2564
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.970055970149254e-05,
+      "loss": 0.0395,
+      "step": 2568
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.9700093283582094e-05,
+      "loss": 0.0542,
+      "step": 2572
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.969962686567164e-05,
+      "loss": 0.0596,
+      "step": 2576
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 4.96991604477612e-05,
+      "loss": 0.0358,
+      "step": 2580
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 4.969869402985075e-05,
+      "loss": 0.0467,
+      "step": 2584
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 4.96982276119403e-05,
+      "loss": 0.0431,
+      "step": 2588
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 4.9697761194029855e-05,
+      "loss": 0.0355,
+      "step": 2592
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 4.969729477611941e-05,
+      "loss": 0.0251,
+      "step": 2596
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 4.969682835820896e-05,
+      "loss": 0.0417,
+      "step": 2600
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 4.9696361940298506e-05,
+      "loss": 0.063,
+      "step": 2604
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 4.969589552238806e-05,
+      "loss": 0.0454,
+      "step": 2608
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 4.9695429104477616e-05,
+      "loss": 0.0351,
+      "step": 2612
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 4.9694962686567164e-05,
+      "loss": 0.0543,
+      "step": 2616
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 4.969449626865672e-05,
+      "loss": 0.0422,
+      "step": 2620
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 4.9694029850746274e-05,
+      "loss": 0.0427,
+      "step": 2624
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 4.969356343283582e-05,
+      "loss": 0.0595,
+      "step": 2628
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 4.9693097014925377e-05,
+      "loss": 0.0437,
+      "step": 2632
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 4.9692630597014925e-05,
+      "loss": 0.0358,
+      "step": 2636
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 4.969216417910448e-05,
+      "loss": 0.041,
+      "step": 2640
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 4.9691697761194035e-05,
+      "loss": 0.0497,
+      "step": 2644
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 4.969123134328358e-05,
+      "loss": 0.0494,
+      "step": 2648
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 4.969076492537314e-05,
+      "loss": 0.0541,
+      "step": 2652
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 4.969029850746269e-05,
+      "loss": 0.0668,
+      "step": 2656
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 4.968983208955224e-05,
+      "loss": 0.0314,
+      "step": 2660
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 4.968936567164179e-05,
+      "loss": 0.0576,
+      "step": 2664
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 4.9688899253731344e-05,
+      "loss": 0.0514,
+      "step": 2668
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 4.96884328358209e-05,
+      "loss": 0.0398,
+      "step": 2672
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 4.968796641791045e-05,
+      "loss": 0.0481,
+      "step": 2676
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 4.96875e-05,
+      "loss": 0.0431,
+      "step": 2680
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 4.9687033582089556e-05,
+      "loss": 0.0419,
+      "step": 2684
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 4.9686567164179105e-05,
+      "loss": 0.0343,
+      "step": 2688
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 4.968610074626866e-05,
+      "loss": 0.0421,
+      "step": 2692
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 4.968563432835821e-05,
+      "loss": 0.0585,
+      "step": 2696
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 4.968516791044777e-05,
+      "loss": 0.0442,
+      "step": 2700
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 4.968470149253732e-05,
+      "loss": 0.0315,
+      "step": 2704
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 4.9684235074626866e-05,
+      "loss": 0.0563,
+      "step": 2708
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 4.968376865671642e-05,
+      "loss": 0.05,
+      "step": 2712
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 4.9683302238805975e-05,
+      "loss": 0.0356,
+      "step": 2716
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 4.9682835820895524e-05,
+      "loss": 0.0399,
+      "step": 2720
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 4.968236940298508e-05,
+      "loss": 0.035,
+      "step": 2724
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 4.9681902985074627e-05,
+      "loss": 0.0222,
+      "step": 2728
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 4.968143656716418e-05,
+      "loss": 0.0527,
+      "step": 2732
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 4.9680970149253736e-05,
+      "loss": 0.0451,
+      "step": 2736
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 4.9680503731343285e-05,
+      "loss": 0.0488,
+      "step": 2740
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 4.968003731343284e-05,
+      "loss": 0.0794,
+      "step": 2744
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 4.9679570895522394e-05,
+      "loss": 0.0571,
+      "step": 2748
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 4.967910447761194e-05,
+      "loss": 0.0391,
+      "step": 2752
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 4.967863805970149e-05,
+      "loss": 0.048,
+      "step": 2756
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 4.967817164179105e-05,
+      "loss": 0.051,
+      "step": 2760
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 4.96777052238806e-05,
+      "loss": 0.0402,
+      "step": 2764
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 4.967723880597015e-05,
+      "loss": 0.0392,
+      "step": 2768
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 4.9676772388059703e-05,
+      "loss": 0.0518,
+      "step": 2772
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 4.967630597014926e-05,
+      "loss": 0.0494,
+      "step": 2776
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 4.9675839552238807e-05,
+      "loss": 0.0324,
+      "step": 2780
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 4.967537313432836e-05,
+      "loss": 0.0357,
+      "step": 2784
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 4.967490671641791e-05,
+      "loss": 0.0458,
+      "step": 2788
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 4.9674440298507464e-05,
+      "loss": 0.0593,
+      "step": 2792
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 4.967397388059702e-05,
+      "loss": 0.0454,
+      "step": 2796
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 4.967350746268657e-05,
+      "loss": 0.0629,
+      "step": 2800
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 4.967304104477612e-05,
+      "loss": 0.0287,
+      "step": 2804
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 4.967257462686568e-05,
+      "loss": 0.0413,
+      "step": 2808
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 4.9672108208955225e-05,
+      "loss": 0.0231,
+      "step": 2812
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 4.9671641791044774e-05,
+      "loss": 0.0362,
+      "step": 2816
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 4.9671175373134335e-05,
+      "loss": 0.055,
+      "step": 2820
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 4.967070895522388e-05,
+      "loss": 0.0448,
+      "step": 2824
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 4.967024253731343e-05,
+      "loss": 0.0337,
+      "step": 2828
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 4.9669776119402986e-05,
+      "loss": 0.048,
+      "step": 2832
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 4.966930970149254e-05,
+      "loss": 0.0534,
+      "step": 2836
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 4.966884328358209e-05,
+      "loss": 0.0296,
+      "step": 2840
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 4.9668376865671644e-05,
+      "loss": 0.0418,
+      "step": 2844
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 4.966791044776119e-05,
+      "loss": 0.0556,
+      "step": 2848
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 4.966744402985075e-05,
+      "loss": 0.0262,
+      "step": 2852
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 4.96669776119403e-05,
+      "loss": 0.0494,
+      "step": 2856
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 4.966651119402985e-05,
+      "loss": 0.0361,
+      "step": 2860
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 4.9666044776119405e-05,
+      "loss": 0.0495,
+      "step": 2864
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 4.966557835820896e-05,
+      "loss": 0.0491,
+      "step": 2868
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 4.966511194029851e-05,
+      "loss": 0.023,
+      "step": 2872
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 4.966464552238806e-05,
+      "loss": 0.0381,
+      "step": 2876
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 4.966417910447762e-05,
+      "loss": 0.0618,
+      "step": 2880
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 4.9663712686567166e-05,
+      "loss": 0.0352,
+      "step": 2884
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 4.966324626865672e-05,
+      "loss": 0.0534,
+      "step": 2888
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.966277985074627e-05,
+      "loss": 0.1242,
+      "step": 2892
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.9662313432835824e-05,
+      "loss": 0.0236,
+      "step": 2896
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.966184701492538e-05,
+      "loss": 0.0396,
+      "step": 2900
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.966138059701493e-05,
+      "loss": 0.0516,
+      "step": 2904
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.9660914179104475e-05,
+      "loss": 0.0378,
+      "step": 2908
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.966044776119404e-05,
+      "loss": 0.049,
+      "step": 2912
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.9659981343283585e-05,
+      "loss": 0.0445,
+      "step": 2916
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.9659514925373133e-05,
+      "loss": 0.0205,
+      "step": 2920
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.965904850746269e-05,
+      "loss": 0.046,
+      "step": 2924
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.965858208955224e-05,
+      "loss": 0.0493,
+      "step": 2928
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.965811567164179e-05,
+      "loss": 0.054,
+      "step": 2932
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 4.9657649253731346e-05,
+      "loss": 0.0409,
+      "step": 2936
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 4.96571828358209e-05,
+      "loss": 0.0396,
+      "step": 2940
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 4.965671641791045e-05,
+      "loss": 0.0362,
+      "step": 2944
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 4.9656250000000004e-05,
+      "loss": 0.0487,
+      "step": 2948
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 4.965578358208955e-05,
+      "loss": 0.0336,
+      "step": 2952
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.965531716417911e-05,
+      "loss": 0.041,
+      "step": 2956
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.965485074626866e-05,
+      "loss": 0.0395,
+      "step": 2960
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.965438432835821e-05,
+      "loss": 0.032,
+      "step": 2964
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 4.965391791044776e-05,
+      "loss": 0.0231,
+      "step": 2968
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 4.965345149253732e-05,
+      "loss": 0.0594,
+      "step": 2972
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 4.965298507462687e-05,
+      "loss": 0.0319,
+      "step": 2976
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 4.9652518656716416e-05,
+      "loss": 0.031,
+      "step": 2980
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 4.965205223880597e-05,
+      "loss": 0.0608,
+      "step": 2984
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 4.9651585820895526e-05,
+      "loss": 0.0273,
+      "step": 2988
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 4.9651119402985074e-05,
+      "loss": 0.0461,
+      "step": 2992
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 4.965065298507463e-05,
+      "loss": 0.0546,
+      "step": 2996
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 4.9650186567164184e-05,
+      "loss": 0.0209,
+      "step": 3000
+    },
+    {
+      "epoch": 2.8,
+      "eval_exact_match": 0.6721470019342359,
+      "eval_exec": 0.7224371373307543,
+      "eval_loss": 0.16579200327396393,
+      "eval_runtime": 1120.3474,
+      "eval_samples_per_second": 0.923,
+      "step": 3000
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 4.964972014925373e-05,
+      "loss": 0.0316,
+      "step": 3004
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 4.964925373134329e-05,
+      "loss": 0.0333,
+      "step": 3008
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 4.9648787313432835e-05,
+      "loss": 0.0722,
+      "step": 3012
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 4.964832089552239e-05,
+      "loss": 0.0709,
+      "step": 3016
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 4.9647854477611945e-05,
+      "loss": 0.0526,
+      "step": 3020
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 4.964738805970149e-05,
+      "loss": 0.0346,
+      "step": 3024
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 4.964692164179105e-05,
+      "loss": 0.0917,
+      "step": 3028
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 4.96464552238806e-05,
+      "loss": 0.0397,
+      "step": 3032
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 4.964598880597015e-05,
+      "loss": 0.0284,
+      "step": 3036
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 4.9645522388059706e-05,
+      "loss": 0.0392,
+      "step": 3040
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 4.9645055970149254e-05,
+      "loss": 0.0406,
+      "step": 3044
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 4.964458955223881e-05,
+      "loss": 0.0423,
+      "step": 3048
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 4.9644123134328364e-05,
+      "loss": 0.0459,
+      "step": 3052
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 4.964365671641791e-05,
+      "loss": 0.0497,
+      "step": 3056
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 4.964319029850747e-05,
+      "loss": 0.0248,
+      "step": 3060
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 4.964272388059702e-05,
+      "loss": 0.0581,
+      "step": 3064
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 4.964225746268657e-05,
+      "loss": 0.0254,
+      "step": 3068
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 4.964179104477612e-05,
+      "loss": 0.029,
+      "step": 3072
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 4.964132462686567e-05,
+      "loss": 0.0354,
+      "step": 3076
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 4.964085820895523e-05,
+      "loss": 0.0398,
+      "step": 3080
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 4.9640391791044776e-05,
+      "loss": 0.0325,
+      "step": 3084
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 4.963992537313433e-05,
+      "loss": 0.0627,
+      "step": 3088
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 4.9639458955223886e-05,
+      "loss": 0.0465,
+      "step": 3092
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.9638992537313434e-05,
+      "loss": 0.0338,
+      "step": 3096
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.963852611940299e-05,
+      "loss": 0.0313,
+      "step": 3100
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.963805970149254e-05,
+      "loss": 0.0247,
+      "step": 3104
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.963759328358209e-05,
+      "loss": 0.029,
+      "step": 3108
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.963712686567165e-05,
+      "loss": 0.0319,
+      "step": 3112
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.9636660447761195e-05,
+      "loss": 0.0358,
+      "step": 3116
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.963619402985074e-05,
+      "loss": 0.0499,
+      "step": 3120
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.9635727611940305e-05,
+      "loss": 0.0388,
+      "step": 3124
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 4.963526119402985e-05,
+      "loss": 0.0304,
+      "step": 3128
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 4.96347947761194e-05,
+      "loss": 0.02,
+      "step": 3132
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 4.9634328358208956e-05,
+      "loss": 0.0344,
+      "step": 3136
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 4.963386194029851e-05,
+      "loss": 0.0321,
+      "step": 3140
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 4.963339552238806e-05,
+      "loss": 0.0496,
+      "step": 3144
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 4.9632929104477614e-05,
+      "loss": 0.0624,
+      "step": 3148
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 4.963246268656717e-05,
+      "loss": 0.039,
+      "step": 3152
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 4.963199626865672e-05,
+      "loss": 0.0444,
+      "step": 3156
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 4.963152985074627e-05,
+      "loss": 0.0342,
+      "step": 3160
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 4.963106343283582e-05,
+      "loss": 0.0329,
+      "step": 3164
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 4.9630597014925375e-05,
+      "loss": 0.0226,
+      "step": 3168
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 4.963013059701493e-05,
+      "loss": 0.0301,
+      "step": 3172
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 4.962966417910448e-05,
+      "loss": 0.019,
+      "step": 3176
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.9629197761194026e-05,
+      "loss": 0.0331,
+      "step": 3180
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.962873134328359e-05,
+      "loss": 0.0497,
+      "step": 3184
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.9628264925373136e-05,
+      "loss": 0.0405,
+      "step": 3188
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 4.962779850746269e-05,
+      "loss": 0.0312,
+      "step": 3192
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 4.962733208955224e-05,
+      "loss": 0.0289,
+      "step": 3196
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 4.9626865671641794e-05,
+      "loss": 0.0514,
+      "step": 3200
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.962639925373135e-05,
+      "loss": 0.0698,
+      "step": 3204
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.96259328358209e-05,
+      "loss": 0.0424,
+      "step": 3208
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.962546641791045e-05,
+      "loss": 0.0357,
+      "step": 3212
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.962500000000001e-05,
+      "loss": 0.0193,
+      "step": 3216
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.9624533582089555e-05,
+      "loss": 0.0471,
+      "step": 3220
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.96240671641791e-05,
+      "loss": 0.0428,
+      "step": 3224
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.9623600746268665e-05,
+      "loss": 0.0268,
+      "step": 3228
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.962313432835821e-05,
+      "loss": 0.0359,
+      "step": 3232
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.962266791044776e-05,
+      "loss": 0.0468,
+      "step": 3236
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.9622201492537316e-05,
+      "loss": 0.0176,
+      "step": 3240
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.962173507462687e-05,
+      "loss": 0.0279,
+      "step": 3244
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.962126865671642e-05,
+      "loss": 0.0193,
+      "step": 3248
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.9620802238805974e-05,
+      "loss": 0.0468,
+      "step": 3252
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.962033582089552e-05,
+      "loss": 0.0284,
+      "step": 3256
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.961986940298508e-05,
+      "loss": 0.0166,
+      "step": 3260
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.961940298507463e-05,
+      "loss": 0.0197,
+      "step": 3264
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.961893656716418e-05,
+      "loss": 0.0305,
+      "step": 3268
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.9618470149253735e-05,
+      "loss": 0.026,
+      "step": 3272
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.961800373134329e-05,
+      "loss": 0.0427,
+      "step": 3276
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.961753731343284e-05,
+      "loss": 0.0206,
+      "step": 3280
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.9617070895522386e-05,
+      "loss": 0.017,
+      "step": 3284
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.961660447761194e-05,
+      "loss": 0.0299,
+      "step": 3288
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.9616138059701496e-05,
+      "loss": 0.0169,
+      "step": 3292
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.9615671641791044e-05,
+      "loss": 0.0345,
+      "step": 3296
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.96152052238806e-05,
+      "loss": 0.0367,
+      "step": 3300
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.9614738805970154e-05,
+      "loss": 0.0288,
+      "step": 3304
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.96142723880597e-05,
+      "loss": 0.0272,
+      "step": 3308
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.961380597014926e-05,
+      "loss": 0.0117,
+      "step": 3312
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.9613339552238805e-05,
+      "loss": 0.0219,
+      "step": 3316
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.961287313432836e-05,
+      "loss": 0.0071,
+      "step": 3320
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.9612406716417915e-05,
+      "loss": 0.0183,
+      "step": 3324
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.961194029850746e-05,
+      "loss": 0.0314,
+      "step": 3328
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.961147388059702e-05,
+      "loss": 0.0123,
+      "step": 3332
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.961100746268657e-05,
+      "loss": 0.0308,
+      "step": 3336
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.961054104477612e-05,
+      "loss": 0.0281,
+      "step": 3340
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.961007462686567e-05,
+      "loss": 0.0215,
+      "step": 3344
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.9609608208955224e-05,
+      "loss": 0.017,
+      "step": 3348
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.960914179104478e-05,
+      "loss": 0.0152,
+      "step": 3352
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.9608675373134334e-05,
+      "loss": 0.044,
+      "step": 3356
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.960820895522388e-05,
+      "loss": 0.0241,
+      "step": 3360
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.960774253731344e-05,
+      "loss": 0.0159,
+      "step": 3364
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.960727611940299e-05,
+      "loss": 0.0282,
+      "step": 3368
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.960680970149254e-05,
+      "loss": 0.0336,
+      "step": 3372
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.960634328358209e-05,
+      "loss": 0.038,
+      "step": 3376
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.960587686567165e-05,
+      "loss": 0.0369,
+      "step": 3380
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.96054104477612e-05,
+      "loss": 0.0246,
+      "step": 3384
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.9604944029850746e-05,
+      "loss": 0.0163,
+      "step": 3388
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.96044776119403e-05,
+      "loss": 0.0241,
+      "step": 3392
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.9604011194029856e-05,
+      "loss": 0.0232,
+      "step": 3396
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.9603544776119404e-05,
+      "loss": 0.0355,
+      "step": 3400
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.960307835820896e-05,
+      "loss": 0.0187,
+      "step": 3404
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.960261194029851e-05,
+      "loss": 0.0208,
+      "step": 3408
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.960214552238806e-05,
+      "loss": 0.0233,
+      "step": 3412
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.960167910447762e-05,
+      "loss": 0.0418,
+      "step": 3416
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.9601212686567165e-05,
+      "loss": 0.0465,
+      "step": 3420
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.960074626865672e-05,
+      "loss": 0.028,
+      "step": 3424
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 4.9600279850746275e-05,
+      "loss": 0.0257,
+      "step": 3428
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 4.959981343283582e-05,
+      "loss": 0.0357,
+      "step": 3432
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 4.959934701492537e-05,
+      "loss": 0.017,
+      "step": 3436
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 4.959888059701493e-05,
+      "loss": 0.0296,
+      "step": 3440
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 4.959841417910448e-05,
+      "loss": 0.0336,
+      "step": 3444
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 4.959794776119403e-05,
+      "loss": 0.0194,
+      "step": 3448
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 4.9597481343283584e-05,
+      "loss": 0.0231,
+      "step": 3452
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 4.959701492537314e-05,
+      "loss": 0.0186,
+      "step": 3456
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 4.959654850746269e-05,
+      "loss": 0.0477,
+      "step": 3460
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 4.959608208955224e-05,
+      "loss": 0.041,
+      "step": 3464
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 4.959561567164179e-05,
+      "loss": 0.0304,
+      "step": 3468
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 4.9595149253731345e-05,
+      "loss": 0.0318,
+      "step": 3472
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 4.95946828358209e-05,
+      "loss": 0.0789,
+      "step": 3476
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 4.959421641791045e-05,
+      "loss": 0.0251,
+      "step": 3480
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 4.959375e-05,
+      "loss": 0.0254,
+      "step": 3484
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 4.959328358208956e-05,
+      "loss": 0.0178,
+      "step": 3488
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 4.9592817164179106e-05,
+      "loss": 0.03,
+      "step": 3492
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 4.9592350746268654e-05,
+      "loss": 0.0202,
+      "step": 3496
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 4.9591884328358216e-05,
+      "loss": 0.0243,
+      "step": 3500
+    },
+    {
+      "epoch": 3.26,
+      "eval_exact_match": 0.7108317214700194,
+      "eval_exec": 0.7543520309477756,
+      "eval_loss": 0.17954690754413605,
+      "eval_runtime": 1070.1188,
+      "eval_samples_per_second": 0.966,
+      "step": 3500
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 4.9591417910447764e-05,
+      "loss": 0.0259,
+      "step": 3504
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 4.959095149253731e-05,
+      "loss": 0.0117,
+      "step": 3508
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 4.959048507462687e-05,
+      "loss": 0.0275,
+      "step": 3512
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 4.959001865671642e-05,
+      "loss": 0.0219,
+      "step": 3516
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 4.9589552238805977e-05,
+      "loss": 0.0259,
+      "step": 3520
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 4.9589085820895525e-05,
+      "loss": 0.0362,
+      "step": 3524
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 4.958861940298507e-05,
+      "loss": 0.0362,
+      "step": 3528
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 4.9588152985074634e-05,
+      "loss": 0.0363,
+      "step": 3532
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 4.958768656716418e-05,
+      "loss": 0.0603,
+      "step": 3536
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 4.958722014925373e-05,
+      "loss": 0.0303,
+      "step": 3540
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 4.9586753731343286e-05,
+      "loss": 0.0152,
+      "step": 3544
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 4.958628731343284e-05,
+      "loss": 0.0177,
+      "step": 3548
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 4.958582089552239e-05,
+      "loss": 0.03,
+      "step": 3552
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 4.9585354477611944e-05,
+      "loss": 0.0267,
+      "step": 3556
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 4.95848880597015e-05,
+      "loss": 0.016,
+      "step": 3560
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 4.958442164179105e-05,
+      "loss": 0.0345,
+      "step": 3564
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 4.95839552238806e-05,
+      "loss": 0.0247,
+      "step": 3568
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 4.958348880597015e-05,
+      "loss": 0.0342,
+      "step": 3572
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 4.9583022388059705e-05,
+      "loss": 0.0234,
+      "step": 3576
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 4.958255597014926e-05,
+      "loss": 0.034,
+      "step": 3580
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 4.958208955223881e-05,
+      "loss": 0.0257,
+      "step": 3584
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 4.9581623134328356e-05,
+      "loss": 0.0194,
+      "step": 3588
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 4.958115671641792e-05,
+      "loss": 0.0289,
+      "step": 3592
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 4.9580690298507466e-05,
+      "loss": 0.0336,
+      "step": 3596
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 4.9580223880597014e-05,
+      "loss": 0.0295,
+      "step": 3600
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 4.957975746268657e-05,
+      "loss": 0.0334,
+      "step": 3604
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 4.9579291044776124e-05,
+      "loss": 0.0198,
+      "step": 3608
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 4.957882462686567e-05,
+      "loss": 0.0195,
+      "step": 3612
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 4.9578358208955227e-05,
+      "loss": 0.0281,
+      "step": 3616
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 4.957789179104478e-05,
+      "loss": 0.0192,
+      "step": 3620
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 4.957742537313433e-05,
+      "loss": 0.0361,
+      "step": 3624
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 4.9576958955223884e-05,
+      "loss": 0.0094,
+      "step": 3628
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 4.957649253731343e-05,
+      "loss": 0.0212,
+      "step": 3632
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 4.957602611940299e-05,
+      "loss": 0.0237,
+      "step": 3636
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 4.957555970149254e-05,
+      "loss": 0.0295,
+      "step": 3640
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 4.957509328358209e-05,
+      "loss": 0.0373,
+      "step": 3644
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 4.957462686567164e-05,
+      "loss": 0.0255,
+      "step": 3648
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 4.95741604477612e-05,
+      "loss": 0.0313,
+      "step": 3652
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 4.957369402985075e-05,
+      "loss": 0.0548,
+      "step": 3656
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 4.95732276119403e-05,
+      "loss": 0.0417,
+      "step": 3660
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 4.957276119402985e-05,
+      "loss": 0.0232,
+      "step": 3664
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 4.9572294776119406e-05,
+      "loss": 0.0203,
+      "step": 3668
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 4.9571828358208955e-05,
+      "loss": 0.0422,
+      "step": 3672
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 4.957136194029851e-05,
+      "loss": 0.0225,
+      "step": 3676
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 4.9570895522388064e-05,
+      "loss": 0.0162,
+      "step": 3680
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 4.957042910447762e-05,
+      "loss": 0.0359,
+      "step": 3684
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 4.956996268656717e-05,
+      "loss": 0.0186,
+      "step": 3688
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 4.9569496268656716e-05,
+      "loss": 0.0165,
+      "step": 3692
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 4.956902985074627e-05,
+      "loss": 0.0148,
+      "step": 3696
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 4.9568563432835825e-05,
+      "loss": 0.026,
+      "step": 3700
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 4.9568097014925374e-05,
+      "loss": 0.0157,
+      "step": 3704
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 4.956763059701493e-05,
+      "loss": 0.0465,
+      "step": 3708
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 4.956716417910448e-05,
+      "loss": 0.0174,
+      "step": 3712
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 4.956669776119403e-05,
+      "loss": 0.0233,
+      "step": 3716
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 4.9566231343283586e-05,
+      "loss": 0.0347,
+      "step": 3720
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 4.9565764925373135e-05,
+      "loss": 0.0201,
+      "step": 3724
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 4.956529850746269e-05,
+      "loss": 0.0182,
+      "step": 3728
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 4.9564832089552244e-05,
+      "loss": 0.0306,
+      "step": 3732
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 4.956436567164179e-05,
+      "loss": 0.0205,
+      "step": 3736
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 4.956389925373135e-05,
+      "loss": 0.0255,
+      "step": 3740
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 4.95634328358209e-05,
+      "loss": 0.0309,
+      "step": 3744
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 4.956296641791045e-05,
+      "loss": 0.018,
+      "step": 3748
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 4.95625e-05,
+      "loss": 0.0175,
+      "step": 3752
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 4.9562033582089553e-05,
+      "loss": 0.0334,
+      "step": 3756
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 4.956156716417911e-05,
+      "loss": 0.0213,
+      "step": 3760
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 4.9561100746268656e-05,
+      "loss": 0.0153,
+      "step": 3764
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 4.956063432835821e-05,
+      "loss": 0.0223,
+      "step": 3768
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 4.9560167910447766e-05,
+      "loss": 0.0178,
+      "step": 3772
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 4.9559701492537314e-05,
+      "loss": 0.0278,
+      "step": 3776
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 4.955923507462687e-05,
+      "loss": 0.0264,
+      "step": 3780
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 4.955876865671642e-05,
+      "loss": 0.0204,
+      "step": 3784
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 4.955830223880597e-05,
+      "loss": 0.0252,
+      "step": 3788
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 4.955783582089553e-05,
+      "loss": 0.044,
+      "step": 3792
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 4.9557369402985075e-05,
+      "loss": 0.0266,
+      "step": 3796
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 4.9556902985074624e-05,
+      "loss": 0.0191,
+      "step": 3800
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 4.9556436567164185e-05,
+      "loss": 0.0422,
+      "step": 3804
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 4.955597014925373e-05,
+      "loss": 0.0284,
+      "step": 3808
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 4.955550373134328e-05,
+      "loss": 0.0344,
+      "step": 3812
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 4.9555037313432836e-05,
+      "loss": 0.0295,
+      "step": 3816
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 4.955457089552239e-05,
+      "loss": 0.0331,
+      "step": 3820
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 4.955410447761194e-05,
+      "loss": 0.0436,
+      "step": 3824
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 4.9553638059701494e-05,
+      "loss": 0.0292,
+      "step": 3828
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 4.955317164179105e-05,
+      "loss": 0.0356,
+      "step": 3832
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 4.95527052238806e-05,
+      "loss": 0.0315,
+      "step": 3836
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 4.955223880597015e-05,
+      "loss": 0.0295,
+      "step": 3840
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 4.95517723880597e-05,
+      "loss": 0.0413,
+      "step": 3844
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 4.955130597014926e-05,
+      "loss": 0.0287,
+      "step": 3848
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 4.955083955223881e-05,
+      "loss": 0.0223,
+      "step": 3852
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 4.955037313432836e-05,
+      "loss": 0.0355,
+      "step": 3856
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 4.954990671641791e-05,
+      "loss": 0.0244,
+      "step": 3860
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 4.954944029850747e-05,
+      "loss": 0.0149,
+      "step": 3864
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 4.9548973880597016e-05,
+      "loss": 0.0225,
+      "step": 3868
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 4.954850746268657e-05,
+      "loss": 0.0261,
+      "step": 3872
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 4.954804104477612e-05,
+      "loss": 0.0264,
+      "step": 3876
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 4.9547574626865674e-05,
+      "loss": 0.0282,
+      "step": 3880
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 4.954710820895523e-05,
+      "loss": 0.0511,
+      "step": 3884
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 4.954664179104478e-05,
+      "loss": 0.0487,
+      "step": 3888
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 4.954617537313433e-05,
+      "loss": 0.0169,
+      "step": 3892
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 4.954570895522389e-05,
+      "loss": 0.0243,
+      "step": 3896
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 4.9545242537313435e-05,
+      "loss": 0.0247,
+      "step": 3900
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 4.954477611940298e-05,
+      "loss": 0.0258,
+      "step": 3904
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 4.9544309701492545e-05,
+      "loss": 0.0328,
+      "step": 3908
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 4.954384328358209e-05,
+      "loss": 0.0328,
+      "step": 3912
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 4.954337686567164e-05,
+      "loss": 0.0265,
+      "step": 3916
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 4.9542910447761196e-05,
+      "loss": 0.0349,
+      "step": 3920
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 4.954244402985075e-05,
+      "loss": 0.0307,
+      "step": 3924
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 4.95419776119403e-05,
+      "loss": 0.0256,
+      "step": 3928
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 4.9541511194029854e-05,
+      "loss": 0.0246,
+      "step": 3932
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 4.95410447761194e-05,
+      "loss": 0.0212,
+      "step": 3936
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 4.954057835820896e-05,
+      "loss": 0.0548,
+      "step": 3940
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 4.954011194029851e-05,
+      "loss": 0.05,
+      "step": 3944
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 4.953964552238806e-05,
+      "loss": 0.036,
+      "step": 3948
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 4.9539179104477615e-05,
+      "loss": 0.0122,
+      "step": 3952
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 4.953871268656717e-05,
+      "loss": 0.0352,
+      "step": 3956
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 4.953824626865672e-05,
+      "loss": 0.0316,
+      "step": 3960
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 4.9537779850746266e-05,
+      "loss": 0.0243,
+      "step": 3964
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 4.953731343283582e-05,
+      "loss": 0.0224,
+      "step": 3968
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 4.9536847014925376e-05,
+      "loss": 0.0249,
+      "step": 3972
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 4.9536380597014924e-05,
+      "loss": 0.0368,
+      "step": 3976
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 4.953591417910448e-05,
+      "loss": 0.0412,
+      "step": 3980
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 4.9535447761194034e-05,
+      "loss": 0.0248,
+      "step": 3984
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 4.953498134328358e-05,
+      "loss": 0.0399,
+      "step": 3988
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 4.953451492537314e-05,
+      "loss": 0.0156,
+      "step": 3992
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 4.9534048507462685e-05,
+      "loss": 0.0109,
+      "step": 3996
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 4.953358208955224e-05,
+      "loss": 0.0412,
+      "step": 4000
+    },
+    {
+      "epoch": 3.73,
+      "eval_exact_match": 0.7030947775628626,
+      "eval_exec": 0.7369439071566731,
+      "eval_loss": 0.189029723405838,
+      "eval_runtime": 1025.0796,
+      "eval_samples_per_second": 1.009,
+      "step": 4000
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 4.9533115671641795e-05,
+      "loss": 0.0135,
+      "step": 4004
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 4.953264925373134e-05,
+      "loss": 0.0238,
+      "step": 4008
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 4.95321828358209e-05,
+      "loss": 0.0143,
+      "step": 4012
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 4.953171641791045e-05,
+      "loss": 0.0691,
+      "step": 4016
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 4.953125e-05,
+      "loss": 0.0458,
+      "step": 4020
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 4.9530783582089556e-05,
+      "loss": 0.0638,
+      "step": 4024
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 4.9530317164179104e-05,
+      "loss": 0.022,
+      "step": 4028
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 4.952985074626866e-05,
+      "loss": 0.0205,
+      "step": 4032
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 4.9529384328358214e-05,
+      "loss": 0.0268,
+      "step": 4036
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 4.952891791044776e-05,
+      "loss": 0.0292,
+      "step": 4040
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 4.952845149253732e-05,
+      "loss": 0.0362,
+      "step": 4044
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 4.952798507462687e-05,
+      "loss": 0.0272,
+      "step": 4048
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 4.952751865671642e-05,
+      "loss": 0.0357,
+      "step": 4052
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 4.952705223880597e-05,
+      "loss": 0.0465,
+      "step": 4056
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 4.952658582089553e-05,
+      "loss": 0.0329,
+      "step": 4060
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 4.952611940298508e-05,
+      "loss": 0.0191,
+      "step": 4064
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 4.9525652985074626e-05,
+      "loss": 0.0204,
+      "step": 4068
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 4.952518656716418e-05,
+      "loss": 0.0362,
+      "step": 4072
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 4.9524720149253736e-05,
+      "loss": 0.0175,
+      "step": 4076
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 4.9524253731343284e-05,
+      "loss": 0.0609,
+      "step": 4080
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 4.952378731343284e-05,
+      "loss": 0.0194,
+      "step": 4084
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 4.952332089552239e-05,
+      "loss": 0.0331,
+      "step": 4088
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 4.952285447761194e-05,
+      "loss": 0.0464,
+      "step": 4092
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 4.95223880597015e-05,
+      "loss": 0.0154,
+      "step": 4096
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 4.9521921641791045e-05,
+      "loss": 0.0402,
+      "step": 4100
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 4.95214552238806e-05,
+      "loss": 0.0201,
+      "step": 4104
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 4.9520988805970155e-05,
+      "loss": 0.0329,
+      "step": 4108
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 4.95205223880597e-05,
+      "loss": 0.0357,
+      "step": 4112
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 4.952005597014925e-05,
+      "loss": 0.0322,
+      "step": 4116
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 4.951958955223881e-05,
+      "loss": 0.0286,
+      "step": 4120
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 4.951912313432836e-05,
+      "loss": 0.0334,
+      "step": 4124
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 4.951865671641791e-05,
+      "loss": 0.0324,
+      "step": 4128
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 4.9518190298507464e-05,
+      "loss": 0.0235,
+      "step": 4132
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 4.951772388059702e-05,
+      "loss": 0.0456,
+      "step": 4136
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 4.951725746268657e-05,
+      "loss": 0.029,
+      "step": 4140
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 4.951679104477612e-05,
+      "loss": 0.039,
+      "step": 4144
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 4.951632462686567e-05,
+      "loss": 0.0576,
+      "step": 4148
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 4.9515858208955225e-05,
+      "loss": 0.0086,
+      "step": 4152
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 4.951539179104478e-05,
+      "loss": 0.0461,
+      "step": 4156
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 4.951492537313433e-05,
+      "loss": 0.0401,
+      "step": 4160
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 4.951445895522388e-05,
+      "loss": 0.0242,
+      "step": 4164
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 4.951399253731344e-05,
+      "loss": 0.0193,
+      "step": 4168
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 4.9513526119402986e-05,
+      "loss": 0.018,
+      "step": 4172
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 4.951305970149254e-05,
+      "loss": 0.0306,
+      "step": 4176
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 4.9512593283582096e-05,
+      "loss": 0.0333,
+      "step": 4180
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 4.9512126865671644e-05,
+      "loss": 0.0168,
+      "step": 4184
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 4.95116604477612e-05,
+      "loss": 0.0275,
+      "step": 4188
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 4.951119402985075e-05,
+      "loss": 0.0228,
+      "step": 4192
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 4.95107276119403e-05,
+      "loss": 0.0351,
+      "step": 4196
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 4.951026119402986e-05,
+      "loss": 0.0217,
+      "step": 4200
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 4.9509794776119405e-05,
+      "loss": 0.0157,
+      "step": 4204
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 4.950932835820895e-05,
+      "loss": 0.031,
+      "step": 4208
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 4.9508861940298515e-05,
+      "loss": 0.0352,
+      "step": 4212
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 4.950839552238806e-05,
+      "loss": 0.0201,
+      "step": 4216
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 4.950792910447761e-05,
+      "loss": 0.03,
+      "step": 4220
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 4.9507462686567166e-05,
+      "loss": 0.0345,
+      "step": 4224
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 4.950699626865672e-05,
+      "loss": 0.043,
+      "step": 4228
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 4.950652985074627e-05,
+      "loss": 0.0389,
+      "step": 4232
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 4.9506063432835824e-05,
+      "loss": 0.0203,
+      "step": 4236
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 4.950559701492538e-05,
+      "loss": 0.0464,
+      "step": 4240
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 4.950513059701493e-05,
+      "loss": 0.0171,
+      "step": 4244
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 4.950466417910448e-05,
+      "loss": 0.0478,
+      "step": 4248
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 4.950419776119403e-05,
+      "loss": 0.0148,
+      "step": 4252
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 4.9503731343283585e-05,
+      "loss": 0.0182,
+      "step": 4256
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 4.950326492537314e-05,
+      "loss": 0.0165,
+      "step": 4260
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 4.950279850746269e-05,
+      "loss": 0.0097,
+      "step": 4264
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 4.9502332089552236e-05,
+      "loss": 0.0343,
+      "step": 4268
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 4.95018656716418e-05,
+      "loss": 0.0079,
+      "step": 4272
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 4.9501399253731346e-05,
+      "loss": 0.0332,
+      "step": 4276
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 4.9500932835820894e-05,
+      "loss": 0.0205,
+      "step": 4280
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 4.950046641791045e-05,
+      "loss": 0.0333,
+      "step": 4284
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 4.9500000000000004e-05,
+      "loss": 0.0344,
+      "step": 4288
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 4.949953358208955e-05,
+      "loss": 0.016,
+      "step": 4292
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 4.949906716417911e-05,
+      "loss": 0.0103,
+      "step": 4296
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 4.949860074626866e-05,
+      "loss": 0.0192,
+      "step": 4300
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 4.949813432835821e-05,
+      "loss": 0.0157,
+      "step": 4304
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 4.9497667910447765e-05,
+      "loss": 0.0178,
+      "step": 4308
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 4.949720149253731e-05,
+      "loss": 0.0073,
+      "step": 4312
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 4.949673507462687e-05,
+      "loss": 0.0299,
+      "step": 4316
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 4.949626865671642e-05,
+      "loss": 0.0182,
+      "step": 4320
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 4.949580223880597e-05,
+      "loss": 0.0246,
+      "step": 4324
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 4.949533582089552e-05,
+      "loss": 0.0322,
+      "step": 4328
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 4.949486940298508e-05,
+      "loss": 0.0391,
+      "step": 4332
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 4.949440298507463e-05,
+      "loss": 0.0207,
+      "step": 4336
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 4.9493936567164184e-05,
+      "loss": 0.0201,
+      "step": 4340
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 4.949347014925373e-05,
+      "loss": 0.0194,
+      "step": 4344
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 4.949300373134329e-05,
+      "loss": 0.0162,
+      "step": 4348
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 4.949253731343284e-05,
+      "loss": 0.0393,
+      "step": 4352
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 4.949207089552239e-05,
+      "loss": 0.0156,
+      "step": 4356
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 4.9491604477611945e-05,
+      "loss": 0.0182,
+      "step": 4360
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 4.94911380597015e-05,
+      "loss": 0.0119,
+      "step": 4364
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 4.949067164179105e-05,
+      "loss": 0.0116,
+      "step": 4368
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 4.9490205223880596e-05,
+      "loss": 0.0179,
+      "step": 4372
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 4.948973880597015e-05,
+      "loss": 0.0153,
+      "step": 4376
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 4.9489272388059706e-05,
+      "loss": 0.0362,
+      "step": 4380
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 4.9488805970149254e-05,
+      "loss": 0.0082,
+      "step": 4384
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 4.948833955223881e-05,
+      "loss": 0.0287,
+      "step": 4388
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 4.9487873134328364e-05,
+      "loss": 0.0125,
+      "step": 4392
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 4.948740671641791e-05,
+      "loss": 0.0094,
+      "step": 4396
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 4.948694029850747e-05,
+      "loss": 0.0187,
+      "step": 4400
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 4.9486473880597015e-05,
+      "loss": 0.0197,
+      "step": 4404
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 4.948600746268657e-05,
+      "loss": 0.0104,
+      "step": 4408
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 4.9485541044776125e-05,
+      "loss": 0.0424,
+      "step": 4412
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 4.948507462686567e-05,
+      "loss": 0.0181,
+      "step": 4416
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 4.948460820895523e-05,
+      "loss": 0.0187,
+      "step": 4420
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 4.948414179104478e-05,
+      "loss": 0.0154,
+      "step": 4424
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 4.948367537313433e-05,
+      "loss": 0.0218,
+      "step": 4428
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 4.948320895522388e-05,
+      "loss": 0.0119,
+      "step": 4432
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 4.9482742537313434e-05,
+      "loss": 0.0215,
+      "step": 4436
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 4.948227611940299e-05,
+      "loss": 0.0112,
+      "step": 4440
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 4.948180970149254e-05,
+      "loss": 0.0187,
+      "step": 4444
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 4.948134328358209e-05,
+      "loss": 0.0083,
+      "step": 4448
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 4.9480876865671647e-05,
+      "loss": 0.0153,
+      "step": 4452
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 4.9480410447761195e-05,
+      "loss": 0.0342,
+      "step": 4456
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 4.947994402985075e-05,
+      "loss": 0.0264,
+      "step": 4460
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 4.94794776119403e-05,
+      "loss": 0.0149,
+      "step": 4464
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 4.947901119402985e-05,
+      "loss": 0.0193,
+      "step": 4468
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 4.947854477611941e-05,
+      "loss": 0.011,
+      "step": 4472
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 4.9478078358208956e-05,
+      "loss": 0.0224,
+      "step": 4476
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 4.9477611940298504e-05,
+      "loss": 0.0225,
+      "step": 4480
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 4.9477145522388066e-05,
+      "loss": 0.0176,
+      "step": 4484
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 4.9476679104477614e-05,
+      "loss": 0.0262,
+      "step": 4488
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 4.947621268656716e-05,
+      "loss": 0.0165,
+      "step": 4492
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 4.947574626865672e-05,
+      "loss": 0.0167,
+      "step": 4496
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 4.947527985074627e-05,
+      "loss": 0.0262,
+      "step": 4500
+    },
+    {
+      "epoch": 4.2,
+      "eval_exact_match": 0.6992263056092843,
+      "eval_exec": 0.7437137330754352,
+      "eval_loss": 0.18752895295619965,
+      "eval_runtime": 1059.8867,
+      "eval_samples_per_second": 0.976,
+      "step": 4500
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 4.9474813432835827e-05,
+      "loss": 0.0147,
+      "step": 4504
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 4.9474347014925375e-05,
+      "loss": 0.0179,
+      "step": 4508
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 4.947388059701493e-05,
+      "loss": 0.0218,
+      "step": 4512
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 4.9473414179104484e-05,
+      "loss": 0.0346,
+      "step": 4516
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 4.947294776119403e-05,
+      "loss": 0.0181,
+      "step": 4520
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 4.947248134328358e-05,
+      "loss": 0.0138,
+      "step": 4524
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 4.947201492537314e-05,
+      "loss": 0.0184,
+      "step": 4528
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 4.947154850746269e-05,
+      "loss": 0.0319,
+      "step": 4532
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 4.947108208955224e-05,
+      "loss": 0.0194,
+      "step": 4536
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 4.9470615671641794e-05,
+      "loss": 0.0148,
+      "step": 4540
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 4.947014925373135e-05,
+      "loss": 0.0233,
+      "step": 4544
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 4.9469682835820897e-05,
+      "loss": 0.0273,
+      "step": 4548
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 4.946921641791045e-05,
+      "loss": 0.0139,
+      "step": 4552
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 4.946875e-05,
+      "loss": 0.0158,
+      "step": 4556
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 4.9468283582089555e-05,
+      "loss": 0.0123,
+      "step": 4560
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 4.946781716417911e-05,
+      "loss": 0.0205,
+      "step": 4564
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 4.946735074626866e-05,
+      "loss": 0.0181,
+      "step": 4568
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 4.946688432835821e-05,
+      "loss": 0.0365,
+      "step": 4572
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 4.946641791044777e-05,
+      "loss": 0.0174,
+      "step": 4576
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 4.9465951492537316e-05,
+      "loss": 0.0239,
+      "step": 4580
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 4.9465485074626864e-05,
+      "loss": 0.0113,
+      "step": 4584
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 4.9465018656716425e-05,
+      "loss": 0.0309,
+      "step": 4588
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 4.9464552238805973e-05,
+      "loss": 0.0245,
+      "step": 4592
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 4.946408582089552e-05,
+      "loss": 0.0252,
+      "step": 4596
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 4.9463619402985077e-05,
+      "loss": 0.0233,
+      "step": 4600
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 4.946315298507463e-05,
+      "loss": 0.0214,
+      "step": 4604
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 4.946268656716418e-05,
+      "loss": 0.0286,
+      "step": 4608
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 4.9462220149253734e-05,
+      "loss": 0.0199,
+      "step": 4612
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 4.946175373134328e-05,
+      "loss": 0.0373,
+      "step": 4616
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 4.946128731343284e-05,
+      "loss": 0.0142,
+      "step": 4620
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 4.946082089552239e-05,
+      "loss": 0.0189,
+      "step": 4624
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 4.946035447761194e-05,
+      "loss": 0.0208,
+      "step": 4628
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 4.9459888059701495e-05,
+      "loss": 0.0197,
+      "step": 4632
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 4.945942164179105e-05,
+      "loss": 0.0124,
+      "step": 4636
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 4.94589552238806e-05,
+      "loss": 0.0308,
+      "step": 4640
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 4.945848880597015e-05,
+      "loss": 0.0192,
+      "step": 4644
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 4.94580223880597e-05,
+      "loss": 0.0239,
+      "step": 4648
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 4.9457555970149256e-05,
+      "loss": 0.0087,
+      "step": 4652
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 4.9457089552238805e-05,
+      "loss": 0.0256,
+      "step": 4656
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 4.945662313432836e-05,
+      "loss": 0.0396,
+      "step": 4660
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 4.9456156716417914e-05,
+      "loss": 0.0162,
+      "step": 4664
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 4.945569029850747e-05,
+      "loss": 0.0233,
+      "step": 4668
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 4.945522388059702e-05,
+      "loss": 0.0246,
+      "step": 4672
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 4.9454757462686566e-05,
+      "loss": 0.0165,
+      "step": 4676
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 4.945429104477613e-05,
+      "loss": 0.027,
+      "step": 4680
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 4.9453824626865675e-05,
+      "loss": 0.0126,
+      "step": 4684
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 4.9453358208955224e-05,
+      "loss": 0.0182,
+      "step": 4688
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 4.945289179104478e-05,
+      "loss": 0.0226,
+      "step": 4692
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 4.945242537313433e-05,
+      "loss": 0.0264,
+      "step": 4696
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 4.945195895522388e-05,
+      "loss": 0.0378,
+      "step": 4700
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 4.9451492537313436e-05,
+      "loss": 0.0161,
+      "step": 4704
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 4.9451026119402984e-05,
+      "loss": 0.012,
+      "step": 4708
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 4.945055970149254e-05,
+      "loss": 0.0132,
+      "step": 4712
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 4.9450093283582094e-05,
+      "loss": 0.0185,
+      "step": 4716
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 4.944962686567164e-05,
+      "loss": 0.0214,
+      "step": 4720
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 4.94491604477612e-05,
+      "loss": 0.0057,
+      "step": 4724
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 4.944869402985075e-05,
+      "loss": 0.0158,
+      "step": 4728
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 4.94482276119403e-05,
+      "loss": 0.0063,
+      "step": 4732
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 4.944776119402985e-05,
+      "loss": 0.0059,
+      "step": 4736
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 4.944729477611941e-05,
+      "loss": 0.03,
+      "step": 4740
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 4.944682835820896e-05,
+      "loss": 0.0185,
+      "step": 4744
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 4.9446361940298506e-05,
+      "loss": 0.0241,
+      "step": 4748
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 4.944589552238806e-05,
+      "loss": 0.0162,
+      "step": 4752
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 4.9445429104477616e-05,
+      "loss": 0.0138,
+      "step": 4756
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 4.9444962686567164e-05,
+      "loss": 0.0128,
+      "step": 4760
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 4.944449626865672e-05,
+      "loss": 0.0107,
+      "step": 4764
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 4.944402985074627e-05,
+      "loss": 0.0171,
+      "step": 4768
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 4.944356343283582e-05,
+      "loss": 0.0238,
+      "step": 4772
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 4.944309701492538e-05,
+      "loss": 0.0075,
+      "step": 4776
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 4.9442630597014925e-05,
+      "loss": 0.0178,
+      "step": 4780
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 4.944216417910448e-05,
+      "loss": 0.0115,
+      "step": 4784
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 4.9441697761194035e-05,
+      "loss": 0.0189,
+      "step": 4788
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 4.944123134328358e-05,
+      "loss": 0.026,
+      "step": 4792
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 4.944076492537313e-05,
+      "loss": 0.0286,
+      "step": 4796
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 4.944029850746269e-05,
+      "loss": 0.0529,
+      "step": 4800
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 4.943983208955224e-05,
+      "loss": 0.0191,
+      "step": 4804
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 4.943936567164179e-05,
+      "loss": 0.0323,
+      "step": 4808
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 4.9438899253731344e-05,
+      "loss": 0.0208,
+      "step": 4812
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 4.94384328358209e-05,
+      "loss": 0.0217,
+      "step": 4816
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 4.943796641791045e-05,
+      "loss": 0.0251,
+      "step": 4820
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 4.94375e-05,
+      "loss": 0.0133,
+      "step": 4824
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 4.943703358208955e-05,
+      "loss": 0.02,
+      "step": 4828
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 4.943656716417911e-05,
+      "loss": 0.0258,
+      "step": 4832
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 4.943610074626866e-05,
+      "loss": 0.0172,
+      "step": 4836
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 4.943563432835821e-05,
+      "loss": 0.0382,
+      "step": 4840
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 4.943516791044776e-05,
+      "loss": 0.014,
+      "step": 4844
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 4.943470149253732e-05,
+      "loss": 0.0078,
+      "step": 4848
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 4.9434235074626866e-05,
+      "loss": 0.0191,
+      "step": 4852
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 4.943376865671642e-05,
+      "loss": 0.025,
+      "step": 4856
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 4.9433302238805976e-05,
+      "loss": 0.0247,
+      "step": 4860
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 4.9432835820895524e-05,
+      "loss": 0.0205,
+      "step": 4864
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 4.943236940298508e-05,
+      "loss": 0.0087,
+      "step": 4868
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 4.943190298507463e-05,
+      "loss": 0.0133,
+      "step": 4872
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 4.943143656716418e-05,
+      "loss": 0.0116,
+      "step": 4876
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 4.943097014925374e-05,
+      "loss": 0.0378,
+      "step": 4880
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 4.9430503731343285e-05,
+      "loss": 0.0138,
+      "step": 4884
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 4.943003731343283e-05,
+      "loss": 0.0109,
+      "step": 4888
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 4.9429570895522395e-05,
+      "loss": 0.0393,
+      "step": 4892
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 4.942910447761194e-05,
+      "loss": 0.0111,
+      "step": 4896
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 4.942863805970149e-05,
+      "loss": 0.0159,
+      "step": 4900
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 4.9428171641791046e-05,
+      "loss": 0.0169,
+      "step": 4904
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 4.94277052238806e-05,
+      "loss": 0.0271,
+      "step": 4908
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 4.942723880597015e-05,
+      "loss": 0.0064,
+      "step": 4912
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 4.9426772388059704e-05,
+      "loss": 0.016,
+      "step": 4916
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 4.942630597014926e-05,
+      "loss": 0.0194,
+      "step": 4920
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 4.942583955223881e-05,
+      "loss": 0.0241,
+      "step": 4924
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 4.942537313432836e-05,
+      "loss": 0.0217,
+      "step": 4928
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 4.942490671641791e-05,
+      "loss": 0.0054,
+      "step": 4932
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 4.9424440298507465e-05,
+      "loss": 0.0156,
+      "step": 4936
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 4.942397388059702e-05,
+      "loss": 0.016,
+      "step": 4940
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 4.942350746268657e-05,
+      "loss": 0.0283,
+      "step": 4944
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 4.9423041044776116e-05,
+      "loss": 0.0189,
+      "step": 4948
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 4.942257462686568e-05,
+      "loss": 0.0236,
+      "step": 4952
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 4.9422108208955226e-05,
+      "loss": 0.0101,
+      "step": 4956
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 4.9421641791044774e-05,
+      "loss": 0.0299,
+      "step": 4960
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 4.942117537313433e-05,
+      "loss": 0.0215,
+      "step": 4964
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 4.9420708955223884e-05,
+      "loss": 0.0109,
+      "step": 4968
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 4.942024253731343e-05,
+      "loss": 0.0157,
+      "step": 4972
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 4.941977611940299e-05,
+      "loss": 0.0241,
+      "step": 4976
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 4.941930970149254e-05,
+      "loss": 0.0279,
+      "step": 4980
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 4.941884328358209e-05,
+      "loss": 0.0318,
+      "step": 4984
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 4.9418376865671645e-05,
+      "loss": 0.022,
+      "step": 4988
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 4.941791044776119e-05,
+      "loss": 0.0137,
+      "step": 4992
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 4.941744402985075e-05,
+      "loss": 0.0226,
+      "step": 4996
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 4.94169776119403e-05,
+      "loss": 0.009,
+      "step": 5000
+    },
+    {
+      "epoch": 4.66,
+      "eval_exact_match": 0.7030947775628626,
+      "eval_exec": 0.746615087040619,
+      "eval_loss": 0.18529821932315826,
+      "eval_runtime": 1152.0827,
+      "eval_samples_per_second": 0.898,
+      "step": 5000
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 4.941651119402985e-05,
+      "loss": 0.0145,
+      "step": 5004
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 4.9416044776119406e-05,
+      "loss": 0.0094,
+      "step": 5008
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 4.941557835820896e-05,
+      "loss": 0.0242,
+      "step": 5012
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 4.941511194029851e-05,
+      "loss": 0.0222,
+      "step": 5016
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 4.9414645522388064e-05,
+      "loss": 0.0191,
+      "step": 5020
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 4.941417910447761e-05,
+      "loss": 0.0174,
+      "step": 5024
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 4.941371268656717e-05,
+      "loss": 0.0283,
+      "step": 5028
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 4.941324626865672e-05,
+      "loss": 0.0104,
+      "step": 5032
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 4.941277985074627e-05,
+      "loss": 0.0104,
+      "step": 5036
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 4.9412313432835825e-05,
+      "loss": 0.0192,
+      "step": 5040
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 4.941184701492538e-05,
+      "loss": 0.0129,
+      "step": 5044
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 4.941138059701493e-05,
+      "loss": 0.0177,
+      "step": 5048
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 4.9410914179104476e-05,
+      "loss": 0.0157,
+      "step": 5052
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 4.941044776119403e-05,
+      "loss": 0.0169,
+      "step": 5056
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 4.9409981343283586e-05,
+      "loss": 0.0124,
+      "step": 5060
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 4.9409514925373134e-05,
+      "loss": 0.0167,
+      "step": 5064
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 4.940904850746269e-05,
+      "loss": 0.0112,
+      "step": 5068
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 4.9408582089552244e-05,
+      "loss": 0.0228,
+      "step": 5072
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 4.940811567164179e-05,
+      "loss": 0.0211,
+      "step": 5076
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 4.940764925373135e-05,
+      "loss": 0.0306,
+      "step": 5080
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 4.9407182835820895e-05,
+      "loss": 0.0173,
+      "step": 5084
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 4.940671641791045e-05,
+      "loss": 0.0061,
+      "step": 5088
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 4.9406250000000005e-05,
+      "loss": 0.0432,
+      "step": 5092
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 4.940578358208955e-05,
+      "loss": 0.0349,
+      "step": 5096
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 4.940531716417911e-05,
+      "loss": 0.0483,
+      "step": 5100
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 4.940485074626866e-05,
+      "loss": 0.0176,
+      "step": 5104
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 4.940438432835821e-05,
+      "loss": 0.0256,
+      "step": 5108
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 4.940391791044776e-05,
+      "loss": 0.0236,
+      "step": 5112
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 4.9403451492537314e-05,
+      "loss": 0.0167,
+      "step": 5116
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 4.940298507462687e-05,
+      "loss": 0.0294,
+      "step": 5120
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 4.940251865671642e-05,
+      "loss": 0.0245,
+      "step": 5124
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 4.940205223880597e-05,
+      "loss": 0.0169,
+      "step": 5128
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 4.940158582089553e-05,
+      "loss": 0.0202,
+      "step": 5132
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 4.9401119402985075e-05,
+      "loss": 0.0286,
+      "step": 5136
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 4.940065298507463e-05,
+      "loss": 0.0245,
+      "step": 5140
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 4.940018656716418e-05,
+      "loss": 0.0287,
+      "step": 5144
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 4.939972014925374e-05,
+      "loss": 0.0185,
+      "step": 5148
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 4.939925373134329e-05,
+      "loss": 0.0274,
+      "step": 5152
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 4.9398787313432836e-05,
+      "loss": 0.0052,
+      "step": 5156
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 4.939832089552239e-05,
+      "loss": 0.0217,
+      "step": 5160
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 4.9397854477611946e-05,
+      "loss": 0.0214,
+      "step": 5164
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 4.9397388059701494e-05,
+      "loss": 0.0243,
+      "step": 5168
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 4.939692164179105e-05,
+      "loss": 0.0253,
+      "step": 5172
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 4.93964552238806e-05,
+      "loss": 0.013,
+      "step": 5176
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 4.939598880597015e-05,
+      "loss": 0.0193,
+      "step": 5180
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 4.939552238805971e-05,
+      "loss": 0.0098,
+      "step": 5184
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 4.9395055970149255e-05,
+      "loss": 0.0193,
+      "step": 5188
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 4.939458955223881e-05,
+      "loss": 0.0238,
+      "step": 5192
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 4.9394123134328365e-05,
+      "loss": 0.0204,
+      "step": 5196
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 4.939365671641791e-05,
+      "loss": 0.0424,
+      "step": 5200
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 4.939319029850746e-05,
+      "loss": 0.012,
+      "step": 5204
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 4.939272388059702e-05,
+      "loss": 0.0168,
+      "step": 5208
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 4.939225746268657e-05,
+      "loss": 0.0328,
+      "step": 5212
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 4.939179104477612e-05,
+      "loss": 0.0151,
+      "step": 5216
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 4.9391324626865674e-05,
+      "loss": 0.0222,
+      "step": 5220
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 4.939085820895523e-05,
+      "loss": 0.0315,
+      "step": 5224
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 4.939039179104478e-05,
+      "loss": 0.0255,
+      "step": 5228
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 4.938992537313433e-05,
+      "loss": 0.0187,
+      "step": 5232
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 4.938945895522388e-05,
+      "loss": 0.0132,
+      "step": 5236
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 4.9388992537313435e-05,
+      "loss": 0.0296,
+      "step": 5240
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 4.938852611940299e-05,
+      "loss": 0.0223,
+      "step": 5244
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 4.938805970149254e-05,
+      "loss": 0.0189,
+      "step": 5248
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 4.938759328358209e-05,
+      "loss": 0.0232,
+      "step": 5252
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 4.938712686567165e-05,
+      "loss": 0.0243,
+      "step": 5256
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 4.9386660447761196e-05,
+      "loss": 0.0218,
+      "step": 5260
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 4.9386194029850744e-05,
+      "loss": 0.0371,
+      "step": 5264
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 4.9385727611940306e-05,
+      "loss": 0.0128,
+      "step": 5268
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 4.9385261194029854e-05,
+      "loss": 0.0266,
+      "step": 5272
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 4.93847947761194e-05,
+      "loss": 0.0152,
+      "step": 5276
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 4.938432835820896e-05,
+      "loss": 0.019,
+      "step": 5280
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 4.938386194029851e-05,
+      "loss": 0.0055,
+      "step": 5284
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 4.938339552238806e-05,
+      "loss": 0.023,
+      "step": 5288
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 4.9382929104477615e-05,
+      "loss": 0.011,
+      "step": 5292
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 4.938246268656716e-05,
+      "loss": 0.0152,
+      "step": 5296
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 4.938199626865672e-05,
+      "loss": 0.0356,
+      "step": 5300
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 4.938152985074627e-05,
+      "loss": 0.0317,
+      "step": 5304
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 4.938106343283582e-05,
+      "loss": 0.0347,
+      "step": 5308
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 4.9380597014925376e-05,
+      "loss": 0.0067,
+      "step": 5312
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 4.938013059701493e-05,
+      "loss": 0.0202,
+      "step": 5316
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 4.937966417910448e-05,
+      "loss": 0.0275,
+      "step": 5320
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 4.9379197761194034e-05,
+      "loss": 0.0165,
+      "step": 5324
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 4.937873134328358e-05,
+      "loss": 0.019,
+      "step": 5328
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 4.937826492537314e-05,
+      "loss": 0.0157,
+      "step": 5332
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 4.937779850746269e-05,
+      "loss": 0.0303,
+      "step": 5336
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 4.937733208955224e-05,
+      "loss": 0.0178,
+      "step": 5340
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 4.9376865671641795e-05,
+      "loss": 0.0162,
+      "step": 5344
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 4.937639925373135e-05,
+      "loss": 0.0235,
+      "step": 5348
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 4.93759328358209e-05,
+      "loss": 0.0154,
+      "step": 5352
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 4.9375466417910446e-05,
+      "loss": 0.0167,
+      "step": 5356
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 4.937500000000001e-05,
+      "loss": 0.0238,
+      "step": 5360
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 4.9374533582089556e-05,
+      "loss": 0.0116,
+      "step": 5364
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 4.9374067164179104e-05,
+      "loss": 0.0073,
+      "step": 5368
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 4.937360074626866e-05,
+      "loss": 0.0178,
+      "step": 5372
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 4.9373134328358214e-05,
+      "loss": 0.0102,
+      "step": 5376
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.937266791044776e-05,
+      "loss": 0.0109,
+      "step": 5380
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.937220149253732e-05,
+      "loss": 0.0288,
+      "step": 5384
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.9371735074626865e-05,
+      "loss": 0.0073,
+      "step": 5388
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.937126865671642e-05,
+      "loss": 0.0181,
+      "step": 5392
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.9370802238805975e-05,
+      "loss": 0.0159,
+      "step": 5396
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.937033582089552e-05,
+      "loss": 0.0234,
+      "step": 5400
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.936986940298508e-05,
+      "loss": 0.0203,
+      "step": 5404
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.936940298507463e-05,
+      "loss": 0.0132,
+      "step": 5408
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.936893656716418e-05,
+      "loss": 0.0069,
+      "step": 5412
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.936847014925373e-05,
+      "loss": 0.0182,
+      "step": 5416
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.936800373134329e-05,
+      "loss": 0.0064,
+      "step": 5420
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.936753731343284e-05,
+      "loss": 0.0057,
+      "step": 5424
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.936707089552239e-05,
+      "loss": 0.0206,
+      "step": 5428
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.936660447761194e-05,
+      "loss": 0.0148,
+      "step": 5432
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.9366138059701497e-05,
+      "loss": 0.0233,
+      "step": 5436
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.9365671641791045e-05,
+      "loss": 0.0297,
+      "step": 5440
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.93652052238806e-05,
+      "loss": 0.0138,
+      "step": 5444
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.936473880597015e-05,
+      "loss": 0.0223,
+      "step": 5448
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.93642723880597e-05,
+      "loss": 0.0104,
+      "step": 5452
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.936380597014926e-05,
+      "loss": 0.0088,
+      "step": 5456
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.9363339552238806e-05,
+      "loss": 0.0276,
+      "step": 5460
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.936287313432836e-05,
+      "loss": 0.0533,
+      "step": 5464
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.9362406716417916e-05,
+      "loss": 0.0108,
+      "step": 5468
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.9361940298507464e-05,
+      "loss": 0.0145,
+      "step": 5472
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.936147388059702e-05,
+      "loss": 0.0045,
+      "step": 5476
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.9361007462686573e-05,
+      "loss": 0.0249,
+      "step": 5480
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.936054104477612e-05,
+      "loss": 0.0071,
+      "step": 5484
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.9360074626865676e-05,
+      "loss": 0.0101,
+      "step": 5488
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.9359608208955225e-05,
+      "loss": 0.0113,
+      "step": 5492
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.935914179104478e-05,
+      "loss": 0.0135,
+      "step": 5496
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.9358675373134334e-05,
+      "loss": 0.0129,
+      "step": 5500
+    },
+    {
+      "epoch": 5.13,
+      "eval_exact_match": 0.723404255319149,
+      "eval_exec": 0.7562862669245648,
+      "eval_loss": 0.20738151669502258,
+      "eval_runtime": 1207.3594,
+      "eval_samples_per_second": 0.856,
+      "step": 5500
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.935820895522388e-05,
+      "loss": 0.0183,
+      "step": 5504
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.935774253731343e-05,
+      "loss": 0.0204,
+      "step": 5508
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.935727611940299e-05,
+      "loss": 0.0059,
+      "step": 5512
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.935680970149254e-05,
+      "loss": 0.017,
+      "step": 5516
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.935634328358209e-05,
+      "loss": 0.0125,
+      "step": 5520
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.9355876865671644e-05,
+      "loss": 0.022,
+      "step": 5524
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.93554104477612e-05,
+      "loss": 0.0079,
+      "step": 5528
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.9354944029850747e-05,
+      "loss": 0.0082,
+      "step": 5532
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.93544776119403e-05,
+      "loss": 0.0078,
+      "step": 5536
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.9354011194029856e-05,
+      "loss": 0.0172,
+      "step": 5540
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.9353544776119405e-05,
+      "loss": 0.0133,
+      "step": 5544
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.935307835820896e-05,
+      "loss": 0.0171,
+      "step": 5548
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.935261194029851e-05,
+      "loss": 0.0192,
+      "step": 5552
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.935214552238806e-05,
+      "loss": 0.0066,
+      "step": 5556
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.935167910447762e-05,
+      "loss": 0.019,
+      "step": 5560
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.9351212686567166e-05,
+      "loss": 0.0155,
+      "step": 5564
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.9350746268656714e-05,
+      "loss": 0.0426,
+      "step": 5568
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.9350279850746275e-05,
+      "loss": 0.0177,
+      "step": 5572
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.9349813432835823e-05,
+      "loss": 0.0092,
+      "step": 5576
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.934934701492537e-05,
+      "loss": 0.0128,
+      "step": 5580
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.9348880597014927e-05,
+      "loss": 0.0208,
+      "step": 5584
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.934841417910448e-05,
+      "loss": 0.0207,
+      "step": 5588
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.934794776119403e-05,
+      "loss": 0.0251,
+      "step": 5592
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.9347481343283584e-05,
+      "loss": 0.0099,
+      "step": 5596
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.934701492537314e-05,
+      "loss": 0.0138,
+      "step": 5600
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.934654850746269e-05,
+      "loss": 0.0153,
+      "step": 5604
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.934608208955224e-05,
+      "loss": 0.0268,
+      "step": 5608
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.934561567164179e-05,
+      "loss": 0.0149,
+      "step": 5612
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.9345149253731345e-05,
+      "loss": 0.0148,
+      "step": 5616
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.93446828358209e-05,
+      "loss": 0.0114,
+      "step": 5620
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.934421641791045e-05,
+      "loss": 0.0064,
+      "step": 5624
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.9343749999999997e-05,
+      "loss": 0.0176,
+      "step": 5628
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.934328358208956e-05,
+      "loss": 0.0175,
+      "step": 5632
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.9342817164179106e-05,
+      "loss": 0.028,
+      "step": 5636
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.934235074626866e-05,
+      "loss": 0.0124,
+      "step": 5640
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.934188432835821e-05,
+      "loss": 0.0196,
+      "step": 5644
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.9341417910447764e-05,
+      "loss": 0.0129,
+      "step": 5648
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.934095149253732e-05,
+      "loss": 0.0173,
+      "step": 5652
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.934048507462687e-05,
+      "loss": 0.042,
+      "step": 5656
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.934001865671642e-05,
+      "loss": 0.0103,
+      "step": 5660
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.933955223880598e-05,
+      "loss": 0.0108,
+      "step": 5664
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.9339085820895525e-05,
+      "loss": 0.0123,
+      "step": 5668
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.9338619402985073e-05,
+      "loss": 0.0061,
+      "step": 5672
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.933815298507463e-05,
+      "loss": 0.0136,
+      "step": 5676
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.933768656716418e-05,
+      "loss": 0.0152,
+      "step": 5680
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.933722014925373e-05,
+      "loss": 0.0209,
+      "step": 5684
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.9336753731343286e-05,
+      "loss": 0.0259,
+      "step": 5688
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.933628731343284e-05,
+      "loss": 0.0155,
+      "step": 5692
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.933582089552239e-05,
+      "loss": 0.0257,
+      "step": 5696
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.9335354477611944e-05,
+      "loss": 0.0359,
+      "step": 5700
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.933488805970149e-05,
+      "loss": 0.0091,
+      "step": 5704
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.933442164179105e-05,
+      "loss": 0.021,
+      "step": 5708
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.93339552238806e-05,
+      "loss": 0.0175,
+      "step": 5712
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.933348880597015e-05,
+      "loss": 0.0134,
+      "step": 5716
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.9333022388059705e-05,
+      "loss": 0.0152,
+      "step": 5720
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.933255597014926e-05,
+      "loss": 0.014,
+      "step": 5724
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.933208955223881e-05,
+      "loss": 0.0205,
+      "step": 5728
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.9331623134328356e-05,
+      "loss": 0.0145,
+      "step": 5732
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.933115671641791e-05,
+      "loss": 0.0054,
+      "step": 5736
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.9330690298507466e-05,
+      "loss": 0.0197,
+      "step": 5740
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.9330223880597014e-05,
+      "loss": 0.0167,
+      "step": 5744
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.932975746268657e-05,
+      "loss": 0.0326,
+      "step": 5748
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.9329291044776124e-05,
+      "loss": 0.0252,
+      "step": 5752
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.932882462686567e-05,
+      "loss": 0.0169,
+      "step": 5756
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.932835820895523e-05,
+      "loss": 0.0097,
+      "step": 5760
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.9327891791044775e-05,
+      "loss": 0.0059,
+      "step": 5764
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.932742537313433e-05,
+      "loss": 0.0128,
+      "step": 5768
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.9326958955223885e-05,
+      "loss": 0.012,
+      "step": 5772
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.932649253731343e-05,
+      "loss": 0.0202,
+      "step": 5776
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.932602611940299e-05,
+      "loss": 0.0112,
+      "step": 5780
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.932555970149254e-05,
+      "loss": 0.0114,
+      "step": 5784
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.932509328358209e-05,
+      "loss": 0.0085,
+      "step": 5788
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.932462686567164e-05,
+      "loss": 0.0115,
+      "step": 5792
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.9324160447761194e-05,
+      "loss": 0.0318,
+      "step": 5796
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.932369402985075e-05,
+      "loss": 0.032,
+      "step": 5800
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.9323227611940304e-05,
+      "loss": 0.015,
+      "step": 5804
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.932276119402985e-05,
+      "loss": 0.0194,
+      "step": 5808
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.932229477611941e-05,
+      "loss": 0.017,
+      "step": 5812
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.932182835820896e-05,
+      "loss": 0.0081,
+      "step": 5816
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.932136194029851e-05,
+      "loss": 0.0084,
+      "step": 5820
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.932089552238806e-05,
+      "loss": 0.0069,
+      "step": 5824
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.932042910447762e-05,
+      "loss": 0.0175,
+      "step": 5828
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.931996268656717e-05,
+      "loss": 0.021,
+      "step": 5832
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.9319496268656716e-05,
+      "loss": 0.0136,
+      "step": 5836
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.931902985074627e-05,
+      "loss": 0.012,
+      "step": 5840
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.9318563432835826e-05,
+      "loss": 0.0068,
+      "step": 5844
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.9318097014925374e-05,
+      "loss": 0.0151,
+      "step": 5848
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.931763059701493e-05,
+      "loss": 0.0189,
+      "step": 5852
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.931716417910448e-05,
+      "loss": 0.0198,
+      "step": 5856
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.931669776119403e-05,
+      "loss": 0.0086,
+      "step": 5860
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.931623134328359e-05,
+      "loss": 0.0126,
+      "step": 5864
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.9315764925373135e-05,
+      "loss": 0.0093,
+      "step": 5868
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.931529850746269e-05,
+      "loss": 0.0232,
+      "step": 5872
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.9314832089552245e-05,
+      "loss": 0.0138,
+      "step": 5876
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.931436567164179e-05,
+      "loss": 0.0332,
+      "step": 5880
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.931389925373134e-05,
+      "loss": 0.0102,
+      "step": 5884
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.93134328358209e-05,
+      "loss": 0.015,
+      "step": 5888
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.931296641791045e-05,
+      "loss": 0.0099,
+      "step": 5892
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.93125e-05,
+      "loss": 0.0109,
+      "step": 5896
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.9312033582089554e-05,
+      "loss": 0.0084,
+      "step": 5900
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.931156716417911e-05,
+      "loss": 0.0145,
+      "step": 5904
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.931110074626866e-05,
+      "loss": 0.0085,
+      "step": 5908
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.931063432835821e-05,
+      "loss": 0.0101,
+      "step": 5912
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.931016791044776e-05,
+      "loss": 0.0131,
+      "step": 5916
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.9309701492537315e-05,
+      "loss": 0.021,
+      "step": 5920
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.930923507462687e-05,
+      "loss": 0.0113,
+      "step": 5924
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.930876865671642e-05,
+      "loss": 0.0208,
+      "step": 5928
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.930830223880597e-05,
+      "loss": 0.0187,
+      "step": 5932
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.930783582089553e-05,
+      "loss": 0.0216,
+      "step": 5936
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.9307369402985076e-05,
+      "loss": 0.0101,
+      "step": 5940
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.9306902985074624e-05,
+      "loss": 0.0108,
+      "step": 5944
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.9306436567164186e-05,
+      "loss": 0.0206,
+      "step": 5948
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.9305970149253734e-05,
+      "loss": 0.0134,
+      "step": 5952
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.930550373134328e-05,
+      "loss": 0.0075,
+      "step": 5956
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.930503731343284e-05,
+      "loss": 0.0154,
+      "step": 5960
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.930457089552239e-05,
+      "loss": 0.0136,
+      "step": 5964
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.930410447761195e-05,
+      "loss": 0.011,
+      "step": 5968
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.9303638059701495e-05,
+      "loss": 0.0114,
+      "step": 5972
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.930317164179104e-05,
+      "loss": 0.0133,
+      "step": 5976
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.9302705223880605e-05,
+      "loss": 0.0109,
+      "step": 5980
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.930223880597015e-05,
+      "loss": 0.0084,
+      "step": 5984
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.93017723880597e-05,
+      "loss": 0.0147,
+      "step": 5988
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.9301305970149256e-05,
+      "loss": 0.0139,
+      "step": 5992
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.930083955223881e-05,
+      "loss": 0.009,
+      "step": 5996
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.930037313432836e-05,
+      "loss": 0.0127,
+      "step": 6000
+    },
+    {
+      "epoch": 5.6,
+      "eval_exact_match": 0.7224371373307543,
+      "eval_exec": 0.7446808510638298,
+      "eval_loss": 0.20770786702632904,
+      "eval_runtime": 1400.4625,
+      "eval_samples_per_second": 0.738,
+      "step": 6000
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.9299906716417914e-05,
+      "loss": 0.0203,
+      "step": 6004
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.929944029850746e-05,
+      "loss": 0.0078,
+      "step": 6008
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.929897388059702e-05,
+      "loss": 0.0127,
+      "step": 6012
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.929850746268657e-05,
+      "loss": 0.0099,
+      "step": 6016
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.929804104477612e-05,
+      "loss": 0.0149,
+      "step": 6020
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.9297574626865675e-05,
+      "loss": 0.0142,
+      "step": 6024
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.929710820895523e-05,
+      "loss": 0.0129,
+      "step": 6028
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.929664179104478e-05,
+      "loss": 0.015,
+      "step": 6032
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.9296175373134326e-05,
+      "loss": 0.0056,
+      "step": 6036
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.929570895522389e-05,
+      "loss": 0.0085,
+      "step": 6040
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.9295242537313436e-05,
+      "loss": 0.0157,
+      "step": 6044
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.9294776119402984e-05,
+      "loss": 0.0248,
+      "step": 6048
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.929430970149254e-05,
+      "loss": 0.0234,
+      "step": 6052
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.9293843283582094e-05,
+      "loss": 0.0103,
+      "step": 6056
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.929337686567164e-05,
+      "loss": 0.0106,
+      "step": 6060
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.92929104477612e-05,
+      "loss": 0.0075,
+      "step": 6064
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.9292444029850745e-05,
+      "loss": 0.0082,
+      "step": 6068
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.92919776119403e-05,
+      "loss": 0.0289,
+      "step": 6072
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.9291511194029855e-05,
+      "loss": 0.0079,
+      "step": 6076
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.92910447761194e-05,
+      "loss": 0.0148,
+      "step": 6080
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.929057835820896e-05,
+      "loss": 0.0072,
+      "step": 6084
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.929011194029851e-05,
+      "loss": 0.0087,
+      "step": 6088
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.928964552238806e-05,
+      "loss": 0.0075,
+      "step": 6092
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.928917910447761e-05,
+      "loss": 0.0202,
+      "step": 6096
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.928871268656717e-05,
+      "loss": 0.0076,
+      "step": 6100
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.928824626865672e-05,
+      "loss": 0.0187,
+      "step": 6104
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.928777985074627e-05,
+      "loss": 0.0158,
+      "step": 6108
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.928731343283582e-05,
+      "loss": 0.0142,
+      "step": 6112
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.928684701492538e-05,
+      "loss": 0.0116,
+      "step": 6116
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.9286380597014925e-05,
+      "loss": 0.0218,
+      "step": 6120
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.928591417910448e-05,
+      "loss": 0.0132,
+      "step": 6124
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.928544776119403e-05,
+      "loss": 0.0084,
+      "step": 6128
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.928498134328359e-05,
+      "loss": 0.0063,
+      "step": 6132
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.928451492537314e-05,
+      "loss": 0.008,
+      "step": 6136
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.9284048507462686e-05,
+      "loss": 0.0157,
+      "step": 6140
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.928358208955224e-05,
+      "loss": 0.0041,
+      "step": 6144
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.9283115671641796e-05,
+      "loss": 0.0168,
+      "step": 6148
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.9282649253731344e-05,
+      "loss": 0.039,
+      "step": 6152
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.92821828358209e-05,
+      "loss": 0.0105,
+      "step": 6156
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.9281716417910454e-05,
+      "loss": 0.0158,
+      "step": 6160
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.928125e-05,
+      "loss": 0.0184,
+      "step": 6164
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.928078358208956e-05,
+      "loss": 0.01,
+      "step": 6168
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.9280317164179105e-05,
+      "loss": 0.0229,
+      "step": 6172
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.927985074626866e-05,
+      "loss": 0.0083,
+      "step": 6176
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.9279384328358215e-05,
+      "loss": 0.016,
+      "step": 6180
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.927891791044776e-05,
+      "loss": 0.0097,
+      "step": 6184
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.927845149253731e-05,
+      "loss": 0.0119,
+      "step": 6188
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.927798507462687e-05,
+      "loss": 0.0089,
+      "step": 6192
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.927751865671642e-05,
+      "loss": 0.0472,
+      "step": 6196
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.927705223880597e-05,
+      "loss": 0.0154,
+      "step": 6200
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.9276585820895524e-05,
+      "loss": 0.0054,
+      "step": 6204
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.927611940298508e-05,
+      "loss": 0.0133,
+      "step": 6208
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.927565298507463e-05,
+      "loss": 0.0163,
+      "step": 6212
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.927518656716418e-05,
+      "loss": 0.0296,
+      "step": 6216
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.927472014925374e-05,
+      "loss": 0.0075,
+      "step": 6220
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.9274253731343285e-05,
+      "loss": 0.0128,
+      "step": 6224
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.927378731343284e-05,
+      "loss": 0.0127,
+      "step": 6228
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.927332089552239e-05,
+      "loss": 0.0246,
+      "step": 6232
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.927285447761194e-05,
+      "loss": 0.0223,
+      "step": 6236
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.92723880597015e-05,
+      "loss": 0.0312,
+      "step": 6240
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.9271921641791046e-05,
+      "loss": 0.0177,
+      "step": 6244
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.9271455223880594e-05,
+      "loss": 0.0184,
+      "step": 6248
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.9270988805970156e-05,
+      "loss": 0.0151,
+      "step": 6252
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.9270522388059704e-05,
+      "loss": 0.0263,
+      "step": 6256
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.927005597014925e-05,
+      "loss": 0.0185,
+      "step": 6260
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.926958955223881e-05,
+      "loss": 0.0235,
+      "step": 6264
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.926912313432836e-05,
+      "loss": 0.0132,
+      "step": 6268
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.926865671641791e-05,
+      "loss": 0.0167,
+      "step": 6272
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.9268190298507465e-05,
+      "loss": 0.0097,
+      "step": 6276
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.926772388059702e-05,
+      "loss": 0.0157,
+      "step": 6280
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.926725746268657e-05,
+      "loss": 0.0161,
+      "step": 6284
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.926679104477612e-05,
+      "loss": 0.0129,
+      "step": 6288
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.926632462686567e-05,
+      "loss": 0.0307,
+      "step": 6292
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.9265858208955226e-05,
+      "loss": 0.0091,
+      "step": 6296
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.926539179104478e-05,
+      "loss": 0.0108,
+      "step": 6300
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.926492537313433e-05,
+      "loss": 0.0164,
+      "step": 6304
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.9264458955223884e-05,
+      "loss": 0.0067,
+      "step": 6308
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.926399253731344e-05,
+      "loss": 0.0078,
+      "step": 6312
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.926352611940299e-05,
+      "loss": 0.0118,
+      "step": 6316
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.926305970149254e-05,
+      "loss": 0.0174,
+      "step": 6320
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.926259328358209e-05,
+      "loss": 0.0226,
+      "step": 6324
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.9262126865671645e-05,
+      "loss": 0.013,
+      "step": 6328
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.92616604477612e-05,
+      "loss": 0.0241,
+      "step": 6332
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.926119402985075e-05,
+      "loss": 0.0154,
+      "step": 6336
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.92607276119403e-05,
+      "loss": 0.0339,
+      "step": 6340
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.926026119402986e-05,
+      "loss": 0.0147,
+      "step": 6344
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.9259794776119406e-05,
+      "loss": 0.0203,
+      "step": 6348
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.9259328358208954e-05,
+      "loss": 0.008,
+      "step": 6352
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.925886194029851e-05,
+      "loss": 0.0371,
+      "step": 6356
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.9258395522388064e-05,
+      "loss": 0.0161,
+      "step": 6360
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.925792910447761e-05,
+      "loss": 0.02,
+      "step": 6364
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.925746268656717e-05,
+      "loss": 0.0117,
+      "step": 6368
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.925699626865672e-05,
+      "loss": 0.0057,
+      "step": 6372
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.925652985074627e-05,
+      "loss": 0.0157,
+      "step": 6376
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.9256063432835825e-05,
+      "loss": 0.0115,
+      "step": 6380
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.925559701492537e-05,
+      "loss": 0.0082,
+      "step": 6384
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.925513059701493e-05,
+      "loss": 0.009,
+      "step": 6388
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.925466417910448e-05,
+      "loss": 0.0169,
+      "step": 6392
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.925419776119403e-05,
+      "loss": 0.0204,
+      "step": 6396
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.9253731343283586e-05,
+      "loss": 0.0356,
+      "step": 6400
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.925326492537314e-05,
+      "loss": 0.0079,
+      "step": 6404
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.925279850746269e-05,
+      "loss": 0.008,
+      "step": 6408
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.925233208955224e-05,
+      "loss": 0.01,
+      "step": 6412
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.925186567164179e-05,
+      "loss": 0.0309,
+      "step": 6416
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.9251399253731347e-05,
+      "loss": 0.023,
+      "step": 6420
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.9250932835820895e-05,
+      "loss": 0.0079,
+      "step": 6424
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.925046641791045e-05,
+      "loss": 0.0145,
+      "step": 6428
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.9250000000000004e-05,
+      "loss": 0.0435,
+      "step": 6432
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.924953358208955e-05,
+      "loss": 0.0114,
+      "step": 6436
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 4.924906716417911e-05,
+      "loss": 0.0078,
+      "step": 6440
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 4.9248600746268656e-05,
+      "loss": 0.0081,
+      "step": 6444
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 4.924813432835821e-05,
+      "loss": 0.013,
+      "step": 6448
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 4.9247667910447765e-05,
+      "loss": 0.0108,
+      "step": 6452
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 4.9247201492537314e-05,
+      "loss": 0.0088,
+      "step": 6456
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 4.924673507462687e-05,
+      "loss": 0.0064,
+      "step": 6460
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 4.9246268656716423e-05,
+      "loss": 0.0124,
+      "step": 6464
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 4.924580223880597e-05,
+      "loss": 0.0052,
+      "step": 6468
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 4.9245335820895526e-05,
+      "loss": 0.0062,
+      "step": 6472
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 4.9244869402985075e-05,
+      "loss": 0.0173,
+      "step": 6476
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 4.924440298507463e-05,
+      "loss": 0.014,
+      "step": 6480
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 4.9243936567164184e-05,
+      "loss": 0.0144,
+      "step": 6484
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 4.924347014925373e-05,
+      "loss": 0.0037,
+      "step": 6488
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 4.924300373134329e-05,
+      "loss": 0.0062,
+      "step": 6492
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 4.924253731343284e-05,
+      "loss": 0.0095,
+      "step": 6496
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 4.924207089552239e-05,
+      "loss": 0.0078,
+      "step": 6500
+    },
+    {
+      "epoch": 6.06,
+      "eval_exact_match": 0.7195357833655706,
+      "eval_exec": 0.741779497098646,
+      "eval_loss": 0.2259366363286972,
+      "eval_runtime": 1072.2544,
+      "eval_samples_per_second": 0.964,
+      "step": 6500
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 4.924160447761194e-05,
+      "loss": 0.0215,
+      "step": 6504
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 4.92411380597015e-05,
+      "loss": 0.005,
+      "step": 6508
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 4.924067164179105e-05,
+      "loss": 0.0141,
+      "step": 6512
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 4.9240205223880597e-05,
+      "loss": 0.0077,
+      "step": 6516
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 4.923973880597015e-05,
+      "loss": 0.0067,
+      "step": 6520
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 4.9239272388059706e-05,
+      "loss": 0.0072,
+      "step": 6524
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 4.9238805970149255e-05,
+      "loss": 0.0193,
+      "step": 6528
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 4.923833955223881e-05,
+      "loss": 0.0103,
+      "step": 6532
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 4.923787313432836e-05,
+      "loss": 0.0157,
+      "step": 6536
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 4.923740671641791e-05,
+      "loss": 0.0103,
+      "step": 6540
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 4.923694029850747e-05,
+      "loss": 0.0052,
+      "step": 6544
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 4.9236473880597016e-05,
+      "loss": 0.0028,
+      "step": 6548
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 4.923600746268657e-05,
+      "loss": 0.0065,
+      "step": 6552
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 4.9235541044776125e-05,
+      "loss": 0.0114,
+      "step": 6556
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 4.9235074626865673e-05,
+      "loss": 0.0147,
+      "step": 6560
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 4.923460820895522e-05,
+      "loss": 0.0049,
+      "step": 6564
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 4.923414179104478e-05,
+      "loss": 0.0079,
+      "step": 6568
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 4.923367537313433e-05,
+      "loss": 0.0099,
+      "step": 6572
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 4.923320895522388e-05,
+      "loss": 0.0079,
+      "step": 6576
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 4.9232742537313434e-05,
+      "loss": 0.0042,
+      "step": 6580
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 4.923227611940299e-05,
+      "loss": 0.0112,
+      "step": 6584
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 4.923180970149254e-05,
+      "loss": 0.006,
+      "step": 6588
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 4.923134328358209e-05,
+      "loss": 0.0107,
+      "step": 6592
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 4.923087686567164e-05,
+      "loss": 0.0112,
+      "step": 6596
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 4.9230410447761195e-05,
+      "loss": 0.0251,
+      "step": 6600
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 4.922994402985075e-05,
+      "loss": 0.023,
+      "step": 6604
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 4.92294776119403e-05,
+      "loss": 0.016,
+      "step": 6608
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 4.922901119402985e-05,
+      "loss": 0.0044,
+      "step": 6612
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 4.922854477611941e-05,
+      "loss": 0.018,
+      "step": 6616
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 4.9228078358208956e-05,
+      "loss": 0.0172,
+      "step": 6620
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 4.922761194029851e-05,
+      "loss": 0.0419,
+      "step": 6624
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 4.9227145522388066e-05,
+      "loss": 0.0044,
+      "step": 6628
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 4.9226679104477614e-05,
+      "loss": 0.0193,
+      "step": 6632
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 4.922621268656717e-05,
+      "loss": 0.0122,
+      "step": 6636
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 4.922574626865672e-05,
+      "loss": 0.009,
+      "step": 6640
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 4.922527985074627e-05,
+      "loss": 0.0184,
+      "step": 6644
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 4.922481343283583e-05,
+      "loss": 0.0084,
+      "step": 6648
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 4.9224347014925375e-05,
+      "loss": 0.0035,
+      "step": 6652
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 4.9223880597014923e-05,
+      "loss": 0.0089,
+      "step": 6656
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 4.9223414179104485e-05,
+      "loss": 0.0073,
+      "step": 6660
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 4.922294776119403e-05,
+      "loss": 0.0126,
+      "step": 6664
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 4.922248134328358e-05,
+      "loss": 0.0028,
+      "step": 6668
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 4.9222014925373136e-05,
+      "loss": 0.0066,
+      "step": 6672
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 4.922154850746269e-05,
+      "loss": 0.0041,
+      "step": 6676
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 4.922108208955224e-05,
+      "loss": 0.0064,
+      "step": 6680
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 4.9220615671641794e-05,
+      "loss": 0.0171,
+      "step": 6684
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 4.922014925373134e-05,
+      "loss": 0.0186,
+      "step": 6688
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 4.92196828358209e-05,
+      "loss": 0.0028,
+      "step": 6692
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 4.921921641791045e-05,
+      "loss": 0.0133,
+      "step": 6696
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 4.921875e-05,
+      "loss": 0.0054,
+      "step": 6700
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 4.9218283582089555e-05,
+      "loss": 0.0071,
+      "step": 6704
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 4.921781716417911e-05,
+      "loss": 0.0216,
+      "step": 6708
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 4.921735074626866e-05,
+      "loss": 0.0213,
+      "step": 6712
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 4.9216884328358206e-05,
+      "loss": 0.0134,
+      "step": 6716
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 4.921641791044777e-05,
+      "loss": 0.0077,
+      "step": 6720
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 4.9215951492537316e-05,
+      "loss": 0.0022,
+      "step": 6724
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 4.9215485074626864e-05,
+      "loss": 0.0227,
+      "step": 6728
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 4.921501865671642e-05,
+      "loss": 0.0049,
+      "step": 6732
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 4.9214552238805974e-05,
+      "loss": 0.0083,
+      "step": 6736
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 4.921408582089552e-05,
+      "loss": 0.0151,
+      "step": 6740
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 4.921361940298508e-05,
+      "loss": 0.0093,
+      "step": 6744
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 4.9213152985074625e-05,
+      "loss": 0.0131,
+      "step": 6748
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 4.921268656716418e-05,
+      "loss": 0.0173,
+      "step": 6752
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 4.9212220149253735e-05,
+      "loss": 0.0198,
+      "step": 6756
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 4.921175373134328e-05,
+      "loss": 0.014,
+      "step": 6760
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 4.921128731343284e-05,
+      "loss": 0.0136,
+      "step": 6764
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 4.921082089552239e-05,
+      "loss": 0.011,
+      "step": 6768
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 4.921035447761194e-05,
+      "loss": 0.0175,
+      "step": 6772
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 4.920988805970149e-05,
+      "loss": 0.0123,
+      "step": 6776
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 4.920942164179105e-05,
+      "loss": 0.0137,
+      "step": 6780
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 4.92089552238806e-05,
+      "loss": 0.0129,
+      "step": 6784
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 4.9208488805970154e-05,
+      "loss": 0.0116,
+      "step": 6788
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 4.92080223880597e-05,
+      "loss": 0.0142,
+      "step": 6792
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 4.920755597014926e-05,
+      "loss": 0.0161,
+      "step": 6796
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 4.920708955223881e-05,
+      "loss": 0.0078,
+      "step": 6800
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 4.920662313432836e-05,
+      "loss": 0.0183,
+      "step": 6804
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 4.920615671641791e-05,
+      "loss": 0.0138,
+      "step": 6808
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 4.920569029850747e-05,
+      "loss": 0.0128,
+      "step": 6812
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 4.920522388059702e-05,
+      "loss": 0.0105,
+      "step": 6816
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 4.9204757462686566e-05,
+      "loss": 0.0246,
+      "step": 6820
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 4.920429104477612e-05,
+      "loss": 0.0076,
+      "step": 6824
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 4.9203824626865676e-05,
+      "loss": 0.0074,
+      "step": 6828
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 4.9203358208955224e-05,
+      "loss": 0.0065,
+      "step": 6832
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 4.920289179104478e-05,
+      "loss": 0.0156,
+      "step": 6836
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 4.9202425373134334e-05,
+      "loss": 0.0128,
+      "step": 6840
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 4.920195895522388e-05,
+      "loss": 0.015,
+      "step": 6844
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 4.920149253731344e-05,
+      "loss": 0.0054,
+      "step": 6848
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 4.9201026119402985e-05,
+      "loss": 0.0092,
+      "step": 6852
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 4.920055970149254e-05,
+      "loss": 0.0159,
+      "step": 6856
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 4.9200093283582095e-05,
+      "loss": 0.0132,
+      "step": 6860
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 4.919962686567164e-05,
+      "loss": 0.0106,
+      "step": 6864
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 4.919916044776119e-05,
+      "loss": 0.0067,
+      "step": 6868
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 4.919869402985075e-05,
+      "loss": 0.0088,
+      "step": 6872
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 4.91982276119403e-05,
+      "loss": 0.0098,
+      "step": 6876
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 4.919776119402985e-05,
+      "loss": 0.0177,
+      "step": 6880
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 4.9197294776119404e-05,
+      "loss": 0.0097,
+      "step": 6884
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 4.919682835820896e-05,
+      "loss": 0.0176,
+      "step": 6888
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 4.919636194029851e-05,
+      "loss": 0.0085,
+      "step": 6892
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 4.919589552238806e-05,
+      "loss": 0.0129,
+      "step": 6896
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 4.919542910447762e-05,
+      "loss": 0.0131,
+      "step": 6900
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 4.9194962686567165e-05,
+      "loss": 0.0155,
+      "step": 6904
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 4.919449626865672e-05,
+      "loss": 0.009,
+      "step": 6908
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 4.919402985074627e-05,
+      "loss": 0.0105,
+      "step": 6912
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 4.919356343283582e-05,
+      "loss": 0.013,
+      "step": 6916
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 4.919309701492538e-05,
+      "loss": 0.0181,
+      "step": 6920
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 4.9192630597014926e-05,
+      "loss": 0.0118,
+      "step": 6924
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 4.9192164179104474e-05,
+      "loss": 0.0109,
+      "step": 6928
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 4.9191697761194036e-05,
+      "loss": 0.011,
+      "step": 6932
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 4.9191231343283584e-05,
+      "loss": 0.0091,
+      "step": 6936
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 4.919076492537313e-05,
+      "loss": 0.0233,
+      "step": 6940
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 4.919029850746269e-05,
+      "loss": 0.0207,
+      "step": 6944
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 4.918983208955224e-05,
+      "loss": 0.0153,
+      "step": 6948
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 4.91893656716418e-05,
+      "loss": 0.0155,
+      "step": 6952
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 4.9188899253731345e-05,
+      "loss": 0.015,
+      "step": 6956
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 4.91884328358209e-05,
+      "loss": 0.0072,
+      "step": 6960
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 4.9187966417910455e-05,
+      "loss": 0.0077,
+      "step": 6964
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 4.91875e-05,
+      "loss": 0.017,
+      "step": 6968
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 4.918703358208955e-05,
+      "loss": 0.0154,
+      "step": 6972
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 4.9186567164179106e-05,
+      "loss": 0.0102,
+      "step": 6976
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 4.918610074626866e-05,
+      "loss": 0.0213,
+      "step": 6980
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 4.918563432835821e-05,
+      "loss": 0.0095,
+      "step": 6984
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 4.9185167910447764e-05,
+      "loss": 0.0187,
+      "step": 6988
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 4.918470149253732e-05,
+      "loss": 0.0093,
+      "step": 6992
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 4.918423507462687e-05,
+      "loss": 0.0095,
+      "step": 6996
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 4.918376865671642e-05,
+      "loss": 0.0084,
+      "step": 7000
+    },
+    {
+      "epoch": 6.53,
+      "eval_exact_match": 0.7030947775628626,
+      "eval_exec": 0.7504835589941973,
+      "eval_loss": 0.20910172164440155,
+      "eval_runtime": 1063.5178,
+      "eval_samples_per_second": 0.972,
+      "step": 7000
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 4.918330223880597e-05,
+      "loss": 0.0181,
+      "step": 7004
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 4.9182835820895525e-05,
+      "loss": 0.011,
+      "step": 7008
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 4.918236940298508e-05,
+      "loss": 0.0015,
+      "step": 7012
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 4.918190298507463e-05,
+      "loss": 0.0112,
+      "step": 7016
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 4.918143656716418e-05,
+      "loss": 0.0163,
+      "step": 7020
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 4.918097014925374e-05,
+      "loss": 0.0113,
+      "step": 7024
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 4.9180503731343286e-05,
+      "loss": 0.0139,
+      "step": 7028
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 4.9180037313432834e-05,
+      "loss": 0.019,
+      "step": 7032
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 4.917957089552239e-05,
+      "loss": 0.0094,
+      "step": 7036
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 4.9179104477611944e-05,
+      "loss": 0.0109,
+      "step": 7040
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 4.917863805970149e-05,
+      "loss": 0.0132,
+      "step": 7044
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 4.917817164179105e-05,
+      "loss": 0.0169,
+      "step": 7048
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 4.91777052238806e-05,
+      "loss": 0.0235,
+      "step": 7052
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 4.917723880597015e-05,
+      "loss": 0.0129,
+      "step": 7056
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 4.9176772388059705e-05,
+      "loss": 0.0266,
+      "step": 7060
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 4.917630597014925e-05,
+      "loss": 0.0084,
+      "step": 7064
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 4.917583955223881e-05,
+      "loss": 0.0167,
+      "step": 7068
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 4.917537313432836e-05,
+      "loss": 0.0045,
+      "step": 7072
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 4.917490671641791e-05,
+      "loss": 0.0207,
+      "step": 7076
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 4.9174440298507466e-05,
+      "loss": 0.0144,
+      "step": 7080
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 4.917397388059702e-05,
+      "loss": 0.0044,
+      "step": 7084
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 4.917350746268657e-05,
+      "loss": 0.0197,
+      "step": 7088
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 4.917304104477612e-05,
+      "loss": 0.0137,
+      "step": 7092
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 4.917257462686567e-05,
+      "loss": 0.0139,
+      "step": 7096
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 4.917210820895523e-05,
+      "loss": 0.0188,
+      "step": 7100
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 4.9171641791044775e-05,
+      "loss": 0.0072,
+      "step": 7104
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 4.917117537313433e-05,
+      "loss": 0.01,
+      "step": 7108
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 4.9170708955223885e-05,
+      "loss": 0.0185,
+      "step": 7112
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 4.917024253731344e-05,
+      "loss": 0.005,
+      "step": 7116
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 4.916977611940299e-05,
+      "loss": 0.0143,
+      "step": 7120
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 4.9169309701492536e-05,
+      "loss": 0.0176,
+      "step": 7124
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 4.91688432835821e-05,
+      "loss": 0.0129,
+      "step": 7128
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 4.9168376865671646e-05,
+      "loss": 0.0038,
+      "step": 7132
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 4.9167910447761194e-05,
+      "loss": 0.0117,
+      "step": 7136
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 4.916744402985075e-05,
+      "loss": 0.0063,
+      "step": 7140
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 4.9166977611940304e-05,
+      "loss": 0.0198,
+      "step": 7144
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 4.916651119402985e-05,
+      "loss": 0.0205,
+      "step": 7148
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 4.916604477611941e-05,
+      "loss": 0.0097,
+      "step": 7152
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 4.9165578358208955e-05,
+      "loss": 0.0118,
+      "step": 7156
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 4.916511194029851e-05,
+      "loss": 0.0137,
+      "step": 7160
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 4.9164645522388065e-05,
+      "loss": 0.0047,
+      "step": 7164
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 4.916417910447761e-05,
+      "loss": 0.0149,
+      "step": 7168
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 4.916371268656717e-05,
+      "loss": 0.0069,
+      "step": 7172
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 4.916324626865672e-05,
+      "loss": 0.0114,
+      "step": 7176
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 4.916277985074627e-05,
+      "loss": 0.0087,
+      "step": 7180
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 4.916231343283582e-05,
+      "loss": 0.0021,
+      "step": 7184
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 4.916184701492538e-05,
+      "loss": 0.0121,
+      "step": 7188
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 4.916138059701493e-05,
+      "loss": 0.0042,
+      "step": 7192
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 4.916091417910448e-05,
+      "loss": 0.0052,
+      "step": 7196
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 4.916044776119403e-05,
+      "loss": 0.0108,
+      "step": 7200
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 4.915998134328359e-05,
+      "loss": 0.0175,
+      "step": 7204
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 4.9159514925373135e-05,
+      "loss": 0.0045,
+      "step": 7208
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 4.915904850746269e-05,
+      "loss": 0.0068,
+      "step": 7212
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 4.915858208955224e-05,
+      "loss": 0.0096,
+      "step": 7216
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 4.915811567164179e-05,
+      "loss": 0.0232,
+      "step": 7220
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 4.915764925373135e-05,
+      "loss": 0.0145,
+      "step": 7224
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 4.9157182835820896e-05,
+      "loss": 0.0063,
+      "step": 7228
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 4.915671641791045e-05,
+      "loss": 0.0228,
+      "step": 7232
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 4.9156250000000006e-05,
+      "loss": 0.0146,
+      "step": 7236
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 4.9155783582089554e-05,
+      "loss": 0.0097,
+      "step": 7240
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 4.91553171641791e-05,
+      "loss": 0.0089,
+      "step": 7244
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 4.9154850746268664e-05,
+      "loss": 0.0091,
+      "step": 7248
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 4.915438432835821e-05,
+      "loss": 0.0064,
+      "step": 7252
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 4.915391791044776e-05,
+      "loss": 0.0121,
+      "step": 7256
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 4.9153451492537315e-05,
+      "loss": 0.0026,
+      "step": 7260
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 4.915298507462687e-05,
+      "loss": 0.0141,
+      "step": 7264
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 4.915251865671642e-05,
+      "loss": 0.004,
+      "step": 7268
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 4.915205223880597e-05,
+      "loss": 0.0124,
+      "step": 7272
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 4.915158582089552e-05,
+      "loss": 0.0115,
+      "step": 7276
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 4.915111940298508e-05,
+      "loss": 0.0193,
+      "step": 7280
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 4.915065298507463e-05,
+      "loss": 0.0088,
+      "step": 7284
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 4.915018656716418e-05,
+      "loss": 0.0076,
+      "step": 7288
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 4.9149720149253734e-05,
+      "loss": 0.0137,
+      "step": 7292
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 4.914925373134329e-05,
+      "loss": 0.0108,
+      "step": 7296
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 4.914878731343284e-05,
+      "loss": 0.008,
+      "step": 7300
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 4.914832089552239e-05,
+      "loss": 0.0146,
+      "step": 7304
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 4.9147854477611947e-05,
+      "loss": 0.0134,
+      "step": 7308
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 4.9147388059701495e-05,
+      "loss": 0.0099,
+      "step": 7312
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 4.914692164179105e-05,
+      "loss": 0.0079,
+      "step": 7316
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 4.91464552238806e-05,
+      "loss": 0.0115,
+      "step": 7320
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 4.914598880597015e-05,
+      "loss": 0.0088,
+      "step": 7324
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 4.914552238805971e-05,
+      "loss": 0.0175,
+      "step": 7328
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 4.9145055970149256e-05,
+      "loss": 0.0159,
+      "step": 7332
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 4.9144589552238804e-05,
+      "loss": 0.0211,
+      "step": 7336
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 4.9144123134328365e-05,
+      "loss": 0.02,
+      "step": 7340
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 4.9143656716417914e-05,
+      "loss": 0.0059,
+      "step": 7344
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 4.914319029850746e-05,
+      "loss": 0.0227,
+      "step": 7348
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 4.9142723880597017e-05,
+      "loss": 0.0125,
+      "step": 7352
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 4.914225746268657e-05,
+      "loss": 0.0047,
+      "step": 7356
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 4.914179104477612e-05,
+      "loss": 0.0119,
+      "step": 7360
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 4.9141324626865675e-05,
+      "loss": 0.0123,
+      "step": 7364
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 4.914085820895522e-05,
+      "loss": 0.0082,
+      "step": 7368
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 4.914039179104478e-05,
+      "loss": 0.0152,
+      "step": 7372
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 4.913992537313433e-05,
+      "loss": 0.0106,
+      "step": 7376
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 4.913945895522388e-05,
+      "loss": 0.0141,
+      "step": 7380
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 4.9138992537313436e-05,
+      "loss": 0.0142,
+      "step": 7384
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 4.913852611940299e-05,
+      "loss": 0.0166,
+      "step": 7388
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 4.913805970149254e-05,
+      "loss": 0.0135,
+      "step": 7392
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 4.913759328358209e-05,
+      "loss": 0.0169,
+      "step": 7396
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 4.913712686567165e-05,
+      "loss": 0.011,
+      "step": 7400
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 4.9136660447761197e-05,
+      "loss": 0.0163,
+      "step": 7404
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 4.9136194029850745e-05,
+      "loss": 0.0102,
+      "step": 7408
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 4.91357276119403e-05,
+      "loss": 0.0066,
+      "step": 7412
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 4.9135261194029854e-05,
+      "loss": 0.0132,
+      "step": 7416
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 4.91347947761194e-05,
+      "loss": 0.0091,
+      "step": 7420
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 4.913432835820896e-05,
+      "loss": 0.0095,
+      "step": 7424
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 4.9133861940298506e-05,
+      "loss": 0.0222,
+      "step": 7428
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 4.913339552238806e-05,
+      "loss": 0.0078,
+      "step": 7432
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 4.9132929104477615e-05,
+      "loss": 0.0157,
+      "step": 7436
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 4.9132462686567164e-05,
+      "loss": 0.0103,
+      "step": 7440
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 4.913199626865672e-05,
+      "loss": 0.0178,
+      "step": 7444
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 4.9131529850746273e-05,
+      "loss": 0.0115,
+      "step": 7448
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 4.913106343283582e-05,
+      "loss": 0.0083,
+      "step": 7452
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 4.9130597014925376e-05,
+      "loss": 0.0242,
+      "step": 7456
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 4.913013059701493e-05,
+      "loss": 0.0086,
+      "step": 7460
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 4.912966417910448e-05,
+      "loss": 0.0056,
+      "step": 7464
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 4.9129197761194034e-05,
+      "loss": 0.0148,
+      "step": 7468
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 4.912873134328358e-05,
+      "loss": 0.0127,
+      "step": 7472
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 4.912826492537314e-05,
+      "loss": 0.0049,
+      "step": 7476
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 4.912779850746269e-05,
+      "loss": 0.0044,
+      "step": 7480
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 4.912733208955224e-05,
+      "loss": 0.0056,
+      "step": 7484
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 4.912686567164179e-05,
+      "loss": 0.0335,
+      "step": 7488
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 4.912639925373135e-05,
+      "loss": 0.0055,
+      "step": 7492
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 4.91259328358209e-05,
+      "loss": 0.0129,
+      "step": 7496
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 4.9125466417910447e-05,
+      "loss": 0.0122,
+      "step": 7500
+    },
+    {
+      "epoch": 7.0,
+      "eval_exact_match": 0.730174081237911,
+      "eval_exec": 0.7591876208897486,
+      "eval_loss": 0.21238785982131958,
+      "eval_runtime": 1160.0553,
+      "eval_samples_per_second": 0.891,
+      "step": 7500
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 4.9125e-05,
+      "loss": 0.0162,
+      "step": 7504
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 4.9124533582089556e-05,
+      "loss": 0.0051,
+      "step": 7508
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 4.9124067164179104e-05,
+      "loss": 0.0151,
+      "step": 7512
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 4.912360074626866e-05,
+      "loss": 0.0096,
+      "step": 7516
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 4.9123134328358214e-05,
+      "loss": 0.0107,
+      "step": 7520
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 4.912266791044776e-05,
+      "loss": 0.0088,
+      "step": 7524
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 4.912220149253732e-05,
+      "loss": 0.0022,
+      "step": 7528
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 4.9121735074626865e-05,
+      "loss": 0.0129,
+      "step": 7532
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 4.912126865671642e-05,
+      "loss": 0.0049,
+      "step": 7536
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 4.9120802238805975e-05,
+      "loss": 0.0068,
+      "step": 7540
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 4.9120335820895523e-05,
+      "loss": 0.0109,
+      "step": 7544
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 4.911986940298507e-05,
+      "loss": 0.0266,
+      "step": 7548
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 4.911940298507463e-05,
+      "loss": 0.0092,
+      "step": 7552
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 4.911893656716418e-05,
+      "loss": 0.0166,
+      "step": 7556
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 4.911847014925373e-05,
+      "loss": 0.007,
+      "step": 7560
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 4.9118003731343284e-05,
+      "loss": 0.0054,
+      "step": 7564
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 4.911753731343284e-05,
+      "loss": 0.0135,
+      "step": 7568
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 4.911707089552239e-05,
+      "loss": 0.0149,
+      "step": 7572
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 4.911660447761194e-05,
+      "loss": 0.0119,
+      "step": 7576
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 4.91161380597015e-05,
+      "loss": 0.0054,
+      "step": 7580
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 4.9115671641791045e-05,
+      "loss": 0.0073,
+      "step": 7584
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 4.91152052238806e-05,
+      "loss": 0.0045,
+      "step": 7588
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 4.911473880597015e-05,
+      "loss": 0.0114,
+      "step": 7592
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 4.91142723880597e-05,
+      "loss": 0.0094,
+      "step": 7596
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 4.911380597014926e-05,
+      "loss": 0.0177,
+      "step": 7600
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 4.9113339552238806e-05,
+      "loss": 0.0134,
+      "step": 7604
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 4.911287313432836e-05,
+      "loss": 0.0129,
+      "step": 7608
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 4.9112406716417916e-05,
+      "loss": 0.0205,
+      "step": 7612
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 4.9111940298507464e-05,
+      "loss": 0.0096,
+      "step": 7616
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 4.911147388059702e-05,
+      "loss": 0.0108,
+      "step": 7620
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 4.911100746268657e-05,
+      "loss": 0.0313,
+      "step": 7624
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 4.911054104477612e-05,
+      "loss": 0.0134,
+      "step": 7628
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 4.911007462686568e-05,
+      "loss": 0.005,
+      "step": 7632
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 4.9109608208955225e-05,
+      "loss": 0.0154,
+      "step": 7636
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 4.910914179104478e-05,
+      "loss": 0.0044,
+      "step": 7640
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 4.9108675373134335e-05,
+      "loss": 0.0115,
+      "step": 7644
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 4.910820895522388e-05,
+      "loss": 0.0098,
+      "step": 7648
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 4.910774253731343e-05,
+      "loss": 0.003,
+      "step": 7652
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 4.9107276119402986e-05,
+      "loss": 0.0067,
+      "step": 7656
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 4.910680970149254e-05,
+      "loss": 0.0123,
+      "step": 7660
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 4.910634328358209e-05,
+      "loss": 0.0124,
+      "step": 7664
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 4.9105876865671644e-05,
+      "loss": 0.0063,
+      "step": 7668
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 4.91054104477612e-05,
+      "loss": 0.0181,
+      "step": 7672
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 4.910494402985075e-05,
+      "loss": 0.005,
+      "step": 7676
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 4.91044776119403e-05,
+      "loss": 0.0052,
+      "step": 7680
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 4.910401119402985e-05,
+      "loss": 0.0135,
+      "step": 7684
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 4.9103544776119405e-05,
+      "loss": 0.0053,
+      "step": 7688
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 4.910307835820896e-05,
+      "loss": 0.0116,
+      "step": 7692
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 4.910261194029851e-05,
+      "loss": 0.0078,
+      "step": 7696
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 4.910214552238806e-05,
+      "loss": 0.0033,
+      "step": 7700
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 4.910167910447762e-05,
+      "loss": 0.0107,
+      "step": 7704
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 4.9101212686567166e-05,
+      "loss": 0.0045,
+      "step": 7708
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 4.9100746268656714e-05,
+      "loss": 0.0127,
+      "step": 7712
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 4.910027985074627e-05,
+      "loss": 0.0106,
+      "step": 7716
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 4.9099813432835824e-05,
+      "loss": 0.0105,
+      "step": 7720
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 4.909934701492537e-05,
+      "loss": 0.0073,
+      "step": 7724
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 4.909888059701493e-05,
+      "loss": 0.0075,
+      "step": 7728
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 4.909841417910448e-05,
+      "loss": 0.013,
+      "step": 7732
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 4.909794776119403e-05,
+      "loss": 0.0041,
+      "step": 7736
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 4.9097481343283585e-05,
+      "loss": 0.0102,
+      "step": 7740
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 4.909701492537313e-05,
+      "loss": 0.0078,
+      "step": 7744
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 4.909654850746269e-05,
+      "loss": 0.0059,
+      "step": 7748
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 4.909608208955224e-05,
+      "loss": 0.0029,
+      "step": 7752
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 4.909561567164179e-05,
+      "loss": 0.0101,
+      "step": 7756
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 4.9095149253731346e-05,
+      "loss": 0.0113,
+      "step": 7760
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 4.90946828358209e-05,
+      "loss": 0.0097,
+      "step": 7764
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 4.909421641791045e-05,
+      "loss": 0.003,
+      "step": 7768
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 4.9093750000000004e-05,
+      "loss": 0.0047,
+      "step": 7772
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 4.909328358208955e-05,
+      "loss": 0.0113,
+      "step": 7776
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 4.909281716417911e-05,
+      "loss": 0.0074,
+      "step": 7780
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 4.909235074626866e-05,
+      "loss": 0.0122,
+      "step": 7784
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 4.909188432835821e-05,
+      "loss": 0.0312,
+      "step": 7788
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 4.9091417910447765e-05,
+      "loss": 0.0104,
+      "step": 7792
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 4.909095149253732e-05,
+      "loss": 0.0124,
+      "step": 7796
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 4.909048507462687e-05,
+      "loss": 0.0217,
+      "step": 7800
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 4.9090018656716416e-05,
+      "loss": 0.0164,
+      "step": 7804
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 4.908955223880598e-05,
+      "loss": 0.0105,
+      "step": 7808
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 4.9089085820895526e-05,
+      "loss": 0.01,
+      "step": 7812
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 4.9088619402985074e-05,
+      "loss": 0.0267,
+      "step": 7816
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 4.908815298507463e-05,
+      "loss": 0.0062,
+      "step": 7820
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 4.9087686567164184e-05,
+      "loss": 0.0099,
+      "step": 7824
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 4.908722014925373e-05,
+      "loss": 0.0089,
+      "step": 7828
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 4.908675373134329e-05,
+      "loss": 0.0067,
+      "step": 7832
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 4.9086287313432835e-05,
+      "loss": 0.0028,
+      "step": 7836
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 4.908582089552239e-05,
+      "loss": 0.0052,
+      "step": 7840
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 4.9085354477611945e-05,
+      "loss": 0.0033,
+      "step": 7844
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 4.908488805970149e-05,
+      "loss": 0.0082,
+      "step": 7848
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 4.908442164179105e-05,
+      "loss": 0.0134,
+      "step": 7852
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 4.90839552238806e-05,
+      "loss": 0.0053,
+      "step": 7856
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 4.908348880597015e-05,
+      "loss": 0.0142,
+      "step": 7860
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 4.90830223880597e-05,
+      "loss": 0.0097,
+      "step": 7864
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 4.908255597014926e-05,
+      "loss": 0.0063,
+      "step": 7868
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 4.908208955223881e-05,
+      "loss": 0.0141,
+      "step": 7872
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 4.908162313432836e-05,
+      "loss": 0.0046,
+      "step": 7876
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 4.908115671641791e-05,
+      "loss": 0.0068,
+      "step": 7880
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 4.908069029850747e-05,
+      "loss": 0.006,
+      "step": 7884
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 4.9080223880597015e-05,
+      "loss": 0.0082,
+      "step": 7888
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 4.907975746268657e-05,
+      "loss": 0.0071,
+      "step": 7892
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 4.907929104477612e-05,
+      "loss": 0.0034,
+      "step": 7896
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 4.907882462686567e-05,
+      "loss": 0.0028,
+      "step": 7900
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 4.907835820895523e-05,
+      "loss": 0.004,
+      "step": 7904
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 4.9077891791044776e-05,
+      "loss": 0.0083,
+      "step": 7908
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 4.907742537313433e-05,
+      "loss": 0.0091,
+      "step": 7912
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 4.9076958955223886e-05,
+      "loss": 0.0068,
+      "step": 7916
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 4.9076492537313434e-05,
+      "loss": 0.0053,
+      "step": 7920
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 4.907602611940298e-05,
+      "loss": 0.0092,
+      "step": 7924
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 4.9075559701492544e-05,
+      "loss": 0.002,
+      "step": 7928
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 4.907509328358209e-05,
+      "loss": 0.0121,
+      "step": 7932
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 4.907462686567165e-05,
+      "loss": 0.0158,
+      "step": 7936
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 4.9074160447761195e-05,
+      "loss": 0.0135,
+      "step": 7940
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 4.907369402985075e-05,
+      "loss": 0.0088,
+      "step": 7944
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 4.9073227611940305e-05,
+      "loss": 0.0071,
+      "step": 7948
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 4.907276119402985e-05,
+      "loss": 0.0072,
+      "step": 7952
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 4.90722947761194e-05,
+      "loss": 0.0344,
+      "step": 7956
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 4.907182835820896e-05,
+      "loss": 0.0059,
+      "step": 7960
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 4.907136194029851e-05,
+      "loss": 0.0101,
+      "step": 7964
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 4.907089552238806e-05,
+      "loss": 0.013,
+      "step": 7968
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 4.9070429104477614e-05,
+      "loss": 0.0268,
+      "step": 7972
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 4.906996268656717e-05,
+      "loss": 0.0088,
+      "step": 7976
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 4.906949626865672e-05,
+      "loss": 0.01,
+      "step": 7980
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 4.906902985074627e-05,
+      "loss": 0.0091,
+      "step": 7984
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 4.906856343283583e-05,
+      "loss": 0.0176,
+      "step": 7988
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 4.9068097014925375e-05,
+      "loss": 0.0144,
+      "step": 7992
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 4.906763059701493e-05,
+      "loss": 0.016,
+      "step": 7996
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 4.906716417910448e-05,
+      "loss": 0.0184,
+      "step": 8000
+    },
+    {
+      "epoch": 7.46,
+      "eval_exact_match": 0.7059961315280464,
+      "eval_exec": 0.7427466150870407,
+      "eval_loss": 0.19857120513916016,
+      "eval_runtime": 1295.3702,
+      "eval_samples_per_second": 0.798,
+      "step": 8000
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 4.906669776119403e-05,
+      "loss": 0.0074,
+      "step": 8004
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 4.906623134328359e-05,
+      "loss": 0.0099,
+      "step": 8008
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 4.9065764925373136e-05,
+      "loss": 0.0031,
+      "step": 8012
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 4.9065298507462684e-05,
+      "loss": 0.0014,
+      "step": 8016
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 4.9064832089552246e-05,
+      "loss": 0.0187,
+      "step": 8020
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 4.9064365671641794e-05,
+      "loss": 0.0078,
+      "step": 8024
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 4.906389925373134e-05,
+      "loss": 0.0109,
+      "step": 8028
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 4.90634328358209e-05,
+      "loss": 0.0088,
+      "step": 8032
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 4.906296641791045e-05,
+      "loss": 0.0195,
+      "step": 8036
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 4.90625e-05,
+      "loss": 0.0075,
+      "step": 8040
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 4.9062033582089555e-05,
+      "loss": 0.0075,
+      "step": 8044
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 4.906156716417911e-05,
+      "loss": 0.0111,
+      "step": 8048
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 4.906110074626866e-05,
+      "loss": 0.0068,
+      "step": 8052
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 4.906063432835821e-05,
+      "loss": 0.0266,
+      "step": 8056
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 4.906016791044776e-05,
+      "loss": 0.0099,
+      "step": 8060
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 4.9059701492537316e-05,
+      "loss": 0.0185,
+      "step": 8064
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 4.905923507462687e-05,
+      "loss": 0.0554,
+      "step": 8068
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 4.905876865671642e-05,
+      "loss": 0.0089,
+      "step": 8072
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 4.905830223880597e-05,
+      "loss": 0.0142,
+      "step": 8076
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 4.905783582089553e-05,
+      "loss": 0.0093,
+      "step": 8080
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 4.905736940298508e-05,
+      "loss": 0.004,
+      "step": 8084
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 4.9056902985074625e-05,
+      "loss": 0.0191,
+      "step": 8088
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 4.905643656716418e-05,
+      "loss": 0.0087,
+      "step": 8092
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 4.9055970149253735e-05,
+      "loss": 0.0024,
+      "step": 8096
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 4.905550373134329e-05,
+      "loss": 0.0119,
+      "step": 8100
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 4.905503731343284e-05,
+      "loss": 0.0183,
+      "step": 8104
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 4.9054570895522386e-05,
+      "loss": 0.0335,
+      "step": 8108
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 4.905410447761195e-05,
+      "loss": 0.0109,
+      "step": 8112
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 4.9053638059701496e-05,
+      "loss": 0.0087,
+      "step": 8116
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 4.9053171641791044e-05,
+      "loss": 0.0147,
+      "step": 8120
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 4.90527052238806e-05,
+      "loss": 0.0057,
+      "step": 8124
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 4.9052238805970154e-05,
+      "loss": 0.0076,
+      "step": 8128
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 4.90517723880597e-05,
+      "loss": 0.0141,
+      "step": 8132
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 4.905130597014926e-05,
+      "loss": 0.011,
+      "step": 8136
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 4.905083955223881e-05,
+      "loss": 0.0131,
+      "step": 8140
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 4.905037313432836e-05,
+      "loss": 0.0042,
+      "step": 8144
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 4.9049906716417915e-05,
+      "loss": 0.0029,
+      "step": 8148
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 4.904944029850746e-05,
+      "loss": 0.0093,
+      "step": 8152
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 4.904897388059702e-05,
+      "loss": 0.0169,
+      "step": 8156
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 4.904850746268657e-05,
+      "loss": 0.0051,
+      "step": 8160
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 4.904804104477612e-05,
+      "loss": 0.0164,
+      "step": 8164
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 4.904757462686567e-05,
+      "loss": 0.0187,
+      "step": 8168
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 4.904710820895523e-05,
+      "loss": 0.0233,
+      "step": 8172
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 4.904664179104478e-05,
+      "loss": 0.0178,
+      "step": 8176
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 4.904617537313433e-05,
+      "loss": 0.0074,
+      "step": 8180
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 4.904570895522388e-05,
+      "loss": 0.0159,
+      "step": 8184
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 4.904524253731344e-05,
+      "loss": 0.0035,
+      "step": 8188
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 4.9044776119402985e-05,
+      "loss": 0.0147,
+      "step": 8192
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 4.904430970149254e-05,
+      "loss": 0.0221,
+      "step": 8196
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 4.9043843283582095e-05,
+      "loss": 0.0231,
+      "step": 8200
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 4.904337686567164e-05,
+      "loss": 0.0097,
+      "step": 8204
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 4.90429104477612e-05,
+      "loss": 0.0051,
+      "step": 8208
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 4.9042444029850746e-05,
+      "loss": 0.0112,
+      "step": 8212
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 4.90419776119403e-05,
+      "loss": 0.0071,
+      "step": 8216
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 4.9041511194029856e-05,
+      "loss": 0.0031,
+      "step": 8220
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 4.9041044776119404e-05,
+      "loss": 0.0922,
+      "step": 8224
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 4.904057835820895e-05,
+      "loss": 0.0062,
+      "step": 8228
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 4.9040111940298514e-05,
+      "loss": 0.0073,
+      "step": 8232
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 4.903964552238806e-05,
+      "loss": 0.008,
+      "step": 8236
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 4.903917910447761e-05,
+      "loss": 0.0109,
+      "step": 8240
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 4.9038712686567165e-05,
+      "loss": 0.0079,
+      "step": 8244
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 4.903824626865672e-05,
+      "loss": 0.0109,
+      "step": 8248
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 4.903777985074627e-05,
+      "loss": 0.0054,
+      "step": 8252
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 4.903731343283582e-05,
+      "loss": 0.005,
+      "step": 8256
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 4.903684701492538e-05,
+      "loss": 0.0063,
+      "step": 8260
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 4.903638059701493e-05,
+      "loss": 0.009,
+      "step": 8264
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 4.903591417910448e-05,
+      "loss": 0.0047,
+      "step": 8268
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 4.903544776119403e-05,
+      "loss": 0.0079,
+      "step": 8272
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 4.9034981343283584e-05,
+      "loss": 0.0158,
+      "step": 8276
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 4.903451492537314e-05,
+      "loss": 0.0092,
+      "step": 8280
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 4.903404850746269e-05,
+      "loss": 0.013,
+      "step": 8284
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 4.903358208955224e-05,
+      "loss": 0.0055,
+      "step": 8288
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 4.9033115671641796e-05,
+      "loss": 0.0108,
+      "step": 8292
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 4.9032649253731345e-05,
+      "loss": 0.0144,
+      "step": 8296
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 4.90321828358209e-05,
+      "loss": 0.0122,
+      "step": 8300
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 4.903171641791045e-05,
+      "loss": 0.0042,
+      "step": 8304
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 4.903125e-05,
+      "loss": 0.0077,
+      "step": 8308
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 4.903078358208956e-05,
+      "loss": 0.0076,
+      "step": 8312
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 4.9030317164179106e-05,
+      "loss": 0.0096,
+      "step": 8316
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 4.902985074626866e-05,
+      "loss": 0.0126,
+      "step": 8320
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 4.9029384328358215e-05,
+      "loss": 0.0178,
+      "step": 8324
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 4.9028917910447764e-05,
+      "loss": 0.0056,
+      "step": 8328
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 4.902845149253731e-05,
+      "loss": 0.0135,
+      "step": 8332
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 4.9027985074626867e-05,
+      "loss": 0.0065,
+      "step": 8336
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 4.902751865671642e-05,
+      "loss": 0.0056,
+      "step": 8340
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 4.902705223880597e-05,
+      "loss": 0.006,
+      "step": 8344
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 4.9026585820895525e-05,
+      "loss": 0.0108,
+      "step": 8348
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 4.902611940298508e-05,
+      "loss": 0.0076,
+      "step": 8352
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 4.902565298507463e-05,
+      "loss": 0.0112,
+      "step": 8356
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 4.902518656716418e-05,
+      "loss": 0.0044,
+      "step": 8360
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 4.902472014925373e-05,
+      "loss": 0.0029,
+      "step": 8364
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 4.9024253731343286e-05,
+      "loss": 0.0062,
+      "step": 8368
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 4.902378731343284e-05,
+      "loss": 0.0114,
+      "step": 8372
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 4.902332089552239e-05,
+      "loss": 0.0344,
+      "step": 8376
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 4.9022854477611943e-05,
+      "loss": 0.0113,
+      "step": 8380
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 4.90223880597015e-05,
+      "loss": 0.0098,
+      "step": 8384
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 4.9021921641791047e-05,
+      "loss": 0.011,
+      "step": 8388
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 4.9021455223880595e-05,
+      "loss": 0.0147,
+      "step": 8392
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 4.902098880597015e-05,
+      "loss": 0.0191,
+      "step": 8396
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 4.9020522388059704e-05,
+      "loss": 0.0161,
+      "step": 8400
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 4.902005597014925e-05,
+      "loss": 0.0097,
+      "step": 8404
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 4.901958955223881e-05,
+      "loss": 0.0107,
+      "step": 8408
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 4.901912313432836e-05,
+      "loss": 0.0112,
+      "step": 8412
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 4.901865671641792e-05,
+      "loss": 0.0135,
+      "step": 8416
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 4.9018190298507465e-05,
+      "loss": 0.0098,
+      "step": 8420
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 4.9017723880597014e-05,
+      "loss": 0.0128,
+      "step": 8424
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 4.9017257462686575e-05,
+      "loss": 0.0111,
+      "step": 8428
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 4.901679104477612e-05,
+      "loss": 0.0231,
+      "step": 8432
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 4.901632462686567e-05,
+      "loss": 0.0146,
+      "step": 8436
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 4.9015858208955226e-05,
+      "loss": 0.0016,
+      "step": 8440
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 4.901539179104478e-05,
+      "loss": 0.0031,
+      "step": 8444
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 4.901492537313433e-05,
+      "loss": 0.0071,
+      "step": 8448
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 4.9014458955223884e-05,
+      "loss": 0.0152,
+      "step": 8452
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 4.901399253731343e-05,
+      "loss": 0.0076,
+      "step": 8456
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 4.901352611940299e-05,
+      "loss": 0.0204,
+      "step": 8460
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 4.901305970149254e-05,
+      "loss": 0.0027,
+      "step": 8464
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 4.901259328358209e-05,
+      "loss": 0.012,
+      "step": 8468
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 4.9012126865671645e-05,
+      "loss": 0.0056,
+      "step": 8472
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 4.90116604477612e-05,
+      "loss": 0.0064,
+      "step": 8476
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 4.901119402985075e-05,
+      "loss": 0.0082,
+      "step": 8480
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 4.9010727611940297e-05,
+      "loss": 0.0071,
+      "step": 8484
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 4.901026119402986e-05,
+      "loss": 0.0152,
+      "step": 8488
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 4.9009794776119406e-05,
+      "loss": 0.008,
+      "step": 8492
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 4.9009328358208954e-05,
+      "loss": 0.0148,
+      "step": 8496
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 4.900886194029851e-05,
+      "loss": 0.0038,
+      "step": 8500
+    },
+    {
+      "epoch": 7.93,
+      "eval_exact_match": 0.7127659574468085,
+      "eval_exec": 0.7504835589941973,
+      "eval_loss": 0.23233897984027863,
+      "eval_runtime": 1425.0133,
+      "eval_samples_per_second": 0.726,
+      "step": 8500
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 4.9008395522388064e-05,
+      "loss": 0.0051,
+      "step": 8504
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 4.900792910447761e-05,
+      "loss": 0.0202,
+      "step": 8508
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 4.900746268656717e-05,
+      "loss": 0.0163,
+      "step": 8512
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 4.9006996268656715e-05,
+      "loss": 0.0055,
+      "step": 8516
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 4.900652985074627e-05,
+      "loss": 0.0053,
+      "step": 8520
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 4.9006063432835825e-05,
+      "loss": 0.0126,
+      "step": 8524
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 4.9005597014925373e-05,
+      "loss": 0.0098,
+      "step": 8528
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 4.900513059701493e-05,
+      "loss": 0.0072,
+      "step": 8532
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 4.900466417910448e-05,
+      "loss": 0.0078,
+      "step": 8536
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 4.900419776119403e-05,
+      "loss": 0.0104,
+      "step": 8540
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 4.900373134328358e-05,
+      "loss": 0.0062,
+      "step": 8544
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 4.900326492537314e-05,
+      "loss": 0.0079,
+      "step": 8548
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 4.900279850746269e-05,
+      "loss": 0.0096,
+      "step": 8552
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 4.900233208955224e-05,
+      "loss": 0.0116,
+      "step": 8556
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 4.900186567164179e-05,
+      "loss": 0.01,
+      "step": 8560
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 4.900139925373135e-05,
+      "loss": 0.0133,
+      "step": 8564
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 4.9000932835820895e-05,
+      "loss": 0.0058,
+      "step": 8568
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 4.900046641791045e-05,
+      "loss": 0.0039,
+      "step": 8572
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 4.9e-05,
+      "loss": 0.0034,
+      "step": 8576
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 4.899953358208956e-05,
+      "loss": 0.0095,
+      "step": 8580
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 4.899906716417911e-05,
+      "loss": 0.0008,
+      "step": 8584
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 4.8998600746268656e-05,
+      "loss": 0.0054,
+      "step": 8588
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 4.899813432835821e-05,
+      "loss": 0.0044,
+      "step": 8592
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 4.8997667910447766e-05,
+      "loss": 0.0024,
+      "step": 8596
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 4.8997201492537314e-05,
+      "loss": 0.0031,
+      "step": 8600
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 4.899673507462687e-05,
+      "loss": 0.0033,
+      "step": 8604
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 4.8996268656716424e-05,
+      "loss": 0.0056,
+      "step": 8608
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 4.899580223880597e-05,
+      "loss": 0.0025,
+      "step": 8612
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 4.899533582089553e-05,
+      "loss": 0.0038,
+      "step": 8616
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 4.8994869402985075e-05,
+      "loss": 0.0088,
+      "step": 8620
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 4.899440298507463e-05,
+      "loss": 0.0041,
+      "step": 8624
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 4.8993936567164185e-05,
+      "loss": 0.0068,
+      "step": 8628
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 4.899347014925373e-05,
+      "loss": 0.011,
+      "step": 8632
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 4.899300373134328e-05,
+      "loss": 0.0037,
+      "step": 8636
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 4.899253731343284e-05,
+      "loss": 0.0043,
+      "step": 8640
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 4.899207089552239e-05,
+      "loss": 0.0058,
+      "step": 8644
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 4.899160447761194e-05,
+      "loss": 0.005,
+      "step": 8648
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 4.8991138059701494e-05,
+      "loss": 0.0084,
+      "step": 8652
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 4.899067164179105e-05,
+      "loss": 0.0252,
+      "step": 8656
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 4.89902052238806e-05,
+      "loss": 0.0063,
+      "step": 8660
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 4.898973880597015e-05,
+      "loss": 0.0119,
+      "step": 8664
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 4.898927238805971e-05,
+      "loss": 0.0016,
+      "step": 8668
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 4.8988805970149255e-05,
+      "loss": 0.0067,
+      "step": 8672
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 4.898833955223881e-05,
+      "loss": 0.0108,
+      "step": 8676
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 4.898787313432836e-05,
+      "loss": 0.0061,
+      "step": 8680
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 4.898740671641791e-05,
+      "loss": 0.0072,
+      "step": 8684
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 4.898694029850747e-05,
+      "loss": 0.0017,
+      "step": 8688
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 4.8986473880597016e-05,
+      "loss": 0.0038,
+      "step": 8692
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 4.8986007462686564e-05,
+      "loss": 0.0169,
+      "step": 8696
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 4.8985541044776126e-05,
+      "loss": 0.0112,
+      "step": 8700
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 4.8985074626865674e-05,
+      "loss": 0.0025,
+      "step": 8704
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 4.898460820895522e-05,
+      "loss": 0.0019,
+      "step": 8708
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 4.898414179104478e-05,
+      "loss": 0.0063,
+      "step": 8712
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 4.898367537313433e-05,
+      "loss": 0.0054,
+      "step": 8716
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 4.898320895522388e-05,
+      "loss": 0.004,
+      "step": 8720
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 4.8982742537313435e-05,
+      "loss": 0.0049,
+      "step": 8724
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 4.898227611940299e-05,
+      "loss": 0.0072,
+      "step": 8728
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 4.898180970149254e-05,
+      "loss": 0.0028,
+      "step": 8732
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 4.898134328358209e-05,
+      "loss": 0.0143,
+      "step": 8736
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 4.898087686567164e-05,
+      "loss": 0.0043,
+      "step": 8740
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 4.8980410447761196e-05,
+      "loss": 0.0136,
+      "step": 8744
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 4.897994402985075e-05,
+      "loss": 0.0029,
+      "step": 8748
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 4.89794776119403e-05,
+      "loss": 0.0028,
+      "step": 8752
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 4.8979011194029854e-05,
+      "loss": 0.0205,
+      "step": 8756
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 4.897854477611941e-05,
+      "loss": 0.005,
+      "step": 8760
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 4.897807835820896e-05,
+      "loss": 0.0037,
+      "step": 8764
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 4.897761194029851e-05,
+      "loss": 0.0069,
+      "step": 8768
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 4.897714552238806e-05,
+      "loss": 0.0075,
+      "step": 8772
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 4.8976679104477615e-05,
+      "loss": 0.0102,
+      "step": 8776
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 4.897621268656717e-05,
+      "loss": 0.0185,
+      "step": 8780
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 4.897574626865672e-05,
+      "loss": 0.0029,
+      "step": 8784
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 4.8975279850746266e-05,
+      "loss": 0.003,
+      "step": 8788
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 4.897481343283583e-05,
+      "loss": 0.0142,
+      "step": 8792
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 4.8974347014925376e-05,
+      "loss": 0.0086,
+      "step": 8796
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 4.8973880597014924e-05,
+      "loss": 0.0037,
+      "step": 8800
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 4.897341417910448e-05,
+      "loss": 0.0189,
+      "step": 8804
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 4.8972947761194034e-05,
+      "loss": 0.0031,
+      "step": 8808
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 4.897248134328358e-05,
+      "loss": 0.0061,
+      "step": 8812
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 4.897201492537314e-05,
+      "loss": 0.0047,
+      "step": 8816
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 4.897154850746269e-05,
+      "loss": 0.0041,
+      "step": 8820
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 4.897108208955224e-05,
+      "loss": 0.0055,
+      "step": 8824
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 4.8970615671641795e-05,
+      "loss": 0.0222,
+      "step": 8828
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 4.897014925373134e-05,
+      "loss": 0.0059,
+      "step": 8832
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 4.89696828358209e-05,
+      "loss": 0.016,
+      "step": 8836
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 4.896921641791045e-05,
+      "loss": 0.0095,
+      "step": 8840
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 4.896875e-05,
+      "loss": 0.0062,
+      "step": 8844
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 4.896828358208955e-05,
+      "loss": 0.0012,
+      "step": 8848
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 4.896781716417911e-05,
+      "loss": 0.0032,
+      "step": 8852
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 4.896735074626866e-05,
+      "loss": 0.0146,
+      "step": 8856
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 4.896688432835821e-05,
+      "loss": 0.0123,
+      "step": 8860
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 4.896641791044776e-05,
+      "loss": 0.0066,
+      "step": 8864
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 4.896595149253732e-05,
+      "loss": 0.0026,
+      "step": 8868
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 4.8965485074626865e-05,
+      "loss": 0.0051,
+      "step": 8872
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 4.896501865671642e-05,
+      "loss": 0.0045,
+      "step": 8876
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 4.8964552238805975e-05,
+      "loss": 0.004,
+      "step": 8880
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 4.896408582089552e-05,
+      "loss": 0.0113,
+      "step": 8884
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 4.896361940298508e-05,
+      "loss": 0.005,
+      "step": 8888
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 4.8963152985074626e-05,
+      "loss": 0.0062,
+      "step": 8892
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 4.896268656716418e-05,
+      "loss": 0.0053,
+      "step": 8896
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 4.8962220149253736e-05,
+      "loss": 0.0049,
+      "step": 8900
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 4.8961753731343284e-05,
+      "loss": 0.0037,
+      "step": 8904
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 4.896128731343284e-05,
+      "loss": 0.0359,
+      "step": 8908
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 4.8960820895522394e-05,
+      "loss": 0.0034,
+      "step": 8912
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 4.896035447761194e-05,
+      "loss": 0.0019,
+      "step": 8916
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 4.89598880597015e-05,
+      "loss": 0.0039,
+      "step": 8920
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 4.8959421641791045e-05,
+      "loss": 0.0091,
+      "step": 8924
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 4.89589552238806e-05,
+      "loss": 0.0103,
+      "step": 8928
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 4.8958488805970155e-05,
+      "loss": 0.0076,
+      "step": 8932
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.89580223880597e-05,
+      "loss": 0.0653,
+      "step": 8936
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.895755597014926e-05,
+      "loss": 0.0164,
+      "step": 8940
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.895708955223881e-05,
+      "loss": 0.0133,
+      "step": 8944
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.895662313432836e-05,
+      "loss": 0.0063,
+      "step": 8948
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.895615671641791e-05,
+      "loss": 0.0046,
+      "step": 8952
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.8955690298507464e-05,
+      "loss": 0.002,
+      "step": 8956
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.895522388059702e-05,
+      "loss": 0.0034,
+      "step": 8960
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.895475746268657e-05,
+      "loss": 0.003,
+      "step": 8964
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.895429104477612e-05,
+      "loss": 0.0112,
+      "step": 8968
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.895382462686568e-05,
+      "loss": 0.0156,
+      "step": 8972
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.8953358208955225e-05,
+      "loss": 0.0045,
+      "step": 8976
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.895289179104478e-05,
+      "loss": 0.003,
+      "step": 8980
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.895242537313433e-05,
+      "loss": 0.0061,
+      "step": 8984
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.895195895522388e-05,
+      "loss": 0.0125,
+      "step": 8988
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.895149253731344e-05,
+      "loss": 0.0068,
+      "step": 8992
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.8951026119402986e-05,
+      "loss": 0.0072,
+      "step": 8996
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.895055970149254e-05,
+      "loss": 0.0054,
+      "step": 9000
+    },
+    {
+      "epoch": 8.4,
+      "eval_exact_match": 0.7321083172147002,
+      "eval_exec": 0.7572533849129593,
+      "eval_loss": 0.24979576468467712,
+      "eval_runtime": 1232.0481,
+      "eval_samples_per_second": 0.839,
+      "step": 9000
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.8950093283582096e-05,
+      "loss": 0.0038,
+      "step": 9004
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.8949626865671644e-05,
+      "loss": 0.007,
+      "step": 9008
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.894916044776119e-05,
+      "loss": 0.0187,
+      "step": 9012
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.894869402985075e-05,
+      "loss": 0.0081,
+      "step": 9016
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.89482276119403e-05,
+      "loss": 0.0047,
+      "step": 9020
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.894776119402985e-05,
+      "loss": 0.0043,
+      "step": 9024
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.8947294776119405e-05,
+      "loss": 0.028,
+      "step": 9028
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.894682835820896e-05,
+      "loss": 0.0063,
+      "step": 9032
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.894636194029851e-05,
+      "loss": 0.0072,
+      "step": 9036
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.894589552238806e-05,
+      "loss": 0.0169,
+      "step": 9040
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.894542910447761e-05,
+      "loss": 0.0325,
+      "step": 9044
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.8944962686567166e-05,
+      "loss": 0.0085,
+      "step": 9048
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.894449626865672e-05,
+      "loss": 0.0022,
+      "step": 9052
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.894402985074627e-05,
+      "loss": 0.0097,
+      "step": 9056
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.8943563432835824e-05,
+      "loss": 0.0049,
+      "step": 9060
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.894309701492538e-05,
+      "loss": 0.0099,
+      "step": 9064
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.894263059701493e-05,
+      "loss": 0.003,
+      "step": 9068
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.894216417910448e-05,
+      "loss": 0.0065,
+      "step": 9072
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.894169776119403e-05,
+      "loss": 0.0048,
+      "step": 9076
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.8941231343283585e-05,
+      "loss": 0.0018,
+      "step": 9080
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.894076492537314e-05,
+      "loss": 0.0039,
+      "step": 9084
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.894029850746269e-05,
+      "loss": 0.0058,
+      "step": 9088
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.893983208955224e-05,
+      "loss": 0.0059,
+      "step": 9092
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.89393656716418e-05,
+      "loss": 0.0059,
+      "step": 9096
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.8938899253731346e-05,
+      "loss": 0.0076,
+      "step": 9100
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.8938432835820894e-05,
+      "loss": 0.0044,
+      "step": 9104
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.8937966417910456e-05,
+      "loss": 0.0052,
+      "step": 9108
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.8937500000000004e-05,
+      "loss": 0.0091,
+      "step": 9112
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.893703358208955e-05,
+      "loss": 0.007,
+      "step": 9116
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.893656716417911e-05,
+      "loss": 0.0019,
+      "step": 9120
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.893610074626866e-05,
+      "loss": 0.0059,
+      "step": 9124
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.893563432835821e-05,
+      "loss": 0.0049,
+      "step": 9128
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.8935167910447765e-05,
+      "loss": 0.0033,
+      "step": 9132
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.893470149253731e-05,
+      "loss": 0.0168,
+      "step": 9136
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.893423507462687e-05,
+      "loss": 0.0114,
+      "step": 9140
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.893376865671642e-05,
+      "loss": 0.0085,
+      "step": 9144
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.893330223880597e-05,
+      "loss": 0.0111,
+      "step": 9148
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.8932835820895526e-05,
+      "loss": 0.0053,
+      "step": 9152
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.893236940298508e-05,
+      "loss": 0.0014,
+      "step": 9156
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.893190298507463e-05,
+      "loss": 0.008,
+      "step": 9160
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.893143656716418e-05,
+      "loss": 0.0079,
+      "step": 9164
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.893097014925374e-05,
+      "loss": 0.0069,
+      "step": 9168
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.893050373134329e-05,
+      "loss": 0.0094,
+      "step": 9172
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.8930037313432835e-05,
+      "loss": 0.0086,
+      "step": 9176
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.892957089552239e-05,
+      "loss": 0.0151,
+      "step": 9180
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.8929104477611945e-05,
+      "loss": 0.0061,
+      "step": 9184
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.892863805970149e-05,
+      "loss": 0.0031,
+      "step": 9188
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.892817164179105e-05,
+      "loss": 0.013,
+      "step": 9192
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.8927705223880596e-05,
+      "loss": 0.0115,
+      "step": 9196
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.892723880597015e-05,
+      "loss": 0.0132,
+      "step": 9200
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.8926772388059706e-05,
+      "loss": 0.016,
+      "step": 9204
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.8926305970149254e-05,
+      "loss": 0.0091,
+      "step": 9208
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.892583955223881e-05,
+      "loss": 0.0022,
+      "step": 9212
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.8925373134328364e-05,
+      "loss": 0.006,
+      "step": 9216
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.892490671641791e-05,
+      "loss": 0.01,
+      "step": 9220
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.892444029850746e-05,
+      "loss": 0.0023,
+      "step": 9224
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.892397388059702e-05,
+      "loss": 0.0046,
+      "step": 9228
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.892350746268657e-05,
+      "loss": 0.0106,
+      "step": 9232
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.8923041044776124e-05,
+      "loss": 0.0152,
+      "step": 9236
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.892257462686567e-05,
+      "loss": 0.0108,
+      "step": 9240
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.892210820895523e-05,
+      "loss": 0.0146,
+      "step": 9244
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.892164179104478e-05,
+      "loss": 0.0046,
+      "step": 9248
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.892117537313433e-05,
+      "loss": 0.0106,
+      "step": 9252
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.892070895522388e-05,
+      "loss": 0.002,
+      "step": 9256
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.892024253731344e-05,
+      "loss": 0.0041,
+      "step": 9260
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.891977611940299e-05,
+      "loss": 0.0086,
+      "step": 9264
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.891930970149254e-05,
+      "loss": 0.0109,
+      "step": 9268
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.891884328358209e-05,
+      "loss": 0.0073,
+      "step": 9272
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.8918376865671646e-05,
+      "loss": 0.0168,
+      "step": 9276
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.8917910447761195e-05,
+      "loss": 0.0122,
+      "step": 9280
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.891744402985075e-05,
+      "loss": 0.0049,
+      "step": 9284
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.8916977611940304e-05,
+      "loss": 0.008,
+      "step": 9288
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.891651119402985e-05,
+      "loss": 0.0071,
+      "step": 9292
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.891604477611941e-05,
+      "loss": 0.0187,
+      "step": 9296
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 4.8915578358208956e-05,
+      "loss": 0.006,
+      "step": 9300
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 4.891511194029851e-05,
+      "loss": 0.0052,
+      "step": 9304
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 4.8914645522388065e-05,
+      "loss": 0.0079,
+      "step": 9308
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 4.8914179104477614e-05,
+      "loss": 0.0029,
+      "step": 9312
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 4.891371268656716e-05,
+      "loss": 0.0057,
+      "step": 9316
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 4.891324626865672e-05,
+      "loss": 0.0256,
+      "step": 9320
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 4.891277985074627e-05,
+      "loss": 0.0075,
+      "step": 9324
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 4.891231343283582e-05,
+      "loss": 0.0223,
+      "step": 9328
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 4.8911847014925375e-05,
+      "loss": 0.013,
+      "step": 9332
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 4.891138059701493e-05,
+      "loss": 0.0038,
+      "step": 9336
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 4.891091417910448e-05,
+      "loss": 0.014,
+      "step": 9340
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 4.891044776119403e-05,
+      "loss": 0.004,
+      "step": 9344
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 4.890998134328359e-05,
+      "loss": 0.0035,
+      "step": 9348
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 4.8909514925373136e-05,
+      "loss": 0.0139,
+      "step": 9352
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 4.890904850746269e-05,
+      "loss": 0.0054,
+      "step": 9356
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 4.890858208955224e-05,
+      "loss": 0.007,
+      "step": 9360
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 4.8908115671641793e-05,
+      "loss": 0.0136,
+      "step": 9364
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 4.890764925373135e-05,
+      "loss": 0.0025,
+      "step": 9368
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 4.8907182835820896e-05,
+      "loss": 0.0116,
+      "step": 9372
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 4.8906716417910445e-05,
+      "loss": 0.013,
+      "step": 9376
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 4.8906250000000006e-05,
+      "loss": 0.0061,
+      "step": 9380
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 4.8905783582089554e-05,
+      "loss": 0.0027,
+      "step": 9384
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 4.89053171641791e-05,
+      "loss": 0.0024,
+      "step": 9388
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 4.890485074626866e-05,
+      "loss": 0.0027,
+      "step": 9392
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 4.890438432835821e-05,
+      "loss": 0.005,
+      "step": 9396
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 4.890391791044777e-05,
+      "loss": 0.0056,
+      "step": 9400
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 4.8903451492537315e-05,
+      "loss": 0.0092,
+      "step": 9404
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 4.890298507462687e-05,
+      "loss": 0.0028,
+      "step": 9408
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 4.8902518656716425e-05,
+      "loss": 0.0263,
+      "step": 9412
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 4.890205223880597e-05,
+      "loss": 0.0037,
+      "step": 9416
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 4.890158582089552e-05,
+      "loss": 0.0013,
+      "step": 9420
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 4.8901119402985076e-05,
+      "loss": 0.0043,
+      "step": 9424
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 4.890065298507463e-05,
+      "loss": 0.016,
+      "step": 9428
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 4.890018656716418e-05,
+      "loss": 0.0099,
+      "step": 9432
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 4.8899720149253734e-05,
+      "loss": 0.0034,
+      "step": 9436
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 4.889925373134329e-05,
+      "loss": 0.0022,
+      "step": 9440
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 4.889878731343284e-05,
+      "loss": 0.0035,
+      "step": 9444
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 4.889832089552239e-05,
+      "loss": 0.0063,
+      "step": 9448
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 4.889785447761194e-05,
+      "loss": 0.0102,
+      "step": 9452
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 4.8897388059701495e-05,
+      "loss": 0.0085,
+      "step": 9456
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 4.889692164179105e-05,
+      "loss": 0.0024,
+      "step": 9460
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 4.88964552238806e-05,
+      "loss": 0.0029,
+      "step": 9464
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 4.8895988805970147e-05,
+      "loss": 0.0111,
+      "step": 9468
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 4.889552238805971e-05,
+      "loss": 0.0018,
+      "step": 9472
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 4.8895055970149256e-05,
+      "loss": 0.005,
+      "step": 9476
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 4.8894589552238804e-05,
+      "loss": 0.011,
+      "step": 9480
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 4.889412313432836e-05,
+      "loss": 0.0065,
+      "step": 9484
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 4.8893656716417914e-05,
+      "loss": 0.0071,
+      "step": 9488
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 4.889319029850746e-05,
+      "loss": 0.0172,
+      "step": 9492
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 4.889272388059702e-05,
+      "loss": 0.0057,
+      "step": 9496
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 4.889225746268657e-05,
+      "loss": 0.0052,
+      "step": 9500
+    },
+    {
+      "epoch": 8.86,
+      "eval_exact_match": 0.718568665377176,
+      "eval_exec": 0.7475822050290135,
+      "eval_loss": 0.24757526814937592,
+      "eval_runtime": 1186.3447,
+      "eval_samples_per_second": 0.872,
+      "step": 9500
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 4.889179104477612e-05,
+      "loss": 0.0078,
+      "step": 9504
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 4.8891324626865675e-05,
+      "loss": 0.0044,
+      "step": 9508
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 4.889085820895522e-05,
+      "loss": 0.0049,
+      "step": 9512
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 4.889039179104478e-05,
+      "loss": 0.0046,
+      "step": 9516
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 4.888992537313433e-05,
+      "loss": 0.0027,
+      "step": 9520
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 4.888945895522388e-05,
+      "loss": 0.0052,
+      "step": 9524
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 4.888899253731343e-05,
+      "loss": 0.0024,
+      "step": 9528
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 4.888852611940299e-05,
+      "loss": 0.0125,
+      "step": 9532
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 4.888805970149254e-05,
+      "loss": 0.0113,
+      "step": 9536
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 4.888759328358209e-05,
+      "loss": 0.0031,
+      "step": 9540
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 4.888712686567164e-05,
+      "loss": 0.0017,
+      "step": 9544
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 4.88866604477612e-05,
+      "loss": 0.0101,
+      "step": 9548
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 4.8886194029850745e-05,
+      "loss": 0.0128,
+      "step": 9552
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 4.88857276119403e-05,
+      "loss": 0.002,
+      "step": 9556
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 4.8885261194029855e-05,
+      "loss": 0.0032,
+      "step": 9560
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 4.888479477611941e-05,
+      "loss": 0.0099,
+      "step": 9564
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 4.888432835820896e-05,
+      "loss": 0.0095,
+      "step": 9568
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 4.8883861940298506e-05,
+      "loss": 0.0131,
+      "step": 9572
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 4.888339552238806e-05,
+      "loss": 0.004,
+      "step": 9576
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 4.8882929104477616e-05,
+      "loss": 0.0112,
+      "step": 9580
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 4.8882462686567164e-05,
+      "loss": 0.0112,
+      "step": 9584
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 4.888199626865672e-05,
+      "loss": 0.0062,
+      "step": 9588
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 4.8881529850746274e-05,
+      "loss": 0.0102,
+      "step": 9592
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 4.888106343283582e-05,
+      "loss": 0.0127,
+      "step": 9596
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 4.888059701492538e-05,
+      "loss": 0.0036,
+      "step": 9600
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 4.8880130597014925e-05,
+      "loss": 0.0038,
+      "step": 9604
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 4.887966417910448e-05,
+      "loss": 0.0052,
+      "step": 9608
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 4.8879197761194035e-05,
+      "loss": 0.0101,
+      "step": 9612
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 4.887873134328358e-05,
+      "loss": 0.0144,
+      "step": 9616
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 4.887826492537314e-05,
+      "loss": 0.0155,
+      "step": 9620
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 4.887779850746269e-05,
+      "loss": 0.0092,
+      "step": 9624
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 4.887733208955224e-05,
+      "loss": 0.0083,
+      "step": 9628
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 4.887686567164179e-05,
+      "loss": 0.0097,
+      "step": 9632
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 4.8876399253731344e-05,
+      "loss": 0.0148,
+      "step": 9636
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 4.88759328358209e-05,
+      "loss": 0.0022,
+      "step": 9640
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 4.887546641791045e-05,
+      "loss": 0.0131,
+      "step": 9644
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 4.8875e-05,
+      "loss": 0.0049,
+      "step": 9648
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 4.887453358208956e-05,
+      "loss": 0.004,
+      "step": 9652
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 4.8874067164179105e-05,
+      "loss": 0.0092,
+      "step": 9656
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 4.887360074626866e-05,
+      "loss": 0.0087,
+      "step": 9660
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 4.887313432835821e-05,
+      "loss": 0.0054,
+      "step": 9664
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 4.887266791044776e-05,
+      "loss": 0.0041,
+      "step": 9668
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 4.887220149253732e-05,
+      "loss": 0.0093,
+      "step": 9672
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 4.8871735074626866e-05,
+      "loss": 0.005,
+      "step": 9676
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 4.887126865671642e-05,
+      "loss": 0.0079,
+      "step": 9680
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 4.8870802238805976e-05,
+      "loss": 0.0252,
+      "step": 9684
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 4.8870335820895524e-05,
+      "loss": 0.0045,
+      "step": 9688
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 4.886986940298507e-05,
+      "loss": 0.0063,
+      "step": 9692
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 4.886940298507463e-05,
+      "loss": 0.0023,
+      "step": 9696
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 4.886893656716418e-05,
+      "loss": 0.0042,
+      "step": 9700
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 4.886847014925373e-05,
+      "loss": 0.0088,
+      "step": 9704
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 4.8868003731343285e-05,
+      "loss": 0.0078,
+      "step": 9708
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 4.886753731343284e-05,
+      "loss": 0.0094,
+      "step": 9712
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 4.886707089552239e-05,
+      "loss": 0.0031,
+      "step": 9716
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 4.886660447761194e-05,
+      "loss": 0.008,
+      "step": 9720
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 4.886613805970149e-05,
+      "loss": 0.0017,
+      "step": 9724
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 4.886567164179105e-05,
+      "loss": 0.007,
+      "step": 9728
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 4.88652052238806e-05,
+      "loss": 0.0045,
+      "step": 9732
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 4.886473880597015e-05,
+      "loss": 0.0098,
+      "step": 9736
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 4.8864272388059704e-05,
+      "loss": 0.0131,
+      "step": 9740
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 4.886380597014926e-05,
+      "loss": 0.0022,
+      "step": 9744
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 4.886333955223881e-05,
+      "loss": 0.0044,
+      "step": 9748
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 4.886287313432836e-05,
+      "loss": 0.0078,
+      "step": 9752
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 4.886240671641791e-05,
+      "loss": 0.0047,
+      "step": 9756
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 4.8861940298507465e-05,
+      "loss": 0.008,
+      "step": 9760
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 4.886147388059702e-05,
+      "loss": 0.0065,
+      "step": 9764
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 4.886100746268657e-05,
+      "loss": 0.0024,
+      "step": 9768
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 4.886054104477612e-05,
+      "loss": 0.0123,
+      "step": 9772
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 4.886007462686568e-05,
+      "loss": 0.0016,
+      "step": 9776
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 4.8859608208955226e-05,
+      "loss": 0.0018,
+      "step": 9780
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 4.8859141791044774e-05,
+      "loss": 0.0096,
+      "step": 9784
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 4.8858675373134336e-05,
+      "loss": 0.0037,
+      "step": 9788
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 4.8858208955223884e-05,
+      "loss": 0.0041,
+      "step": 9792
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 4.885774253731343e-05,
+      "loss": 0.0031,
+      "step": 9796
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 4.885727611940299e-05,
+      "loss": 0.0084,
+      "step": 9800
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 4.885680970149254e-05,
+      "loss": 0.0042,
+      "step": 9804
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 4.885634328358209e-05,
+      "loss": 0.0035,
+      "step": 9808
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 4.8855876865671645e-05,
+      "loss": 0.008,
+      "step": 9812
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 4.885541044776119e-05,
+      "loss": 0.0092,
+      "step": 9816
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 4.885494402985075e-05,
+      "loss": 0.0018,
+      "step": 9820
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 4.88544776119403e-05,
+      "loss": 0.0074,
+      "step": 9824
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 4.885401119402985e-05,
+      "loss": 0.0033,
+      "step": 9828
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 4.8853544776119406e-05,
+      "loss": 0.0062,
+      "step": 9832
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 4.885307835820896e-05,
+      "loss": 0.0108,
+      "step": 9836
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 4.885261194029851e-05,
+      "loss": 0.0013,
+      "step": 9840
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 4.885214552238806e-05,
+      "loss": 0.0041,
+      "step": 9844
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 4.885167910447762e-05,
+      "loss": 0.0033,
+      "step": 9848
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 4.885121268656717e-05,
+      "loss": 0.0097,
+      "step": 9852
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 4.8850746268656715e-05,
+      "loss": 0.0038,
+      "step": 9856
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 4.885027985074627e-05,
+      "loss": 0.0038,
+      "step": 9860
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 4.8849813432835825e-05,
+      "loss": 0.0096,
+      "step": 9864
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 4.884934701492537e-05,
+      "loss": 0.0028,
+      "step": 9868
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 4.884888059701493e-05,
+      "loss": 0.0026,
+      "step": 9872
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 4.8848414179104476e-05,
+      "loss": 0.0077,
+      "step": 9876
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 4.884794776119403e-05,
+      "loss": 0.0042,
+      "step": 9880
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 4.8847481343283586e-05,
+      "loss": 0.0126,
+      "step": 9884
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 4.8847014925373134e-05,
+      "loss": 0.0051,
+      "step": 9888
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 4.884654850746269e-05,
+      "loss": 0.0021,
+      "step": 9892
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 4.8846082089552244e-05,
+      "loss": 0.0047,
+      "step": 9896
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 4.884561567164179e-05,
+      "loss": 0.0127,
+      "step": 9900
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 4.884514925373135e-05,
+      "loss": 0.0092,
+      "step": 9904
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 4.88446828358209e-05,
+      "loss": 0.0014,
+      "step": 9908
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 4.884421641791045e-05,
+      "loss": 0.0029,
+      "step": 9912
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 4.8843750000000005e-05,
+      "loss": 0.0095,
+      "step": 9916
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 4.884328358208955e-05,
+      "loss": 0.0046,
+      "step": 9920
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 4.884281716417911e-05,
+      "loss": 0.0055,
+      "step": 9924
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 4.884235074626866e-05,
+      "loss": 0.0018,
+      "step": 9928
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 4.884188432835821e-05,
+      "loss": 0.0035,
+      "step": 9932
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 4.884141791044776e-05,
+      "loss": 0.0005,
+      "step": 9936
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 4.884095149253732e-05,
+      "loss": 0.0007,
+      "step": 9940
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 4.884048507462687e-05,
+      "loss": 0.0067,
+      "step": 9944
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 4.884001865671642e-05,
+      "loss": 0.0027,
+      "step": 9948
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 4.883955223880597e-05,
+      "loss": 0.0144,
+      "step": 9952
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 4.883908582089553e-05,
+      "loss": 0.0074,
+      "step": 9956
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 4.8838619402985075e-05,
+      "loss": 0.0047,
+      "step": 9960
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 4.883815298507463e-05,
+      "loss": 0.0044,
+      "step": 9964
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 4.8837686567164185e-05,
+      "loss": 0.0054,
+      "step": 9968
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 4.883722014925373e-05,
+      "loss": 0.0024,
+      "step": 9972
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 4.883675373134329e-05,
+      "loss": 0.0095,
+      "step": 9976
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 4.8836287313432836e-05,
+      "loss": 0.0029,
+      "step": 9980
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 4.883582089552239e-05,
+      "loss": 0.0036,
+      "step": 9984
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 4.8835354477611946e-05,
+      "loss": 0.0026,
+      "step": 9988
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 4.8834888059701494e-05,
+      "loss": 0.0053,
+      "step": 9992
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 4.883442164179104e-05,
+      "loss": 0.0005,
+      "step": 9996
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 4.8833955223880604e-05,
+      "loss": 0.005,
+      "step": 10000
+    },
+    {
+      "epoch": 9.33,
+      "eval_exact_match": 0.7166344294003868,
+      "eval_exec": 0.7543520309477756,
+      "eval_loss": 0.2650771737098694,
+      "eval_runtime": 1372.3967,
+      "eval_samples_per_second": 0.753,
+      "step": 10000
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 4.883348880597015e-05,
+      "loss": 0.0067,
+      "step": 10004
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 4.88330223880597e-05,
+      "loss": 0.0106,
+      "step": 10008
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 4.8832555970149255e-05,
+      "loss": 0.003,
+      "step": 10012
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 4.883208955223881e-05,
+      "loss": 0.0035,
+      "step": 10016
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 4.883162313432836e-05,
+      "loss": 0.0091,
+      "step": 10020
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 4.883115671641791e-05,
+      "loss": 0.0061,
+      "step": 10024
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 4.883069029850747e-05,
+      "loss": 0.0104,
+      "step": 10028
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 4.8830223880597016e-05,
+      "loss": 0.0064,
+      "step": 10032
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 4.882975746268657e-05,
+      "loss": 0.0049,
+      "step": 10036
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 4.882929104477612e-05,
+      "loss": 0.0095,
+      "step": 10040
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 4.8828824626865674e-05,
+      "loss": 0.0055,
+      "step": 10044
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 4.882835820895523e-05,
+      "loss": 0.0069,
+      "step": 10048
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 4.882789179104478e-05,
+      "loss": 0.0087,
+      "step": 10052
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 4.882742537313433e-05,
+      "loss": 0.006,
+      "step": 10056
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 4.8826958955223887e-05,
+      "loss": 0.0081,
+      "step": 10060
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 4.8826492537313435e-05,
+      "loss": 0.0057,
+      "step": 10064
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 4.882602611940299e-05,
+      "loss": 0.0143,
+      "step": 10068
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 4.882555970149254e-05,
+      "loss": 0.0006,
+      "step": 10072
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 4.882509328358209e-05,
+      "loss": 0.0026,
+      "step": 10076
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 4.882462686567165e-05,
+      "loss": 0.0029,
+      "step": 10080
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 4.8824160447761196e-05,
+      "loss": 0.0095,
+      "step": 10084
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 4.882369402985075e-05,
+      "loss": 0.0035,
+      "step": 10088
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 4.8823227611940306e-05,
+      "loss": 0.005,
+      "step": 10092
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 4.8822761194029854e-05,
+      "loss": 0.0078,
+      "step": 10096
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 4.88222947761194e-05,
+      "loss": 0.0091,
+      "step": 10100
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 4.882182835820896e-05,
+      "loss": 0.0016,
+      "step": 10104
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 4.882136194029851e-05,
+      "loss": 0.0048,
+      "step": 10108
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 4.882089552238806e-05,
+      "loss": 0.0057,
+      "step": 10112
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 4.8820429104477615e-05,
+      "loss": 0.0101,
+      "step": 10116
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 4.881996268656717e-05,
+      "loss": 0.0057,
+      "step": 10120
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 4.881949626865672e-05,
+      "loss": 0.0196,
+      "step": 10124
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 4.881902985074627e-05,
+      "loss": 0.0033,
+      "step": 10128
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 4.881856343283582e-05,
+      "loss": 0.0071,
+      "step": 10132
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 4.8818097014925376e-05,
+      "loss": 0.007,
+      "step": 10136
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 4.881763059701493e-05,
+      "loss": 0.0107,
+      "step": 10140
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 4.881716417910448e-05,
+      "loss": 0.0095,
+      "step": 10144
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 4.881669776119403e-05,
+      "loss": 0.0059,
+      "step": 10148
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 4.881623134328359e-05,
+      "loss": 0.0046,
+      "step": 10152
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 4.881576492537314e-05,
+      "loss": 0.01,
+      "step": 10156
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 4.8815298507462685e-05,
+      "loss": 0.0111,
+      "step": 10160
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 4.881483208955224e-05,
+      "loss": 0.0019,
+      "step": 10164
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 4.8814365671641795e-05,
+      "loss": 0.0036,
+      "step": 10168
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 4.881389925373134e-05,
+      "loss": 0.0077,
+      "step": 10172
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 4.88134328358209e-05,
+      "loss": 0.0048,
+      "step": 10176
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 4.881296641791045e-05,
+      "loss": 0.0053,
+      "step": 10180
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 4.88125e-05,
+      "loss": 0.009,
+      "step": 10184
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 4.8812033582089556e-05,
+      "loss": 0.0019,
+      "step": 10188
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 4.8811567164179104e-05,
+      "loss": 0.0117,
+      "step": 10192
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 4.881110074626866e-05,
+      "loss": 0.0055,
+      "step": 10196
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 4.8810634328358213e-05,
+      "loss": 0.0021,
+      "step": 10200
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 4.881016791044776e-05,
+      "loss": 0.0033,
+      "step": 10204
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 4.880970149253731e-05,
+      "loss": 0.0086,
+      "step": 10208
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 4.880923507462687e-05,
+      "loss": 0.002,
+      "step": 10212
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 4.880876865671642e-05,
+      "loss": 0.0155,
+      "step": 10216
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 4.8808302238805974e-05,
+      "loss": 0.0059,
+      "step": 10220
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 4.880783582089552e-05,
+      "loss": 0.0176,
+      "step": 10224
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 4.880736940298508e-05,
+      "loss": 0.0054,
+      "step": 10228
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 4.880690298507463e-05,
+      "loss": 0.0152,
+      "step": 10232
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 4.880643656716418e-05,
+      "loss": 0.0048,
+      "step": 10236
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 4.8805970149253735e-05,
+      "loss": 0.0032,
+      "step": 10240
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 4.880550373134329e-05,
+      "loss": 0.0011,
+      "step": 10244
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 4.880503731343284e-05,
+      "loss": 0.0076,
+      "step": 10248
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 4.880457089552239e-05,
+      "loss": 0.0022,
+      "step": 10252
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 4.880410447761195e-05,
+      "loss": 0.0075,
+      "step": 10256
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 4.8803638059701496e-05,
+      "loss": 0.0046,
+      "step": 10260
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 4.8803171641791045e-05,
+      "loss": 0.0122,
+      "step": 10264
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 4.88027052238806e-05,
+      "loss": 0.0022,
+      "step": 10268
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 4.8802238805970154e-05,
+      "loss": 0.0089,
+      "step": 10272
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 4.88017723880597e-05,
+      "loss": 0.0036,
+      "step": 10276
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 4.880130597014926e-05,
+      "loss": 0.0034,
+      "step": 10280
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 4.8800839552238806e-05,
+      "loss": 0.0018,
+      "step": 10284
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 4.880037313432836e-05,
+      "loss": 0.013,
+      "step": 10288
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 4.8799906716417915e-05,
+      "loss": 0.0064,
+      "step": 10292
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 4.8799440298507464e-05,
+      "loss": 0.0188,
+      "step": 10296
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 4.879897388059702e-05,
+      "loss": 0.0066,
+      "step": 10300
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 4.879850746268657e-05,
+      "loss": 0.002,
+      "step": 10304
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 4.879804104477612e-05,
+      "loss": 0.0185,
+      "step": 10308
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 4.879757462686567e-05,
+      "loss": 0.0018,
+      "step": 10312
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 4.8797108208955224e-05,
+      "loss": 0.0054,
+      "step": 10316
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 4.879664179104478e-05,
+      "loss": 0.0053,
+      "step": 10320
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 4.879617537313433e-05,
+      "loss": 0.0041,
+      "step": 10324
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 4.879570895522388e-05,
+      "loss": 0.0044,
+      "step": 10328
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 4.879524253731344e-05,
+      "loss": 0.0053,
+      "step": 10332
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 4.8794776119402985e-05,
+      "loss": 0.0015,
+      "step": 10336
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 4.879430970149254e-05,
+      "loss": 0.0031,
+      "step": 10340
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 4.879384328358209e-05,
+      "loss": 0.0079,
+      "step": 10344
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 4.8793376865671643e-05,
+      "loss": 0.0049,
+      "step": 10348
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 4.87929104477612e-05,
+      "loss": 0.0065,
+      "step": 10352
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 4.8792444029850746e-05,
+      "loss": 0.0036,
+      "step": 10356
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 4.87919776119403e-05,
+      "loss": 0.004,
+      "step": 10360
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 4.8791511194029856e-05,
+      "loss": 0.0037,
+      "step": 10364
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 4.8791044776119404e-05,
+      "loss": 0.007,
+      "step": 10368
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 4.879057835820895e-05,
+      "loss": 0.0059,
+      "step": 10372
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 4.879011194029851e-05,
+      "loss": 0.0114,
+      "step": 10376
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 4.878964552238806e-05,
+      "loss": 0.0061,
+      "step": 10380
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 4.878917910447762e-05,
+      "loss": 0.0068,
+      "step": 10384
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 4.8788712686567165e-05,
+      "loss": 0.0023,
+      "step": 10388
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 4.878824626865672e-05,
+      "loss": 0.0013,
+      "step": 10392
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 4.8787779850746275e-05,
+      "loss": 0.018,
+      "step": 10396
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 4.878731343283582e-05,
+      "loss": 0.0046,
+      "step": 10400
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 4.878684701492537e-05,
+      "loss": 0.0023,
+      "step": 10404
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 4.878638059701493e-05,
+      "loss": 0.0012,
+      "step": 10408
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 4.878591417910448e-05,
+      "loss": 0.0094,
+      "step": 10412
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 4.878544776119403e-05,
+      "loss": 0.0019,
+      "step": 10416
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 4.8784981343283584e-05,
+      "loss": 0.0021,
+      "step": 10420
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 4.878451492537314e-05,
+      "loss": 0.0051,
+      "step": 10424
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 4.878404850746269e-05,
+      "loss": 0.0066,
+      "step": 10428
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 4.878358208955224e-05,
+      "loss": 0.0152,
+      "step": 10432
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 4.878311567164179e-05,
+      "loss": 0.0071,
+      "step": 10436
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 4.8782649253731345e-05,
+      "loss": 0.0122,
+      "step": 10440
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 4.87821828358209e-05,
+      "loss": 0.0071,
+      "step": 10444
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 4.878171641791045e-05,
+      "loss": 0.0062,
+      "step": 10448
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 4.878125e-05,
+      "loss": 0.0091,
+      "step": 10452
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 4.878078358208956e-05,
+      "loss": 0.0056,
+      "step": 10456
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 4.8780317164179106e-05,
+      "loss": 0.0059,
+      "step": 10460
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 4.8779850746268654e-05,
+      "loss": 0.0024,
+      "step": 10464
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 4.8779384328358216e-05,
+      "loss": 0.0057,
+      "step": 10468
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 4.8778917910447764e-05,
+      "loss": 0.0105,
+      "step": 10472
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 4.877845149253731e-05,
+      "loss": 0.0117,
+      "step": 10476
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 4.877798507462687e-05,
+      "loss": 0.0049,
+      "step": 10480
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 4.877751865671642e-05,
+      "loss": 0.0039,
+      "step": 10484
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 4.877705223880597e-05,
+      "loss": 0.0077,
+      "step": 10488
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 4.8776585820895525e-05,
+      "loss": 0.0077,
+      "step": 10492
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 4.877611940298507e-05,
+      "loss": 0.0067,
+      "step": 10496
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 4.877565298507463e-05,
+      "loss": 0.0076,
+      "step": 10500
+    },
+    {
+      "epoch": 9.79,
+      "eval_exact_match": 0.7263056092843327,
+      "eval_exec": 0.746615087040619,
+      "eval_loss": 0.25435617566108704,
+      "eval_runtime": 1163.3507,
+      "eval_samples_per_second": 0.889,
+      "step": 10500
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 4.877518656716418e-05,
+      "loss": 0.0073,
+      "step": 10504
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 4.877472014925373e-05,
+      "loss": 0.0052,
+      "step": 10508
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 4.8774253731343286e-05,
+      "loss": 0.0037,
+      "step": 10512
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 4.877378731343284e-05,
+      "loss": 0.0038,
+      "step": 10516
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 4.877332089552239e-05,
+      "loss": 0.0067,
+      "step": 10520
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 4.877285447761194e-05,
+      "loss": 0.0172,
+      "step": 10524
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 4.87723880597015e-05,
+      "loss": 0.0064,
+      "step": 10528
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 4.877192164179105e-05,
+      "loss": 0.0025,
+      "step": 10532
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 4.8771455223880595e-05,
+      "loss": 0.0071,
+      "step": 10536
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 4.877098880597015e-05,
+      "loss": 0.0036,
+      "step": 10540
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.8770522388059705e-05,
+      "loss": 0.002,
+      "step": 10544
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.877005597014926e-05,
+      "loss": 0.0026,
+      "step": 10548
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.876958955223881e-05,
+      "loss": 0.0103,
+      "step": 10552
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.8769123134328356e-05,
+      "loss": 0.022,
+      "step": 10556
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.876865671641792e-05,
+      "loss": 0.0156,
+      "step": 10560
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.8768190298507466e-05,
+      "loss": 0.0061,
+      "step": 10564
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.8767723880597014e-05,
+      "loss": 0.0035,
+      "step": 10568
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.876725746268657e-05,
+      "loss": 0.0074,
+      "step": 10572
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 4.8766791044776124e-05,
+      "loss": 0.0039,
+      "step": 10576
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 4.876632462686567e-05,
+      "loss": 0.0038,
+      "step": 10580
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 4.876585820895523e-05,
+      "loss": 0.0179,
+      "step": 10584
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 4.876539179104478e-05,
+      "loss": 0.0066,
+      "step": 10588
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 4.876492537313433e-05,
+      "loss": 0.0219,
+      "step": 10592
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 4.8764458955223885e-05,
+      "loss": 0.0091,
+      "step": 10596
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 4.876399253731343e-05,
+      "loss": 0.0163,
+      "step": 10600
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 4.876352611940299e-05,
+      "loss": 0.0063,
+      "step": 10604
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 4.876305970149254e-05,
+      "loss": 0.0016,
+      "step": 10608
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 4.876259328358209e-05,
+      "loss": 0.0107,
+      "step": 10612
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 4.876212686567164e-05,
+      "loss": 0.0038,
+      "step": 10616
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 4.87616604477612e-05,
+      "loss": 0.0051,
+      "step": 10620
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 4.876119402985075e-05,
+      "loss": 0.0046,
+      "step": 10624
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 4.87607276119403e-05,
+      "loss": 0.0087,
+      "step": 10628
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 4.876026119402985e-05,
+      "loss": 0.0055,
+      "step": 10632
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 4.875979477611941e-05,
+      "loss": 0.0079,
+      "step": 10636
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 4.8759328358208955e-05,
+      "loss": 0.0017,
+      "step": 10640
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 4.875886194029851e-05,
+      "loss": 0.0048,
+      "step": 10644
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 4.8758395522388065e-05,
+      "loss": 0.0052,
+      "step": 10648
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 4.875792910447761e-05,
+      "loss": 0.0089,
+      "step": 10652
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 4.875746268656717e-05,
+      "loss": 0.0057,
+      "step": 10656
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 4.8756996268656716e-05,
+      "loss": 0.0079,
+      "step": 10660
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 4.875652985074627e-05,
+      "loss": 0.0027,
+      "step": 10664
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 4.8756063432835826e-05,
+      "loss": 0.0058,
+      "step": 10668
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 4.8755597014925374e-05,
+      "loss": 0.0024,
+      "step": 10672
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 4.875513059701492e-05,
+      "loss": 0.0095,
+      "step": 10676
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 4.8754664179104484e-05,
+      "loss": 0.0044,
+      "step": 10680
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 4.875419776119403e-05,
+      "loss": 0.0091,
+      "step": 10684
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 4.875373134328358e-05,
+      "loss": 0.0044,
+      "step": 10688
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 4.8753264925373135e-05,
+      "loss": 0.015,
+      "step": 10692
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 4.875279850746269e-05,
+      "loss": 0.0052,
+      "step": 10696
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 4.875233208955224e-05,
+      "loss": 0.0128,
+      "step": 10700
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 4.875186567164179e-05,
+      "loss": 0.0034,
+      "step": 10704
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 4.875139925373135e-05,
+      "loss": 0.0081,
+      "step": 10708
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 4.87509328358209e-05,
+      "loss": 0.0177,
+      "step": 10712
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 4.875046641791045e-05,
+      "loss": 0.0081,
+      "step": 10716
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 4.875e-05,
+      "loss": 0.0112,
+      "step": 10720
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 4.8749533582089554e-05,
+      "loss": 0.0034,
+      "step": 10724
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 4.874906716417911e-05,
+      "loss": 0.0058,
+      "step": 10728
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 4.874860074626866e-05,
+      "loss": 0.0068,
+      "step": 10732
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 4.874813432835821e-05,
+      "loss": 0.003,
+      "step": 10736
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 4.874766791044777e-05,
+      "loss": 0.0021,
+      "step": 10740
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 4.8747201492537315e-05,
+      "loss": 0.0023,
+      "step": 10744
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 4.874673507462687e-05,
+      "loss": 0.0125,
+      "step": 10748
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 4.874626865671642e-05,
+      "loss": 0.0033,
+      "step": 10752
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 4.874580223880597e-05,
+      "loss": 0.0033,
+      "step": 10756
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 4.874533582089553e-05,
+      "loss": 0.004,
+      "step": 10760
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 4.8744869402985076e-05,
+      "loss": 0.0077,
+      "step": 10764
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 4.874440298507463e-05,
+      "loss": 0.0011,
+      "step": 10768
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 4.8743936567164186e-05,
+      "loss": 0.018,
+      "step": 10772
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 4.8743470149253734e-05,
+      "loss": 0.0127,
+      "step": 10776
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 4.874300373134328e-05,
+      "loss": 0.0034,
+      "step": 10780
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 4.874253731343284e-05,
+      "loss": 0.0034,
+      "step": 10784
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 4.874207089552239e-05,
+      "loss": 0.0028,
+      "step": 10788
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 4.874160447761194e-05,
+      "loss": 0.0033,
+      "step": 10792
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 4.8741138059701495e-05,
+      "loss": 0.0063,
+      "step": 10796
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 4.874067164179105e-05,
+      "loss": 0.0076,
+      "step": 10800
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 4.87402052238806e-05,
+      "loss": 0.0073,
+      "step": 10804
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 4.873973880597015e-05,
+      "loss": 0.0182,
+      "step": 10808
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 4.87392723880597e-05,
+      "loss": 0.0039,
+      "step": 10812
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 4.8738805970149256e-05,
+      "loss": 0.0044,
+      "step": 10816
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 4.873833955223881e-05,
+      "loss": 0.0031,
+      "step": 10820
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 4.873787313432836e-05,
+      "loss": 0.004,
+      "step": 10824
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 4.873740671641791e-05,
+      "loss": 0.0014,
+      "step": 10828
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 4.873694029850747e-05,
+      "loss": 0.0011,
+      "step": 10832
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 4.873647388059702e-05,
+      "loss": 0.0087,
+      "step": 10836
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 4.8736007462686565e-05,
+      "loss": 0.0035,
+      "step": 10840
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 4.873554104477612e-05,
+      "loss": 0.0023,
+      "step": 10844
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 4.8735074626865675e-05,
+      "loss": 0.0029,
+      "step": 10848
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 4.873460820895522e-05,
+      "loss": 0.0066,
+      "step": 10852
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 4.873414179104478e-05,
+      "loss": 0.0055,
+      "step": 10856
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 4.873367537313433e-05,
+      "loss": 0.0012,
+      "step": 10860
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 4.873320895522388e-05,
+      "loss": 0.0036,
+      "step": 10864
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 4.8732742537313436e-05,
+      "loss": 0.0055,
+      "step": 10868
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 4.8732276119402984e-05,
+      "loss": 0.006,
+      "step": 10872
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 4.8731809701492546e-05,
+      "loss": 0.0038,
+      "step": 10876
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 4.8731343283582094e-05,
+      "loss": 0.0035,
+      "step": 10880
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 4.873087686567164e-05,
+      "loss": 0.0044,
+      "step": 10884
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 4.87304104477612e-05,
+      "loss": 0.0012,
+      "step": 10888
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 4.872994402985075e-05,
+      "loss": 0.0067,
+      "step": 10892
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 4.87294776119403e-05,
+      "loss": 0.0057,
+      "step": 10896
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 4.8729011194029855e-05,
+      "loss": 0.0012,
+      "step": 10900
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 4.87285447761194e-05,
+      "loss": 0.0095,
+      "step": 10904
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 4.872807835820896e-05,
+      "loss": 0.003,
+      "step": 10908
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 4.872761194029851e-05,
+      "loss": 0.0048,
+      "step": 10912
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 4.872714552238806e-05,
+      "loss": 0.0052,
+      "step": 10916
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 4.8726679104477616e-05,
+      "loss": 0.0047,
+      "step": 10920
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 4.872621268656717e-05,
+      "loss": 0.0015,
+      "step": 10924
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 4.872574626865672e-05,
+      "loss": 0.0007,
+      "step": 10928
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 4.872527985074627e-05,
+      "loss": 0.0024,
+      "step": 10932
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 4.872481343283583e-05,
+      "loss": 0.003,
+      "step": 10936
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 4.872434701492538e-05,
+      "loss": 0.0018,
+      "step": 10940
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 4.8723880597014925e-05,
+      "loss": 0.0084,
+      "step": 10944
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 4.872341417910448e-05,
+      "loss": 0.0087,
+      "step": 10948
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 4.8722947761194035e-05,
+      "loss": 0.0147,
+      "step": 10952
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 4.872248134328358e-05,
+      "loss": 0.0021,
+      "step": 10956
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 4.872201492537314e-05,
+      "loss": 0.0037,
+      "step": 10960
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 4.8721548507462686e-05,
+      "loss": 0.0021,
+      "step": 10964
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 4.872108208955224e-05,
+      "loss": 0.0016,
+      "step": 10968
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 4.8720615671641796e-05,
+      "loss": 0.003,
+      "step": 10972
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 4.8720149253731344e-05,
+      "loss": 0.0034,
+      "step": 10976
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 4.87196828358209e-05,
+      "loss": 0.0032,
+      "step": 10980
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 4.8719216417910454e-05,
+      "loss": 0.0095,
+      "step": 10984
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 4.871875e-05,
+      "loss": 0.0083,
+      "step": 10988
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 4.871828358208955e-05,
+      "loss": 0.0156,
+      "step": 10992
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 4.8717817164179105e-05,
+      "loss": 0.0089,
+      "step": 10996
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 4.871735074626866e-05,
+      "loss": 0.0049,
+      "step": 11000
+    },
+    {
+      "epoch": 10.26,
+      "eval_exact_match": 0.7388781431334622,
+      "eval_exec": 0.7514506769825918,
+      "eval_loss": 0.25768712162971497,
+      "eval_runtime": 1364.5681,
+      "eval_samples_per_second": 0.758,
+      "step": 11000
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 4.871688432835821e-05,
+      "loss": 0.0159,
+      "step": 11004
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 4.871641791044776e-05,
+      "loss": 0.0124,
+      "step": 11008
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 4.871595149253732e-05,
+      "loss": 0.0136,
+      "step": 11012
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 4.8715485074626866e-05,
+      "loss": 0.0046,
+      "step": 11016
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 4.871501865671642e-05,
+      "loss": 0.0018,
+      "step": 11020
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 4.871455223880597e-05,
+      "loss": 0.0029,
+      "step": 11024
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 4.8714085820895524e-05,
+      "loss": 0.002,
+      "step": 11028
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 4.871361940298508e-05,
+      "loss": 0.0022,
+      "step": 11032
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 4.871315298507463e-05,
+      "loss": 0.006,
+      "step": 11036
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 4.871268656716418e-05,
+      "loss": 0.005,
+      "step": 11040
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 4.8712220149253737e-05,
+      "loss": 0.0033,
+      "step": 11044
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 4.8711753731343285e-05,
+      "loss": 0.0016,
+      "step": 11048
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 4.871128731343284e-05,
+      "loss": 0.0011,
+      "step": 11052
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 4.871082089552239e-05,
+      "loss": 0.0031,
+      "step": 11056
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 4.871035447761194e-05,
+      "loss": 0.0041,
+      "step": 11060
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 4.87098880597015e-05,
+      "loss": 0.0029,
+      "step": 11064
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 4.8709421641791046e-05,
+      "loss": 0.0194,
+      "step": 11068
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 4.87089552238806e-05,
+      "loss": 0.0036,
+      "step": 11072
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 4.8708488805970156e-05,
+      "loss": 0.0092,
+      "step": 11076
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 4.8708022388059704e-05,
+      "loss": 0.0026,
+      "step": 11080
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 4.870755597014925e-05,
+      "loss": 0.0023,
+      "step": 11084
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 4.8707089552238813e-05,
+      "loss": 0.0055,
+      "step": 11088
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 4.870662313432836e-05,
+      "loss": 0.0009,
+      "step": 11092
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 4.870615671641791e-05,
+      "loss": 0.0041,
+      "step": 11096
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 4.8705690298507465e-05,
+      "loss": 0.0043,
+      "step": 11100
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 4.870522388059702e-05,
+      "loss": 0.0057,
+      "step": 11104
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 4.870475746268657e-05,
+      "loss": 0.0052,
+      "step": 11108
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 4.870429104477612e-05,
+      "loss": 0.003,
+      "step": 11112
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 4.870382462686567e-05,
+      "loss": 0.0026,
+      "step": 11116
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 4.8703358208955226e-05,
+      "loss": 0.0012,
+      "step": 11120
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 4.870289179104478e-05,
+      "loss": 0.0201,
+      "step": 11124
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 4.870242537313433e-05,
+      "loss": 0.0014,
+      "step": 11128
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 4.8701958955223884e-05,
+      "loss": 0.0032,
+      "step": 11132
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 4.870149253731344e-05,
+      "loss": 0.0047,
+      "step": 11136
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 4.8701026119402987e-05,
+      "loss": 0.0053,
+      "step": 11140
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 4.8700559701492535e-05,
+      "loss": 0.009,
+      "step": 11144
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 4.8700093283582096e-05,
+      "loss": 0.0018,
+      "step": 11148
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 4.8699626865671645e-05,
+      "loss": 0.0007,
+      "step": 11152
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 4.869916044776119e-05,
+      "loss": 0.005,
+      "step": 11156
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 4.869869402985075e-05,
+      "loss": 0.0029,
+      "step": 11160
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 4.86982276119403e-05,
+      "loss": 0.002,
+      "step": 11164
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 4.869776119402985e-05,
+      "loss": 0.0049,
+      "step": 11168
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 4.8697294776119406e-05,
+      "loss": 0.0034,
+      "step": 11172
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 4.8696828358208954e-05,
+      "loss": 0.0078,
+      "step": 11176
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 4.869636194029851e-05,
+      "loss": 0.0016,
+      "step": 11180
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 4.8695895522388063e-05,
+      "loss": 0.0114,
+      "step": 11184
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 4.869542910447761e-05,
+      "loss": 0.0186,
+      "step": 11188
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 4.8694962686567167e-05,
+      "loss": 0.0076,
+      "step": 11192
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 4.869449626865672e-05,
+      "loss": 0.0042,
+      "step": 11196
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 4.869402985074627e-05,
+      "loss": 0.019,
+      "step": 11200
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 4.8693563432835824e-05,
+      "loss": 0.0029,
+      "step": 11204
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 4.869309701492538e-05,
+      "loss": 0.0135,
+      "step": 11208
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 4.869263059701493e-05,
+      "loss": 0.002,
+      "step": 11212
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 4.869216417910448e-05,
+      "loss": 0.0021,
+      "step": 11216
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 4.869169776119403e-05,
+      "loss": 0.003,
+      "step": 11220
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 4.8691231343283585e-05,
+      "loss": 0.0094,
+      "step": 11224
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 4.869076492537314e-05,
+      "loss": 0.0035,
+      "step": 11228
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 4.869029850746269e-05,
+      "loss": 0.0061,
+      "step": 11232
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 4.868983208955224e-05,
+      "loss": 0.0027,
+      "step": 11236
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 4.86893656716418e-05,
+      "loss": 0.003,
+      "step": 11240
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 4.8688899253731346e-05,
+      "loss": 0.0033,
+      "step": 11244
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 4.8688432835820895e-05,
+      "loss": 0.004,
+      "step": 11248
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 4.868796641791045e-05,
+      "loss": 0.0024,
+      "step": 11252
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 4.8687500000000004e-05,
+      "loss": 0.0114,
+      "step": 11256
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 4.868703358208955e-05,
+      "loss": 0.0017,
+      "step": 11260
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 4.868656716417911e-05,
+      "loss": 0.0173,
+      "step": 11264
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 4.868610074626866e-05,
+      "loss": 0.0101,
+      "step": 11268
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 4.868563432835821e-05,
+      "loss": 0.0054,
+      "step": 11272
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 4.8685167910447765e-05,
+      "loss": 0.0031,
+      "step": 11276
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 4.8684701492537313e-05,
+      "loss": 0.0042,
+      "step": 11280
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 4.868423507462687e-05,
+      "loss": 0.01,
+      "step": 11284
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 4.868376865671642e-05,
+      "loss": 0.0073,
+      "step": 11288
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 4.868330223880597e-05,
+      "loss": 0.002,
+      "step": 11292
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 4.868283582089552e-05,
+      "loss": 0.0045,
+      "step": 11296
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 4.868236940298508e-05,
+      "loss": 0.0062,
+      "step": 11300
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 4.868190298507463e-05,
+      "loss": 0.0057,
+      "step": 11304
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 4.868143656716418e-05,
+      "loss": 0.0037,
+      "step": 11308
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 4.868097014925373e-05,
+      "loss": 0.0071,
+      "step": 11312
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 4.868050373134329e-05,
+      "loss": 0.0073,
+      "step": 11316
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 4.8680037313432835e-05,
+      "loss": 0.0121,
+      "step": 11320
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 4.867957089552239e-05,
+      "loss": 0.0091,
+      "step": 11324
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 4.8679104477611945e-05,
+      "loss": 0.0058,
+      "step": 11328
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 4.8678638059701493e-05,
+      "loss": 0.0047,
+      "step": 11332
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 4.867817164179105e-05,
+      "loss": 0.002,
+      "step": 11336
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 4.8677705223880596e-05,
+      "loss": 0.0051,
+      "step": 11340
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 4.867723880597015e-05,
+      "loss": 0.0096,
+      "step": 11344
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 4.8676772388059706e-05,
+      "loss": 0.0019,
+      "step": 11348
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 4.8676305970149254e-05,
+      "loss": 0.006,
+      "step": 11352
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 4.86758395522388e-05,
+      "loss": 0.0018,
+      "step": 11356
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 4.8675373134328364e-05,
+      "loss": 0.0004,
+      "step": 11360
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 4.867490671641791e-05,
+      "loss": 0.0032,
+      "step": 11364
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 4.867444029850747e-05,
+      "loss": 0.0023,
+      "step": 11368
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 4.8673973880597015e-05,
+      "loss": 0.0312,
+      "step": 11372
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 4.867350746268657e-05,
+      "loss": 0.0034,
+      "step": 11376
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 4.8673041044776125e-05,
+      "loss": 0.0044,
+      "step": 11380
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 4.867257462686567e-05,
+      "loss": 0.0065,
+      "step": 11384
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 4.867210820895523e-05,
+      "loss": 0.0032,
+      "step": 11388
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 4.867164179104478e-05,
+      "loss": 0.0054,
+      "step": 11392
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 4.867117537313433e-05,
+      "loss": 0.0187,
+      "step": 11396
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 4.867070895522388e-05,
+      "loss": 0.0059,
+      "step": 11400
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 4.8670242537313434e-05,
+      "loss": 0.0104,
+      "step": 11404
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 4.866977611940299e-05,
+      "loss": 0.0055,
+      "step": 11408
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 4.866930970149254e-05,
+      "loss": 0.0007,
+      "step": 11412
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 4.866884328358209e-05,
+      "loss": 0.0154,
+      "step": 11416
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 4.866837686567165e-05,
+      "loss": 0.0075,
+      "step": 11420
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 4.8667910447761195e-05,
+      "loss": 0.0022,
+      "step": 11424
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 4.866744402985075e-05,
+      "loss": 0.0005,
+      "step": 11428
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 4.86669776119403e-05,
+      "loss": 0.0053,
+      "step": 11432
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 4.866651119402985e-05,
+      "loss": 0.0014,
+      "step": 11436
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 4.866604477611941e-05,
+      "loss": 0.0055,
+      "step": 11440
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 4.8665578358208956e-05,
+      "loss": 0.0081,
+      "step": 11444
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 4.866511194029851e-05,
+      "loss": 0.0083,
+      "step": 11448
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 4.8664645522388066e-05,
+      "loss": 0.0017,
+      "step": 11452
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 4.8664179104477614e-05,
+      "loss": 0.0027,
+      "step": 11456
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 4.866371268656716e-05,
+      "loss": 0.0168,
+      "step": 11460
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 4.866324626865672e-05,
+      "loss": 0.0013,
+      "step": 11464
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 4.866277985074627e-05,
+      "loss": 0.0057,
+      "step": 11468
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 4.866231343283582e-05,
+      "loss": 0.0095,
+      "step": 11472
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 4.8661847014925375e-05,
+      "loss": 0.0029,
+      "step": 11476
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 4.866138059701493e-05,
+      "loss": 0.0151,
+      "step": 11480
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 4.866091417910448e-05,
+      "loss": 0.0043,
+      "step": 11484
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 4.866044776119403e-05,
+      "loss": 0.001,
+      "step": 11488
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 4.865998134328358e-05,
+      "loss": 0.0074,
+      "step": 11492
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 4.8659514925373136e-05,
+      "loss": 0.0067,
+      "step": 11496
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 4.865904850746269e-05,
+      "loss": 0.0019,
+      "step": 11500
+    },
+    {
+      "epoch": 10.73,
+      "eval_exact_match": 0.7321083172147002,
+      "eval_exec": 0.7620889748549323,
+      "eval_loss": 0.27294236421585083,
+      "eval_runtime": 4988.8833,
+      "eval_samples_per_second": 0.207,
+      "step": 11500
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 4.865858208955224e-05,
+      "loss": 0.0054,
+      "step": 11504
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 4.865811567164179e-05,
+      "loss": 0.0097,
+      "step": 11508
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 4.865764925373135e-05,
+      "loss": 0.0018,
+      "step": 11512
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 4.86571828358209e-05,
+      "loss": 0.0024,
+      "step": 11516
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 4.8656716417910445e-05,
+      "loss": 0.0051,
+      "step": 11520
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 4.865625e-05,
+      "loss": 0.0032,
+      "step": 11524
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 4.8655783582089555e-05,
+      "loss": 0.0028,
+      "step": 11528
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 4.865531716417911e-05,
+      "loss": 0.0004,
+      "step": 11532
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 4.865485074626866e-05,
+      "loss": 0.0465,
+      "step": 11536
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 4.865438432835821e-05,
+      "loss": 0.0039,
+      "step": 11540
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 4.865391791044777e-05,
+      "loss": 0.0051,
+      "step": 11544
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 4.8653451492537316e-05,
+      "loss": 0.0056,
+      "step": 11548
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 4.8652985074626864e-05,
+      "loss": 0.0043,
+      "step": 11552
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 4.8652518656716426e-05,
+      "loss": 0.0012,
+      "step": 11556
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 4.8652052238805974e-05,
+      "loss": 0.0004,
+      "step": 11560
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 4.865158582089552e-05,
+      "loss": 0.0015,
+      "step": 11564
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 4.865111940298508e-05,
+      "loss": 0.0031,
+      "step": 11568
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 4.865065298507463e-05,
+      "loss": 0.0018,
+      "step": 11572
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 4.865018656716418e-05,
+      "loss": 0.0042,
+      "step": 11576
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 4.8649720149253735e-05,
+      "loss": 0.0076,
+      "step": 11580
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 4.864925373134328e-05,
+      "loss": 0.0032,
+      "step": 11584
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 4.864878731343284e-05,
+      "loss": 0.0011,
+      "step": 11588
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 4.864832089552239e-05,
+      "loss": 0.0061,
+      "step": 11592
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 4.864785447761194e-05,
+      "loss": 0.0019,
+      "step": 11596
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 4.8647388059701496e-05,
+      "loss": 0.0035,
+      "step": 11600
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 4.864692164179105e-05,
+      "loss": 0.0057,
+      "step": 11604
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 4.86464552238806e-05,
+      "loss": 0.0037,
+      "step": 11608
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 4.864598880597015e-05,
+      "loss": 0.0068,
+      "step": 11612
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 4.864552238805971e-05,
+      "loss": 0.0087,
+      "step": 11616
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 4.864505597014926e-05,
+      "loss": 0.0074,
+      "step": 11620
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 4.8644589552238805e-05,
+      "loss": 0.0016,
+      "step": 11624
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 4.864412313432836e-05,
+      "loss": 0.0041,
+      "step": 11628
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 4.8643656716417915e-05,
+      "loss": 0.0054,
+      "step": 11632
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 4.864319029850746e-05,
+      "loss": 0.0204,
+      "step": 11636
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 4.864272388059702e-05,
+      "loss": 0.003,
+      "step": 11640
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 4.8642257462686566e-05,
+      "loss": 0.0014,
+      "step": 11644
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 4.864179104477612e-05,
+      "loss": 0.0107,
+      "step": 11648
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 4.8641324626865676e-05,
+      "loss": 0.0018,
+      "step": 11652
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 4.8640858208955224e-05,
+      "loss": 0.0025,
+      "step": 11656
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 4.864039179104478e-05,
+      "loss": 0.0021,
+      "step": 11660
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 4.8639925373134334e-05,
+      "loss": 0.0039,
+      "step": 11664
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 4.863945895522388e-05,
+      "loss": 0.0034,
+      "step": 11668
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 4.863899253731343e-05,
+      "loss": 0.0057,
+      "step": 11672
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 4.8638526119402985e-05,
+      "loss": 0.0024,
+      "step": 11676
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 4.863805970149254e-05,
+      "loss": 0.0056,
+      "step": 11680
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 4.863759328358209e-05,
+      "loss": 0.0083,
+      "step": 11684
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 4.863712686567164e-05,
+      "loss": 0.0061,
+      "step": 11688
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 4.86366604477612e-05,
+      "loss": 0.0049,
+      "step": 11692
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 4.863619402985075e-05,
+      "loss": 0.0035,
+      "step": 11696
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 4.86357276119403e-05,
+      "loss": 0.0211,
+      "step": 11700
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 4.863526119402985e-05,
+      "loss": 0.012,
+      "step": 11704
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 4.863479477611941e-05,
+      "loss": 0.0067,
+      "step": 11708
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 4.863432835820896e-05,
+      "loss": 0.0035,
+      "step": 11712
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 4.863386194029851e-05,
+      "loss": 0.0078,
+      "step": 11716
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 4.863339552238806e-05,
+      "loss": 0.0042,
+      "step": 11720
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 4.863292910447762e-05,
+      "loss": 0.0036,
+      "step": 11724
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 4.8632462686567165e-05,
+      "loss": 0.0036,
+      "step": 11728
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 4.863199626865672e-05,
+      "loss": 0.0044,
+      "step": 11732
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 4.863152985074627e-05,
+      "loss": 0.0137,
+      "step": 11736
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 4.863106343283582e-05,
+      "loss": 0.0015,
+      "step": 11740
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 4.863059701492538e-05,
+      "loss": 0.0024,
+      "step": 11744
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 4.8630130597014926e-05,
+      "loss": 0.0024,
+      "step": 11748
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 4.862966417910448e-05,
+      "loss": 0.0022,
+      "step": 11752
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 4.8629197761194036e-05,
+      "loss": 0.0009,
+      "step": 11756
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 4.8628731343283584e-05,
+      "loss": 0.0066,
+      "step": 11760
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 4.862826492537313e-05,
+      "loss": 0.0165,
+      "step": 11764
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 4.8627798507462694e-05,
+      "loss": 0.0085,
+      "step": 11768
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 4.862733208955224e-05,
+      "loss": 0.0094,
+      "step": 11772
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 4.862686567164179e-05,
+      "loss": 0.0009,
+      "step": 11776
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 4.8626399253731345e-05,
+      "loss": 0.0039,
+      "step": 11780
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 4.86259328358209e-05,
+      "loss": 0.0113,
+      "step": 11784
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 4.862546641791045e-05,
+      "loss": 0.0044,
+      "step": 11788
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 4.8625e-05,
+      "loss": 0.0027,
+      "step": 11792
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 4.862453358208955e-05,
+      "loss": 0.0014,
+      "step": 11796
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 4.8624067164179106e-05,
+      "loss": 0.0013,
+      "step": 11800
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 4.862360074626866e-05,
+      "loss": 0.0007,
+      "step": 11804
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 4.862313432835821e-05,
+      "loss": 0.004,
+      "step": 11808
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 4.8622667910447764e-05,
+      "loss": 0.0021,
+      "step": 11812
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 4.862220149253732e-05,
+      "loss": 0.0034,
+      "step": 11816
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 4.862173507462687e-05,
+      "loss": 0.0059,
+      "step": 11820
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 4.8621268656716415e-05,
+      "loss": 0.0025,
+      "step": 11824
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 4.862080223880598e-05,
+      "loss": 0.0009,
+      "step": 11828
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 4.8620335820895525e-05,
+      "loss": 0.0025,
+      "step": 11832
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 4.861986940298507e-05,
+      "loss": 0.0071,
+      "step": 11836
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 4.861940298507463e-05,
+      "loss": 0.0012,
+      "step": 11840
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 4.861893656716418e-05,
+      "loss": 0.001,
+      "step": 11844
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 4.861847014925374e-05,
+      "loss": 0.0013,
+      "step": 11848
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 4.8618003731343286e-05,
+      "loss": 0.0019,
+      "step": 11852
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 4.8617537313432834e-05,
+      "loss": 0.0056,
+      "step": 11856
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 4.8617070895522396e-05,
+      "loss": 0.0038,
+      "step": 11860
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 4.8616604477611944e-05,
+      "loss": 0.0036,
+      "step": 11864
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 4.861613805970149e-05,
+      "loss": 0.0014,
+      "step": 11868
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 4.861567164179105e-05,
+      "loss": 0.0019,
+      "step": 11872
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 4.86152052238806e-05,
+      "loss": 0.0009,
+      "step": 11876
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 4.861473880597015e-05,
+      "loss": 0.0039,
+      "step": 11880
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 4.8614272388059705e-05,
+      "loss": 0.0012,
+      "step": 11884
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 4.861380597014926e-05,
+      "loss": 0.0022,
+      "step": 11888
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 4.861333955223881e-05,
+      "loss": 0.0032,
+      "step": 11892
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 4.861287313432836e-05,
+      "loss": 0.002,
+      "step": 11896
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 4.861240671641791e-05,
+      "loss": 0.0079,
+      "step": 11900
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 4.8611940298507466e-05,
+      "loss": 0.0097,
+      "step": 11904
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 4.861147388059702e-05,
+      "loss": 0.0023,
+      "step": 11908
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 4.861100746268657e-05,
+      "loss": 0.0062,
+      "step": 11912
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 4.861054104477612e-05,
+      "loss": 0.0034,
+      "step": 11916
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 4.861007462686568e-05,
+      "loss": 0.001,
+      "step": 11920
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 4.860960820895523e-05,
+      "loss": 0.0027,
+      "step": 11924
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 4.8609141791044775e-05,
+      "loss": 0.0022,
+      "step": 11928
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 4.860867537313433e-05,
+      "loss": 0.0078,
+      "step": 11932
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 4.8608208955223885e-05,
+      "loss": 0.0028,
+      "step": 11936
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 4.860774253731343e-05,
+      "loss": 0.0079,
+      "step": 11940
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 4.860727611940299e-05,
+      "loss": 0.0014,
+      "step": 11944
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 4.860680970149254e-05,
+      "loss": 0.0053,
+      "step": 11948
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 4.860634328358209e-05,
+      "loss": 0.0048,
+      "step": 11952
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 4.8605876865671646e-05,
+      "loss": 0.0078,
+      "step": 11956
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 4.8605410447761194e-05,
+      "loss": 0.0035,
+      "step": 11960
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 4.860494402985075e-05,
+      "loss": 0.0044,
+      "step": 11964
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 4.8604477611940304e-05,
+      "loss": 0.0121,
+      "step": 11968
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 4.860401119402985e-05,
+      "loss": 0.0056,
+      "step": 11972
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 4.86035447761194e-05,
+      "loss": 0.0039,
+      "step": 11976
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 4.860307835820896e-05,
+      "loss": 0.0053,
+      "step": 11980
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 4.860261194029851e-05,
+      "loss": 0.0029,
+      "step": 11984
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 4.860214552238806e-05,
+      "loss": 0.0017,
+      "step": 11988
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 4.860167910447761e-05,
+      "loss": 0.0027,
+      "step": 11992
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 4.860121268656717e-05,
+      "loss": 0.0033,
+      "step": 11996
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 4.8600746268656716e-05,
+      "loss": 0.0026,
+      "step": 12000
+    },
+    {
+      "epoch": 11.19,
+      "eval_exact_match": 0.7292069632495164,
+      "eval_exec": 0.7553191489361702,
+      "eval_loss": 0.28799450397491455,
+      "eval_runtime": 1479.0438,
+      "eval_samples_per_second": 0.699,
+      "step": 12000
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 4.860027985074627e-05,
+      "loss": 0.0073,
+      "step": 12004
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 4.8599813432835826e-05,
+      "loss": 0.0067,
+      "step": 12008
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 4.859934701492538e-05,
+      "loss": 0.0064,
+      "step": 12012
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 4.859888059701493e-05,
+      "loss": 0.041,
+      "step": 12016
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 4.859841417910448e-05,
+      "loss": 0.0005,
+      "step": 12020
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 4.859794776119403e-05,
+      "loss": 0.004,
+      "step": 12024
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 4.8597481343283587e-05,
+      "loss": 0.0102,
+      "step": 12028
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 4.8597014925373135e-05,
+      "loss": 0.0062,
+      "step": 12032
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 4.859654850746269e-05,
+      "loss": 0.0041,
+      "step": 12036
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 4.8596082089552244e-05,
+      "loss": 0.0034,
+      "step": 12040
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 4.859561567164179e-05,
+      "loss": 0.0085,
+      "step": 12044
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 4.859514925373135e-05,
+      "loss": 0.0106,
+      "step": 12048
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 4.8594682835820896e-05,
+      "loss": 0.0103,
+      "step": 12052
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 4.859421641791045e-05,
+      "loss": 0.006,
+      "step": 12056
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 4.8593750000000005e-05,
+      "loss": 0.0047,
+      "step": 12060
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 4.8593283582089554e-05,
+      "loss": 0.0104,
+      "step": 12064
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 4.859281716417911e-05,
+      "loss": 0.0043,
+      "step": 12068
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 4.8592350746268663e-05,
+      "loss": 0.0063,
+      "step": 12072
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 4.859188432835821e-05,
+      "loss": 0.0032,
+      "step": 12076
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 4.859141791044776e-05,
+      "loss": 0.0088,
+      "step": 12080
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 4.8590951492537315e-05,
+      "loss": 0.0046,
+      "step": 12084
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 4.859048507462687e-05,
+      "loss": 0.0052,
+      "step": 12088
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 4.859001865671642e-05,
+      "loss": 0.0082,
+      "step": 12092
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 4.858955223880597e-05,
+      "loss": 0.002,
+      "step": 12096
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 4.858908582089553e-05,
+      "loss": 0.0014,
+      "step": 12100
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 4.8588619402985076e-05,
+      "loss": 0.0029,
+      "step": 12104
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 4.858815298507463e-05,
+      "loss": 0.0057,
+      "step": 12108
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 4.858768656716418e-05,
+      "loss": 0.0019,
+      "step": 12112
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 4.8587220149253734e-05,
+      "loss": 0.011,
+      "step": 12116
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 4.858675373134329e-05,
+      "loss": 0.0029,
+      "step": 12120
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 4.8586287313432837e-05,
+      "loss": 0.0103,
+      "step": 12124
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 4.858582089552239e-05,
+      "loss": 0.0034,
+      "step": 12128
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 4.8585354477611946e-05,
+      "loss": 0.0027,
+      "step": 12132
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 4.8584888059701495e-05,
+      "loss": 0.0065,
+      "step": 12136
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 4.858442164179104e-05,
+      "loss": 0.0038,
+      "step": 12140
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 4.85839552238806e-05,
+      "loss": 0.0018,
+      "step": 12144
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 4.858348880597015e-05,
+      "loss": 0.003,
+      "step": 12148
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 4.85830223880597e-05,
+      "loss": 0.0033,
+      "step": 12152
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 4.8582555970149256e-05,
+      "loss": 0.0043,
+      "step": 12156
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 4.858208955223881e-05,
+      "loss": 0.0072,
+      "step": 12160
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 4.858162313432836e-05,
+      "loss": 0.0037,
+      "step": 12164
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 4.8581156716417913e-05,
+      "loss": 0.0056,
+      "step": 12168
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 4.858069029850746e-05,
+      "loss": 0.0006,
+      "step": 12172
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 4.858022388059702e-05,
+      "loss": 0.0046,
+      "step": 12176
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 4.857975746268657e-05,
+      "loss": 0.003,
+      "step": 12180
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 4.857929104477612e-05,
+      "loss": 0.0136,
+      "step": 12184
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 4.8578824626865674e-05,
+      "loss": 0.0166,
+      "step": 12188
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 4.857835820895523e-05,
+      "loss": 0.0101,
+      "step": 12192
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 4.857789179104478e-05,
+      "loss": 0.0007,
+      "step": 12196
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 4.857742537313433e-05,
+      "loss": 0.0022,
+      "step": 12200
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 4.857695895522388e-05,
+      "loss": 0.0046,
+      "step": 12204
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 4.8576492537313435e-05,
+      "loss": 0.0103,
+      "step": 12208
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 4.857602611940299e-05,
+      "loss": 0.0057,
+      "step": 12212
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 4.857555970149254e-05,
+      "loss": 0.001,
+      "step": 12216
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 4.857509328358209e-05,
+      "loss": 0.0023,
+      "step": 12220
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 4.857462686567165e-05,
+      "loss": 0.0158,
+      "step": 12224
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 4.8574160447761196e-05,
+      "loss": 0.0056,
+      "step": 12228
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 4.8573694029850745e-05,
+      "loss": 0.0043,
+      "step": 12232
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 4.8573227611940306e-05,
+      "loss": 0.007,
+      "step": 12236
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 4.8572761194029854e-05,
+      "loss": 0.01,
+      "step": 12240
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 4.85722947761194e-05,
+      "loss": 0.0065,
+      "step": 12244
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 4.857182835820896e-05,
+      "loss": 0.0031,
+      "step": 12248
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 4.857136194029851e-05,
+      "loss": 0.0028,
+      "step": 12252
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 4.857089552238806e-05,
+      "loss": 0.0023,
+      "step": 12256
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 4.8570429104477615e-05,
+      "loss": 0.0022,
+      "step": 12260
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 4.8569962686567163e-05,
+      "loss": 0.0092,
+      "step": 12264
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 4.856949626865672e-05,
+      "loss": 0.0047,
+      "step": 12268
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 4.856902985074627e-05,
+      "loss": 0.0081,
+      "step": 12272
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 4.856856343283582e-05,
+      "loss": 0.002,
+      "step": 12276
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 4.8568097014925376e-05,
+      "loss": 0.0027,
+      "step": 12280
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 4.856763059701493e-05,
+      "loss": 0.0015,
+      "step": 12284
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 4.856716417910448e-05,
+      "loss": 0.003,
+      "step": 12288
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 4.856669776119403e-05,
+      "loss": 0.0054,
+      "step": 12292
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 4.856623134328359e-05,
+      "loss": 0.0237,
+      "step": 12296
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 4.856576492537314e-05,
+      "loss": 0.006,
+      "step": 12300
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 4.8565298507462685e-05,
+      "loss": 0.0008,
+      "step": 12304
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 4.856483208955224e-05,
+      "loss": 0.0012,
+      "step": 12308
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 4.8564365671641795e-05,
+      "loss": 0.0025,
+      "step": 12312
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 4.8563899253731343e-05,
+      "loss": 0.0073,
+      "step": 12316
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 4.85634328358209e-05,
+      "loss": 0.0037,
+      "step": 12320
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 4.8562966417910446e-05,
+      "loss": 0.0057,
+      "step": 12324
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 4.85625e-05,
+      "loss": 0.006,
+      "step": 12328
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 4.8562033582089556e-05,
+      "loss": 0.0032,
+      "step": 12332
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 4.8561567164179104e-05,
+      "loss": 0.0137,
+      "step": 12336
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 4.856110074626866e-05,
+      "loss": 0.0022,
+      "step": 12340
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 4.8560634328358214e-05,
+      "loss": 0.0046,
+      "step": 12344
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 4.856016791044776e-05,
+      "loss": 0.0016,
+      "step": 12348
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 4.855970149253732e-05,
+      "loss": 0.0054,
+      "step": 12352
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 4.8559235074626865e-05,
+      "loss": 0.015,
+      "step": 12356
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 4.855876865671642e-05,
+      "loss": 0.0028,
+      "step": 12360
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 4.8558302238805975e-05,
+      "loss": 0.004,
+      "step": 12364
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 4.855783582089552e-05,
+      "loss": 0.0046,
+      "step": 12368
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 4.855736940298508e-05,
+      "loss": 0.0031,
+      "step": 12372
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 4.855690298507463e-05,
+      "loss": 0.0008,
+      "step": 12376
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 4.855643656716418e-05,
+      "loss": 0.0011,
+      "step": 12380
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 4.855597014925373e-05,
+      "loss": 0.0039,
+      "step": 12384
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 4.855550373134329e-05,
+      "loss": 0.0028,
+      "step": 12388
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 4.855503731343284e-05,
+      "loss": 0.0044,
+      "step": 12392
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 4.855457089552239e-05,
+      "loss": 0.0022,
+      "step": 12396
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 4.855410447761194e-05,
+      "loss": 0.013,
+      "step": 12400
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 4.85536380597015e-05,
+      "loss": 0.0035,
+      "step": 12404
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 4.8553171641791045e-05,
+      "loss": 0.0038,
+      "step": 12408
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 4.85527052238806e-05,
+      "loss": 0.0004,
+      "step": 12412
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 4.855223880597015e-05,
+      "loss": 0.0019,
+      "step": 12416
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 4.85517723880597e-05,
+      "loss": 0.0123,
+      "step": 12420
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 4.855130597014926e-05,
+      "loss": 0.0087,
+      "step": 12424
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 4.8550839552238806e-05,
+      "loss": 0.0008,
+      "step": 12428
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 4.855037313432836e-05,
+      "loss": 0.0046,
+      "step": 12432
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 4.8549906716417916e-05,
+      "loss": 0.0057,
+      "step": 12436
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 4.8549440298507464e-05,
+      "loss": 0.0011,
+      "step": 12440
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 4.854897388059701e-05,
+      "loss": 0.0024,
+      "step": 12444
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 4.8548507462686574e-05,
+      "loss": 0.0038,
+      "step": 12448
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 4.854804104477612e-05,
+      "loss": 0.0114,
+      "step": 12452
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 4.854757462686567e-05,
+      "loss": 0.007,
+      "step": 12456
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 4.8547108208955225e-05,
+      "loss": 0.0028,
+      "step": 12460
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 4.854664179104478e-05,
+      "loss": 0.0039,
+      "step": 12464
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 4.854617537313433e-05,
+      "loss": 0.0057,
+      "step": 12468
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 4.854570895522388e-05,
+      "loss": 0.0133,
+      "step": 12472
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 4.854524253731343e-05,
+      "loss": 0.0005,
+      "step": 12476
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 4.8544776119402986e-05,
+      "loss": 0.0009,
+      "step": 12480
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 4.854430970149254e-05,
+      "loss": 0.0055,
+      "step": 12484
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 4.854384328358209e-05,
+      "loss": 0.0012,
+      "step": 12488
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 4.8543376865671644e-05,
+      "loss": 0.0037,
+      "step": 12492
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 4.85429104477612e-05,
+      "loss": 0.0054,
+      "step": 12496
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 4.854244402985075e-05,
+      "loss": 0.0035,
+      "step": 12500
+    },
+    {
+      "epoch": 11.66,
+      "eval_exact_match": 0.7340425531914894,
+      "eval_exec": 0.7611218568665378,
+      "eval_loss": 0.2715718448162079,
+      "eval_runtime": 1831.9341,
+      "eval_samples_per_second": 0.564,
+      "step": 12500
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 4.85419776119403e-05,
+      "loss": 0.004,
+      "step": 12504
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 4.854151119402986e-05,
+      "loss": 0.0061,
+      "step": 12508
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 4.8541044776119405e-05,
+      "loss": 0.0124,
+      "step": 12512
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 4.854057835820896e-05,
+      "loss": 0.004,
+      "step": 12516
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 4.854011194029851e-05,
+      "loss": 0.0018,
+      "step": 12520
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 4.853964552238806e-05,
+      "loss": 0.0019,
+      "step": 12524
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 4.853917910447762e-05,
+      "loss": 0.0036,
+      "step": 12528
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 4.8538712686567166e-05,
+      "loss": 0.0024,
+      "step": 12532
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 4.8538246268656714e-05,
+      "loss": 0.0188,
+      "step": 12536
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 4.8537779850746276e-05,
+      "loss": 0.0057,
+      "step": 12540
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 4.8537313432835824e-05,
+      "loss": 0.0071,
+      "step": 12544
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 4.853684701492537e-05,
+      "loss": 0.004,
+      "step": 12548
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 4.853638059701493e-05,
+      "loss": 0.0049,
+      "step": 12552
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 4.853591417910448e-05,
+      "loss": 0.0012,
+      "step": 12556
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 4.853544776119403e-05,
+      "loss": 0.0031,
+      "step": 12560
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 4.8534981343283585e-05,
+      "loss": 0.0016,
+      "step": 12564
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 4.853451492537314e-05,
+      "loss": 0.0092,
+      "step": 12568
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 4.853404850746269e-05,
+      "loss": 0.0102,
+      "step": 12572
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 4.853358208955224e-05,
+      "loss": 0.0059,
+      "step": 12576
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 4.853311567164179e-05,
+      "loss": 0.0085,
+      "step": 12580
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 4.8532649253731346e-05,
+      "loss": 0.0012,
+      "step": 12584
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 4.85321828358209e-05,
+      "loss": 0.0033,
+      "step": 12588
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 4.853171641791045e-05,
+      "loss": 0.0124,
+      "step": 12592
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 4.853125e-05,
+      "loss": 0.0054,
+      "step": 12596
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 4.853078358208956e-05,
+      "loss": 0.0055,
+      "step": 12600
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 4.853031716417911e-05,
+      "loss": 0.0073,
+      "step": 12604
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 4.8529850746268655e-05,
+      "loss": 0.0036,
+      "step": 12608
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 4.852938432835821e-05,
+      "loss": 0.01,
+      "step": 12612
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 4.8528917910447765e-05,
+      "loss": 0.0048,
+      "step": 12616
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 4.852845149253731e-05,
+      "loss": 0.0093,
+      "step": 12620
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 4.852798507462687e-05,
+      "loss": 0.0096,
+      "step": 12624
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 4.852751865671642e-05,
+      "loss": 0.0075,
+      "step": 12628
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 4.852705223880597e-05,
+      "loss": 0.0062,
+      "step": 12632
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 4.8526585820895526e-05,
+      "loss": 0.0034,
+      "step": 12636
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 4.8526119402985074e-05,
+      "loss": 0.0076,
+      "step": 12640
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 4.852565298507463e-05,
+      "loss": 0.0033,
+      "step": 12644
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 4.8525186567164184e-05,
+      "loss": 0.0045,
+      "step": 12648
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 4.852472014925373e-05,
+      "loss": 0.0009,
+      "step": 12652
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 4.852425373134328e-05,
+      "loss": 0.0036,
+      "step": 12656
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 4.852378731343284e-05,
+      "loss": 0.0031,
+      "step": 12660
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 4.852332089552239e-05,
+      "loss": 0.0046,
+      "step": 12664
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 4.8522854477611945e-05,
+      "loss": 0.0096,
+      "step": 12668
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 4.852238805970149e-05,
+      "loss": 0.0056,
+      "step": 12672
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 4.852192164179105e-05,
+      "loss": 0.001,
+      "step": 12676
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 4.85214552238806e-05,
+      "loss": 0.0037,
+      "step": 12680
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 4.852098880597015e-05,
+      "loss": 0.0024,
+      "step": 12684
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 4.8520522388059706e-05,
+      "loss": 0.0048,
+      "step": 12688
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 4.852005597014926e-05,
+      "loss": 0.0019,
+      "step": 12692
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 4.851958955223881e-05,
+      "loss": 0.0137,
+      "step": 12696
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 4.851912313432836e-05,
+      "loss": 0.0009,
+      "step": 12700
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 4.851865671641791e-05,
+      "loss": 0.0021,
+      "step": 12704
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 4.851819029850747e-05,
+      "loss": 0.0109,
+      "step": 12708
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 4.8517723880597015e-05,
+      "loss": 0.0013,
+      "step": 12712
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 4.851725746268657e-05,
+      "loss": 0.0026,
+      "step": 12716
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 4.8516791044776125e-05,
+      "loss": 0.0037,
+      "step": 12720
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 4.851632462686567e-05,
+      "loss": 0.007,
+      "step": 12724
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 4.851585820895523e-05,
+      "loss": 0.0014,
+      "step": 12728
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 4.8515391791044776e-05,
+      "loss": 0.0035,
+      "step": 12732
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 4.851492537313433e-05,
+      "loss": 0.0008,
+      "step": 12736
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 4.8514458955223886e-05,
+      "loss": 0.0022,
+      "step": 12740
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 4.8513992537313434e-05,
+      "loss": 0.0071,
+      "step": 12744
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 4.851352611940299e-05,
+      "loss": 0.0099,
+      "step": 12748
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 4.8513059701492544e-05,
+      "loss": 0.0049,
+      "step": 12752
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 4.851259328358209e-05,
+      "loss": 0.0013,
+      "step": 12756
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 4.851212686567164e-05,
+      "loss": 0.0058,
+      "step": 12760
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 4.8511660447761195e-05,
+      "loss": 0.0041,
+      "step": 12764
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 4.851119402985075e-05,
+      "loss": 0.0051,
+      "step": 12768
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 4.85107276119403e-05,
+      "loss": 0.0015,
+      "step": 12772
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 4.851026119402985e-05,
+      "loss": 0.0047,
+      "step": 12776
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 4.850979477611941e-05,
+      "loss": 0.0118,
+      "step": 12780
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 4.8509328358208956e-05,
+      "loss": 0.0017,
+      "step": 12784
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 4.850886194029851e-05,
+      "loss": 0.0031,
+      "step": 12788
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 4.850839552238806e-05,
+      "loss": 0.001,
+      "step": 12792
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 4.8507929104477614e-05,
+      "loss": 0.0053,
+      "step": 12796
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 4.850746268656717e-05,
+      "loss": 0.0016,
+      "step": 12800
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 4.850699626865672e-05,
+      "loss": 0.0067,
+      "step": 12804
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 4.850652985074627e-05,
+      "loss": 0.0027,
+      "step": 12808
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 4.850606343283583e-05,
+      "loss": 0.0021,
+      "step": 12812
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 4.8505597014925375e-05,
+      "loss": 0.0151,
+      "step": 12816
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 4.850513059701492e-05,
+      "loss": 0.0073,
+      "step": 12820
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 4.850466417910448e-05,
+      "loss": 0.0051,
+      "step": 12824
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 4.850419776119403e-05,
+      "loss": 0.0044,
+      "step": 12828
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 4.850373134328359e-05,
+      "loss": 0.0032,
+      "step": 12832
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 4.8503264925373136e-05,
+      "loss": 0.0065,
+      "step": 12836
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 4.850279850746269e-05,
+      "loss": 0.0048,
+      "step": 12840
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 4.8502332089552246e-05,
+      "loss": 0.0049,
+      "step": 12844
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 4.8501865671641794e-05,
+      "loss": 0.0006,
+      "step": 12848
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 4.850139925373134e-05,
+      "loss": 0.0011,
+      "step": 12852
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 4.8500932835820904e-05,
+      "loss": 0.0028,
+      "step": 12856
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 4.850046641791045e-05,
+      "loss": 0.0016,
+      "step": 12860
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 4.85e-05,
+      "loss": 0.004,
+      "step": 12864
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 4.8499533582089555e-05,
+      "loss": 0.0025,
+      "step": 12868
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 4.849906716417911e-05,
+      "loss": 0.0103,
+      "step": 12872
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 4.849860074626866e-05,
+      "loss": 0.0008,
+      "step": 12876
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 4.849813432835821e-05,
+      "loss": 0.0046,
+      "step": 12880
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 4.849766791044776e-05,
+      "loss": 0.0017,
+      "step": 12884
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 4.8497201492537316e-05,
+      "loss": 0.002,
+      "step": 12888
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 4.849673507462687e-05,
+      "loss": 0.0037,
+      "step": 12892
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 4.849626865671642e-05,
+      "loss": 0.0072,
+      "step": 12896
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 4.8495802238805974e-05,
+      "loss": 0.0018,
+      "step": 12900
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 4.849533582089553e-05,
+      "loss": 0.0022,
+      "step": 12904
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 4.849486940298508e-05,
+      "loss": 0.0014,
+      "step": 12908
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 4.8494402985074625e-05,
+      "loss": 0.0023,
+      "step": 12912
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 4.8493936567164187e-05,
+      "loss": 0.0034,
+      "step": 12916
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 4.8493470149253735e-05,
+      "loss": 0.0044,
+      "step": 12920
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 4.849300373134328e-05,
+      "loss": 0.0052,
+      "step": 12924
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 4.849253731343284e-05,
+      "loss": 0.0061,
+      "step": 12928
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 4.849207089552239e-05,
+      "loss": 0.0022,
+      "step": 12932
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 4.849160447761194e-05,
+      "loss": 0.0068,
+      "step": 12936
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 4.8491138059701496e-05,
+      "loss": 0.005,
+      "step": 12940
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 4.8490671641791044e-05,
+      "loss": 0.015,
+      "step": 12944
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 4.84902052238806e-05,
+      "loss": 0.0009,
+      "step": 12948
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 4.8489738805970154e-05,
+      "loss": 0.002,
+      "step": 12952
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 4.84892723880597e-05,
+      "loss": 0.0019,
+      "step": 12956
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 4.848880597014926e-05,
+      "loss": 0.0029,
+      "step": 12960
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 4.848833955223881e-05,
+      "loss": 0.0012,
+      "step": 12964
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 4.848787313432836e-05,
+      "loss": 0.0198,
+      "step": 12968
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 4.848740671641791e-05,
+      "loss": 0.0013,
+      "step": 12972
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 4.848694029850747e-05,
+      "loss": 0.0032,
+      "step": 12976
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 4.848647388059702e-05,
+      "loss": 0.0141,
+      "step": 12980
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 4.8486007462686566e-05,
+      "loss": 0.0038,
+      "step": 12984
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 4.848554104477612e-05,
+      "loss": 0.0037,
+      "step": 12988
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 4.8485074626865676e-05,
+      "loss": 0.0028,
+      "step": 12992
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 4.848460820895523e-05,
+      "loss": 0.0139,
+      "step": 12996
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 4.848414179104478e-05,
+      "loss": 0.0063,
+      "step": 13000
+    },
+    {
+      "epoch": 12.13,
+      "eval_exact_match": 0.730174081237911,
+      "eval_exec": 0.7572533849129593,
+      "eval_loss": 0.28556960821151733,
+      "eval_runtime": 1682.081,
+      "eval_samples_per_second": 0.615,
+      "step": 13000
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 4.848367537313433e-05,
+      "loss": 0.0037,
+      "step": 13004
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 4.848320895522389e-05,
+      "loss": 0.0027,
+      "step": 13008
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 4.8482742537313437e-05,
+      "loss": 0.004,
+      "step": 13012
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 4.8482276119402985e-05,
+      "loss": 0.0039,
+      "step": 13016
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 4.848180970149254e-05,
+      "loss": 0.0056,
+      "step": 13020
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 4.8481343283582094e-05,
+      "loss": 0.0067,
+      "step": 13024
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 4.848087686567164e-05,
+      "loss": 0.0021,
+      "step": 13028
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 4.84804104477612e-05,
+      "loss": 0.0053,
+      "step": 13032
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 4.8479944029850746e-05,
+      "loss": 0.0016,
+      "step": 13036
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 4.84794776119403e-05,
+      "loss": 0.0004,
+      "step": 13040
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 4.8479011194029855e-05,
+      "loss": 0.0014,
+      "step": 13044
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 4.8478544776119404e-05,
+      "loss": 0.0163,
+      "step": 13048
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 4.847807835820896e-05,
+      "loss": 0.0108,
+      "step": 13052
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 4.8477611940298513e-05,
+      "loss": 0.006,
+      "step": 13056
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 4.847714552238806e-05,
+      "loss": 0.0039,
+      "step": 13060
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 4.847667910447761e-05,
+      "loss": 0.002,
+      "step": 13064
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 4.847621268656717e-05,
+      "loss": 0.0018,
+      "step": 13068
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 4.847574626865672e-05,
+      "loss": 0.0013,
+      "step": 13072
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 4.847527985074627e-05,
+      "loss": 0.0022,
+      "step": 13076
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 4.847481343283582e-05,
+      "loss": 0.0071,
+      "step": 13080
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 4.847434701492538e-05,
+      "loss": 0.0003,
+      "step": 13084
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 4.8473880597014926e-05,
+      "loss": 0.0025,
+      "step": 13088
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 4.847341417910448e-05,
+      "loss": 0.0033,
+      "step": 13092
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 4.847294776119403e-05,
+      "loss": 0.001,
+      "step": 13096
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 4.8472481343283584e-05,
+      "loss": 0.0064,
+      "step": 13100
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 4.847201492537314e-05,
+      "loss": 0.0025,
+      "step": 13104
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 4.8471548507462687e-05,
+      "loss": 0.0011,
+      "step": 13108
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 4.847108208955224e-05,
+      "loss": 0.0005,
+      "step": 13112
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 4.8470615671641796e-05,
+      "loss": 0.0029,
+      "step": 13116
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 4.8470149253731344e-05,
+      "loss": 0.0021,
+      "step": 13120
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 4.846968283582089e-05,
+      "loss": 0.0043,
+      "step": 13124
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 4.8469216417910454e-05,
+      "loss": 0.0226,
+      "step": 13128
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 4.846875e-05,
+      "loss": 0.0006,
+      "step": 13132
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 4.846828358208955e-05,
+      "loss": 0.0005,
+      "step": 13136
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 4.8467817164179105e-05,
+      "loss": 0.0057,
+      "step": 13140
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 4.846735074626866e-05,
+      "loss": 0.006,
+      "step": 13144
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 4.846688432835821e-05,
+      "loss": 0.0037,
+      "step": 13148
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 4.8466417910447763e-05,
+      "loss": 0.0027,
+      "step": 13152
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 4.846595149253731e-05,
+      "loss": 0.0009,
+      "step": 13156
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 4.846548507462687e-05,
+      "loss": 0.0017,
+      "step": 13160
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 4.846501865671642e-05,
+      "loss": 0.004,
+      "step": 13164
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 4.846455223880597e-05,
+      "loss": 0.0093,
+      "step": 13168
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 4.8464085820895524e-05,
+      "loss": 0.0036,
+      "step": 13172
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 4.846361940298508e-05,
+      "loss": 0.0018,
+      "step": 13176
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 4.846315298507463e-05,
+      "loss": 0.0036,
+      "step": 13180
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 4.846268656716418e-05,
+      "loss": 0.0029,
+      "step": 13184
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 4.846222014925374e-05,
+      "loss": 0.0133,
+      "step": 13188
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 4.8461753731343285e-05,
+      "loss": 0.0097,
+      "step": 13192
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 4.846128731343284e-05,
+      "loss": 0.0028,
+      "step": 13196
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 4.846082089552239e-05,
+      "loss": 0.0009,
+      "step": 13200
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 4.846035447761194e-05,
+      "loss": 0.0009,
+      "step": 13204
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 4.84598880597015e-05,
+      "loss": 0.0101,
+      "step": 13208
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 4.8459421641791046e-05,
+      "loss": 0.0021,
+      "step": 13212
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 4.8458955223880595e-05,
+      "loss": 0.0066,
+      "step": 13216
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 4.8458488805970156e-05,
+      "loss": 0.0027,
+      "step": 13220
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 4.8458022388059704e-05,
+      "loss": 0.0007,
+      "step": 13224
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 4.845755597014925e-05,
+      "loss": 0.0025,
+      "step": 13228
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 4.845708955223881e-05,
+      "loss": 0.0007,
+      "step": 13232
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 4.845662313432836e-05,
+      "loss": 0.003,
+      "step": 13236
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 4.845615671641791e-05,
+      "loss": 0.0023,
+      "step": 13240
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 4.8455690298507465e-05,
+      "loss": 0.0033,
+      "step": 13244
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 4.845522388059702e-05,
+      "loss": 0.0045,
+      "step": 13248
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 4.845475746268657e-05,
+      "loss": 0.0027,
+      "step": 13252
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 4.845429104477612e-05,
+      "loss": 0.0039,
+      "step": 13256
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 4.845382462686567e-05,
+      "loss": 0.0004,
+      "step": 13260
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 4.8453358208955226e-05,
+      "loss": 0.0049,
+      "step": 13264
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 4.845289179104478e-05,
+      "loss": 0.0062,
+      "step": 13268
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 4.845242537313433e-05,
+      "loss": 0.0009,
+      "step": 13272
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 4.845195895522388e-05,
+      "loss": 0.005,
+      "step": 13276
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 4.845149253731344e-05,
+      "loss": 0.0025,
+      "step": 13280
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 4.845102611940299e-05,
+      "loss": 0.0037,
+      "step": 13284
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 4.8450559701492535e-05,
+      "loss": 0.0039,
+      "step": 13288
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 4.845009328358209e-05,
+      "loss": 0.002,
+      "step": 13292
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 4.8449626865671645e-05,
+      "loss": 0.0055,
+      "step": 13296
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 4.844916044776119e-05,
+      "loss": 0.001,
+      "step": 13300
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 4.844869402985075e-05,
+      "loss": 0.0008,
+      "step": 13304
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 4.84482276119403e-05,
+      "loss": 0.0092,
+      "step": 13308
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 4.844776119402985e-05,
+      "loss": 0.0028,
+      "step": 13312
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 4.8447294776119406e-05,
+      "loss": 0.0009,
+      "step": 13316
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 4.8446828358208954e-05,
+      "loss": 0.0088,
+      "step": 13320
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 4.844636194029851e-05,
+      "loss": 0.0014,
+      "step": 13324
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 4.8445895522388064e-05,
+      "loss": 0.0012,
+      "step": 13328
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 4.844542910447761e-05,
+      "loss": 0.002,
+      "step": 13332
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 4.844496268656717e-05,
+      "loss": 0.0031,
+      "step": 13336
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 4.844449626865672e-05,
+      "loss": 0.0013,
+      "step": 13340
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 4.844402985074627e-05,
+      "loss": 0.0059,
+      "step": 13344
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 4.8443563432835825e-05,
+      "loss": 0.0025,
+      "step": 13348
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 4.844309701492537e-05,
+      "loss": 0.0002,
+      "step": 13352
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 4.844263059701493e-05,
+      "loss": 0.002,
+      "step": 13356
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 4.844216417910448e-05,
+      "loss": 0.0007,
+      "step": 13360
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 4.844169776119403e-05,
+      "loss": 0.009,
+      "step": 13364
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 4.8441231343283586e-05,
+      "loss": 0.0065,
+      "step": 13368
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 4.844076492537314e-05,
+      "loss": 0.0114,
+      "step": 13372
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 4.844029850746269e-05,
+      "loss": 0.0021,
+      "step": 13376
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 4.843983208955224e-05,
+      "loss": 0.0036,
+      "step": 13380
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 4.843936567164179e-05,
+      "loss": 0.0027,
+      "step": 13384
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 4.843889925373135e-05,
+      "loss": 0.0042,
+      "step": 13388
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 4.8438432835820895e-05,
+      "loss": 0.0013,
+      "step": 13392
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 4.843796641791045e-05,
+      "loss": 0.0008,
+      "step": 13396
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 4.8437500000000005e-05,
+      "loss": 0.0047,
+      "step": 13400
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 4.843703358208955e-05,
+      "loss": 0.0012,
+      "step": 13404
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 4.843656716417911e-05,
+      "loss": 0.0024,
+      "step": 13408
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 4.8436100746268656e-05,
+      "loss": 0.0128,
+      "step": 13412
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 4.843563432835821e-05,
+      "loss": 0.0007,
+      "step": 13416
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 4.8435167910447766e-05,
+      "loss": 0.0055,
+      "step": 13420
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 4.8434701492537314e-05,
+      "loss": 0.0122,
+      "step": 13424
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 4.843423507462687e-05,
+      "loss": 0.0029,
+      "step": 13428
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 4.8433768656716424e-05,
+      "loss": 0.0019,
+      "step": 13432
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 4.843330223880597e-05,
+      "loss": 0.0009,
+      "step": 13436
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 4.843283582089552e-05,
+      "loss": 0.0006,
+      "step": 13440
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 4.8432369402985075e-05,
+      "loss": 0.004,
+      "step": 13444
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 4.843190298507463e-05,
+      "loss": 0.0022,
+      "step": 13448
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 4.843143656716418e-05,
+      "loss": 0.005,
+      "step": 13452
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 4.843097014925373e-05,
+      "loss": 0.0061,
+      "step": 13456
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 4.843050373134329e-05,
+      "loss": 0.0062,
+      "step": 13460
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 4.8430037313432836e-05,
+      "loss": 0.0023,
+      "step": 13464
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 4.842957089552239e-05,
+      "loss": 0.0005,
+      "step": 13468
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 4.842910447761194e-05,
+      "loss": 0.008,
+      "step": 13472
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 4.8428638059701494e-05,
+      "loss": 0.0039,
+      "step": 13476
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 4.842817164179105e-05,
+      "loss": 0.0017,
+      "step": 13480
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 4.84277052238806e-05,
+      "loss": 0.0037,
+      "step": 13484
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 4.842723880597015e-05,
+      "loss": 0.0091,
+      "step": 13488
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 4.842677238805971e-05,
+      "loss": 0.0006,
+      "step": 13492
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 4.8426305970149255e-05,
+      "loss": 0.0026,
+      "step": 13496
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 4.842583955223881e-05,
+      "loss": 0.0058,
+      "step": 13500
+    },
+    {
+      "epoch": 12.59,
+      "eval_exact_match": 0.7156673114119922,
+      "eval_exec": 0.7562862669245648,
+      "eval_loss": 0.295368492603302,
+      "eval_runtime": 1541.5565,
+      "eval_samples_per_second": 0.671,
+      "step": 13500
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 4.842537313432836e-05,
+      "loss": 0.0075,
+      "step": 13504
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 4.842490671641791e-05,
+      "loss": 0.002,
+      "step": 13508
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 4.842444029850747e-05,
+      "loss": 0.0156,
+      "step": 13512
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 4.8423973880597016e-05,
+      "loss": 0.0082,
+      "step": 13516
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 4.842350746268657e-05,
+      "loss": 0.0057,
+      "step": 13520
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 4.8423041044776126e-05,
+      "loss": 0.0022,
+      "step": 13524
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 4.8422574626865674e-05,
+      "loss": 0.0029,
+      "step": 13528
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 4.842210820895522e-05,
+      "loss": 0.0014,
+      "step": 13532
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 4.8421641791044784e-05,
+      "loss": 0.0013,
+      "step": 13536
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 4.842117537313433e-05,
+      "loss": 0.002,
+      "step": 13540
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 4.842070895522388e-05,
+      "loss": 0.0019,
+      "step": 13544
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 4.8420242537313435e-05,
+      "loss": 0.0006,
+      "step": 13548
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 4.841977611940299e-05,
+      "loss": 0.0011,
+      "step": 13552
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 4.841930970149254e-05,
+      "loss": 0.0063,
+      "step": 13556
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 4.841884328358209e-05,
+      "loss": 0.0033,
+      "step": 13560
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 4.841837686567164e-05,
+      "loss": 0.0013,
+      "step": 13564
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 4.8417910447761196e-05,
+      "loss": 0.0014,
+      "step": 13568
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 4.841744402985075e-05,
+      "loss": 0.0036,
+      "step": 13572
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 4.84169776119403e-05,
+      "loss": 0.0034,
+      "step": 13576
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 4.8416511194029854e-05,
+      "loss": 0.0062,
+      "step": 13580
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 4.841604477611941e-05,
+      "loss": 0.0032,
+      "step": 13584
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 4.841557835820896e-05,
+      "loss": 0.005,
+      "step": 13588
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 4.8415111940298505e-05,
+      "loss": 0.008,
+      "step": 13592
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 4.841464552238807e-05,
+      "loss": 0.0038,
+      "step": 13596
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 4.8414179104477615e-05,
+      "loss": 0.0033,
+      "step": 13600
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 4.841371268656716e-05,
+      "loss": 0.0008,
+      "step": 13604
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 4.841324626865672e-05,
+      "loss": 0.0041,
+      "step": 13608
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 4.841277985074627e-05,
+      "loss": 0.006,
+      "step": 13612
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 4.841231343283582e-05,
+      "loss": 0.0007,
+      "step": 13616
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 4.8411847014925376e-05,
+      "loss": 0.0035,
+      "step": 13620
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 4.8411380597014924e-05,
+      "loss": 0.0067,
+      "step": 13624
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 4.841091417910448e-05,
+      "loss": 0.0056,
+      "step": 13628
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 4.8410447761194034e-05,
+      "loss": 0.0396,
+      "step": 13632
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 4.840998134328358e-05,
+      "loss": 0.0055,
+      "step": 13636
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 4.840951492537314e-05,
+      "loss": 0.0048,
+      "step": 13640
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 4.840904850746269e-05,
+      "loss": 0.0046,
+      "step": 13644
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 4.840858208955224e-05,
+      "loss": 0.0037,
+      "step": 13648
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 4.8408115671641795e-05,
+      "loss": 0.0061,
+      "step": 13652
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 4.840764925373135e-05,
+      "loss": 0.0053,
+      "step": 13656
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 4.84071828358209e-05,
+      "loss": 0.0048,
+      "step": 13660
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 4.840671641791045e-05,
+      "loss": 0.0038,
+      "step": 13664
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 4.840625e-05,
+      "loss": 0.0069,
+      "step": 13668
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 4.8405783582089556e-05,
+      "loss": 0.0086,
+      "step": 13672
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 4.840531716417911e-05,
+      "loss": 0.0017,
+      "step": 13676
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 4.840485074626866e-05,
+      "loss": 0.0021,
+      "step": 13680
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 4.840438432835821e-05,
+      "loss": 0.004,
+      "step": 13684
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 4.840391791044777e-05,
+      "loss": 0.0007,
+      "step": 13688
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 4.840345149253732e-05,
+      "loss": 0.0118,
+      "step": 13692
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 4.8402985074626865e-05,
+      "loss": 0.0022,
+      "step": 13696
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 4.840251865671642e-05,
+      "loss": 0.0033,
+      "step": 13700
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 4.8402052238805975e-05,
+      "loss": 0.0048,
+      "step": 13704
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 4.840158582089552e-05,
+      "loss": 0.004,
+      "step": 13708
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 4.840111940298508e-05,
+      "loss": 0.0018,
+      "step": 13712
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 4.8400652985074626e-05,
+      "loss": 0.0021,
+      "step": 13716
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 4.840018656716418e-05,
+      "loss": 0.007,
+      "step": 13720
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 4.8399720149253736e-05,
+      "loss": 0.0012,
+      "step": 13724
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 4.8399253731343284e-05,
+      "loss": 0.0032,
+      "step": 13728
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 4.839878731343284e-05,
+      "loss": 0.0031,
+      "step": 13732
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 4.8398320895522394e-05,
+      "loss": 0.0064,
+      "step": 13736
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 4.839785447761194e-05,
+      "loss": 0.0105,
+      "step": 13740
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 4.839738805970149e-05,
+      "loss": 0.004,
+      "step": 13744
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 4.839692164179105e-05,
+      "loss": 0.0012,
+      "step": 13748
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 4.83964552238806e-05,
+      "loss": 0.0024,
+      "step": 13752
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 4.839598880597015e-05,
+      "loss": 0.0046,
+      "step": 13756
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 4.83955223880597e-05,
+      "loss": 0.003,
+      "step": 13760
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 4.839505597014926e-05,
+      "loss": 0.0049,
+      "step": 13764
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 4.8394589552238806e-05,
+      "loss": 0.0031,
+      "step": 13768
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 4.839412313432836e-05,
+      "loss": 0.0045,
+      "step": 13772
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 4.839365671641791e-05,
+      "loss": 0.0035,
+      "step": 13776
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 4.8393190298507464e-05,
+      "loss": 0.0107,
+      "step": 13780
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 4.839272388059702e-05,
+      "loss": 0.0014,
+      "step": 13784
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 4.839225746268657e-05,
+      "loss": 0.0063,
+      "step": 13788
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 4.839179104477612e-05,
+      "loss": 0.001,
+      "step": 13792
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 4.839132462686568e-05,
+      "loss": 0.0022,
+      "step": 13796
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 4.8390858208955225e-05,
+      "loss": 0.0021,
+      "step": 13800
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 4.839039179104477e-05,
+      "loss": 0.0031,
+      "step": 13804
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 4.8389925373134335e-05,
+      "loss": 0.0099,
+      "step": 13808
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 4.838945895522388e-05,
+      "loss": 0.0008,
+      "step": 13812
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 4.838899253731344e-05,
+      "loss": 0.009,
+      "step": 13816
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 4.8388526119402986e-05,
+      "loss": 0.01,
+      "step": 13820
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 4.838805970149254e-05,
+      "loss": 0.0051,
+      "step": 13824
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 4.8387593283582096e-05,
+      "loss": 0.0097,
+      "step": 13828
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 4.8387126865671644e-05,
+      "loss": 0.0004,
+      "step": 13832
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 4.838666044776119e-05,
+      "loss": 0.001,
+      "step": 13836
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 4.8386194029850754e-05,
+      "loss": 0.0049,
+      "step": 13840
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 4.83857276119403e-05,
+      "loss": 0.0015,
+      "step": 13844
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 4.838526119402985e-05,
+      "loss": 0.0044,
+      "step": 13848
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 4.8384794776119405e-05,
+      "loss": 0.002,
+      "step": 13852
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 4.838432835820896e-05,
+      "loss": 0.0018,
+      "step": 13856
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 4.838386194029851e-05,
+      "loss": 0.0216,
+      "step": 13860
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 4.838339552238806e-05,
+      "loss": 0.0043,
+      "step": 13864
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 4.838292910447762e-05,
+      "loss": 0.0007,
+      "step": 13868
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 4.8382462686567166e-05,
+      "loss": 0.0054,
+      "step": 13872
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 4.838199626865672e-05,
+      "loss": 0.0009,
+      "step": 13876
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 4.838152985074627e-05,
+      "loss": 0.0131,
+      "step": 13880
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 4.8381063432835824e-05,
+      "loss": 0.0049,
+      "step": 13884
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 4.838059701492538e-05,
+      "loss": 0.0013,
+      "step": 13888
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 4.838013059701493e-05,
+      "loss": 0.0049,
+      "step": 13892
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 4.8379664179104475e-05,
+      "loss": 0.004,
+      "step": 13896
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 4.8379197761194036e-05,
+      "loss": 0.0028,
+      "step": 13900
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 4.8378731343283585e-05,
+      "loss": 0.0072,
+      "step": 13904
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 4.837826492537313e-05,
+      "loss": 0.0047,
+      "step": 13908
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 4.837779850746269e-05,
+      "loss": 0.0005,
+      "step": 13912
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 4.837733208955224e-05,
+      "loss": 0.0068,
+      "step": 13916
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 4.837686567164179e-05,
+      "loss": 0.0106,
+      "step": 13920
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 4.8376399253731346e-05,
+      "loss": 0.0077,
+      "step": 13924
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 4.83759328358209e-05,
+      "loss": 0.003,
+      "step": 13928
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 4.837546641791045e-05,
+      "loss": 0.0073,
+      "step": 13932
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 4.8375000000000004e-05,
+      "loss": 0.0009,
+      "step": 13936
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 4.837453358208955e-05,
+      "loss": 0.0026,
+      "step": 13940
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 4.8374067164179107e-05,
+      "loss": 0.0009,
+      "step": 13944
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 4.837360074626866e-05,
+      "loss": 0.0012,
+      "step": 13948
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 4.837313432835821e-05,
+      "loss": 0.0035,
+      "step": 13952
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 4.837266791044776e-05,
+      "loss": 0.0041,
+      "step": 13956
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 4.837220149253732e-05,
+      "loss": 0.0033,
+      "step": 13960
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 4.837173507462687e-05,
+      "loss": 0.0029,
+      "step": 13964
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 4.8371268656716416e-05,
+      "loss": 0.0017,
+      "step": 13968
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 4.837080223880597e-05,
+      "loss": 0.0047,
+      "step": 13972
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 4.8370335820895526e-05,
+      "loss": 0.0019,
+      "step": 13976
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 4.836986940298508e-05,
+      "loss": 0.0022,
+      "step": 13980
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 4.836940298507463e-05,
+      "loss": 0.0007,
+      "step": 13984
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 4.8368936567164183e-05,
+      "loss": 0.0045,
+      "step": 13988
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 4.836847014925374e-05,
+      "loss": 0.0013,
+      "step": 13992
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 4.8368003731343287e-05,
+      "loss": 0.0051,
+      "step": 13996
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 4.8367537313432835e-05,
+      "loss": 0.0061,
+      "step": 14000
+    },
+    {
+      "epoch": 13.06,
+      "eval_exact_match": 0.7108317214700194,
+      "eval_exec": 0.7456479690522244,
+      "eval_loss": 0.29429176449775696,
+      "eval_runtime": 1353.2304,
+      "eval_samples_per_second": 0.764,
+      "step": 14000
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 4.836707089552239e-05,
+      "loss": 0.0029,
+      "step": 14004
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 4.8366604477611944e-05,
+      "loss": 0.0007,
+      "step": 14008
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 4.836613805970149e-05,
+      "loss": 0.0118,
+      "step": 14012
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 4.836567164179105e-05,
+      "loss": 0.0014,
+      "step": 14016
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 4.83652052238806e-05,
+      "loss": 0.0061,
+      "step": 14020
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 4.836473880597015e-05,
+      "loss": 0.0007,
+      "step": 14024
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 4.8364272388059705e-05,
+      "loss": 0.0039,
+      "step": 14028
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 4.8363805970149254e-05,
+      "loss": 0.0063,
+      "step": 14032
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 4.836333955223881e-05,
+      "loss": 0.0004,
+      "step": 14036
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 4.8362873134328363e-05,
+      "loss": 0.0054,
+      "step": 14040
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 4.836240671641791e-05,
+      "loss": 0.0014,
+      "step": 14044
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 4.8361940298507466e-05,
+      "loss": 0.002,
+      "step": 14048
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 4.836147388059702e-05,
+      "loss": 0.006,
+      "step": 14052
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 4.836100746268657e-05,
+      "loss": 0.0012,
+      "step": 14056
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 4.836054104477612e-05,
+      "loss": 0.0007,
+      "step": 14060
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 4.836007462686567e-05,
+      "loss": 0.0023,
+      "step": 14064
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 4.835960820895523e-05,
+      "loss": 0.001,
+      "step": 14068
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 4.8359141791044776e-05,
+      "loss": 0.0075,
+      "step": 14072
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 4.835867537313433e-05,
+      "loss": 0.0018,
+      "step": 14076
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 4.8358208955223885e-05,
+      "loss": 0.0024,
+      "step": 14080
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 4.8357742537313433e-05,
+      "loss": 0.0012,
+      "step": 14084
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 4.835727611940299e-05,
+      "loss": 0.002,
+      "step": 14088
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 4.8356809701492537e-05,
+      "loss": 0.0019,
+      "step": 14092
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 4.835634328358209e-05,
+      "loss": 0.0041,
+      "step": 14096
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 4.8355876865671646e-05,
+      "loss": 0.006,
+      "step": 14100
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 4.8355410447761194e-05,
+      "loss": 0.0048,
+      "step": 14104
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 4.835494402985075e-05,
+      "loss": 0.0118,
+      "step": 14108
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 4.8354477611940304e-05,
+      "loss": 0.0067,
+      "step": 14112
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 4.835401119402985e-05,
+      "loss": 0.0075,
+      "step": 14116
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 4.83535447761194e-05,
+      "loss": 0.0042,
+      "step": 14120
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 4.8353078358208955e-05,
+      "loss": 0.0119,
+      "step": 14124
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 4.835261194029851e-05,
+      "loss": 0.0032,
+      "step": 14128
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 4.835214552238806e-05,
+      "loss": 0.0019,
+      "step": 14132
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 4.8351679104477613e-05,
+      "loss": 0.0052,
+      "step": 14136
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 4.835121268656717e-05,
+      "loss": 0.0137,
+      "step": 14140
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 4.835074626865672e-05,
+      "loss": 0.0016,
+      "step": 14144
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 4.835027985074627e-05,
+      "loss": 0.0104,
+      "step": 14148
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 4.834981343283582e-05,
+      "loss": 0.0018,
+      "step": 14152
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 4.834934701492538e-05,
+      "loss": 0.0009,
+      "step": 14156
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 4.834888059701493e-05,
+      "loss": 0.0051,
+      "step": 14160
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 4.834841417910448e-05,
+      "loss": 0.0082,
+      "step": 14164
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 4.834794776119403e-05,
+      "loss": 0.0052,
+      "step": 14168
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 4.834748134328359e-05,
+      "loss": 0.0013,
+      "step": 14172
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 4.8347014925373135e-05,
+      "loss": 0.0037,
+      "step": 14176
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 4.834654850746269e-05,
+      "loss": 0.0014,
+      "step": 14180
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 4.834608208955224e-05,
+      "loss": 0.0066,
+      "step": 14184
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 4.834561567164179e-05,
+      "loss": 0.0038,
+      "step": 14188
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 4.834514925373135e-05,
+      "loss": 0.0055,
+      "step": 14192
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 4.8344682835820896e-05,
+      "loss": 0.0034,
+      "step": 14196
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 4.834421641791045e-05,
+      "loss": 0.0021,
+      "step": 14200
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 4.8343750000000006e-05,
+      "loss": 0.006,
+      "step": 14204
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 4.8343283582089554e-05,
+      "loss": 0.001,
+      "step": 14208
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 4.83428171641791e-05,
+      "loss": 0.0006,
+      "step": 14212
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 4.8342350746268664e-05,
+      "loss": 0.0009,
+      "step": 14216
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 4.834188432835821e-05,
+      "loss": 0.0015,
+      "step": 14220
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 4.834141791044776e-05,
+      "loss": 0.0008,
+      "step": 14224
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 4.8340951492537315e-05,
+      "loss": 0.0043,
+      "step": 14228
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 4.834048507462687e-05,
+      "loss": 0.0044,
+      "step": 14232
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 4.834001865671642e-05,
+      "loss": 0.002,
+      "step": 14236
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 4.833955223880597e-05,
+      "loss": 0.0025,
+      "step": 14240
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 4.833908582089552e-05,
+      "loss": 0.007,
+      "step": 14244
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 4.8338619402985076e-05,
+      "loss": 0.0061,
+      "step": 14248
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 4.833815298507463e-05,
+      "loss": 0.0038,
+      "step": 14252
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 4.833768656716418e-05,
+      "loss": 0.0027,
+      "step": 14256
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 4.8337220149253734e-05,
+      "loss": 0.0008,
+      "step": 14260
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 4.833675373134329e-05,
+      "loss": 0.0024,
+      "step": 14264
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 4.833628731343284e-05,
+      "loss": 0.0032,
+      "step": 14268
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 4.8335820895522385e-05,
+      "loss": 0.0079,
+      "step": 14272
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 4.833535447761195e-05,
+      "loss": 0.0065,
+      "step": 14276
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 4.8334888059701495e-05,
+      "loss": 0.0045,
+      "step": 14280
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 4.833442164179104e-05,
+      "loss": 0.0006,
+      "step": 14284
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 4.83339552238806e-05,
+      "loss": 0.0027,
+      "step": 14288
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 4.833348880597015e-05,
+      "loss": 0.0089,
+      "step": 14292
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 4.83330223880597e-05,
+      "loss": 0.0009,
+      "step": 14296
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 4.8332555970149256e-05,
+      "loss": 0.0052,
+      "step": 14300
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 4.8332089552238804e-05,
+      "loss": 0.0019,
+      "step": 14304
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 4.8331623134328366e-05,
+      "loss": 0.0033,
+      "step": 14308
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 4.8331156716417914e-05,
+      "loss": 0.0011,
+      "step": 14312
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 4.833069029850746e-05,
+      "loss": 0.0017,
+      "step": 14316
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 4.833022388059702e-05,
+      "loss": 0.0008,
+      "step": 14320
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 4.832975746268657e-05,
+      "loss": 0.001,
+      "step": 14324
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 4.832929104477612e-05,
+      "loss": 0.0044,
+      "step": 14328
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 4.8328824626865675e-05,
+      "loss": 0.0029,
+      "step": 14332
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 4.832835820895523e-05,
+      "loss": 0.0106,
+      "step": 14336
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 4.832789179104478e-05,
+      "loss": 0.016,
+      "step": 14340
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 4.832742537313433e-05,
+      "loss": 0.0078,
+      "step": 14344
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 4.832695895522388e-05,
+      "loss": 0.0097,
+      "step": 14348
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 4.8326492537313436e-05,
+      "loss": 0.0039,
+      "step": 14352
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 4.832602611940299e-05,
+      "loss": 0.0068,
+      "step": 14356
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 4.832555970149254e-05,
+      "loss": 0.0163,
+      "step": 14360
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 4.832509328358209e-05,
+      "loss": 0.0011,
+      "step": 14364
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 4.832462686567165e-05,
+      "loss": 0.0113,
+      "step": 14368
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 4.83241604477612e-05,
+      "loss": 0.0027,
+      "step": 14372
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 4.8323694029850745e-05,
+      "loss": 0.0008,
+      "step": 14376
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 4.83232276119403e-05,
+      "loss": 0.003,
+      "step": 14380
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 4.8322761194029855e-05,
+      "loss": 0.0014,
+      "step": 14384
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 4.83222947761194e-05,
+      "loss": 0.0042,
+      "step": 14388
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 4.832182835820896e-05,
+      "loss": 0.0017,
+      "step": 14392
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 4.8321361940298506e-05,
+      "loss": 0.0041,
+      "step": 14396
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 4.832089552238806e-05,
+      "loss": 0.0012,
+      "step": 14400
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 4.8320429104477616e-05,
+      "loss": 0.0005,
+      "step": 14404
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 4.8319962686567164e-05,
+      "loss": 0.0005,
+      "step": 14408
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 4.831949626865672e-05,
+      "loss": 0.0031,
+      "step": 14412
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 4.8319029850746274e-05,
+      "loss": 0.007,
+      "step": 14416
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 4.831856343283582e-05,
+      "loss": 0.0006,
+      "step": 14420
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 4.831809701492537e-05,
+      "loss": 0.0038,
+      "step": 14424
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 4.831763059701493e-05,
+      "loss": 0.0033,
+      "step": 14428
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 4.831716417910448e-05,
+      "loss": 0.0012,
+      "step": 14432
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 4.831669776119403e-05,
+      "loss": 0.005,
+      "step": 14436
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 4.831623134328358e-05,
+      "loss": 0.01,
+      "step": 14440
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 4.831576492537314e-05,
+      "loss": 0.0034,
+      "step": 14444
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 4.8315298507462686e-05,
+      "loss": 0.0139,
+      "step": 14448
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 4.831483208955224e-05,
+      "loss": 0.0069,
+      "step": 14452
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 4.831436567164179e-05,
+      "loss": 0.0012,
+      "step": 14456
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 4.8313899253731344e-05,
+      "loss": 0.004,
+      "step": 14460
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 4.83134328358209e-05,
+      "loss": 0.0009,
+      "step": 14464
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 4.831296641791045e-05,
+      "loss": 0.0009,
+      "step": 14468
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 4.83125e-05,
+      "loss": 0.0009,
+      "step": 14472
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 4.831203358208956e-05,
+      "loss": 0.0011,
+      "step": 14476
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 4.8311567164179105e-05,
+      "loss": 0.0003,
+      "step": 14480
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 4.831110074626866e-05,
+      "loss": 0.0001,
+      "step": 14484
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 4.8310634328358215e-05,
+      "loss": 0.0037,
+      "step": 14488
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 4.831016791044776e-05,
+      "loss": 0.0035,
+      "step": 14492
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 4.830970149253732e-05,
+      "loss": 0.0006,
+      "step": 14496
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 4.8309235074626866e-05,
+      "loss": 0.0156,
+      "step": 14500
+    },
+    {
+      "epoch": 13.53,
+      "eval_exact_match": 0.7224371373307543,
+      "eval_exec": 0.7591876208897486,
+      "eval_loss": 0.2937206029891968,
+      "eval_runtime": 1346.8122,
+      "eval_samples_per_second": 0.768,
+      "step": 14500
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 4.830876865671642e-05,
+      "loss": 0.0009,
+      "step": 14504
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 4.8308302238805976e-05,
+      "loss": 0.0014,
+      "step": 14508
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 4.8307835820895524e-05,
+      "loss": 0.0018,
+      "step": 14512
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 4.830736940298507e-05,
+      "loss": 0.0028,
+      "step": 14516
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 4.8306902985074634e-05,
+      "loss": 0.0014,
+      "step": 14520
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 4.830643656716418e-05,
+      "loss": 0.0018,
+      "step": 14524
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 4.830597014925373e-05,
+      "loss": 0.0011,
+      "step": 14528
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 4.8305503731343285e-05,
+      "loss": 0.0041,
+      "step": 14532
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 4.830503731343284e-05,
+      "loss": 0.0035,
+      "step": 14536
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 4.830457089552239e-05,
+      "loss": 0.0096,
+      "step": 14540
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 4.830410447761194e-05,
+      "loss": 0.0062,
+      "step": 14544
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 4.83036380597015e-05,
+      "loss": 0.0106,
+      "step": 14548
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 4.8303171641791046e-05,
+      "loss": 0.0049,
+      "step": 14552
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 4.83027052238806e-05,
+      "loss": 0.0069,
+      "step": 14556
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 4.830223880597015e-05,
+      "loss": 0.0088,
+      "step": 14560
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 4.8301772388059704e-05,
+      "loss": 0.0012,
+      "step": 14564
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 4.830130597014926e-05,
+      "loss": 0.0029,
+      "step": 14568
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 4.830083955223881e-05,
+      "loss": 0.0005,
+      "step": 14572
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 4.8300373134328355e-05,
+      "loss": 0.001,
+      "step": 14576
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 4.829990671641792e-05,
+      "loss": 0.0041,
+      "step": 14580
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 4.8299440298507465e-05,
+      "loss": 0.0037,
+      "step": 14584
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 4.829897388059701e-05,
+      "loss": 0.0035,
+      "step": 14588
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 4.829850746268657e-05,
+      "loss": 0.0068,
+      "step": 14592
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 4.829804104477612e-05,
+      "loss": 0.0018,
+      "step": 14596
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 4.829757462686567e-05,
+      "loss": 0.0029,
+      "step": 14600
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 4.8297108208955226e-05,
+      "loss": 0.0013,
+      "step": 14604
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 4.829664179104478e-05,
+      "loss": 0.0051,
+      "step": 14608
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 4.829617537313433e-05,
+      "loss": 0.0012,
+      "step": 14612
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 4.8295708955223884e-05,
+      "loss": 0.0021,
+      "step": 14616
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 4.829524253731343e-05,
+      "loss": 0.0019,
+      "step": 14620
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 4.829477611940299e-05,
+      "loss": 0.0023,
+      "step": 14624
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 4.829430970149254e-05,
+      "loss": 0.0015,
+      "step": 14628
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 4.829384328358209e-05,
+      "loss": 0.002,
+      "step": 14632
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 4.8293376865671645e-05,
+      "loss": 0.0095,
+      "step": 14636
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 4.82929104477612e-05,
+      "loss": 0.006,
+      "step": 14640
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 4.829244402985075e-05,
+      "loss": 0.0024,
+      "step": 14644
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 4.82919776119403e-05,
+      "loss": 0.0022,
+      "step": 14648
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 4.829151119402985e-05,
+      "loss": 0.0016,
+      "step": 14652
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 4.8291044776119406e-05,
+      "loss": 0.0011,
+      "step": 14656
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 4.829057835820896e-05,
+      "loss": 0.001,
+      "step": 14660
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 4.829011194029851e-05,
+      "loss": 0.0068,
+      "step": 14664
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 4.8289645522388064e-05,
+      "loss": 0.0007,
+      "step": 14668
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 4.828917910447762e-05,
+      "loss": 0.0044,
+      "step": 14672
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 4.828871268656717e-05,
+      "loss": 0.0097,
+      "step": 14676
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 4.8288246268656715e-05,
+      "loss": 0.0004,
+      "step": 14680
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 4.828777985074627e-05,
+      "loss": 0.0077,
+      "step": 14684
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 4.8287313432835825e-05,
+      "loss": 0.0011,
+      "step": 14688
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 4.828684701492537e-05,
+      "loss": 0.0008,
+      "step": 14692
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 4.828638059701493e-05,
+      "loss": 0.0111,
+      "step": 14696
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 4.828591417910448e-05,
+      "loss": 0.0005,
+      "step": 14700
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 4.828544776119403e-05,
+      "loss": 0.0037,
+      "step": 14704
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 4.8284981343283586e-05,
+      "loss": 0.0024,
+      "step": 14708
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 4.8284514925373134e-05,
+      "loss": 0.0024,
+      "step": 14712
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 4.828404850746269e-05,
+      "loss": 0.0052,
+      "step": 14716
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 4.8283582089552244e-05,
+      "loss": 0.0005,
+      "step": 14720
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 4.828311567164179e-05,
+      "loss": 0.006,
+      "step": 14724
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 4.828264925373135e-05,
+      "loss": 0.0023,
+      "step": 14728
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 4.82821828358209e-05,
+      "loss": 0.0033,
+      "step": 14732
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 4.828171641791045e-05,
+      "loss": 0.0012,
+      "step": 14736
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 4.828125e-05,
+      "loss": 0.0076,
+      "step": 14740
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 4.828078358208955e-05,
+      "loss": 0.0063,
+      "step": 14744
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 4.828031716417911e-05,
+      "loss": 0.0086,
+      "step": 14748
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 4.8279850746268656e-05,
+      "loss": 0.0102,
+      "step": 14752
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 4.827938432835821e-05,
+      "loss": 0.0011,
+      "step": 14756
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 4.8278917910447766e-05,
+      "loss": 0.0056,
+      "step": 14760
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 4.8278451492537314e-05,
+      "loss": 0.0025,
+      "step": 14764
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 4.827798507462687e-05,
+      "loss": 0.0076,
+      "step": 14768
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 4.827751865671642e-05,
+      "loss": 0.0081,
+      "step": 14772
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 4.827705223880597e-05,
+      "loss": 0.0039,
+      "step": 14776
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 4.827658582089553e-05,
+      "loss": 0.0013,
+      "step": 14780
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 4.8276119402985075e-05,
+      "loss": 0.0019,
+      "step": 14784
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 4.827565298507463e-05,
+      "loss": 0.0004,
+      "step": 14788
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 4.8275186567164185e-05,
+      "loss": 0.0035,
+      "step": 14792
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 4.827472014925373e-05,
+      "loss": 0.0006,
+      "step": 14796
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 4.827425373134329e-05,
+      "loss": 0.0058,
+      "step": 14800
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 4.8273787313432836e-05,
+      "loss": 0.0014,
+      "step": 14804
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 4.827332089552239e-05,
+      "loss": 0.0023,
+      "step": 14808
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 4.8272854477611946e-05,
+      "loss": 0.0006,
+      "step": 14812
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 4.8272388059701494e-05,
+      "loss": 0.0057,
+      "step": 14816
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 4.827192164179105e-05,
+      "loss": 0.0039,
+      "step": 14820
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 4.8271455223880604e-05,
+      "loss": 0.0086,
+      "step": 14824
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 4.827098880597015e-05,
+      "loss": 0.0012,
+      "step": 14828
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 4.82705223880597e-05,
+      "loss": 0.0007,
+      "step": 14832
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 4.827005597014926e-05,
+      "loss": 0.0006,
+      "step": 14836
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 4.826958955223881e-05,
+      "loss": 0.0049,
+      "step": 14840
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 4.826912313432836e-05,
+      "loss": 0.0008,
+      "step": 14844
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 4.826865671641791e-05,
+      "loss": 0.001,
+      "step": 14848
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 4.826819029850747e-05,
+      "loss": 0.0003,
+      "step": 14852
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 4.8267723880597016e-05,
+      "loss": 0.0022,
+      "step": 14856
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 4.826725746268657e-05,
+      "loss": 0.0009,
+      "step": 14860
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 4.826679104477612e-05,
+      "loss": 0.0005,
+      "step": 14864
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 4.8266324626865674e-05,
+      "loss": 0.0038,
+      "step": 14868
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 4.826585820895523e-05,
+      "loss": 0.002,
+      "step": 14872
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 4.826539179104478e-05,
+      "loss": 0.01,
+      "step": 14876
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 4.826492537313433e-05,
+      "loss": 0.006,
+      "step": 14880
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 4.8264458955223886e-05,
+      "loss": 0.011,
+      "step": 14884
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 4.8263992537313435e-05,
+      "loss": 0.0027,
+      "step": 14888
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 4.826352611940298e-05,
+      "loss": 0.0025,
+      "step": 14892
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 4.8263059701492544e-05,
+      "loss": 0.0008,
+      "step": 14896
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 4.826259328358209e-05,
+      "loss": 0.0017,
+      "step": 14900
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 4.826212686567164e-05,
+      "loss": 0.001,
+      "step": 14904
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 4.8261660447761196e-05,
+      "loss": 0.0004,
+      "step": 14908
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 4.826119402985075e-05,
+      "loss": 0.0025,
+      "step": 14912
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 4.82607276119403e-05,
+      "loss": 0.0011,
+      "step": 14916
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 4.8260261194029854e-05,
+      "loss": 0.0007,
+      "step": 14920
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 4.82597947761194e-05,
+      "loss": 0.0015,
+      "step": 14924
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 4.8259328358208957e-05,
+      "loss": 0.0101,
+      "step": 14928
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 4.825886194029851e-05,
+      "loss": 0.0003,
+      "step": 14932
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 4.825839552238806e-05,
+      "loss": 0.0052,
+      "step": 14936
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 4.8257929104477615e-05,
+      "loss": 0.0044,
+      "step": 14940
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 4.825746268656717e-05,
+      "loss": 0.0005,
+      "step": 14944
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 4.825699626865672e-05,
+      "loss": 0.0054,
+      "step": 14948
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 4.8256529850746266e-05,
+      "loss": 0.0016,
+      "step": 14952
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 4.825606343283583e-05,
+      "loss": 0.0017,
+      "step": 14956
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 4.8255597014925376e-05,
+      "loss": 0.0065,
+      "step": 14960
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 4.825513059701493e-05,
+      "loss": 0.001,
+      "step": 14964
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 4.825466417910448e-05,
+      "loss": 0.012,
+      "step": 14968
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 4.8254197761194033e-05,
+      "loss": 0.0023,
+      "step": 14972
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 4.825373134328359e-05,
+      "loss": 0.0053,
+      "step": 14976
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 4.8253264925373136e-05,
+      "loss": 0.0021,
+      "step": 14980
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 4.8252798507462685e-05,
+      "loss": 0.0019,
+      "step": 14984
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 4.8252332089552246e-05,
+      "loss": 0.002,
+      "step": 14988
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 4.8251865671641794e-05,
+      "loss": 0.0069,
+      "step": 14992
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 4.825139925373134e-05,
+      "loss": 0.0035,
+      "step": 14996
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 4.82509328358209e-05,
+      "loss": 0.0026,
+      "step": 15000
+    },
+    {
+      "epoch": 13.99,
+      "eval_exact_match": 0.7156673114119922,
+      "eval_exec": 0.7524177949709865,
+      "eval_loss": 0.31172001361846924,
+      "eval_runtime": 1126.7509,
+      "eval_samples_per_second": 0.918,
+      "step": 15000
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 4.825046641791045e-05,
+      "loss": 0.0005,
+      "step": 15004
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 4.825e-05,
+      "loss": 0.0089,
+      "step": 15008
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 4.8249533582089555e-05,
+      "loss": 0.0092,
+      "step": 15012
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 4.824906716417911e-05,
+      "loss": 0.001,
+      "step": 15016
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 4.824860074626866e-05,
+      "loss": 0.0052,
+      "step": 15020
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 4.824813432835821e-05,
+      "loss": 0.0005,
+      "step": 15024
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 4.824766791044776e-05,
+      "loss": 0.0014,
+      "step": 15028
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 4.8247201492537316e-05,
+      "loss": 0.0028,
+      "step": 15032
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 4.824673507462687e-05,
+      "loss": 0.0008,
+      "step": 15036
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 4.824626865671642e-05,
+      "loss": 0.0031,
+      "step": 15040
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 4.824580223880597e-05,
+      "loss": 0.0055,
+      "step": 15044
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 4.824533582089553e-05,
+      "loss": 0.0049,
+      "step": 15048
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 4.824486940298508e-05,
+      "loss": 0.0045,
+      "step": 15052
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 4.8244402985074626e-05,
+      "loss": 0.0033,
+      "step": 15056
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 4.824393656716418e-05,
+      "loss": 0.0015,
+      "step": 15060
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 4.8243470149253735e-05,
+      "loss": 0.0027,
+      "step": 15064
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 4.8243003731343283e-05,
+      "loss": 0.0007,
+      "step": 15068
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 4.824253731343284e-05,
+      "loss": 0.001,
+      "step": 15072
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 4.8242070895522387e-05,
+      "loss": 0.001,
+      "step": 15076
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 4.824160447761194e-05,
+      "loss": 0.0022,
+      "step": 15080
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 4.8241138059701496e-05,
+      "loss": 0.0005,
+      "step": 15084
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 4.8240671641791044e-05,
+      "loss": 0.0004,
+      "step": 15088
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 4.82402052238806e-05,
+      "loss": 0.0027,
+      "step": 15092
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 4.8239738805970154e-05,
+      "loss": 0.0034,
+      "step": 15096
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 4.82392723880597e-05,
+      "loss": 0.0063,
+      "step": 15100
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 4.823880597014925e-05,
+      "loss": 0.0004,
+      "step": 15104
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 4.823833955223881e-05,
+      "loss": 0.0002,
+      "step": 15108
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 4.823787313432836e-05,
+      "loss": 0.0029,
+      "step": 15112
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 4.8237406716417915e-05,
+      "loss": 0.0011,
+      "step": 15116
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 4.8236940298507463e-05,
+      "loss": 0.0043,
+      "step": 15120
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 4.823647388059702e-05,
+      "loss": 0.0084,
+      "step": 15124
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 4.823600746268657e-05,
+      "loss": 0.0062,
+      "step": 15128
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 4.823554104477612e-05,
+      "loss": 0.0088,
+      "step": 15132
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 4.823507462686567e-05,
+      "loss": 0.0005,
+      "step": 15136
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 4.823460820895523e-05,
+      "loss": 0.0078,
+      "step": 15140
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 4.823414179104478e-05,
+      "loss": 0.0007,
+      "step": 15144
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 4.823367537313433e-05,
+      "loss": 0.0056,
+      "step": 15148
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 4.823320895522388e-05,
+      "loss": 0.0005,
+      "step": 15152
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 4.823274253731344e-05,
+      "loss": 0.0117,
+      "step": 15156
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 4.8232276119402985e-05,
+      "loss": 0.0032,
+      "step": 15160
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 4.823180970149254e-05,
+      "loss": 0.0032,
+      "step": 15164
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 4.8231343283582095e-05,
+      "loss": 0.0061,
+      "step": 15168
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 4.823087686567164e-05,
+      "loss": 0.0018,
+      "step": 15172
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 4.82304104477612e-05,
+      "loss": 0.0101,
+      "step": 15176
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 4.8229944029850746e-05,
+      "loss": 0.0014,
+      "step": 15180
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 4.82294776119403e-05,
+      "loss": 0.0058,
+      "step": 15184
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 4.8229011194029856e-05,
+      "loss": 0.0036,
+      "step": 15188
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 4.8228544776119404e-05,
+      "loss": 0.0013,
+      "step": 15192
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 4.822807835820895e-05,
+      "loss": 0.0029,
+      "step": 15196
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 4.8227611940298514e-05,
+      "loss": 0.0063,
+      "step": 15200
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 4.822714552238806e-05,
+      "loss": 0.0028,
+      "step": 15204
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 4.822667910447761e-05,
+      "loss": 0.0006,
+      "step": 15208
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 4.8226212686567165e-05,
+      "loss": 0.0014,
+      "step": 15212
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 4.822574626865672e-05,
+      "loss": 0.0005,
+      "step": 15216
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 4.822527985074627e-05,
+      "loss": 0.0019,
+      "step": 15220
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 4.822481343283582e-05,
+      "loss": 0.0003,
+      "step": 15224
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 4.822434701492538e-05,
+      "loss": 0.001,
+      "step": 15228
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 4.8223880597014926e-05,
+      "loss": 0.0025,
+      "step": 15232
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 4.822341417910448e-05,
+      "loss": 0.0007,
+      "step": 15236
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 4.822294776119403e-05,
+      "loss": 0.0043,
+      "step": 15240
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 4.8222481343283584e-05,
+      "loss": 0.0014,
+      "step": 15244
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 4.822201492537314e-05,
+      "loss": 0.0025,
+      "step": 15248
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 4.822154850746269e-05,
+      "loss": 0.0002,
+      "step": 15252
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 4.8221082089552235e-05,
+      "loss": 0.0012,
+      "step": 15256
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 4.82206156716418e-05,
+      "loss": 0.0026,
+      "step": 15260
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 4.8220149253731345e-05,
+      "loss": 0.0005,
+      "step": 15264
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 4.821968283582089e-05,
+      "loss": 0.0004,
+      "step": 15268
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 4.821921641791045e-05,
+      "loss": 0.0002,
+      "step": 15272
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 4.821875e-05,
+      "loss": 0.0069,
+      "step": 15276
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 4.821828358208956e-05,
+      "loss": 0.0019,
+      "step": 15280
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 4.8217817164179106e-05,
+      "loss": 0.0076,
+      "step": 15284
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 4.821735074626866e-05,
+      "loss": 0.0015,
+      "step": 15288
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 4.8216884328358216e-05,
+      "loss": 0.0013,
+      "step": 15292
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 4.8216417910447764e-05,
+      "loss": 0.0014,
+      "step": 15296
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 4.821595149253731e-05,
+      "loss": 0.0048,
+      "step": 15300
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 4.821548507462687e-05,
+      "loss": 0.002,
+      "step": 15304
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 4.821501865671642e-05,
+      "loss": 0.0023,
+      "step": 15308
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 4.821455223880597e-05,
+      "loss": 0.0034,
+      "step": 15312
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 4.8214085820895525e-05,
+      "loss": 0.0047,
+      "step": 15316
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 4.821361940298508e-05,
+      "loss": 0.0092,
+      "step": 15320
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 4.821315298507463e-05,
+      "loss": 0.0144,
+      "step": 15324
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 4.821268656716418e-05,
+      "loss": 0.0068,
+      "step": 15328
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 4.821222014925373e-05,
+      "loss": 0.0042,
+      "step": 15332
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 4.8211753731343286e-05,
+      "loss": 0.0021,
+      "step": 15336
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 4.821128731343284e-05,
+      "loss": 0.004,
+      "step": 15340
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 4.821082089552239e-05,
+      "loss": 0.0018,
+      "step": 15344
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 4.8210354477611944e-05,
+      "loss": 0.0032,
+      "step": 15348
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 4.82098880597015e-05,
+      "loss": 0.0047,
+      "step": 15352
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 4.820942164179105e-05,
+      "loss": 0.0024,
+      "step": 15356
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 4.8208955223880595e-05,
+      "loss": 0.0012,
+      "step": 15360
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 4.820848880597015e-05,
+      "loss": 0.0011,
+      "step": 15364
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 4.8208022388059705e-05,
+      "loss": 0.0041,
+      "step": 15368
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 4.820755597014925e-05,
+      "loss": 0.0026,
+      "step": 15372
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 4.820708955223881e-05,
+      "loss": 0.0012,
+      "step": 15376
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 4.820662313432836e-05,
+      "loss": 0.0038,
+      "step": 15380
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 4.820615671641791e-05,
+      "loss": 0.0011,
+      "step": 15384
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 4.8205690298507466e-05,
+      "loss": 0.0007,
+      "step": 15388
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 4.8205223880597014e-05,
+      "loss": 0.0003,
+      "step": 15392
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 4.820475746268657e-05,
+      "loss": 0.0002,
+      "step": 15396
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 4.8204291044776124e-05,
+      "loss": 0.0008,
+      "step": 15400
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 4.820382462686567e-05,
+      "loss": 0.0011,
+      "step": 15404
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 4.820335820895523e-05,
+      "loss": 0.0026,
+      "step": 15408
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 4.820289179104478e-05,
+      "loss": 0.0019,
+      "step": 15412
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 4.820242537313433e-05,
+      "loss": 0.0036,
+      "step": 15416
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 4.820195895522388e-05,
+      "loss": 0.003,
+      "step": 15420
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 4.820149253731343e-05,
+      "loss": 0.0109,
+      "step": 15424
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 4.820102611940299e-05,
+      "loss": 0.0034,
+      "step": 15428
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 4.8200559701492536e-05,
+      "loss": 0.0005,
+      "step": 15432
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 4.820009328358209e-05,
+      "loss": 0.0084,
+      "step": 15436
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 4.8199626865671646e-05,
+      "loss": 0.0008,
+      "step": 15440
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 4.81991604477612e-05,
+      "loss": 0.009,
+      "step": 15444
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 4.819869402985075e-05,
+      "loss": 0.0021,
+      "step": 15448
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 4.81982276119403e-05,
+      "loss": 0.0029,
+      "step": 15452
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 4.819776119402986e-05,
+      "loss": 0.0013,
+      "step": 15456
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 4.819729477611941e-05,
+      "loss": 0.0014,
+      "step": 15460
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 4.8196828358208955e-05,
+      "loss": 0.0073,
+      "step": 15464
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 4.819636194029851e-05,
+      "loss": 0.0027,
+      "step": 15468
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 4.8195895522388065e-05,
+      "loss": 0.0027,
+      "step": 15472
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 4.819542910447761e-05,
+      "loss": 0.0005,
+      "step": 15476
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 4.819496268656717e-05,
+      "loss": 0.0028,
+      "step": 15480
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 4.8194496268656716e-05,
+      "loss": 0.0005,
+      "step": 15484
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 4.819402985074627e-05,
+      "loss": 0.0063,
+      "step": 15488
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 4.8193563432835826e-05,
+      "loss": 0.0037,
+      "step": 15492
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 4.8193097014925374e-05,
+      "loss": 0.0025,
+      "step": 15496
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 4.819263059701493e-05,
+      "loss": 0.0044,
+      "step": 15500
+    },
+    {
+      "epoch": 14.46,
+      "eval_exact_match": 0.7340425531914894,
+      "eval_exec": 0.7727272727272727,
+      "eval_loss": 0.30817216634750366,
+      "eval_runtime": 1067.8388,
+      "eval_samples_per_second": 0.968,
+      "step": 15500
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 4.8192164179104484e-05,
+      "loss": 0.0109,
+      "step": 15504
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 4.819169776119403e-05,
+      "loss": 0.0041,
+      "step": 15508
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 4.819123134328358e-05,
+      "loss": 0.0005,
+      "step": 15512
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 4.819076492537314e-05,
+      "loss": 0.0005,
+      "step": 15516
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 4.819029850746269e-05,
+      "loss": 0.0044,
+      "step": 15520
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 4.818983208955224e-05,
+      "loss": 0.0005,
+      "step": 15524
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 4.818936567164179e-05,
+      "loss": 0.0006,
+      "step": 15528
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 4.818889925373135e-05,
+      "loss": 0.0026,
+      "step": 15532
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 4.8188432835820896e-05,
+      "loss": 0.0042,
+      "step": 15536
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 4.818796641791045e-05,
+      "loss": 0.0017,
+      "step": 15540
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 4.81875e-05,
+      "loss": 0.0013,
+      "step": 15544
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 4.8187033582089554e-05,
+      "loss": 0.0041,
+      "step": 15548
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 4.818656716417911e-05,
+      "loss": 0.0098,
+      "step": 15552
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 4.818610074626866e-05,
+      "loss": 0.0025,
+      "step": 15556
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 4.818563432835821e-05,
+      "loss": 0.0019,
+      "step": 15560
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 4.818516791044777e-05,
+      "loss": 0.0021,
+      "step": 15564
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 4.8184701492537315e-05,
+      "loss": 0.0054,
+      "step": 15568
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 4.818423507462686e-05,
+      "loss": 0.0031,
+      "step": 15572
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 4.8183768656716425e-05,
+      "loss": 0.0079,
+      "step": 15576
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 4.818330223880597e-05,
+      "loss": 0.0037,
+      "step": 15580
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 4.818283582089552e-05,
+      "loss": 0.0011,
+      "step": 15584
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 4.8182369402985076e-05,
+      "loss": 0.0073,
+      "step": 15588
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 4.818190298507463e-05,
+      "loss": 0.0069,
+      "step": 15592
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 4.818143656716418e-05,
+      "loss": 0.006,
+      "step": 15596
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 4.8180970149253734e-05,
+      "loss": 0.0053,
+      "step": 15600
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 4.818050373134328e-05,
+      "loss": 0.0039,
+      "step": 15604
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 4.8180037313432844e-05,
+      "loss": 0.0132,
+      "step": 15608
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 4.817957089552239e-05,
+      "loss": 0.0041,
+      "step": 15612
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 4.817910447761194e-05,
+      "loss": 0.0063,
+      "step": 15616
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 4.8178638059701495e-05,
+      "loss": 0.0051,
+      "step": 15620
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 4.817817164179105e-05,
+      "loss": 0.0036,
+      "step": 15624
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 4.81777052238806e-05,
+      "loss": 0.0027,
+      "step": 15628
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 4.817723880597015e-05,
+      "loss": 0.0041,
+      "step": 15632
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 4.817677238805971e-05,
+      "loss": 0.0017,
+      "step": 15636
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 4.8176305970149256e-05,
+      "loss": 0.0036,
+      "step": 15640
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 4.817583955223881e-05,
+      "loss": 0.0112,
+      "step": 15644
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 4.817537313432836e-05,
+      "loss": 0.0043,
+      "step": 15648
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 4.8174906716417914e-05,
+      "loss": 0.0023,
+      "step": 15652
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 4.817444029850747e-05,
+      "loss": 0.0029,
+      "step": 15656
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 4.817397388059702e-05,
+      "loss": 0.001,
+      "step": 15660
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 4.8173507462686565e-05,
+      "loss": 0.001,
+      "step": 15664
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 4.8173041044776127e-05,
+      "loss": 0.0011,
+      "step": 15668
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 4.8172574626865675e-05,
+      "loss": 0.009,
+      "step": 15672
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 4.817210820895522e-05,
+      "loss": 0.0054,
+      "step": 15676
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 4.817164179104478e-05,
+      "loss": 0.0043,
+      "step": 15680
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 4.817117537313433e-05,
+      "loss": 0.0026,
+      "step": 15684
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 4.817070895522388e-05,
+      "loss": 0.0013,
+      "step": 15688
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 4.8170242537313436e-05,
+      "loss": 0.0018,
+      "step": 15692
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 4.816977611940299e-05,
+      "loss": 0.0052,
+      "step": 15696
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 4.816930970149254e-05,
+      "loss": 0.0049,
+      "step": 15700
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 4.8168843283582094e-05,
+      "loss": 0.0006,
+      "step": 15704
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 4.816837686567164e-05,
+      "loss": 0.0119,
+      "step": 15708
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 4.81679104477612e-05,
+      "loss": 0.0032,
+      "step": 15712
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 4.816744402985075e-05,
+      "loss": 0.0048,
+      "step": 15716
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 4.81669776119403e-05,
+      "loss": 0.0005,
+      "step": 15720
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 4.816651119402985e-05,
+      "loss": 0.001,
+      "step": 15724
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 4.816604477611941e-05,
+      "loss": 0.0078,
+      "step": 15728
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 4.816557835820896e-05,
+      "loss": 0.0037,
+      "step": 15732
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 4.8165111940298506e-05,
+      "loss": 0.0027,
+      "step": 15736
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 4.816464552238806e-05,
+      "loss": 0.0025,
+      "step": 15740
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 4.8164179104477616e-05,
+      "loss": 0.0017,
+      "step": 15744
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 4.8163712686567164e-05,
+      "loss": 0.0008,
+      "step": 15748
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 4.816324626865672e-05,
+      "loss": 0.0007,
+      "step": 15752
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 4.8162779850746274e-05,
+      "loss": 0.0105,
+      "step": 15756
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 4.816231343283582e-05,
+      "loss": 0.0056,
+      "step": 15760
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 4.816184701492538e-05,
+      "loss": 0.0056,
+      "step": 15764
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 4.8161380597014925e-05,
+      "loss": 0.009,
+      "step": 15768
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 4.816091417910448e-05,
+      "loss": 0.0036,
+      "step": 15772
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 4.8160447761194035e-05,
+      "loss": 0.0007,
+      "step": 15776
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 4.815998134328358e-05,
+      "loss": 0.002,
+      "step": 15780
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 4.815951492537314e-05,
+      "loss": 0.0027,
+      "step": 15784
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 4.815904850746269e-05,
+      "loss": 0.002,
+      "step": 15788
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 4.815858208955224e-05,
+      "loss": 0.0004,
+      "step": 15792
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 4.8158115671641796e-05,
+      "loss": 0.0035,
+      "step": 15796
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 4.8157649253731344e-05,
+      "loss": 0.0032,
+      "step": 15800
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 4.81571828358209e-05,
+      "loss": 0.0004,
+      "step": 15804
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 4.8156716417910453e-05,
+      "loss": 0.0007,
+      "step": 15808
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 4.815625e-05,
+      "loss": 0.002,
+      "step": 15812
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 4.815578358208955e-05,
+      "loss": 0.0007,
+      "step": 15816
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 4.815531716417911e-05,
+      "loss": 0.0001,
+      "step": 15820
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 4.815485074626866e-05,
+      "loss": 0.0011,
+      "step": 15824
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 4.815438432835821e-05,
+      "loss": 0.0041,
+      "step": 15828
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 4.815391791044776e-05,
+      "loss": 0.0061,
+      "step": 15832
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 4.815345149253732e-05,
+      "loss": 0.0028,
+      "step": 15836
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 4.8152985074626866e-05,
+      "loss": 0.0028,
+      "step": 15840
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 4.815251865671642e-05,
+      "loss": 0.0051,
+      "step": 15844
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 4.8152052238805975e-05,
+      "loss": 0.0054,
+      "step": 15848
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 4.8151585820895524e-05,
+      "loss": 0.0016,
+      "step": 15852
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 4.815111940298508e-05,
+      "loss": 0.0024,
+      "step": 15856
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 4.815065298507463e-05,
+      "loss": 0.0041,
+      "step": 15860
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 4.815018656716418e-05,
+      "loss": 0.0025,
+      "step": 15864
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 4.8149720149253736e-05,
+      "loss": 0.003,
+      "step": 15868
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 4.8149253731343285e-05,
+      "loss": 0.0043,
+      "step": 15872
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 4.814878731343283e-05,
+      "loss": 0.014,
+      "step": 15876
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 4.8148320895522394e-05,
+      "loss": 0.0046,
+      "step": 15880
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 4.814785447761194e-05,
+      "loss": 0.0089,
+      "step": 15884
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 4.814738805970149e-05,
+      "loss": 0.0005,
+      "step": 15888
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 4.8146921641791046e-05,
+      "loss": 0.0025,
+      "step": 15892
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 4.81464552238806e-05,
+      "loss": 0.0027,
+      "step": 15896
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 4.814598880597015e-05,
+      "loss": 0.0082,
+      "step": 15900
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 4.8145522388059704e-05,
+      "loss": 0.0016,
+      "step": 15904
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 4.814505597014926e-05,
+      "loss": 0.0556,
+      "step": 15908
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 4.8144589552238807e-05,
+      "loss": 0.0032,
+      "step": 15912
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 4.814412313432836e-05,
+      "loss": 0.0014,
+      "step": 15916
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 4.814365671641791e-05,
+      "loss": 0.0013,
+      "step": 15920
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 4.8143190298507465e-05,
+      "loss": 0.0037,
+      "step": 15924
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 4.814272388059702e-05,
+      "loss": 0.002,
+      "step": 15928
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 4.814225746268657e-05,
+      "loss": 0.0032,
+      "step": 15932
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 4.814179104477612e-05,
+      "loss": 0.0133,
+      "step": 15936
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 4.814132462686568e-05,
+      "loss": 0.0004,
+      "step": 15940
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 4.8140858208955225e-05,
+      "loss": 0.0026,
+      "step": 15944
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 4.814039179104478e-05,
+      "loss": 0.0004,
+      "step": 15948
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 4.813992537313433e-05,
+      "loss": 0.0043,
+      "step": 15952
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 4.8139458955223883e-05,
+      "loss": 0.0121,
+      "step": 15956
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 4.813899253731344e-05,
+      "loss": 0.0007,
+      "step": 15960
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 4.8138526119402986e-05,
+      "loss": 0.0046,
+      "step": 15964
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 4.813805970149254e-05,
+      "loss": 0.0008,
+      "step": 15968
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 4.8137593283582096e-05,
+      "loss": 0.0106,
+      "step": 15972
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 4.8137126865671644e-05,
+      "loss": 0.0034,
+      "step": 15976
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 4.813666044776119e-05,
+      "loss": 0.0062,
+      "step": 15980
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 4.813619402985075e-05,
+      "loss": 0.0059,
+      "step": 15984
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 4.81357276119403e-05,
+      "loss": 0.0028,
+      "step": 15988
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 4.813526119402985e-05,
+      "loss": 0.0056,
+      "step": 15992
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 4.8134794776119405e-05,
+      "loss": 0.0029,
+      "step": 15996
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 4.813432835820896e-05,
+      "loss": 0.0031,
+      "step": 16000
+    },
+    {
+      "epoch": 14.93,
+      "eval_exact_match": 0.7166344294003868,
+      "eval_exec": 0.7553191489361702,
+      "eval_loss": 0.2952513098716736,
+      "eval_runtime": 1078.2591,
+      "eval_samples_per_second": 0.959,
+      "step": 16000
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 4.813386194029851e-05,
+      "loss": 0.005,
+      "step": 16004
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 4.813339552238806e-05,
+      "loss": 0.0055,
+      "step": 16008
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 4.813292910447761e-05,
+      "loss": 0.0085,
+      "step": 16012
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 4.8132462686567166e-05,
+      "loss": 0.0029,
+      "step": 16016
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 4.813199626865672e-05,
+      "loss": 0.0035,
+      "step": 16020
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 4.813152985074627e-05,
+      "loss": 0.0051,
+      "step": 16024
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 4.8131063432835824e-05,
+      "loss": 0.0015,
+      "step": 16028
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 4.813059701492538e-05,
+      "loss": 0.005,
+      "step": 16032
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 4.813013059701493e-05,
+      "loss": 0.0011,
+      "step": 16036
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 4.8129664179104476e-05,
+      "loss": 0.002,
+      "step": 16040
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 4.812919776119403e-05,
+      "loss": 0.0034,
+      "step": 16044
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 4.8128731343283585e-05,
+      "loss": 0.0025,
+      "step": 16048
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 4.8128264925373133e-05,
+      "loss": 0.002,
+      "step": 16052
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 4.812779850746269e-05,
+      "loss": 0.0081,
+      "step": 16056
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 4.812733208955224e-05,
+      "loss": 0.0041,
+      "step": 16060
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 4.812686567164179e-05,
+      "loss": 0.0021,
+      "step": 16064
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 4.8126399253731346e-05,
+      "loss": 0.0031,
+      "step": 16068
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 4.8125932835820894e-05,
+      "loss": 0.0066,
+      "step": 16072
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 4.812546641791045e-05,
+      "loss": 0.0015,
+      "step": 16076
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 4.8125000000000004e-05,
+      "loss": 0.0011,
+      "step": 16080
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 4.812453358208955e-05,
+      "loss": 0.0049,
+      "step": 16084
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 4.812406716417911e-05,
+      "loss": 0.0013,
+      "step": 16088
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 4.812360074626866e-05,
+      "loss": 0.0014,
+      "step": 16092
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 4.812313432835821e-05,
+      "loss": 0.0009,
+      "step": 16096
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 4.8122667910447765e-05,
+      "loss": 0.0009,
+      "step": 16100
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 4.812220149253731e-05,
+      "loss": 0.0004,
+      "step": 16104
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 4.812173507462687e-05,
+      "loss": 0.015,
+      "step": 16108
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 4.812126865671642e-05,
+      "loss": 0.0018,
+      "step": 16112
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 4.812080223880597e-05,
+      "loss": 0.002,
+      "step": 16116
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 4.8120335820895526e-05,
+      "loss": 0.0027,
+      "step": 16120
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 4.811986940298508e-05,
+      "loss": 0.0018,
+      "step": 16124
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 4.811940298507463e-05,
+      "loss": 0.0019,
+      "step": 16128
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 4.811893656716418e-05,
+      "loss": 0.0064,
+      "step": 16132
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 4.811847014925374e-05,
+      "loss": 0.0063,
+      "step": 16136
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 4.811800373134329e-05,
+      "loss": 0.0016,
+      "step": 16140
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 4.8117537313432835e-05,
+      "loss": 0.0042,
+      "step": 16144
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 4.811707089552239e-05,
+      "loss": 0.0011,
+      "step": 16148
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 4.8116604477611945e-05,
+      "loss": 0.0019,
+      "step": 16152
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 4.811613805970149e-05,
+      "loss": 0.0003,
+      "step": 16156
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 4.811567164179105e-05,
+      "loss": 0.0019,
+      "step": 16160
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 4.8115205223880596e-05,
+      "loss": 0.0049,
+      "step": 16164
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 4.811473880597015e-05,
+      "loss": 0.0036,
+      "step": 16168
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 4.8114272388059706e-05,
+      "loss": 0.0012,
+      "step": 16172
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 4.8113805970149254e-05,
+      "loss": 0.0004,
+      "step": 16176
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 4.811333955223881e-05,
+      "loss": 0.0018,
+      "step": 16180
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 4.8112873134328364e-05,
+      "loss": 0.0003,
+      "step": 16184
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 4.811240671641791e-05,
+      "loss": 0.003,
+      "step": 16188
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 4.811194029850746e-05,
+      "loss": 0.0029,
+      "step": 16192
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 4.811147388059702e-05,
+      "loss": 0.0081,
+      "step": 16196
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 4.811100746268657e-05,
+      "loss": 0.0094,
+      "step": 16200
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 4.811054104477612e-05,
+      "loss": 0.0017,
+      "step": 16204
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 4.811007462686567e-05,
+      "loss": 0.0049,
+      "step": 16208
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 4.810960820895523e-05,
+      "loss": 0.0044,
+      "step": 16212
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 4.8109141791044776e-05,
+      "loss": 0.0012,
+      "step": 16216
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 4.810867537313433e-05,
+      "loss": 0.0013,
+      "step": 16220
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 4.810820895522388e-05,
+      "loss": 0.0024,
+      "step": 16224
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 4.8107742537313434e-05,
+      "loss": 0.0004,
+      "step": 16228
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 4.810727611940299e-05,
+      "loss": 0.0072,
+      "step": 16232
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 4.810680970149254e-05,
+      "loss": 0.0043,
+      "step": 16236
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 4.810634328358209e-05,
+      "loss": 0.0004,
+      "step": 16240
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 4.810587686567165e-05,
+      "loss": 0.0024,
+      "step": 16244
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 4.8105410447761195e-05,
+      "loss": 0.0038,
+      "step": 16248
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 4.810494402985074e-05,
+      "loss": 0.0016,
+      "step": 16252
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 4.8104477611940305e-05,
+      "loss": 0.0045,
+      "step": 16256
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 4.810401119402985e-05,
+      "loss": 0.0002,
+      "step": 16260
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 4.810354477611941e-05,
+      "loss": 0.0052,
+      "step": 16264
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 4.8103078358208956e-05,
+      "loss": 0.0047,
+      "step": 16268
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 4.810261194029851e-05,
+      "loss": 0.0022,
+      "step": 16272
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 4.8102145522388066e-05,
+      "loss": 0.0205,
+      "step": 16276
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 4.8101679104477614e-05,
+      "loss": 0.0053,
+      "step": 16280
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 4.810121268656716e-05,
+      "loss": 0.0004,
+      "step": 16284
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 4.8100746268656724e-05,
+      "loss": 0.0042,
+      "step": 16288
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 4.810027985074627e-05,
+      "loss": 0.0027,
+      "step": 16292
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 4.809981343283582e-05,
+      "loss": 0.0143,
+      "step": 16296
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 4.8099347014925375e-05,
+      "loss": 0.0028,
+      "step": 16300
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 4.809888059701493e-05,
+      "loss": 0.0093,
+      "step": 16304
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 4.809841417910448e-05,
+      "loss": 0.0039,
+      "step": 16308
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 4.809794776119403e-05,
+      "loss": 0.0009,
+      "step": 16312
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 4.809748134328359e-05,
+      "loss": 0.0022,
+      "step": 16316
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 4.8097014925373136e-05,
+      "loss": 0.0019,
+      "step": 16320
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 4.809654850746269e-05,
+      "loss": 0.0058,
+      "step": 16324
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 4.809608208955224e-05,
+      "loss": 0.0013,
+      "step": 16328
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 4.8095615671641794e-05,
+      "loss": 0.0025,
+      "step": 16332
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 4.809514925373135e-05,
+      "loss": 0.0045,
+      "step": 16336
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 4.80946828358209e-05,
+      "loss": 0.0026,
+      "step": 16340
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 4.8094216417910445e-05,
+      "loss": 0.0062,
+      "step": 16344
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 4.809375000000001e-05,
+      "loss": 0.0044,
+      "step": 16348
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 4.8093283582089555e-05,
+      "loss": 0.0026,
+      "step": 16352
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 4.80928171641791e-05,
+      "loss": 0.003,
+      "step": 16356
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 4.809235074626866e-05,
+      "loss": 0.0039,
+      "step": 16360
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 4.809188432835821e-05,
+      "loss": 0.012,
+      "step": 16364
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 4.809141791044776e-05,
+      "loss": 0.0023,
+      "step": 16368
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 4.8090951492537316e-05,
+      "loss": 0.0057,
+      "step": 16372
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 4.809048507462687e-05,
+      "loss": 0.0011,
+      "step": 16376
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 4.809001865671642e-05,
+      "loss": 0.009,
+      "step": 16380
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 4.8089552238805974e-05,
+      "loss": 0.0026,
+      "step": 16384
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 4.808908582089552e-05,
+      "loss": 0.0005,
+      "step": 16388
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 4.808861940298508e-05,
+      "loss": 0.0054,
+      "step": 16392
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 4.808815298507463e-05,
+      "loss": 0.0035,
+      "step": 16396
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 4.808768656716418e-05,
+      "loss": 0.0049,
+      "step": 16400
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 4.808722014925373e-05,
+      "loss": 0.0028,
+      "step": 16404
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 4.808675373134329e-05,
+      "loss": 0.0002,
+      "step": 16408
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 4.808628731343284e-05,
+      "loss": 0.0003,
+      "step": 16412
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 4.8085820895522386e-05,
+      "loss": 0.0014,
+      "step": 16416
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 4.808535447761194e-05,
+      "loss": 0.009,
+      "step": 16420
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 4.8084888059701496e-05,
+      "loss": 0.0002,
+      "step": 16424
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 4.808442164179105e-05,
+      "loss": 0.0004,
+      "step": 16428
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 4.80839552238806e-05,
+      "loss": 0.0043,
+      "step": 16432
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 4.8083488805970154e-05,
+      "loss": 0.0091,
+      "step": 16436
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 4.808302238805971e-05,
+      "loss": 0.0007,
+      "step": 16440
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 4.808255597014926e-05,
+      "loss": 0.0021,
+      "step": 16444
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 4.8082089552238805e-05,
+      "loss": 0.0036,
+      "step": 16448
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 4.808162313432836e-05,
+      "loss": 0.0018,
+      "step": 16452
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 4.8081156716417915e-05,
+      "loss": 0.0222,
+      "step": 16456
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 4.808069029850746e-05,
+      "loss": 0.0022,
+      "step": 16460
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 4.808022388059702e-05,
+      "loss": 0.0023,
+      "step": 16464
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 4.807975746268657e-05,
+      "loss": 0.0022,
+      "step": 16468
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 4.807929104477612e-05,
+      "loss": 0.0056,
+      "step": 16472
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 4.8078824626865676e-05,
+      "loss": 0.0005,
+      "step": 16476
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 4.8078358208955224e-05,
+      "loss": 0.0017,
+      "step": 16480
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 4.807789179104478e-05,
+      "loss": 0.0005,
+      "step": 16484
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 4.8077425373134334e-05,
+      "loss": 0.0085,
+      "step": 16488
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 4.807695895522388e-05,
+      "loss": 0.0005,
+      "step": 16492
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 4.807649253731343e-05,
+      "loss": 0.0032,
+      "step": 16496
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 4.807602611940299e-05,
+      "loss": 0.0007,
+      "step": 16500
+    },
+    {
+      "epoch": 15.39,
+      "eval_exact_match": 0.7272727272727273,
+      "eval_exec": 0.7669245647969052,
+      "eval_loss": 0.3177631199359894,
+      "eval_runtime": 1071.5811,
+      "eval_samples_per_second": 0.965,
+      "step": 16500
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 4.807555970149254e-05,
+      "loss": 0.0019,
+      "step": 16504
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 4.807509328358209e-05,
+      "loss": 0.0072,
+      "step": 16508
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 4.807462686567164e-05,
+      "loss": 0.0004,
+      "step": 16512
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 4.80741604477612e-05,
+      "loss": 0.0006,
+      "step": 16516
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 4.8073694029850746e-05,
+      "loss": 0.0023,
+      "step": 16520
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 4.80732276119403e-05,
+      "loss": 0.0017,
+      "step": 16524
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 4.8072761194029856e-05,
+      "loss": 0.0038,
+      "step": 16528
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 4.8072294776119404e-05,
+      "loss": 0.0014,
+      "step": 16532
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 4.807182835820896e-05,
+      "loss": 0.0032,
+      "step": 16536
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 4.807136194029851e-05,
+      "loss": 0.0007,
+      "step": 16540
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 4.807089552238806e-05,
+      "loss": 0.0019,
+      "step": 16544
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 4.807042910447762e-05,
+      "loss": 0.0026,
+      "step": 16548
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 4.8069962686567165e-05,
+      "loss": 0.004,
+      "step": 16552
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 4.806949626865671e-05,
+      "loss": 0.0006,
+      "step": 16556
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 4.8069029850746275e-05,
+      "loss": 0.0016,
+      "step": 16560
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 4.806856343283582e-05,
+      "loss": 0.0025,
+      "step": 16564
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 4.806809701492537e-05,
+      "loss": 0.0006,
+      "step": 16568
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 4.8067630597014926e-05,
+      "loss": 0.0008,
+      "step": 16572
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 4.806716417910448e-05,
+      "loss": 0.0043,
+      "step": 16576
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 4.806669776119403e-05,
+      "loss": 0.0013,
+      "step": 16580
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 4.8066231343283584e-05,
+      "loss": 0.0018,
+      "step": 16584
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 4.806576492537314e-05,
+      "loss": 0.0001,
+      "step": 16588
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 4.8065298507462694e-05,
+      "loss": 0.0008,
+      "step": 16592
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 4.806483208955224e-05,
+      "loss": 0.0017,
+      "step": 16596
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 4.806436567164179e-05,
+      "loss": 0.0005,
+      "step": 16600
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 4.8063899253731345e-05,
+      "loss": 0.0004,
+      "step": 16604
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 4.80634328358209e-05,
+      "loss": 0.0003,
+      "step": 16608
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 4.806296641791045e-05,
+      "loss": 0.0063,
+      "step": 16612
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 4.80625e-05,
+      "loss": 0.0138,
+      "step": 16616
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 4.806203358208956e-05,
+      "loss": 0.0003,
+      "step": 16620
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 4.8061567164179106e-05,
+      "loss": 0.0004,
+      "step": 16624
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 4.806110074626866e-05,
+      "loss": 0.0017,
+      "step": 16628
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 4.806063432835821e-05,
+      "loss": 0.0007,
+      "step": 16632
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 4.8060167910447764e-05,
+      "loss": 0.0026,
+      "step": 16636
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 4.805970149253732e-05,
+      "loss": 0.0024,
+      "step": 16640
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 4.805923507462687e-05,
+      "loss": 0.0096,
+      "step": 16644
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 4.805876865671642e-05,
+      "loss": 0.0074,
+      "step": 16648
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 4.8058302238805977e-05,
+      "loss": 0.0007,
+      "step": 16652
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 4.8057835820895525e-05,
+      "loss": 0.0049,
+      "step": 16656
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 4.805736940298507e-05,
+      "loss": 0.001,
+      "step": 16660
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 4.805690298507463e-05,
+      "loss": 0.007,
+      "step": 16664
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 4.805643656716418e-05,
+      "loss": 0.0027,
+      "step": 16668
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 4.805597014925373e-05,
+      "loss": 0.0018,
+      "step": 16672
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 4.8055503731343286e-05,
+      "loss": 0.0017,
+      "step": 16676
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 4.805503731343284e-05,
+      "loss": 0.0009,
+      "step": 16680
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 4.805457089552239e-05,
+      "loss": 0.005,
+      "step": 16684
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 4.8054104477611944e-05,
+      "loss": 0.0016,
+      "step": 16688
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 4.805363805970149e-05,
+      "loss": 0.0005,
+      "step": 16692
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 4.805317164179105e-05,
+      "loss": 0.0041,
+      "step": 16696
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 4.80527052238806e-05,
+      "loss": 0.0019,
+      "step": 16700
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 4.805223880597015e-05,
+      "loss": 0.0024,
+      "step": 16704
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 4.8051772388059705e-05,
+      "loss": 0.0122,
+      "step": 16708
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 4.805130597014926e-05,
+      "loss": 0.0028,
+      "step": 16712
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 4.805083955223881e-05,
+      "loss": 0.003,
+      "step": 16716
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 4.8050373134328356e-05,
+      "loss": 0.0032,
+      "step": 16720
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 4.804990671641791e-05,
+      "loss": 0.0014,
+      "step": 16724
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 4.8049440298507466e-05,
+      "loss": 0.0014,
+      "step": 16728
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 4.8048973880597014e-05,
+      "loss": 0.001,
+      "step": 16732
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 4.804850746268657e-05,
+      "loss": 0.0025,
+      "step": 16736
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 4.8048041044776124e-05,
+      "loss": 0.0017,
+      "step": 16740
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 4.804757462686567e-05,
+      "loss": 0.0124,
+      "step": 16744
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 4.8047108208955227e-05,
+      "loss": 0.0022,
+      "step": 16748
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 4.8046641791044775e-05,
+      "loss": 0.0093,
+      "step": 16752
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 4.8046175373134336e-05,
+      "loss": 0.0019,
+      "step": 16756
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 4.8045708955223885e-05,
+      "loss": 0.003,
+      "step": 16760
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 4.804524253731343e-05,
+      "loss": 0.0019,
+      "step": 16764
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 4.804477611940299e-05,
+      "loss": 0.0017,
+      "step": 16768
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 4.804430970149254e-05,
+      "loss": 0.0004,
+      "step": 16772
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 4.804384328358209e-05,
+      "loss": 0.0017,
+      "step": 16776
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 4.8043376865671646e-05,
+      "loss": 0.0031,
+      "step": 16780
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 4.8042910447761194e-05,
+      "loss": 0.0042,
+      "step": 16784
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 4.804244402985075e-05,
+      "loss": 0.002,
+      "step": 16788
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 4.8041977611940303e-05,
+      "loss": 0.0007,
+      "step": 16792
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 4.804151119402985e-05,
+      "loss": 0.0008,
+      "step": 16796
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 4.8041044776119407e-05,
+      "loss": 0.0008,
+      "step": 16800
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 4.804057835820896e-05,
+      "loss": 0.0017,
+      "step": 16804
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 4.804011194029851e-05,
+      "loss": 0.0055,
+      "step": 16808
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 4.803964552238806e-05,
+      "loss": 0.0021,
+      "step": 16812
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 4.803917910447762e-05,
+      "loss": 0.0046,
+      "step": 16816
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 4.803871268656717e-05,
+      "loss": 0.0013,
+      "step": 16820
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 4.8038246268656716e-05,
+      "loss": 0.0049,
+      "step": 16824
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 4.803777985074627e-05,
+      "loss": 0.0023,
+      "step": 16828
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 4.8037313432835825e-05,
+      "loss": 0.0031,
+      "step": 16832
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 4.8036847014925374e-05,
+      "loss": 0.0127,
+      "step": 16836
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 4.803638059701493e-05,
+      "loss": 0.0041,
+      "step": 16840
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 4.803591417910448e-05,
+      "loss": 0.0102,
+      "step": 16844
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 4.803544776119403e-05,
+      "loss": 0.0031,
+      "step": 16848
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 4.8034981343283586e-05,
+      "loss": 0.0038,
+      "step": 16852
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 4.8034514925373135e-05,
+      "loss": 0.0079,
+      "step": 16856
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 4.803404850746269e-05,
+      "loss": 0.0113,
+      "step": 16860
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 4.8033582089552244e-05,
+      "loss": 0.0009,
+      "step": 16864
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 4.803311567164179e-05,
+      "loss": 0.0019,
+      "step": 16868
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 4.803264925373134e-05,
+      "loss": 0.0053,
+      "step": 16872
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 4.80321828358209e-05,
+      "loss": 0.0009,
+      "step": 16876
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 4.803171641791045e-05,
+      "loss": 0.001,
+      "step": 16880
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 4.803125e-05,
+      "loss": 0.0125,
+      "step": 16884
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 4.8030783582089553e-05,
+      "loss": 0.0018,
+      "step": 16888
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 4.803031716417911e-05,
+      "loss": 0.0032,
+      "step": 16892
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 4.8029850746268657e-05,
+      "loss": 0.0013,
+      "step": 16896
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 4.802938432835821e-05,
+      "loss": 0.0017,
+      "step": 16900
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 4.802891791044776e-05,
+      "loss": 0.0011,
+      "step": 16904
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 4.8028451492537314e-05,
+      "loss": 0.0013,
+      "step": 16908
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 4.802798507462687e-05,
+      "loss": 0.0009,
+      "step": 16912
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 4.802751865671642e-05,
+      "loss": 0.0007,
+      "step": 16916
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 4.802705223880597e-05,
+      "loss": 0.001,
+      "step": 16920
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 4.802658582089553e-05,
+      "loss": 0.0007,
+      "step": 16924
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 4.8026119402985075e-05,
+      "loss": 0.0075,
+      "step": 16928
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 4.802565298507463e-05,
+      "loss": 0.0008,
+      "step": 16932
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 4.8025186567164185e-05,
+      "loss": 0.0015,
+      "step": 16936
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 4.8024720149253733e-05,
+      "loss": 0.0034,
+      "step": 16940
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 4.802425373134329e-05,
+      "loss": 0.0054,
+      "step": 16944
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 4.8023787313432836e-05,
+      "loss": 0.0013,
+      "step": 16948
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 4.802332089552239e-05,
+      "loss": 0.001,
+      "step": 16952
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 4.8022854477611946e-05,
+      "loss": 0.0023,
+      "step": 16956
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 4.8022388059701494e-05,
+      "loss": 0.0005,
+      "step": 16960
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 4.802192164179104e-05,
+      "loss": 0.0018,
+      "step": 16964
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 4.8021455223880604e-05,
+      "loss": 0.0038,
+      "step": 16968
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 4.802098880597015e-05,
+      "loss": 0.0024,
+      "step": 16972
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 4.80205223880597e-05,
+      "loss": 0.0078,
+      "step": 16976
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 4.8020055970149255e-05,
+      "loss": 0.0027,
+      "step": 16980
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 4.801958955223881e-05,
+      "loss": 0.0014,
+      "step": 16984
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 4.801912313432836e-05,
+      "loss": 0.0037,
+      "step": 16988
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 4.801865671641791e-05,
+      "loss": 0.0042,
+      "step": 16992
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 4.801819029850747e-05,
+      "loss": 0.0025,
+      "step": 16996
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 4.8017723880597016e-05,
+      "loss": 0.0119,
+      "step": 17000
+    },
+    {
+      "epoch": 15.86,
+      "eval_exact_match": 0.7321083172147002,
+      "eval_exec": 0.7611218568665378,
+      "eval_loss": 0.2985568642616272,
+      "eval_runtime": 1073.3502,
+      "eval_samples_per_second": 0.963,
+      "step": 17000
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 4.801725746268657e-05,
+      "loss": 0.0016,
+      "step": 17004
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 4.801679104477612e-05,
+      "loss": 0.0039,
+      "step": 17008
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 4.8016324626865674e-05,
+      "loss": 0.0053,
+      "step": 17012
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 4.801585820895523e-05,
+      "loss": 0.0005,
+      "step": 17016
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 4.801539179104478e-05,
+      "loss": 0.0023,
+      "step": 17020
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 4.8014925373134325e-05,
+      "loss": 0.0009,
+      "step": 17024
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 4.801445895522389e-05,
+      "loss": 0.0106,
+      "step": 17028
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 4.8013992537313435e-05,
+      "loss": 0.0086,
+      "step": 17032
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 4.8013526119402983e-05,
+      "loss": 0.0005,
+      "step": 17036
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 4.801305970149254e-05,
+      "loss": 0.0011,
+      "step": 17040
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 4.801259328358209e-05,
+      "loss": 0.0035,
+      "step": 17044
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 4.801212686567164e-05,
+      "loss": 0.0006,
+      "step": 17048
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 4.8011660447761196e-05,
+      "loss": 0.0013,
+      "step": 17052
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 4.801119402985075e-05,
+      "loss": 0.005,
+      "step": 17056
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 4.80107276119403e-05,
+      "loss": 0.0029,
+      "step": 17060
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 4.8010261194029854e-05,
+      "loss": 0.0006,
+      "step": 17064
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 4.80097947761194e-05,
+      "loss": 0.0028,
+      "step": 17068
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 4.800932835820896e-05,
+      "loss": 0.0086,
+      "step": 17072
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 4.800886194029851e-05,
+      "loss": 0.0046,
+      "step": 17076
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 4.800839552238806e-05,
+      "loss": 0.0011,
+      "step": 17080
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 4.8007929104477615e-05,
+      "loss": 0.0025,
+      "step": 17084
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 4.800746268656717e-05,
+      "loss": 0.0012,
+      "step": 17088
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 4.800699626865672e-05,
+      "loss": 0.006,
+      "step": 17092
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 4.800652985074627e-05,
+      "loss": 0.0031,
+      "step": 17096
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 4.800606343283582e-05,
+      "loss": 0.0008,
+      "step": 17100
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 4.8005597014925376e-05,
+      "loss": 0.0068,
+      "step": 17104
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 4.800513059701493e-05,
+      "loss": 0.0005,
+      "step": 17108
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 4.800466417910448e-05,
+      "loss": 0.0005,
+      "step": 17112
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 4.8004197761194034e-05,
+      "loss": 0.0024,
+      "step": 17116
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 4.800373134328359e-05,
+      "loss": 0.0016,
+      "step": 17120
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 4.800326492537314e-05,
+      "loss": 0.0004,
+      "step": 17124
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 4.8002798507462685e-05,
+      "loss": 0.0037,
+      "step": 17128
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 4.800233208955224e-05,
+      "loss": 0.0018,
+      "step": 17132
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 4.8001865671641795e-05,
+      "loss": 0.0022,
+      "step": 17136
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 4.800139925373134e-05,
+      "loss": 0.0007,
+      "step": 17140
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 4.80009328358209e-05,
+      "loss": 0.0013,
+      "step": 17144
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 4.800046641791045e-05,
+      "loss": 0.0039,
+      "step": 17148
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 4.8e-05,
+      "loss": 0.0049,
+      "step": 17152
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 4.7999533582089556e-05,
+      "loss": 0.0026,
+      "step": 17156
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 4.7999067164179104e-05,
+      "loss": 0.0088,
+      "step": 17160
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 4.799860074626866e-05,
+      "loss": 0.0007,
+      "step": 17164
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 4.7998134328358214e-05,
+      "loss": 0.0012,
+      "step": 17168
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 4.799766791044776e-05,
+      "loss": 0.0009,
+      "step": 17172
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 4.799720149253731e-05,
+      "loss": 0.0035,
+      "step": 17176
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 4.799673507462687e-05,
+      "loss": 0.0006,
+      "step": 17180
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 4.799626865671642e-05,
+      "loss": 0.0043,
+      "step": 17184
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 4.799580223880597e-05,
+      "loss": 0.0002,
+      "step": 17188
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 4.799533582089552e-05,
+      "loss": 0.0001,
+      "step": 17192
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 4.799486940298508e-05,
+      "loss": 0.0037,
+      "step": 17196
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 4.7994402985074626e-05,
+      "loss": 0.0033,
+      "step": 17200
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 4.799393656716418e-05,
+      "loss": 0.0002,
+      "step": 17204
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 4.7993470149253736e-05,
+      "loss": 0.0011,
+      "step": 17208
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 4.7993003731343284e-05,
+      "loss": 0.0001,
+      "step": 17212
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 4.799253731343284e-05,
+      "loss": 0.0005,
+      "step": 17216
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 4.799207089552239e-05,
+      "loss": 0.0023,
+      "step": 17220
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 4.799160447761194e-05,
+      "loss": 0.0112,
+      "step": 17224
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 4.79911380597015e-05,
+      "loss": 0.0062,
+      "step": 17228
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 4.7990671641791045e-05,
+      "loss": 0.001,
+      "step": 17232
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 4.799020522388059e-05,
+      "loss": 0.0012,
+      "step": 17236
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 4.7989738805970155e-05,
+      "loss": 0.0004,
+      "step": 17240
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 4.79892723880597e-05,
+      "loss": 0.0001,
+      "step": 17244
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 4.798880597014926e-05,
+      "loss": 0.0014,
+      "step": 17248
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 4.7988339552238806e-05,
+      "loss": 0.003,
+      "step": 17252
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 4.798787313432836e-05,
+      "loss": 0.0062,
+      "step": 17256
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 4.7987406716417916e-05,
+      "loss": 0.0018,
+      "step": 17260
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 4.7986940298507464e-05,
+      "loss": 0.001,
+      "step": 17264
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 4.798647388059702e-05,
+      "loss": 0.0036,
+      "step": 17268
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 4.7986007462686574e-05,
+      "loss": 0.0008,
+      "step": 17272
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 4.798554104477612e-05,
+      "loss": 0.0004,
+      "step": 17276
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 4.798507462686567e-05,
+      "loss": 0.0041,
+      "step": 17280
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 4.7984608208955225e-05,
+      "loss": 0.0003,
+      "step": 17284
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 4.798414179104478e-05,
+      "loss": 0.0012,
+      "step": 17288
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 4.798367537313433e-05,
+      "loss": 0.0057,
+      "step": 17292
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 4.798320895522388e-05,
+      "loss": 0.0005,
+      "step": 17296
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 4.798274253731344e-05,
+      "loss": 0.0002,
+      "step": 17300
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 4.7982276119402986e-05,
+      "loss": 0.0036,
+      "step": 17304
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 4.798180970149254e-05,
+      "loss": 0.0022,
+      "step": 17308
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 4.798134328358209e-05,
+      "loss": 0.0042,
+      "step": 17312
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 4.7980876865671644e-05,
+      "loss": 0.0023,
+      "step": 17316
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 4.79804104477612e-05,
+      "loss": 0.0008,
+      "step": 17320
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 4.797994402985075e-05,
+      "loss": 0.0009,
+      "step": 17324
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 4.79794776119403e-05,
+      "loss": 0.0075,
+      "step": 17328
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 4.797901119402986e-05,
+      "loss": 0.0036,
+      "step": 17332
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 4.7978544776119405e-05,
+      "loss": 0.0057,
+      "step": 17336
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 4.797807835820895e-05,
+      "loss": 0.0024,
+      "step": 17340
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 4.797761194029851e-05,
+      "loss": 0.0023,
+      "step": 17344
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 4.797714552238806e-05,
+      "loss": 0.0398,
+      "step": 17348
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 4.797667910447761e-05,
+      "loss": 0.0049,
+      "step": 17352
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 4.7976212686567166e-05,
+      "loss": 0.0004,
+      "step": 17356
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 4.797574626865672e-05,
+      "loss": 0.0033,
+      "step": 17360
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 4.797527985074627e-05,
+      "loss": 0.0007,
+      "step": 17364
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 4.7974813432835824e-05,
+      "loss": 0.0017,
+      "step": 17368
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 4.797434701492537e-05,
+      "loss": 0.0024,
+      "step": 17372
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 4.797388059701493e-05,
+      "loss": 0.0026,
+      "step": 17376
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 4.797341417910448e-05,
+      "loss": 0.0003,
+      "step": 17380
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 4.797294776119403e-05,
+      "loss": 0.0032,
+      "step": 17384
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 4.7972481343283585e-05,
+      "loss": 0.0009,
+      "step": 17388
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 4.797201492537314e-05,
+      "loss": 0.002,
+      "step": 17392
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 4.797154850746269e-05,
+      "loss": 0.001,
+      "step": 17396
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 4.7971082089552236e-05,
+      "loss": 0.0016,
+      "step": 17400
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 4.797061567164179e-05,
+      "loss": 0.0039,
+      "step": 17404
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 4.7970149253731346e-05,
+      "loss": 0.002,
+      "step": 17408
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 4.79696828358209e-05,
+      "loss": 0.0006,
+      "step": 17412
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 4.796921641791045e-05,
+      "loss": 0.0004,
+      "step": 17416
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 4.7968750000000004e-05,
+      "loss": 0.0054,
+      "step": 17420
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 4.796828358208956e-05,
+      "loss": 0.0098,
+      "step": 17424
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 4.796781716417911e-05,
+      "loss": 0.0011,
+      "step": 17428
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 4.7967350746268655e-05,
+      "loss": 0.0047,
+      "step": 17432
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 4.796688432835822e-05,
+      "loss": 0.007,
+      "step": 17436
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 4.7966417910447765e-05,
+      "loss": 0.0004,
+      "step": 17440
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 4.796595149253731e-05,
+      "loss": 0.0062,
+      "step": 17444
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 4.796548507462687e-05,
+      "loss": 0.0034,
+      "step": 17448
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 4.796501865671642e-05,
+      "loss": 0.0049,
+      "step": 17452
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 4.796455223880597e-05,
+      "loss": 0.0079,
+      "step": 17456
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 4.7964085820895526e-05,
+      "loss": 0.0007,
+      "step": 17460
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 4.7963619402985074e-05,
+      "loss": 0.0051,
+      "step": 17464
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 4.796315298507463e-05,
+      "loss": 0.0125,
+      "step": 17468
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 4.7962686567164184e-05,
+      "loss": 0.0018,
+      "step": 17472
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 4.796222014925373e-05,
+      "loss": 0.0008,
+      "step": 17476
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 4.796175373134329e-05,
+      "loss": 0.0007,
+      "step": 17480
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 4.796128731343284e-05,
+      "loss": 0.0022,
+      "step": 17484
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 4.796082089552239e-05,
+      "loss": 0.0008,
+      "step": 17488
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 4.796035447761194e-05,
+      "loss": 0.0038,
+      "step": 17492
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 4.79598880597015e-05,
+      "loss": 0.0036,
+      "step": 17496
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 4.795942164179105e-05,
+      "loss": 0.0002,
+      "step": 17500
+    },
+    {
+      "epoch": 16.32,
+      "eval_exact_match": 0.7330754352030948,
+      "eval_exec": 0.7475822050290135,
+      "eval_loss": 0.3184237778186798,
+      "eval_runtime": 1056.9157,
+      "eval_samples_per_second": 0.978,
+      "step": 17500
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 4.7958955223880596e-05,
+      "loss": 0.0139,
+      "step": 17504
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 4.795848880597015e-05,
+      "loss": 0.0024,
+      "step": 17508
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 4.7958022388059706e-05,
+      "loss": 0.005,
+      "step": 17512
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 4.7957555970149254e-05,
+      "loss": 0.001,
+      "step": 17516
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 4.795708955223881e-05,
+      "loss": 0.0053,
+      "step": 17520
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 4.795662313432836e-05,
+      "loss": 0.0021,
+      "step": 17524
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 4.795615671641791e-05,
+      "loss": 0.0059,
+      "step": 17528
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 4.795569029850747e-05,
+      "loss": 0.0008,
+      "step": 17532
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 4.7955223880597015e-05,
+      "loss": 0.0151,
+      "step": 17536
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 4.795475746268657e-05,
+      "loss": 0.0023,
+      "step": 17540
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 4.7954291044776125e-05,
+      "loss": 0.0022,
+      "step": 17544
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 4.795382462686567e-05,
+      "loss": 0.0004,
+      "step": 17548
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 4.795335820895522e-05,
+      "loss": 0.0005,
+      "step": 17552
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 4.795289179104478e-05,
+      "loss": 0.0019,
+      "step": 17556
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 4.795242537313433e-05,
+      "loss": 0.0057,
+      "step": 17560
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 4.795195895522388e-05,
+      "loss": 0.0046,
+      "step": 17564
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 4.7951492537313434e-05,
+      "loss": 0.0009,
+      "step": 17568
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 4.795102611940299e-05,
+      "loss": 0.0035,
+      "step": 17572
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 4.7950559701492544e-05,
+      "loss": 0.0062,
+      "step": 17576
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 4.795009328358209e-05,
+      "loss": 0.0036,
+      "step": 17580
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 4.794962686567164e-05,
+      "loss": 0.0003,
+      "step": 17584
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 4.79491604477612e-05,
+      "loss": 0.011,
+      "step": 17588
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 4.794869402985075e-05,
+      "loss": 0.0021,
+      "step": 17592
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 4.79482276119403e-05,
+      "loss": 0.0015,
+      "step": 17596
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 4.794776119402985e-05,
+      "loss": 0.0022,
+      "step": 17600
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 4.794729477611941e-05,
+      "loss": 0.0033,
+      "step": 17604
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 4.7946828358208956e-05,
+      "loss": 0.0012,
+      "step": 17608
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 4.794636194029851e-05,
+      "loss": 0.0006,
+      "step": 17612
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 4.7945895522388066e-05,
+      "loss": 0.0052,
+      "step": 17616
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 4.7945429104477614e-05,
+      "loss": 0.0002,
+      "step": 17620
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 4.794496268656717e-05,
+      "loss": 0.0016,
+      "step": 17624
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 4.794449626865672e-05,
+      "loss": 0.0009,
+      "step": 17628
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 4.794402985074627e-05,
+      "loss": 0.0031,
+      "step": 17632
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 4.7943563432835827e-05,
+      "loss": 0.0077,
+      "step": 17636
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 4.7943097014925375e-05,
+      "loss": 0.0037,
+      "step": 17640
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 4.794263059701492e-05,
+      "loss": 0.0023,
+      "step": 17644
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 4.7942164179104485e-05,
+      "loss": 0.0008,
+      "step": 17648
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 4.794169776119403e-05,
+      "loss": 0.0013,
+      "step": 17652
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 4.794123134328358e-05,
+      "loss": 0.0015,
+      "step": 17656
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 4.7940764925373136e-05,
+      "loss": 0.0071,
+      "step": 17660
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 4.794029850746269e-05,
+      "loss": 0.0004,
+      "step": 17664
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 4.793983208955224e-05,
+      "loss": 0.0041,
+      "step": 17668
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 4.7939365671641794e-05,
+      "loss": 0.0022,
+      "step": 17672
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 4.793889925373135e-05,
+      "loss": 0.0052,
+      "step": 17676
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 4.79384328358209e-05,
+      "loss": 0.0029,
+      "step": 17680
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 4.793796641791045e-05,
+      "loss": 0.0017,
+      "step": 17684
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 4.79375e-05,
+      "loss": 0.0033,
+      "step": 17688
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 4.7937033582089555e-05,
+      "loss": 0.0054,
+      "step": 17692
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 4.793656716417911e-05,
+      "loss": 0.0005,
+      "step": 17696
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 4.793610074626866e-05,
+      "loss": 0.0047,
+      "step": 17700
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 4.7935634328358206e-05,
+      "loss": 0.0014,
+      "step": 17704
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 4.793516791044777e-05,
+      "loss": 0.0014,
+      "step": 17708
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 4.7934701492537316e-05,
+      "loss": 0.0051,
+      "step": 17712
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 4.7934235074626864e-05,
+      "loss": 0.0008,
+      "step": 17716
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 4.793376865671642e-05,
+      "loss": 0.0046,
+      "step": 17720
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 4.7933302238805974e-05,
+      "loss": 0.0013,
+      "step": 17724
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 4.793283582089552e-05,
+      "loss": 0.0003,
+      "step": 17728
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 4.7932369402985077e-05,
+      "loss": 0.0048,
+      "step": 17732
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 4.793190298507463e-05,
+      "loss": 0.0008,
+      "step": 17736
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 4.7931436567164186e-05,
+      "loss": 0.0007,
+      "step": 17740
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 4.7930970149253735e-05,
+      "loss": 0.001,
+      "step": 17744
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 4.793050373134328e-05,
+      "loss": 0.005,
+      "step": 17748
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 4.793003731343284e-05,
+      "loss": 0.0033,
+      "step": 17752
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 4.792957089552239e-05,
+      "loss": 0.0031,
+      "step": 17756
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 4.792910447761194e-05,
+      "loss": 0.0004,
+      "step": 17760
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 4.7928638059701496e-05,
+      "loss": 0.0024,
+      "step": 17764
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 4.792817164179105e-05,
+      "loss": 0.0059,
+      "step": 17768
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 4.79277052238806e-05,
+      "loss": 0.0031,
+      "step": 17772
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 4.7927238805970153e-05,
+      "loss": 0.0009,
+      "step": 17776
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 4.79267723880597e-05,
+      "loss": 0.0044,
+      "step": 17780
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 4.7926305970149256e-05,
+      "loss": 0.0076,
+      "step": 17784
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 4.792583955223881e-05,
+      "loss": 0.0002,
+      "step": 17788
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 4.792537313432836e-05,
+      "loss": 0.0005,
+      "step": 17792
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 4.7924906716417914e-05,
+      "loss": 0.0018,
+      "step": 17796
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 4.792444029850747e-05,
+      "loss": 0.0041,
+      "step": 17800
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 4.792397388059702e-05,
+      "loss": 0.0115,
+      "step": 17804
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 4.7923507462686566e-05,
+      "loss": 0.0022,
+      "step": 17808
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 4.792304104477612e-05,
+      "loss": 0.0012,
+      "step": 17812
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 4.7922574626865675e-05,
+      "loss": 0.0002,
+      "step": 17816
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 4.7922108208955224e-05,
+      "loss": 0.0014,
+      "step": 17820
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 4.792164179104478e-05,
+      "loss": 0.0007,
+      "step": 17824
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 4.792117537313433e-05,
+      "loss": 0.0051,
+      "step": 17828
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 4.792070895522388e-05,
+      "loss": 0.0007,
+      "step": 17832
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 4.7920242537313436e-05,
+      "loss": 0.0012,
+      "step": 17836
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 4.7919776119402985e-05,
+      "loss": 0.0037,
+      "step": 17840
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 4.791930970149254e-05,
+      "loss": 0.0002,
+      "step": 17844
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 4.7918843283582094e-05,
+      "loss": 0.0069,
+      "step": 17848
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 4.791837686567164e-05,
+      "loss": 0.0016,
+      "step": 17852
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 4.791791044776119e-05,
+      "loss": 0.0029,
+      "step": 17856
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 4.791744402985075e-05,
+      "loss": 0.005,
+      "step": 17860
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 4.79169776119403e-05,
+      "loss": 0.0005,
+      "step": 17864
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 4.791651119402985e-05,
+      "loss": 0.0014,
+      "step": 17868
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 4.7916044776119403e-05,
+      "loss": 0.0003,
+      "step": 17872
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 4.791557835820896e-05,
+      "loss": 0.0012,
+      "step": 17876
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 4.7915111940298507e-05,
+      "loss": 0.0041,
+      "step": 17880
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 4.791464552238806e-05,
+      "loss": 0.0016,
+      "step": 17884
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 4.7914179104477616e-05,
+      "loss": 0.0005,
+      "step": 17888
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 4.7913712686567164e-05,
+      "loss": 0.0002,
+      "step": 17892
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 4.791324626865672e-05,
+      "loss": 0.0017,
+      "step": 17896
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 4.791277985074627e-05,
+      "loss": 0.0069,
+      "step": 17900
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 4.791231343283583e-05,
+      "loss": 0.0039,
+      "step": 17904
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 4.791184701492538e-05,
+      "loss": 0.0006,
+      "step": 17908
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 4.7911380597014925e-05,
+      "loss": 0.0,
+      "step": 17912
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 4.791091417910448e-05,
+      "loss": 0.0031,
+      "step": 17916
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 4.7910447761194035e-05,
+      "loss": 0.0014,
+      "step": 17920
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 4.7909981343283583e-05,
+      "loss": 0.0024,
+      "step": 17924
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 4.790951492537314e-05,
+      "loss": 0.0006,
+      "step": 17928
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 4.7909048507462686e-05,
+      "loss": 0.0009,
+      "step": 17932
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 4.790858208955224e-05,
+      "loss": 0.0012,
+      "step": 17936
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 4.7908115671641796e-05,
+      "loss": 0.0021,
+      "step": 17940
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 4.7907649253731344e-05,
+      "loss": 0.0082,
+      "step": 17944
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 4.79071828358209e-05,
+      "loss": 0.0014,
+      "step": 17948
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 4.7906716417910454e-05,
+      "loss": 0.0013,
+      "step": 17952
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 4.790625e-05,
+      "loss": 0.0008,
+      "step": 17956
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 4.790578358208955e-05,
+      "loss": 0.0026,
+      "step": 17960
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 4.790531716417911e-05,
+      "loss": 0.0003,
+      "step": 17964
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 4.790485074626866e-05,
+      "loss": 0.0055,
+      "step": 17968
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 4.790438432835821e-05,
+      "loss": 0.0009,
+      "step": 17972
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 4.790391791044776e-05,
+      "loss": 0.0008,
+      "step": 17976
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 4.790345149253732e-05,
+      "loss": 0.0027,
+      "step": 17980
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 4.7902985074626866e-05,
+      "loss": 0.0003,
+      "step": 17984
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 4.790251865671642e-05,
+      "loss": 0.0023,
+      "step": 17988
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 4.790205223880597e-05,
+      "loss": 0.0005,
+      "step": 17992
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 4.7901585820895524e-05,
+      "loss": 0.0024,
+      "step": 17996
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 4.790111940298508e-05,
+      "loss": 0.0005,
+      "step": 18000
+    },
+    {
+      "epoch": 16.79,
+      "eval_exact_match": 0.730174081237911,
+      "eval_exec": 0.7553191489361702,
+      "eval_loss": 0.3473448157310486,
+      "eval_runtime": 1359.5008,
+      "eval_samples_per_second": 0.761,
+      "step": 18000
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 4.790065298507463e-05,
+      "loss": 0.0041,
+      "step": 18004
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 4.790018656716418e-05,
+      "loss": 0.0034,
+      "step": 18008
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 4.789972014925374e-05,
+      "loss": 0.0004,
+      "step": 18012
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 4.7899253731343285e-05,
+      "loss": 0.0046,
+      "step": 18016
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 4.7898787313432833e-05,
+      "loss": 0.001,
+      "step": 18020
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 4.789832089552239e-05,
+      "loss": 0.0069,
+      "step": 18024
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 4.789785447761194e-05,
+      "loss": 0.0081,
+      "step": 18028
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 4.789738805970149e-05,
+      "loss": 0.0012,
+      "step": 18032
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 4.7896921641791046e-05,
+      "loss": 0.0023,
+      "step": 18036
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 4.78964552238806e-05,
+      "loss": 0.0049,
+      "step": 18040
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 4.789598880597015e-05,
+      "loss": 0.0011,
+      "step": 18044
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 4.7895522388059704e-05,
+      "loss": 0.0028,
+      "step": 18048
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 4.789505597014925e-05,
+      "loss": 0.0016,
+      "step": 18052
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 4.789458955223881e-05,
+      "loss": 0.0007,
+      "step": 18056
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 4.789412313432836e-05,
+      "loss": 0.0008,
+      "step": 18060
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 4.789365671641791e-05,
+      "loss": 0.0013,
+      "step": 18064
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 4.7893190298507465e-05,
+      "loss": 0.0038,
+      "step": 18068
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 4.789272388059702e-05,
+      "loss": 0.0002,
+      "step": 18072
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 4.789225746268657e-05,
+      "loss": 0.0016,
+      "step": 18076
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 4.789179104477612e-05,
+      "loss": 0.0003,
+      "step": 18080
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 4.789132462686567e-05,
+      "loss": 0.0013,
+      "step": 18084
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 4.7890858208955226e-05,
+      "loss": 0.0124,
+      "step": 18088
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 4.789039179104478e-05,
+      "loss": 0.0009,
+      "step": 18092
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 4.788992537313433e-05,
+      "loss": 0.0034,
+      "step": 18096
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 4.7889458955223884e-05,
+      "loss": 0.0118,
+      "step": 18100
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 4.788899253731344e-05,
+      "loss": 0.003,
+      "step": 18104
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 4.788852611940299e-05,
+      "loss": 0.0029,
+      "step": 18108
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 4.7888059701492535e-05,
+      "loss": 0.0013,
+      "step": 18112
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 4.78875932835821e-05,
+      "loss": 0.0057,
+      "step": 18116
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 4.7887126865671645e-05,
+      "loss": 0.0001,
+      "step": 18120
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 4.788666044776119e-05,
+      "loss": 0.001,
+      "step": 18124
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 4.788619402985075e-05,
+      "loss": 0.011,
+      "step": 18128
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 4.78857276119403e-05,
+      "loss": 0.0044,
+      "step": 18132
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 4.788526119402985e-05,
+      "loss": 0.0017,
+      "step": 18136
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 4.7884794776119406e-05,
+      "loss": 0.0047,
+      "step": 18140
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 4.7884328358208954e-05,
+      "loss": 0.0044,
+      "step": 18144
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 4.788386194029851e-05,
+      "loss": 0.0012,
+      "step": 18148
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 4.7883395522388064e-05,
+      "loss": 0.0003,
+      "step": 18152
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 4.788292910447761e-05,
+      "loss": 0.0006,
+      "step": 18156
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 4.788246268656717e-05,
+      "loss": 0.0041,
+      "step": 18160
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 4.788199626865672e-05,
+      "loss": 0.0014,
+      "step": 18164
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 4.788152985074627e-05,
+      "loss": 0.0039,
+      "step": 18168
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 4.788106343283582e-05,
+      "loss": 0.0021,
+      "step": 18172
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 4.788059701492538e-05,
+      "loss": 0.0052,
+      "step": 18176
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 4.788013059701493e-05,
+      "loss": 0.0009,
+      "step": 18180
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 4.7879664179104476e-05,
+      "loss": 0.0008,
+      "step": 18184
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 4.787919776119403e-05,
+      "loss": 0.0023,
+      "step": 18188
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 4.7878731343283586e-05,
+      "loss": 0.0028,
+      "step": 18192
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 4.7878264925373134e-05,
+      "loss": 0.0007,
+      "step": 18196
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 4.787779850746269e-05,
+      "loss": 0.0004,
+      "step": 18200
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 4.787733208955224e-05,
+      "loss": 0.0063,
+      "step": 18204
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 4.787686567164179e-05,
+      "loss": 0.0018,
+      "step": 18208
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 4.787639925373135e-05,
+      "loss": 0.0034,
+      "step": 18212
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 4.7875932835820895e-05,
+      "loss": 0.0019,
+      "step": 18216
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 4.787546641791045e-05,
+      "loss": 0.0065,
+      "step": 18220
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 4.7875000000000005e-05,
+      "loss": 0.0032,
+      "step": 18224
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 4.787453358208955e-05,
+      "loss": 0.0003,
+      "step": 18228
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 4.787406716417911e-05,
+      "loss": 0.0036,
+      "step": 18232
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 4.787360074626866e-05,
+      "loss": 0.0004,
+      "step": 18236
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 4.787313432835821e-05,
+      "loss": 0.0008,
+      "step": 18240
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 4.7872667910447766e-05,
+      "loss": 0.0037,
+      "step": 18244
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 4.7872201492537314e-05,
+      "loss": 0.0003,
+      "step": 18248
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 4.787173507462687e-05,
+      "loss": 0.0202,
+      "step": 18252
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 4.7871268656716424e-05,
+      "loss": 0.0002,
+      "step": 18256
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 4.787080223880597e-05,
+      "loss": 0.0032,
+      "step": 18260
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 4.787033582089552e-05,
+      "loss": 0.0015,
+      "step": 18264
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 4.786986940298508e-05,
+      "loss": 0.0018,
+      "step": 18268
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 4.786940298507463e-05,
+      "loss": 0.0034,
+      "step": 18272
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 4.786893656716418e-05,
+      "loss": 0.0008,
+      "step": 18276
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 4.786847014925373e-05,
+      "loss": 0.0004,
+      "step": 18280
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 4.786800373134329e-05,
+      "loss": 0.0002,
+      "step": 18284
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 4.7867537313432836e-05,
+      "loss": 0.0006,
+      "step": 18288
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 4.786707089552239e-05,
+      "loss": 0.0016,
+      "step": 18292
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 4.7866604477611946e-05,
+      "loss": 0.0019,
+      "step": 18296
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 4.7866138059701494e-05,
+      "loss": 0.003,
+      "step": 18300
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 4.786567164179105e-05,
+      "loss": 0.0014,
+      "step": 18304
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 4.78652052238806e-05,
+      "loss": 0.0056,
+      "step": 18308
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 4.786473880597015e-05,
+      "loss": 0.0004,
+      "step": 18312
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 4.786427238805971e-05,
+      "loss": 0.0002,
+      "step": 18316
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 4.7863805970149255e-05,
+      "loss": 0.0025,
+      "step": 18320
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 4.78633395522388e-05,
+      "loss": 0.0019,
+      "step": 18324
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 4.7862873134328365e-05,
+      "loss": 0.0006,
+      "step": 18328
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 4.786240671641791e-05,
+      "loss": 0.0006,
+      "step": 18332
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 4.786194029850746e-05,
+      "loss": 0.0004,
+      "step": 18336
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 4.7861473880597016e-05,
+      "loss": 0.0008,
+      "step": 18340
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 4.786100746268657e-05,
+      "loss": 0.0019,
+      "step": 18344
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 4.786054104477612e-05,
+      "loss": 0.0002,
+      "step": 18348
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 4.7860074626865674e-05,
+      "loss": 0.0008,
+      "step": 18352
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 4.785960820895523e-05,
+      "loss": 0.0015,
+      "step": 18356
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 4.785914179104478e-05,
+      "loss": 0.0041,
+      "step": 18360
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 4.785867537313433e-05,
+      "loss": 0.0029,
+      "step": 18364
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 4.785820895522388e-05,
+      "loss": 0.0002,
+      "step": 18368
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 4.7857742537313435e-05,
+      "loss": 0.0004,
+      "step": 18372
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 4.785727611940299e-05,
+      "loss": 0.0019,
+      "step": 18376
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 4.785680970149254e-05,
+      "loss": 0.0011,
+      "step": 18380
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 4.7856343283582086e-05,
+      "loss": 0.0027,
+      "step": 18384
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 4.785587686567165e-05,
+      "loss": 0.0055,
+      "step": 18388
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 4.7855410447761196e-05,
+      "loss": 0.0001,
+      "step": 18392
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 4.785494402985075e-05,
+      "loss": 0.0004,
+      "step": 18396
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 4.78544776119403e-05,
+      "loss": 0.0002,
+      "step": 18400
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 4.7854011194029854e-05,
+      "loss": 0.006,
+      "step": 18404
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 4.785354477611941e-05,
+      "loss": 0.0001,
+      "step": 18408
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 4.785307835820896e-05,
+      "loss": 0.0029,
+      "step": 18412
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 4.785261194029851e-05,
+      "loss": 0.0004,
+      "step": 18416
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 4.785214552238807e-05,
+      "loss": 0.0008,
+      "step": 18420
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 4.7851679104477615e-05,
+      "loss": 0.0004,
+      "step": 18424
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 4.785121268656716e-05,
+      "loss": 0.0022,
+      "step": 18428
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 4.785074626865672e-05,
+      "loss": 0.0009,
+      "step": 18432
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 4.785027985074627e-05,
+      "loss": 0.0009,
+      "step": 18436
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 4.784981343283582e-05,
+      "loss": 0.0053,
+      "step": 18440
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 4.7849347014925376e-05,
+      "loss": 0.0019,
+      "step": 18444
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 4.784888059701493e-05,
+      "loss": 0.0028,
+      "step": 18448
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 4.784841417910448e-05,
+      "loss": 0.0016,
+      "step": 18452
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 4.7847947761194034e-05,
+      "loss": 0.0012,
+      "step": 18456
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 4.784748134328358e-05,
+      "loss": 0.0004,
+      "step": 18460
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 4.784701492537314e-05,
+      "loss": 0.0002,
+      "step": 18464
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 4.784654850746269e-05,
+      "loss": 0.0049,
+      "step": 18468
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 4.784608208955224e-05,
+      "loss": 0.0037,
+      "step": 18472
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 4.7845615671641795e-05,
+      "loss": 0.0102,
+      "step": 18476
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 4.784514925373135e-05,
+      "loss": 0.0021,
+      "step": 18480
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 4.78446828358209e-05,
+      "loss": 0.0025,
+      "step": 18484
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 4.7844216417910446e-05,
+      "loss": 0.0012,
+      "step": 18488
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 4.784375e-05,
+      "loss": 0.0019,
+      "step": 18492
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 4.7843283582089556e-05,
+      "loss": 0.0006,
+      "step": 18496
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 4.7842817164179104e-05,
+      "loss": 0.007,
+      "step": 18500
+    },
+    {
+      "epoch": 17.26,
+      "eval_exact_match": 0.7292069632495164,
+      "eval_exec": 0.7562862669245648,
+      "eval_loss": 0.35152459144592285,
+      "eval_runtime": 1127.9193,
+      "eval_samples_per_second": 0.917,
+      "step": 18500
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 4.784235074626866e-05,
+      "loss": 0.0043,
+      "step": 18504
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 4.7841884328358214e-05,
+      "loss": 0.0025,
+      "step": 18508
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 4.784141791044776e-05,
+      "loss": 0.0137,
+      "step": 18512
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 4.784095149253732e-05,
+      "loss": 0.002,
+      "step": 18516
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 4.7840485074626865e-05,
+      "loss": 0.0014,
+      "step": 18520
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 4.784001865671642e-05,
+      "loss": 0.0019,
+      "step": 18524
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 4.7839552238805975e-05,
+      "loss": 0.0002,
+      "step": 18528
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 4.783908582089552e-05,
+      "loss": 0.0006,
+      "step": 18532
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 4.783861940298507e-05,
+      "loss": 0.0048,
+      "step": 18536
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 4.783815298507463e-05,
+      "loss": 0.0017,
+      "step": 18540
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 4.783768656716418e-05,
+      "loss": 0.0053,
+      "step": 18544
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 4.7837220149253736e-05,
+      "loss": 0.0008,
+      "step": 18548
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 4.7836753731343284e-05,
+      "loss": 0.0023,
+      "step": 18552
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 4.783628731343284e-05,
+      "loss": 0.0049,
+      "step": 18556
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 4.7835820895522394e-05,
+      "loss": 0.0028,
+      "step": 18560
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 4.783535447761194e-05,
+      "loss": 0.0002,
+      "step": 18564
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 4.78348880597015e-05,
+      "loss": 0.001,
+      "step": 18568
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 4.783442164179105e-05,
+      "loss": 0.0032,
+      "step": 18572
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 4.78339552238806e-05,
+      "loss": 0.0009,
+      "step": 18576
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 4.783348880597015e-05,
+      "loss": 0.0019,
+      "step": 18580
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 4.783302238805971e-05,
+      "loss": 0.0008,
+      "step": 18584
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 4.783255597014926e-05,
+      "loss": 0.0029,
+      "step": 18588
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 4.7832089552238806e-05,
+      "loss": 0.0024,
+      "step": 18592
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 4.783162313432836e-05,
+      "loss": 0.0064,
+      "step": 18596
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 4.7831156716417916e-05,
+      "loss": 0.0005,
+      "step": 18600
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 4.7830690298507464e-05,
+      "loss": 0.0026,
+      "step": 18604
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 4.783022388059702e-05,
+      "loss": 0.0002,
+      "step": 18608
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 4.782975746268657e-05,
+      "loss": 0.0007,
+      "step": 18612
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 4.782929104477612e-05,
+      "loss": 0.0005,
+      "step": 18616
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 4.7828824626865677e-05,
+      "loss": 0.0015,
+      "step": 18620
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 4.7828358208955225e-05,
+      "loss": 0.002,
+      "step": 18624
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 4.782789179104478e-05,
+      "loss": 0.0133,
+      "step": 18628
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 4.7827425373134334e-05,
+      "loss": 0.0004,
+      "step": 18632
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 4.782695895522388e-05,
+      "loss": 0.0002,
+      "step": 18636
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 4.782649253731343e-05,
+      "loss": 0.0052,
+      "step": 18640
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 4.782602611940299e-05,
+      "loss": 0.0012,
+      "step": 18644
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 4.782555970149254e-05,
+      "loss": 0.0045,
+      "step": 18648
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 4.782509328358209e-05,
+      "loss": 0.0024,
+      "step": 18652
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 4.7824626865671644e-05,
+      "loss": 0.0025,
+      "step": 18656
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 4.78241604477612e-05,
+      "loss": 0.004,
+      "step": 18660
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 4.782369402985075e-05,
+      "loss": 0.0038,
+      "step": 18664
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 4.78232276119403e-05,
+      "loss": 0.0013,
+      "step": 18668
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 4.782276119402985e-05,
+      "loss": 0.002,
+      "step": 18672
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 4.7822294776119405e-05,
+      "loss": 0.0007,
+      "step": 18676
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 4.782182835820896e-05,
+      "loss": 0.0028,
+      "step": 18680
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 4.782136194029851e-05,
+      "loss": 0.0001,
+      "step": 18684
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 4.782089552238806e-05,
+      "loss": 0.0058,
+      "step": 18688
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 4.782042910447762e-05,
+      "loss": 0.0038,
+      "step": 18692
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 4.7819962686567166e-05,
+      "loss": 0.0014,
+      "step": 18696
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 4.7819496268656714e-05,
+      "loss": 0.0005,
+      "step": 18700
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 4.781902985074627e-05,
+      "loss": 0.0137,
+      "step": 18704
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 4.7818563432835824e-05,
+      "loss": 0.0038,
+      "step": 18708
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 4.781809701492538e-05,
+      "loss": 0.0009,
+      "step": 18712
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 4.7817630597014927e-05,
+      "loss": 0.0006,
+      "step": 18716
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 4.781716417910448e-05,
+      "loss": 0.0013,
+      "step": 18720
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 4.7816697761194036e-05,
+      "loss": 0.0049,
+      "step": 18724
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 4.7816231343283585e-05,
+      "loss": 0.0006,
+      "step": 18728
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 4.781576492537313e-05,
+      "loss": 0.004,
+      "step": 18732
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 4.7815298507462694e-05,
+      "loss": 0.0003,
+      "step": 18736
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 4.781483208955224e-05,
+      "loss": 0.0011,
+      "step": 18740
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 4.781436567164179e-05,
+      "loss": 0.0006,
+      "step": 18744
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 4.7813899253731345e-05,
+      "loss": 0.0025,
+      "step": 18748
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 4.78134328358209e-05,
+      "loss": 0.0009,
+      "step": 18752
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 4.781296641791045e-05,
+      "loss": 0.0027,
+      "step": 18756
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 4.7812500000000003e-05,
+      "loss": 0.0025,
+      "step": 18760
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 4.781203358208955e-05,
+      "loss": 0.0008,
+      "step": 18764
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 4.7811567164179106e-05,
+      "loss": 0.0051,
+      "step": 18768
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 4.781110074626866e-05,
+      "loss": 0.0012,
+      "step": 18772
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 4.781063432835821e-05,
+      "loss": 0.0003,
+      "step": 18776
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 4.7810167910447764e-05,
+      "loss": 0.0024,
+      "step": 18780
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 4.780970149253732e-05,
+      "loss": 0.0002,
+      "step": 18784
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 4.780923507462687e-05,
+      "loss": 0.0074,
+      "step": 18788
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 4.7808768656716416e-05,
+      "loss": 0.0023,
+      "step": 18792
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 4.780830223880598e-05,
+      "loss": 0.0009,
+      "step": 18796
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 4.7807835820895525e-05,
+      "loss": 0.0157,
+      "step": 18800
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 4.7807369402985074e-05,
+      "loss": 0.0032,
+      "step": 18804
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 4.780690298507463e-05,
+      "loss": 0.0005,
+      "step": 18808
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 4.780643656716418e-05,
+      "loss": 0.0012,
+      "step": 18812
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 4.780597014925373e-05,
+      "loss": 0.0014,
+      "step": 18816
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 4.7805503731343286e-05,
+      "loss": 0.0055,
+      "step": 18820
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 4.7805037313432835e-05,
+      "loss": 0.0026,
+      "step": 18824
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 4.780457089552239e-05,
+      "loss": 0.0003,
+      "step": 18828
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 4.7804104477611944e-05,
+      "loss": 0.0013,
+      "step": 18832
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 4.780363805970149e-05,
+      "loss": 0.0062,
+      "step": 18836
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 4.780317164179105e-05,
+      "loss": 0.0007,
+      "step": 18840
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 4.78027052238806e-05,
+      "loss": 0.001,
+      "step": 18844
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 4.780223880597015e-05,
+      "loss": 0.0005,
+      "step": 18848
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 4.78017723880597e-05,
+      "loss": 0.0013,
+      "step": 18852
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 4.780130597014926e-05,
+      "loss": 0.0033,
+      "step": 18856
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 4.780083955223881e-05,
+      "loss": 0.0028,
+      "step": 18860
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 4.7800373134328356e-05,
+      "loss": 0.0045,
+      "step": 18864
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 4.779990671641791e-05,
+      "loss": 0.001,
+      "step": 18868
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 4.7799440298507466e-05,
+      "loss": 0.0095,
+      "step": 18872
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 4.779897388059702e-05,
+      "loss": 0.0001,
+      "step": 18876
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 4.779850746268657e-05,
+      "loss": 0.0029,
+      "step": 18880
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 4.779804104477612e-05,
+      "loss": 0.0007,
+      "step": 18884
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 4.779757462686568e-05,
+      "loss": 0.0042,
+      "step": 18888
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 4.779710820895523e-05,
+      "loss": 0.0029,
+      "step": 18892
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 4.7796641791044775e-05,
+      "loss": 0.0007,
+      "step": 18896
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 4.779617537313433e-05,
+      "loss": 0.001,
+      "step": 18900
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 4.7795708955223885e-05,
+      "loss": 0.003,
+      "step": 18904
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 4.779524253731343e-05,
+      "loss": 0.0074,
+      "step": 18908
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 4.779477611940299e-05,
+      "loss": 0.0013,
+      "step": 18912
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 4.779430970149254e-05,
+      "loss": 0.0011,
+      "step": 18916
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 4.779384328358209e-05,
+      "loss": 0.004,
+      "step": 18920
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 4.7793376865671646e-05,
+      "loss": 0.0014,
+      "step": 18924
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 4.7792910447761194e-05,
+      "loss": 0.0015,
+      "step": 18928
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 4.779244402985075e-05,
+      "loss": 0.0012,
+      "step": 18932
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 4.7791977611940304e-05,
+      "loss": 0.0003,
+      "step": 18936
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 4.779151119402985e-05,
+      "loss": 0.0001,
+      "step": 18940
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 4.77910447761194e-05,
+      "loss": 0.0025,
+      "step": 18944
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 4.779057835820896e-05,
+      "loss": 0.0009,
+      "step": 18948
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 4.779011194029851e-05,
+      "loss": 0.0003,
+      "step": 18952
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 4.778964552238806e-05,
+      "loss": 0.0022,
+      "step": 18956
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 4.778917910447761e-05,
+      "loss": 0.0035,
+      "step": 18960
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 4.778871268656717e-05,
+      "loss": 0.0004,
+      "step": 18964
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 4.7788246268656716e-05,
+      "loss": 0.0001,
+      "step": 18968
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 4.778777985074627e-05,
+      "loss": 0.0002,
+      "step": 18972
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 4.7787313432835826e-05,
+      "loss": 0.0026,
+      "step": 18976
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 4.7786847014925374e-05,
+      "loss": 0.0026,
+      "step": 18980
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 4.778638059701493e-05,
+      "loss": 0.0017,
+      "step": 18984
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 4.778591417910448e-05,
+      "loss": 0.0053,
+      "step": 18988
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 4.778544776119403e-05,
+      "loss": 0.0042,
+      "step": 18992
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 4.778498134328359e-05,
+      "loss": 0.0018,
+      "step": 18996
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 4.7784514925373135e-05,
+      "loss": 0.0001,
+      "step": 19000
+    },
+    {
+      "epoch": 17.72,
+      "eval_exact_match": 0.7282398452611218,
+      "eval_exec": 0.7601547388781431,
+      "eval_loss": 0.3401549458503723,
+      "eval_runtime": 1055.5133,
+      "eval_samples_per_second": 0.98,
+      "step": 19000
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 4.7784048507462683e-05,
+      "loss": 0.0016,
+      "step": 19004
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 4.7783582089552245e-05,
+      "loss": 0.0027,
+      "step": 19008
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 4.778311567164179e-05,
+      "loss": 0.0041,
+      "step": 19012
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 4.778264925373134e-05,
+      "loss": 0.0009,
+      "step": 19016
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 4.7782182835820896e-05,
+      "loss": 0.0073,
+      "step": 19020
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 4.778171641791045e-05,
+      "loss": 0.0015,
+      "step": 19024
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 4.778125e-05,
+      "loss": 0.0036,
+      "step": 19028
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 4.7780783582089554e-05,
+      "loss": 0.0074,
+      "step": 19032
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 4.778031716417911e-05,
+      "loss": 0.0003,
+      "step": 19036
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 4.7779850746268664e-05,
+      "loss": 0.0006,
+      "step": 19040
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 4.777938432835821e-05,
+      "loss": 0.001,
+      "step": 19044
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 4.777891791044776e-05,
+      "loss": 0.0008,
+      "step": 19048
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 4.7778451492537315e-05,
+      "loss": 0.0005,
+      "step": 19052
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 4.777798507462687e-05,
+      "loss": 0.0007,
+      "step": 19056
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 4.777751865671642e-05,
+      "loss": 0.0002,
+      "step": 19060
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 4.777705223880597e-05,
+      "loss": 0.0003,
+      "step": 19064
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 4.777658582089553e-05,
+      "loss": 0.001,
+      "step": 19068
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 4.7776119402985076e-05,
+      "loss": 0.0006,
+      "step": 19072
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 4.777565298507463e-05,
+      "loss": 0.0043,
+      "step": 19076
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 4.777518656716418e-05,
+      "loss": 0.002,
+      "step": 19080
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 4.7774720149253734e-05,
+      "loss": 0.0013,
+      "step": 19084
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 4.777425373134329e-05,
+      "loss": 0.0001,
+      "step": 19088
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 4.777378731343284e-05,
+      "loss": 0.004,
+      "step": 19092
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 4.777332089552239e-05,
+      "loss": 0.0023,
+      "step": 19096
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 4.777285447761195e-05,
+      "loss": 0.0003,
+      "step": 19100
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 4.7772388059701495e-05,
+      "loss": 0.001,
+      "step": 19104
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 4.777192164179104e-05,
+      "loss": 0.0035,
+      "step": 19108
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 4.77714552238806e-05,
+      "loss": 0.0021,
+      "step": 19112
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 4.777098880597015e-05,
+      "loss": 0.0005,
+      "step": 19116
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 4.77705223880597e-05,
+      "loss": 0.0005,
+      "step": 19120
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 4.7770055970149256e-05,
+      "loss": 0.0004,
+      "step": 19124
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 4.776958955223881e-05,
+      "loss": 0.0001,
+      "step": 19128
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 4.776912313432836e-05,
+      "loss": 0.0002,
+      "step": 19132
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 4.7768656716417914e-05,
+      "loss": 0.0117,
+      "step": 19136
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 4.776819029850746e-05,
+      "loss": 0.0008,
+      "step": 19140
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 4.776772388059702e-05,
+      "loss": 0.0037,
+      "step": 19144
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 4.776725746268657e-05,
+      "loss": 0.0002,
+      "step": 19148
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 4.776679104477612e-05,
+      "loss": 0.0005,
+      "step": 19152
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 4.7766324626865675e-05,
+      "loss": 0.0026,
+      "step": 19156
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 4.776585820895523e-05,
+      "loss": 0.0021,
+      "step": 19160
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 4.776539179104478e-05,
+      "loss": 0.0062,
+      "step": 19164
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 4.7764925373134326e-05,
+      "loss": 0.002,
+      "step": 19168
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 4.776445895522388e-05,
+      "loss": 0.0011,
+      "step": 19172
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 4.7763992537313436e-05,
+      "loss": 0.0026,
+      "step": 19176
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 4.7763526119402984e-05,
+      "loss": 0.0024,
+      "step": 19180
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 4.776305970149254e-05,
+      "loss": 0.0022,
+      "step": 19184
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 4.7762593283582094e-05,
+      "loss": 0.0018,
+      "step": 19188
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 4.776212686567164e-05,
+      "loss": 0.0019,
+      "step": 19192
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 4.77616604477612e-05,
+      "loss": 0.0001,
+      "step": 19196
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 4.7761194029850745e-05,
+      "loss": 0.0007,
+      "step": 19200
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 4.776072761194031e-05,
+      "loss": 0.0002,
+      "step": 19204
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 4.7760261194029855e-05,
+      "loss": 0.0001,
+      "step": 19208
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 4.77597947761194e-05,
+      "loss": 0.0004,
+      "step": 19212
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 4.775932835820896e-05,
+      "loss": 0.0008,
+      "step": 19216
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 4.775886194029851e-05,
+      "loss": 0.001,
+      "step": 19220
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 4.775839552238806e-05,
+      "loss": 0.001,
+      "step": 19224
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 4.7757929104477616e-05,
+      "loss": 0.0003,
+      "step": 19228
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 4.7757462686567164e-05,
+      "loss": 0.0016,
+      "step": 19232
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 4.775699626865672e-05,
+      "loss": 0.002,
+      "step": 19236
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 4.7756529850746274e-05,
+      "loss": 0.0033,
+      "step": 19240
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 4.775606343283582e-05,
+      "loss": 0.0007,
+      "step": 19244
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 4.775559701492538e-05,
+      "loss": 0.0015,
+      "step": 19248
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 4.775513059701493e-05,
+      "loss": 0.0029,
+      "step": 19252
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 4.775466417910448e-05,
+      "loss": 0.0053,
+      "step": 19256
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 4.775419776119403e-05,
+      "loss": 0.0033,
+      "step": 19260
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 4.775373134328359e-05,
+      "loss": 0.0094,
+      "step": 19264
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 4.775326492537314e-05,
+      "loss": 0.0032,
+      "step": 19268
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 4.7752798507462686e-05,
+      "loss": 0.0007,
+      "step": 19272
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 4.775233208955224e-05,
+      "loss": 0.0167,
+      "step": 19276
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 4.7751865671641796e-05,
+      "loss": 0.0062,
+      "step": 19280
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 4.7751399253731344e-05,
+      "loss": 0.005,
+      "step": 19284
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 4.77509328358209e-05,
+      "loss": 0.0014,
+      "step": 19288
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 4.775046641791045e-05,
+      "loss": 0.0047,
+      "step": 19292
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 4.775e-05,
+      "loss": 0.0049,
+      "step": 19296
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 4.774953358208956e-05,
+      "loss": 0.0003,
+      "step": 19300
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 4.7749067164179105e-05,
+      "loss": 0.0048,
+      "step": 19304
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 4.774860074626866e-05,
+      "loss": 0.001,
+      "step": 19308
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 4.7748134328358215e-05,
+      "loss": 0.0018,
+      "step": 19312
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 4.774766791044776e-05,
+      "loss": 0.0004,
+      "step": 19316
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 4.774720149253731e-05,
+      "loss": 0.0014,
+      "step": 19320
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 4.774673507462687e-05,
+      "loss": 0.0046,
+      "step": 19324
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 4.774626865671642e-05,
+      "loss": 0.0002,
+      "step": 19328
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 4.774580223880597e-05,
+      "loss": 0.0082,
+      "step": 19332
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 4.7745335820895524e-05,
+      "loss": 0.0046,
+      "step": 19336
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 4.774486940298508e-05,
+      "loss": 0.0008,
+      "step": 19340
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 4.774440298507463e-05,
+      "loss": 0.0003,
+      "step": 19344
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 4.774393656716418e-05,
+      "loss": 0.001,
+      "step": 19348
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 4.774347014925373e-05,
+      "loss": 0.002,
+      "step": 19352
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 4.7743003731343285e-05,
+      "loss": 0.0033,
+      "step": 19356
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 4.774253731343284e-05,
+      "loss": 0.0005,
+      "step": 19360
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 4.774207089552239e-05,
+      "loss": 0.0007,
+      "step": 19364
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 4.774160447761194e-05,
+      "loss": 0.0035,
+      "step": 19368
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 4.77411380597015e-05,
+      "loss": 0.0001,
+      "step": 19372
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 4.7740671641791046e-05,
+      "loss": 0.0002,
+      "step": 19376
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 4.77402052238806e-05,
+      "loss": 0.0058,
+      "step": 19380
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 4.773973880597015e-05,
+      "loss": 0.0005,
+      "step": 19384
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 4.7739272388059704e-05,
+      "loss": 0.002,
+      "step": 19388
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 4.773880597014926e-05,
+      "loss": 0.0095,
+      "step": 19392
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 4.773833955223881e-05,
+      "loss": 0.0022,
+      "step": 19396
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 4.773787313432836e-05,
+      "loss": 0.0041,
+      "step": 19400
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 4.773740671641792e-05,
+      "loss": 0.0015,
+      "step": 19404
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 4.7736940298507465e-05,
+      "loss": 0.0038,
+      "step": 19408
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 4.773647388059701e-05,
+      "loss": 0.0002,
+      "step": 19412
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 4.7736007462686575e-05,
+      "loss": 0.0033,
+      "step": 19416
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 4.773554104477612e-05,
+      "loss": 0.0003,
+      "step": 19420
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 4.773507462686567e-05,
+      "loss": 0.0005,
+      "step": 19424
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 4.7734608208955226e-05,
+      "loss": 0.0003,
+      "step": 19428
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 4.773414179104478e-05,
+      "loss": 0.0009,
+      "step": 19432
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 4.773367537313433e-05,
+      "loss": 0.0189,
+      "step": 19436
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 4.7733208955223884e-05,
+      "loss": 0.0036,
+      "step": 19440
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 4.773274253731343e-05,
+      "loss": 0.0008,
+      "step": 19444
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 4.773227611940299e-05,
+      "loss": 0.0014,
+      "step": 19448
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 4.773180970149254e-05,
+      "loss": 0.001,
+      "step": 19452
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 4.773134328358209e-05,
+      "loss": 0.0032,
+      "step": 19456
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 4.7730876865671645e-05,
+      "loss": 0.0025,
+      "step": 19460
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 4.77304104477612e-05,
+      "loss": 0.0018,
+      "step": 19464
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 4.772994402985075e-05,
+      "loss": 0.0002,
+      "step": 19468
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 4.7729477611940296e-05,
+      "loss": 0.0008,
+      "step": 19472
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 4.772901119402986e-05,
+      "loss": 0.001,
+      "step": 19476
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 4.7728544776119406e-05,
+      "loss": 0.0002,
+      "step": 19480
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 4.7728078358208954e-05,
+      "loss": 0.0,
+      "step": 19484
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 4.772761194029851e-05,
+      "loss": 0.0005,
+      "step": 19488
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 4.7727145522388064e-05,
+      "loss": 0.0026,
+      "step": 19492
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 4.772667910447761e-05,
+      "loss": 0.001,
+      "step": 19496
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 4.772621268656717e-05,
+      "loss": 0.0037,
+      "step": 19500
+    },
+    {
+      "epoch": 18.19,
+      "eval_exact_match": 0.7292069632495164,
+      "eval_exec": 0.7562862669245648,
+      "eval_loss": 0.34727153182029724,
+      "eval_runtime": 1047.0187,
+      "eval_samples_per_second": 0.988,
+      "step": 19500
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 4.7725746268656715e-05,
+      "loss": 0.0025,
+      "step": 19504
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 4.772527985074627e-05,
+      "loss": 0.0047,
+      "step": 19508
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 4.7724813432835825e-05,
+      "loss": 0.0004,
+      "step": 19512
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 4.772434701492537e-05,
+      "loss": 0.0015,
+      "step": 19516
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 4.772388059701493e-05,
+      "loss": 0.0003,
+      "step": 19520
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 4.772341417910448e-05,
+      "loss": 0.0003,
+      "step": 19524
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 4.772294776119403e-05,
+      "loss": 0.001,
+      "step": 19528
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 4.7722481343283586e-05,
+      "loss": 0.003,
+      "step": 19532
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 4.772201492537314e-05,
+      "loss": 0.0019,
+      "step": 19536
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 4.772154850746269e-05,
+      "loss": 0.0001,
+      "step": 19540
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 4.7721082089552244e-05,
+      "loss": 0.0022,
+      "step": 19544
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 4.772061567164179e-05,
+      "loss": 0.0008,
+      "step": 19548
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 4.7720149253731347e-05,
+      "loss": 0.001,
+      "step": 19552
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 4.77196828358209e-05,
+      "loss": 0.0003,
+      "step": 19556
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 4.771921641791045e-05,
+      "loss": 0.0001,
+      "step": 19560
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 4.771875e-05,
+      "loss": 0.0001,
+      "step": 19564
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 4.771828358208956e-05,
+      "loss": 0.0018,
+      "step": 19568
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 4.771781716417911e-05,
+      "loss": 0.0018,
+      "step": 19572
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 4.7717350746268656e-05,
+      "loss": 0.001,
+      "step": 19576
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 4.771688432835821e-05,
+      "loss": 0.0007,
+      "step": 19580
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 4.7716417910447766e-05,
+      "loss": 0.0019,
+      "step": 19584
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 4.7715951492537314e-05,
+      "loss": 0.0018,
+      "step": 19588
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 4.771548507462687e-05,
+      "loss": 0.0076,
+      "step": 19592
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 4.7715018656716423e-05,
+      "loss": 0.0002,
+      "step": 19596
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 4.771455223880597e-05,
+      "loss": 0.0102,
+      "step": 19600
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 4.7714085820895527e-05,
+      "loss": 0.0018,
+      "step": 19604
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 4.7713619402985075e-05,
+      "loss": 0.0008,
+      "step": 19608
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 4.771315298507463e-05,
+      "loss": 0.0012,
+      "step": 19612
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 4.7712686567164184e-05,
+      "loss": 0.002,
+      "step": 19616
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 4.771222014925373e-05,
+      "loss": 0.0046,
+      "step": 19620
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 4.771175373134328e-05,
+      "loss": 0.0061,
+      "step": 19624
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 4.771128731343284e-05,
+      "loss": 0.001,
+      "step": 19628
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 4.771082089552239e-05,
+      "loss": 0.002,
+      "step": 19632
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 4.771035447761194e-05,
+      "loss": 0.0012,
+      "step": 19636
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 4.7709888059701494e-05,
+      "loss": 0.0,
+      "step": 19640
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 4.770942164179105e-05,
+      "loss": 0.0019,
+      "step": 19644
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 4.77089552238806e-05,
+      "loss": 0.0005,
+      "step": 19648
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 4.770848880597015e-05,
+      "loss": 0.0058,
+      "step": 19652
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 4.7708022388059706e-05,
+      "loss": 0.0007,
+      "step": 19656
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 4.7707555970149255e-05,
+      "loss": 0.001,
+      "step": 19660
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 4.770708955223881e-05,
+      "loss": 0.0023,
+      "step": 19664
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 4.770662313432836e-05,
+      "loss": 0.0056,
+      "step": 19668
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 4.770615671641791e-05,
+      "loss": 0.0024,
+      "step": 19672
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 4.770569029850747e-05,
+      "loss": 0.0051,
+      "step": 19676
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 4.7705223880597016e-05,
+      "loss": 0.0016,
+      "step": 19680
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 4.7704757462686564e-05,
+      "loss": 0.0104,
+      "step": 19684
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 4.7704291044776125e-05,
+      "loss": 0.0003,
+      "step": 19688
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 4.7703824626865673e-05,
+      "loss": 0.0033,
+      "step": 19692
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 4.770335820895523e-05,
+      "loss": 0.0005,
+      "step": 19696
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 4.7702891791044777e-05,
+      "loss": 0.0015,
+      "step": 19700
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 4.770242537313433e-05,
+      "loss": 0.0002,
+      "step": 19704
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 4.7701958955223886e-05,
+      "loss": 0.0003,
+      "step": 19708
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 4.7701492537313434e-05,
+      "loss": 0.0054,
+      "step": 19712
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 4.770102611940299e-05,
+      "loss": 0.0025,
+      "step": 19716
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 4.7700559701492544e-05,
+      "loss": 0.0014,
+      "step": 19720
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 4.770009328358209e-05,
+      "loss": 0.0006,
+      "step": 19724
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 4.769962686567164e-05,
+      "loss": 0.0008,
+      "step": 19728
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 4.7699160447761195e-05,
+      "loss": 0.0006,
+      "step": 19732
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 4.769869402985075e-05,
+      "loss": 0.0002,
+      "step": 19736
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 4.76982276119403e-05,
+      "loss": 0.0044,
+      "step": 19740
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 4.7697761194029853e-05,
+      "loss": 0.001,
+      "step": 19744
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 4.769729477611941e-05,
+      "loss": 0.0011,
+      "step": 19748
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 4.7696828358208956e-05,
+      "loss": 0.0009,
+      "step": 19752
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 4.769636194029851e-05,
+      "loss": 0.0032,
+      "step": 19756
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 4.769589552238806e-05,
+      "loss": 0.001,
+      "step": 19760
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 4.7695429104477614e-05,
+      "loss": 0.0008,
+      "step": 19764
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 4.769496268656717e-05,
+      "loss": 0.0022,
+      "step": 19768
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 4.769449626865672e-05,
+      "loss": 0.0001,
+      "step": 19772
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 4.769402985074627e-05,
+      "loss": 0.0003,
+      "step": 19776
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 4.769356343283583e-05,
+      "loss": 0.0043,
+      "step": 19780
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 4.7693097014925375e-05,
+      "loss": 0.0003,
+      "step": 19784
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 4.7692630597014924e-05,
+      "loss": 0.0007,
+      "step": 19788
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 4.769216417910448e-05,
+      "loss": 0.0013,
+      "step": 19792
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 4.769169776119403e-05,
+      "loss": 0.0001,
+      "step": 19796
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 4.769123134328358e-05,
+      "loss": 0.0005,
+      "step": 19800
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 4.7690764925373136e-05,
+      "loss": 0.0015,
+      "step": 19804
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 4.769029850746269e-05,
+      "loss": 0.0018,
+      "step": 19808
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 4.768983208955224e-05,
+      "loss": 0.0104,
+      "step": 19812
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 4.7689365671641794e-05,
+      "loss": 0.0005,
+      "step": 19816
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 4.768889925373134e-05,
+      "loss": 0.0006,
+      "step": 19820
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 4.76884328358209e-05,
+      "loss": 0.0001,
+      "step": 19824
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 4.768796641791045e-05,
+      "loss": 0.0131,
+      "step": 19828
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 4.76875e-05,
+      "loss": 0.0006,
+      "step": 19832
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 4.7687033582089555e-05,
+      "loss": 0.0004,
+      "step": 19836
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 4.768656716417911e-05,
+      "loss": 0.0004,
+      "step": 19840
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 4.768610074626866e-05,
+      "loss": 0.0006,
+      "step": 19844
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 4.7685634328358206e-05,
+      "loss": 0.0004,
+      "step": 19848
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 4.768516791044776e-05,
+      "loss": 0.003,
+      "step": 19852
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 4.7684701492537316e-05,
+      "loss": 0.0117,
+      "step": 19856
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 4.768423507462687e-05,
+      "loss": 0.0017,
+      "step": 19860
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 4.768376865671642e-05,
+      "loss": 0.003,
+      "step": 19864
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 4.7683302238805974e-05,
+      "loss": 0.0009,
+      "step": 19868
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 4.768283582089553e-05,
+      "loss": 0.0057,
+      "step": 19872
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 4.768236940298508e-05,
+      "loss": 0.0018,
+      "step": 19876
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 4.7681902985074625e-05,
+      "loss": 0.0009,
+      "step": 19880
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 4.768143656716419e-05,
+      "loss": 0.0014,
+      "step": 19884
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 4.7680970149253735e-05,
+      "loss": 0.0004,
+      "step": 19888
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 4.768050373134328e-05,
+      "loss": 0.0036,
+      "step": 19892
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 4.768003731343284e-05,
+      "loss": 0.0022,
+      "step": 19896
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 4.767957089552239e-05,
+      "loss": 0.0007,
+      "step": 19900
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 4.767910447761194e-05,
+      "loss": 0.0018,
+      "step": 19904
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 4.7678638059701496e-05,
+      "loss": 0.0004,
+      "step": 19908
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 4.7678171641791044e-05,
+      "loss": 0.0002,
+      "step": 19912
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 4.76777052238806e-05,
+      "loss": 0.0021,
+      "step": 19916
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 4.7677238805970154e-05,
+      "loss": 0.0014,
+      "step": 19920
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 4.76767723880597e-05,
+      "loss": 0.014,
+      "step": 19924
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 4.767630597014926e-05,
+      "loss": 0.0028,
+      "step": 19928
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 4.767583955223881e-05,
+      "loss": 0.0004,
+      "step": 19932
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 4.767537313432836e-05,
+      "loss": 0.0009,
+      "step": 19936
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 4.767490671641791e-05,
+      "loss": 0.0011,
+      "step": 19940
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 4.767444029850747e-05,
+      "loss": 0.0016,
+      "step": 19944
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 4.767397388059702e-05,
+      "loss": 0.0238,
+      "step": 19948
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 4.7673507462686566e-05,
+      "loss": 0.0036,
+      "step": 19952
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 4.767304104477612e-05,
+      "loss": 0.0089,
+      "step": 19956
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 4.7672574626865676e-05,
+      "loss": 0.0041,
+      "step": 19960
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 4.7672108208955224e-05,
+      "loss": 0.0005,
+      "step": 19964
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 4.767164179104478e-05,
+      "loss": 0.0002,
+      "step": 19968
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 4.767117537313433e-05,
+      "loss": 0.0013,
+      "step": 19972
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 4.767070895522388e-05,
+      "loss": 0.0007,
+      "step": 19976
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 4.767024253731344e-05,
+      "loss": 0.0004,
+      "step": 19980
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 4.7669776119402985e-05,
+      "loss": 0.001,
+      "step": 19984
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 4.766930970149254e-05,
+      "loss": 0.0006,
+      "step": 19988
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 4.7668843283582095e-05,
+      "loss": 0.0007,
+      "step": 19992
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 4.766837686567164e-05,
+      "loss": 0.0006,
+      "step": 19996
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 4.766791044776119e-05,
+      "loss": 0.0025,
+      "step": 20000
+    },
+    {
+      "epoch": 18.66,
+      "eval_exact_match": 0.7263056092843327,
+      "eval_exec": 0.7543520309477756,
+      "eval_loss": 0.3301263451576233,
+      "eval_runtime": 1520.1707,
+      "eval_samples_per_second": 0.68,
+      "step": 20000
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 4.766744402985075e-05,
+      "loss": 0.0024,
+      "step": 20004
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 4.76669776119403e-05,
+      "loss": 0.0036,
+      "step": 20008
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 4.766651119402985e-05,
+      "loss": 0.0009,
+      "step": 20012
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 4.7666044776119404e-05,
+      "loss": 0.001,
+      "step": 20016
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 4.766557835820896e-05,
+      "loss": 0.0146,
+      "step": 20020
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 4.7665111940298514e-05,
+      "loss": 0.0006,
+      "step": 20024
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 4.766464552238806e-05,
+      "loss": 0.0068,
+      "step": 20028
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 4.766417910447761e-05,
+      "loss": 0.0021,
+      "step": 20032
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 4.766371268656717e-05,
+      "loss": 0.0053,
+      "step": 20036
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 4.766324626865672e-05,
+      "loss": 0.0006,
+      "step": 20040
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 4.766277985074627e-05,
+      "loss": 0.0008,
+      "step": 20044
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 4.766231343283582e-05,
+      "loss": 0.0032,
+      "step": 20048
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 4.766184701492538e-05,
+      "loss": 0.0026,
+      "step": 20052
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 4.7661380597014926e-05,
+      "loss": 0.0032,
+      "step": 20056
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 4.766091417910448e-05,
+      "loss": 0.0131,
+      "step": 20060
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 4.766044776119403e-05,
+      "loss": 0.0026,
+      "step": 20064
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 4.7659981343283584e-05,
+      "loss": 0.0009,
+      "step": 20068
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 4.765951492537314e-05,
+      "loss": 0.0003,
+      "step": 20072
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 4.765904850746269e-05,
+      "loss": 0.0007,
+      "step": 20076
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 4.765858208955224e-05,
+      "loss": 0.0048,
+      "step": 20080
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 4.76581156716418e-05,
+      "loss": 0.0032,
+      "step": 20084
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 4.7657649253731345e-05,
+      "loss": 0.0023,
+      "step": 20088
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 4.765718283582089e-05,
+      "loss": 0.0005,
+      "step": 20092
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 4.7656716417910455e-05,
+      "loss": 0.0006,
+      "step": 20096
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 4.765625e-05,
+      "loss": 0.0026,
+      "step": 20100
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 4.765578358208955e-05,
+      "loss": 0.0002,
+      "step": 20104
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 4.7655317164179106e-05,
+      "loss": 0.0031,
+      "step": 20108
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 4.765485074626866e-05,
+      "loss": 0.0003,
+      "step": 20112
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 4.765438432835821e-05,
+      "loss": 0.0017,
+      "step": 20116
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 4.7653917910447764e-05,
+      "loss": 0.0003,
+      "step": 20120
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 4.765345149253731e-05,
+      "loss": 0.0011,
+      "step": 20124
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 4.765298507462687e-05,
+      "loss": 0.0013,
+      "step": 20128
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 4.765251865671642e-05,
+      "loss": 0.0071,
+      "step": 20132
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 4.765205223880597e-05,
+      "loss": 0.0006,
+      "step": 20136
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 4.7651585820895525e-05,
+      "loss": 0.0009,
+      "step": 20140
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 4.765111940298508e-05,
+      "loss": 0.0018,
+      "step": 20144
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 4.765065298507463e-05,
+      "loss": 0.0144,
+      "step": 20148
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 4.7650186567164176e-05,
+      "loss": 0.0011,
+      "step": 20152
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 4.764972014925374e-05,
+      "loss": 0.0033,
+      "step": 20156
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 4.7649253731343286e-05,
+      "loss": 0.0064,
+      "step": 20160
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 4.7648787313432834e-05,
+      "loss": 0.0016,
+      "step": 20164
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 4.764832089552239e-05,
+      "loss": 0.0014,
+      "step": 20168
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 4.7647854477611944e-05,
+      "loss": 0.0002,
+      "step": 20172
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 4.764738805970149e-05,
+      "loss": 0.0019,
+      "step": 20176
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 4.764692164179105e-05,
+      "loss": 0.0024,
+      "step": 20180
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 4.7646455223880595e-05,
+      "loss": 0.005,
+      "step": 20184
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 4.764598880597016e-05,
+      "loss": 0.0004,
+      "step": 20188
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 4.7645522388059705e-05,
+      "loss": 0.0003,
+      "step": 20192
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 4.764505597014925e-05,
+      "loss": 0.0008,
+      "step": 20196
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 4.764458955223881e-05,
+      "loss": 0.0009,
+      "step": 20200
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 4.764412313432836e-05,
+      "loss": 0.0004,
+      "step": 20204
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 4.764365671641791e-05,
+      "loss": 0.0062,
+      "step": 20208
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 4.7643190298507466e-05,
+      "loss": 0.0042,
+      "step": 20212
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 4.764272388059702e-05,
+      "loss": 0.0016,
+      "step": 20216
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 4.764225746268657e-05,
+      "loss": 0.0016,
+      "step": 20220
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 4.7641791044776124e-05,
+      "loss": 0.0009,
+      "step": 20224
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 4.764132462686567e-05,
+      "loss": 0.0018,
+      "step": 20228
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 4.764085820895523e-05,
+      "loss": 0.0009,
+      "step": 20232
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 4.764039179104478e-05,
+      "loss": 0.0021,
+      "step": 20236
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 4.763992537313433e-05,
+      "loss": 0.0036,
+      "step": 20240
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 4.763945895522388e-05,
+      "loss": 0.0006,
+      "step": 20244
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 4.763899253731344e-05,
+      "loss": 0.0013,
+      "step": 20248
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 4.763852611940299e-05,
+      "loss": 0.0019,
+      "step": 20252
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 4.7638059701492536e-05,
+      "loss": 0.0003,
+      "step": 20256
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 4.763759328358209e-05,
+      "loss": 0.0237,
+      "step": 20260
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 4.7637126865671646e-05,
+      "loss": 0.0001,
+      "step": 20264
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 4.7636660447761194e-05,
+      "loss": 0.0005,
+      "step": 20268
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 4.763619402985075e-05,
+      "loss": 0.0016,
+      "step": 20272
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 4.7635727611940304e-05,
+      "loss": 0.0026,
+      "step": 20276
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 4.763526119402985e-05,
+      "loss": 0.0013,
+      "step": 20280
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 4.763479477611941e-05,
+      "loss": 0.0007,
+      "step": 20284
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 4.7634328358208955e-05,
+      "loss": 0.0037,
+      "step": 20288
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 4.763386194029851e-05,
+      "loss": 0.0016,
+      "step": 20292
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 4.7633395522388065e-05,
+      "loss": 0.001,
+      "step": 20296
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 4.763292910447761e-05,
+      "loss": 0.0004,
+      "step": 20300
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 4.763246268656716e-05,
+      "loss": 0.0015,
+      "step": 20304
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 4.763199626865672e-05,
+      "loss": 0.0057,
+      "step": 20308
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 4.763152985074627e-05,
+      "loss": 0.0001,
+      "step": 20312
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 4.763106343283582e-05,
+      "loss": 0.0113,
+      "step": 20316
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 4.7630597014925374e-05,
+      "loss": 0.0017,
+      "step": 20320
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 4.763013059701493e-05,
+      "loss": 0.0024,
+      "step": 20324
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 4.762966417910448e-05,
+      "loss": 0.0005,
+      "step": 20328
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 4.762919776119403e-05,
+      "loss": 0.0051,
+      "step": 20332
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 4.762873134328359e-05,
+      "loss": 0.0017,
+      "step": 20336
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 4.7628264925373135e-05,
+      "loss": 0.0002,
+      "step": 20340
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 4.762779850746269e-05,
+      "loss": 0.0024,
+      "step": 20344
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 4.762733208955224e-05,
+      "loss": 0.007,
+      "step": 20348
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 4.762686567164179e-05,
+      "loss": 0.0002,
+      "step": 20352
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 4.762639925373135e-05,
+      "loss": 0.0005,
+      "step": 20356
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 4.7625932835820896e-05,
+      "loss": 0.0012,
+      "step": 20360
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 4.762546641791045e-05,
+      "loss": 0.0012,
+      "step": 20364
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 4.7625000000000006e-05,
+      "loss": 0.0002,
+      "step": 20368
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 4.7624533582089554e-05,
+      "loss": 0.0021,
+      "step": 20372
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 4.762406716417911e-05,
+      "loss": 0.0027,
+      "step": 20376
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 4.762360074626866e-05,
+      "loss": 0.0007,
+      "step": 20380
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 4.762313432835821e-05,
+      "loss": 0.004,
+      "step": 20384
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 4.762266791044777e-05,
+      "loss": 0.0004,
+      "step": 20388
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 4.7622201492537315e-05,
+      "loss": 0.0001,
+      "step": 20392
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 4.762173507462687e-05,
+      "loss": 0.0009,
+      "step": 20396
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 4.7621268656716425e-05,
+      "loss": 0.0089,
+      "step": 20400
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 4.762080223880597e-05,
+      "loss": 0.0002,
+      "step": 20404
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 4.762033582089552e-05,
+      "loss": 0.0019,
+      "step": 20408
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 4.7619869402985076e-05,
+      "loss": 0.0003,
+      "step": 20412
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 4.761940298507463e-05,
+      "loss": 0.0013,
+      "step": 20416
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 4.761893656716418e-05,
+      "loss": 0.0034,
+      "step": 20420
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 4.7618470149253734e-05,
+      "loss": 0.0009,
+      "step": 20424
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 4.761800373134329e-05,
+      "loss": 0.0014,
+      "step": 20428
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 4.761753731343284e-05,
+      "loss": 0.0013,
+      "step": 20432
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 4.761707089552239e-05,
+      "loss": 0.0007,
+      "step": 20436
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 4.761660447761194e-05,
+      "loss": 0.0006,
+      "step": 20440
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 4.7616138059701495e-05,
+      "loss": 0.0049,
+      "step": 20444
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 4.761567164179105e-05,
+      "loss": 0.0066,
+      "step": 20448
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 4.76152052238806e-05,
+      "loss": 0.0008,
+      "step": 20452
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 4.761473880597015e-05,
+      "loss": 0.0011,
+      "step": 20456
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 4.761427238805971e-05,
+      "loss": 0.0002,
+      "step": 20460
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 4.7613805970149256e-05,
+      "loss": 0.0002,
+      "step": 20464
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 4.7613339552238804e-05,
+      "loss": 0.0,
+      "step": 20468
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 4.761287313432836e-05,
+      "loss": 0.0014,
+      "step": 20472
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 4.7612406716417914e-05,
+      "loss": 0.0011,
+      "step": 20476
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 4.761194029850746e-05,
+      "loss": 0.0005,
+      "step": 20480
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 4.761147388059702e-05,
+      "loss": 0.0002,
+      "step": 20484
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 4.761100746268657e-05,
+      "loss": 0.0027,
+      "step": 20488
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 4.761054104477612e-05,
+      "loss": 0.0006,
+      "step": 20492
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 4.7610074626865675e-05,
+      "loss": 0.0019,
+      "step": 20496
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 4.760960820895522e-05,
+      "loss": 0.0001,
+      "step": 20500
+    },
+    {
+      "epoch": 19.12,
+      "eval_exact_match": 0.7340425531914894,
+      "eval_exec": 0.7601547388781431,
+      "eval_loss": 0.35038140416145325,
+      "eval_runtime": 1096.2977,
+      "eval_samples_per_second": 0.943,
+      "step": 20500
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 4.760914179104478e-05,
+      "loss": 0.001,
+      "step": 20504
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 4.760867537313433e-05,
+      "loss": 0.0015,
+      "step": 20508
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 4.760820895522388e-05,
+      "loss": 0.0236,
+      "step": 20512
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 4.7607742537313436e-05,
+      "loss": 0.0016,
+      "step": 20516
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 4.760727611940299e-05,
+      "loss": 0.0006,
+      "step": 20520
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 4.760680970149254e-05,
+      "loss": 0.0004,
+      "step": 20524
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 4.7606343283582094e-05,
+      "loss": 0.0001,
+      "step": 20528
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 4.760587686567164e-05,
+      "loss": 0.0006,
+      "step": 20532
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 4.7605410447761197e-05,
+      "loss": 0.0021,
+      "step": 20536
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 4.760494402985075e-05,
+      "loss": 0.0039,
+      "step": 20540
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 4.76044776119403e-05,
+      "loss": 0.0032,
+      "step": 20544
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 4.7604011194029855e-05,
+      "loss": 0.0026,
+      "step": 20548
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 4.760354477611941e-05,
+      "loss": 0.0011,
+      "step": 20552
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 4.760307835820896e-05,
+      "loss": 0.0015,
+      "step": 20556
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 4.7602611940298506e-05,
+      "loss": 0.0033,
+      "step": 20560
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 4.760214552238807e-05,
+      "loss": 0.0002,
+      "step": 20564
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 4.7601679104477616e-05,
+      "loss": 0.002,
+      "step": 20568
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 4.7601212686567164e-05,
+      "loss": 0.009,
+      "step": 20572
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 4.760074626865672e-05,
+      "loss": 0.0006,
+      "step": 20576
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 4.7600279850746273e-05,
+      "loss": 0.0002,
+      "step": 20580
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 4.759981343283582e-05,
+      "loss": 0.0012,
+      "step": 20584
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 4.7599347014925376e-05,
+      "loss": 0.0019,
+      "step": 20588
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 4.7598880597014925e-05,
+      "loss": 0.0005,
+      "step": 20592
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 4.759841417910448e-05,
+      "loss": 0.0133,
+      "step": 20596
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 4.7597947761194034e-05,
+      "loss": 0.0001,
+      "step": 20600
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 4.759748134328358e-05,
+      "loss": 0.0013,
+      "step": 20604
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 4.759701492537314e-05,
+      "loss": 0.0002,
+      "step": 20608
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 4.759654850746269e-05,
+      "loss": 0.0044,
+      "step": 20612
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 4.759608208955224e-05,
+      "loss": 0.0002,
+      "step": 20616
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 4.759561567164179e-05,
+      "loss": 0.0007,
+      "step": 20620
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 4.759514925373135e-05,
+      "loss": 0.0018,
+      "step": 20624
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 4.75946828358209e-05,
+      "loss": 0.0061,
+      "step": 20628
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 4.7594216417910447e-05,
+      "loss": 0.0014,
+      "step": 20632
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 4.759375e-05,
+      "loss": 0.0011,
+      "step": 20636
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 4.7593283582089556e-05,
+      "loss": 0.0002,
+      "step": 20640
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 4.7592817164179105e-05,
+      "loss": 0.0028,
+      "step": 20644
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 4.759235074626866e-05,
+      "loss": 0.0021,
+      "step": 20648
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 4.759188432835821e-05,
+      "loss": 0.0002,
+      "step": 20652
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 4.759141791044776e-05,
+      "loss": 0.0008,
+      "step": 20656
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 4.759095149253732e-05,
+      "loss": 0.0021,
+      "step": 20660
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 4.7590485074626866e-05,
+      "loss": 0.0035,
+      "step": 20664
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 4.759001865671642e-05,
+      "loss": 0.0002,
+      "step": 20668
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 4.7589552238805975e-05,
+      "loss": 0.0004,
+      "step": 20672
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 4.7589085820895523e-05,
+      "loss": 0.0007,
+      "step": 20676
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 4.758861940298508e-05,
+      "loss": 0.0005,
+      "step": 20680
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 4.758815298507463e-05,
+      "loss": 0.0002,
+      "step": 20684
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 4.758768656716418e-05,
+      "loss": 0.0005,
+      "step": 20688
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 4.7587220149253736e-05,
+      "loss": 0.0001,
+      "step": 20692
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 4.7586753731343284e-05,
+      "loss": 0.0017,
+      "step": 20696
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 4.758628731343284e-05,
+      "loss": 0.0006,
+      "step": 20700
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 4.7585820895522394e-05,
+      "loss": 0.0003,
+      "step": 20704
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 4.758535447761194e-05,
+      "loss": 0.0007,
+      "step": 20708
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 4.758488805970149e-05,
+      "loss": 0.0018,
+      "step": 20712
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 4.758442164179105e-05,
+      "loss": 0.0003,
+      "step": 20716
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 4.75839552238806e-05,
+      "loss": 0.0039,
+      "step": 20720
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 4.758348880597015e-05,
+      "loss": 0.002,
+      "step": 20724
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 4.7583022388059703e-05,
+      "loss": 0.0006,
+      "step": 20728
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 4.758255597014926e-05,
+      "loss": 0.0009,
+      "step": 20732
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 4.7582089552238806e-05,
+      "loss": 0.0011,
+      "step": 20736
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 4.758162313432836e-05,
+      "loss": 0.0001,
+      "step": 20740
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 4.758115671641791e-05,
+      "loss": 0.0014,
+      "step": 20744
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 4.7580690298507464e-05,
+      "loss": 0.0003,
+      "step": 20748
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 4.758022388059702e-05,
+      "loss": 0.0003,
+      "step": 20752
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 4.757975746268657e-05,
+      "loss": 0.0028,
+      "step": 20756
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 4.757929104477612e-05,
+      "loss": 0.0005,
+      "step": 20760
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 4.757882462686568e-05,
+      "loss": 0.0004,
+      "step": 20764
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 4.7578358208955225e-05,
+      "loss": 0.0011,
+      "step": 20768
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 4.7577891791044773e-05,
+      "loss": 0.0003,
+      "step": 20772
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 4.7577425373134335e-05,
+      "loss": 0.0002,
+      "step": 20776
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 4.757695895522388e-05,
+      "loss": 0.0001,
+      "step": 20780
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 4.757649253731343e-05,
+      "loss": 0.0002,
+      "step": 20784
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 4.7576026119402986e-05,
+      "loss": 0.003,
+      "step": 20788
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 4.757555970149254e-05,
+      "loss": 0.0002,
+      "step": 20792
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 4.757509328358209e-05,
+      "loss": 0.0017,
+      "step": 20796
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 4.7574626865671644e-05,
+      "loss": 0.0032,
+      "step": 20800
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 4.757416044776119e-05,
+      "loss": 0.0023,
+      "step": 20804
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 4.757369402985075e-05,
+      "loss": 0.0003,
+      "step": 20808
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 4.75732276119403e-05,
+      "loss": 0.0037,
+      "step": 20812
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 4.757276119402985e-05,
+      "loss": 0.0008,
+      "step": 20816
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 4.7572294776119405e-05,
+      "loss": 0.0021,
+      "step": 20820
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 4.757182835820896e-05,
+      "loss": 0.0005,
+      "step": 20824
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 4.757136194029851e-05,
+      "loss": 0.0001,
+      "step": 20828
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 4.7570895522388056e-05,
+      "loss": 0.0001,
+      "step": 20832
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 4.757042910447762e-05,
+      "loss": 0.0005,
+      "step": 20836
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 4.7569962686567166e-05,
+      "loss": 0.002,
+      "step": 20840
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 4.756949626865672e-05,
+      "loss": 0.0021,
+      "step": 20844
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 4.756902985074627e-05,
+      "loss": 0.0001,
+      "step": 20848
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 4.7568563432835824e-05,
+      "loss": 0.0032,
+      "step": 20852
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 4.756809701492538e-05,
+      "loss": 0.0003,
+      "step": 20856
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 4.756763059701493e-05,
+      "loss": 0.0038,
+      "step": 20860
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 4.7567164179104475e-05,
+      "loss": 0.0087,
+      "step": 20864
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 4.756669776119404e-05,
+      "loss": 0.0025,
+      "step": 20868
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 4.7566231343283585e-05,
+      "loss": 0.0016,
+      "step": 20872
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 4.756576492537313e-05,
+      "loss": 0.0013,
+      "step": 20876
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 4.756529850746269e-05,
+      "loss": 0.0016,
+      "step": 20880
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 4.756483208955224e-05,
+      "loss": 0.0001,
+      "step": 20884
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 4.756436567164179e-05,
+      "loss": 0.0016,
+      "step": 20888
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 4.7563899253731346e-05,
+      "loss": 0.0056,
+      "step": 20892
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 4.75634328358209e-05,
+      "loss": 0.0002,
+      "step": 20896
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 4.756296641791045e-05,
+      "loss": 0.0001,
+      "step": 20900
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 4.7562500000000004e-05,
+      "loss": 0.0047,
+      "step": 20904
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 4.756203358208955e-05,
+      "loss": 0.0059,
+      "step": 20908
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 4.756156716417911e-05,
+      "loss": 0.0003,
+      "step": 20912
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 4.756110074626866e-05,
+      "loss": 0.0013,
+      "step": 20916
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 4.756063432835821e-05,
+      "loss": 0.0007,
+      "step": 20920
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 4.756016791044776e-05,
+      "loss": 0.0058,
+      "step": 20924
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 4.755970149253732e-05,
+      "loss": 0.0018,
+      "step": 20928
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 4.755923507462687e-05,
+      "loss": 0.0001,
+      "step": 20932
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 4.7558768656716416e-05,
+      "loss": 0.0082,
+      "step": 20936
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 4.755830223880597e-05,
+      "loss": 0.0066,
+      "step": 20940
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 4.7557835820895526e-05,
+      "loss": 0.0064,
+      "step": 20944
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 4.7557369402985074e-05,
+      "loss": 0.0015,
+      "step": 20948
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 4.755690298507463e-05,
+      "loss": 0.0042,
+      "step": 20952
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 4.7556436567164184e-05,
+      "loss": 0.0006,
+      "step": 20956
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 4.755597014925373e-05,
+      "loss": 0.0001,
+      "step": 20960
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 4.755550373134329e-05,
+      "loss": 0.0042,
+      "step": 20964
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 4.7555037313432835e-05,
+      "loss": 0.0001,
+      "step": 20968
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 4.755457089552239e-05,
+      "loss": 0.0044,
+      "step": 20972
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 4.7554104477611945e-05,
+      "loss": 0.0087,
+      "step": 20976
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 4.755363805970149e-05,
+      "loss": 0.0023,
+      "step": 20980
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 4.755317164179104e-05,
+      "loss": 0.0004,
+      "step": 20984
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 4.75527052238806e-05,
+      "loss": 0.0004,
+      "step": 20988
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 4.755223880597015e-05,
+      "loss": 0.0007,
+      "step": 20992
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 4.75517723880597e-05,
+      "loss": 0.0063,
+      "step": 20996
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 4.7551305970149254e-05,
+      "loss": 0.0003,
+      "step": 21000
+    },
+    {
+      "epoch": 19.59,
+      "eval_exact_match": 0.7253384912959381,
+      "eval_exec": 0.7553191489361702,
+      "eval_loss": 0.33756983280181885,
+      "eval_runtime": 1072.0397,
+      "eval_samples_per_second": 0.965,
+      "step": 21000
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 4.755083955223881e-05,
+      "loss": 0.0037,
+      "step": 21004
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 4.7550373134328364e-05,
+      "loss": 0.0005,
+      "step": 21008
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 4.754990671641791e-05,
+      "loss": 0.0002,
+      "step": 21012
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 4.754944029850747e-05,
+      "loss": 0.0008,
+      "step": 21016
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 4.754897388059702e-05,
+      "loss": 0.0061,
+      "step": 21020
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 4.754850746268657e-05,
+      "loss": 0.0014,
+      "step": 21024
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 4.754804104477612e-05,
+      "loss": 0.0056,
+      "step": 21028
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 4.754757462686567e-05,
+      "loss": 0.0009,
+      "step": 21032
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 4.754710820895523e-05,
+      "loss": 0.0099,
+      "step": 21036
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 4.7546641791044776e-05,
+      "loss": 0.0005,
+      "step": 21040
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 4.754617537313433e-05,
+      "loss": 0.0065,
+      "step": 21044
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 4.7545708955223886e-05,
+      "loss": 0.0012,
+      "step": 21048
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 4.7545242537313434e-05,
+      "loss": 0.0002,
+      "step": 21052
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 4.754477611940299e-05,
+      "loss": 0.0021,
+      "step": 21056
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 4.754430970149254e-05,
+      "loss": 0.0005,
+      "step": 21060
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 4.754384328358209e-05,
+      "loss": 0.0007,
+      "step": 21064
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 4.754337686567165e-05,
+      "loss": 0.0002,
+      "step": 21068
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 4.7542910447761195e-05,
+      "loss": 0.0032,
+      "step": 21072
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 4.754244402985075e-05,
+      "loss": 0.0073,
+      "step": 21076
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 4.7541977611940305e-05,
+      "loss": 0.0034,
+      "step": 21080
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 4.754151119402985e-05,
+      "loss": 0.0011,
+      "step": 21084
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 4.75410447761194e-05,
+      "loss": 0.0085,
+      "step": 21088
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 4.7540578358208956e-05,
+      "loss": 0.0041,
+      "step": 21092
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 4.754011194029851e-05,
+      "loss": 0.0002,
+      "step": 21096
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 4.753964552238806e-05,
+      "loss": 0.0033,
+      "step": 21100
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 4.7539179104477614e-05,
+      "loss": 0.0009,
+      "step": 21104
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 4.753871268656717e-05,
+      "loss": 0.0004,
+      "step": 21108
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 4.753824626865672e-05,
+      "loss": 0.0017,
+      "step": 21112
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 4.753777985074627e-05,
+      "loss": 0.0024,
+      "step": 21116
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 4.753731343283582e-05,
+      "loss": 0.0003,
+      "step": 21120
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 4.7536847014925375e-05,
+      "loss": 0.0007,
+      "step": 21124
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 4.753638059701493e-05,
+      "loss": 0.0053,
+      "step": 21128
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 4.753591417910448e-05,
+      "loss": 0.0009,
+      "step": 21132
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 4.753544776119403e-05,
+      "loss": 0.0008,
+      "step": 21136
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 4.753498134328359e-05,
+      "loss": 0.0035,
+      "step": 21140
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 4.7534514925373136e-05,
+      "loss": 0.0008,
+      "step": 21144
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 4.7534048507462684e-05,
+      "loss": 0.0029,
+      "step": 21148
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 4.753358208955224e-05,
+      "loss": 0.0089,
+      "step": 21152
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 4.7533115671641794e-05,
+      "loss": 0.0034,
+      "step": 21156
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 4.753264925373134e-05,
+      "loss": 0.0016,
+      "step": 21160
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 4.75321828358209e-05,
+      "loss": 0.0006,
+      "step": 21164
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 4.753171641791045e-05,
+      "loss": 0.0003,
+      "step": 21168
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 4.753125000000001e-05,
+      "loss": 0.0036,
+      "step": 21172
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 4.7530783582089555e-05,
+      "loss": 0.0001,
+      "step": 21176
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 4.75303171641791e-05,
+      "loss": 0.0003,
+      "step": 21180
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 4.7529850746268665e-05,
+      "loss": 0.001,
+      "step": 21184
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 4.752938432835821e-05,
+      "loss": 0.0008,
+      "step": 21188
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 4.752891791044776e-05,
+      "loss": 0.0002,
+      "step": 21192
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 4.7528451492537316e-05,
+      "loss": 0.0001,
+      "step": 21196
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 4.752798507462687e-05,
+      "loss": 0.0001,
+      "step": 21200
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 4.752751865671642e-05,
+      "loss": 0.0011,
+      "step": 21204
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 4.7527052238805974e-05,
+      "loss": 0.0002,
+      "step": 21208
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 4.752658582089552e-05,
+      "loss": 0.0024,
+      "step": 21212
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 4.752611940298508e-05,
+      "loss": 0.0023,
+      "step": 21216
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 4.752565298507463e-05,
+      "loss": 0.0044,
+      "step": 21220
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 4.752518656716418e-05,
+      "loss": 0.001,
+      "step": 21224
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 4.7524720149253735e-05,
+      "loss": 0.001,
+      "step": 21228
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 4.752425373134329e-05,
+      "loss": 0.0075,
+      "step": 21232
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 4.752378731343284e-05,
+      "loss": 0.0005,
+      "step": 21236
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 4.7523320895522386e-05,
+      "loss": 0.0036,
+      "step": 21240
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 4.752285447761195e-05,
+      "loss": 0.0015,
+      "step": 21244
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 4.7522388059701496e-05,
+      "loss": 0.0062,
+      "step": 21248
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 4.7521921641791044e-05,
+      "loss": 0.0002,
+      "step": 21252
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 4.75214552238806e-05,
+      "loss": 0.0007,
+      "step": 21256
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 4.7520988805970154e-05,
+      "loss": 0.0002,
+      "step": 21260
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 4.75205223880597e-05,
+      "loss": 0.0016,
+      "step": 21264
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 4.752005597014926e-05,
+      "loss": 0.0025,
+      "step": 21268
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 4.7519589552238805e-05,
+      "loss": 0.0002,
+      "step": 21272
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 4.751912313432836e-05,
+      "loss": 0.0004,
+      "step": 21276
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 4.7518656716417915e-05,
+      "loss": 0.0004,
+      "step": 21280
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 4.751819029850746e-05,
+      "loss": 0.0002,
+      "step": 21284
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 4.751772388059702e-05,
+      "loss": 0.0125,
+      "step": 21288
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 4.751725746268657e-05,
+      "loss": 0.0006,
+      "step": 21292
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 4.751679104477612e-05,
+      "loss": 0.0034,
+      "step": 21296
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 4.751632462686567e-05,
+      "loss": 0.011,
+      "step": 21300
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 4.751585820895523e-05,
+      "loss": 0.0026,
+      "step": 21304
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 4.751539179104478e-05,
+      "loss": 0.0015,
+      "step": 21308
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 4.751492537313433e-05,
+      "loss": 0.0013,
+      "step": 21312
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 4.751445895522388e-05,
+      "loss": 0.0002,
+      "step": 21316
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 4.751399253731344e-05,
+      "loss": 0.0001,
+      "step": 21320
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 4.7513526119402985e-05,
+      "loss": 0.0036,
+      "step": 21324
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 4.751305970149254e-05,
+      "loss": 0.0013,
+      "step": 21328
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 4.751259328358209e-05,
+      "loss": 0.0009,
+      "step": 21332
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 4.751212686567165e-05,
+      "loss": 0.0024,
+      "step": 21336
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 4.75116604477612e-05,
+      "loss": 0.008,
+      "step": 21340
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 4.7511194029850746e-05,
+      "loss": 0.0018,
+      "step": 21344
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 4.75107276119403e-05,
+      "loss": 0.0031,
+      "step": 21348
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 4.7510261194029856e-05,
+      "loss": 0.0002,
+      "step": 21352
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 4.7509794776119404e-05,
+      "loss": 0.0017,
+      "step": 21356
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 4.750932835820896e-05,
+      "loss": 0.0004,
+      "step": 21360
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 4.7508861940298514e-05,
+      "loss": 0.0022,
+      "step": 21364
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 4.750839552238806e-05,
+      "loss": 0.0005,
+      "step": 21368
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 4.750792910447762e-05,
+      "loss": 0.0028,
+      "step": 21372
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 4.7507462686567165e-05,
+      "loss": 0.0007,
+      "step": 21376
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 4.750699626865672e-05,
+      "loss": 0.0053,
+      "step": 21380
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 4.7506529850746275e-05,
+      "loss": 0.0003,
+      "step": 21384
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 4.750606343283582e-05,
+      "loss": 0.0004,
+      "step": 21388
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 4.750559701492537e-05,
+      "loss": 0.0046,
+      "step": 21392
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 4.750513059701493e-05,
+      "loss": 0.0003,
+      "step": 21396
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 4.750466417910448e-05,
+      "loss": 0.0001,
+      "step": 21400
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 4.750419776119403e-05,
+      "loss": 0.0012,
+      "step": 21404
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 4.7503731343283584e-05,
+      "loss": 0.0019,
+      "step": 21408
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 4.750326492537314e-05,
+      "loss": 0.0003,
+      "step": 21412
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 4.750279850746269e-05,
+      "loss": 0.0007,
+      "step": 21416
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 4.750233208955224e-05,
+      "loss": 0.0089,
+      "step": 21420
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 4.750186567164179e-05,
+      "loss": 0.0011,
+      "step": 21424
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 4.7501399253731345e-05,
+      "loss": 0.0011,
+      "step": 21428
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 4.75009328358209e-05,
+      "loss": 0.001,
+      "step": 21432
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 4.750046641791045e-05,
+      "loss": 0.0058,
+      "step": 21436
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 4.75e-05,
+      "loss": 0.0003,
+      "step": 21440
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 4.749953358208956e-05,
+      "loss": 0.0001,
+      "step": 21444
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 4.7499067164179106e-05,
+      "loss": 0.0005,
+      "step": 21448
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 4.7498600746268654e-05,
+      "loss": 0.0018,
+      "step": 21452
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 4.7498134328358215e-05,
+      "loss": 0.0005,
+      "step": 21456
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 4.7497667910447764e-05,
+      "loss": 0.0027,
+      "step": 21460
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 4.749720149253731e-05,
+      "loss": 0.0005,
+      "step": 21464
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 4.749673507462687e-05,
+      "loss": 0.0005,
+      "step": 21468
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 4.749626865671642e-05,
+      "loss": 0.0014,
+      "step": 21472
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 4.749580223880597e-05,
+      "loss": 0.001,
+      "step": 21476
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 4.7495335820895525e-05,
+      "loss": 0.0003,
+      "step": 21480
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 4.749486940298507e-05,
+      "loss": 0.0005,
+      "step": 21484
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 4.749440298507463e-05,
+      "loss": 0.0003,
+      "step": 21488
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 4.749393656716418e-05,
+      "loss": 0.0037,
+      "step": 21492
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 4.749347014925373e-05,
+      "loss": 0.0005,
+      "step": 21496
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 4.7493003731343286e-05,
+      "loss": 0.0023,
+      "step": 21500
+    },
+    {
+      "epoch": 20.06,
+      "eval_exact_match": 0.7224371373307543,
+      "eval_exec": 0.7562862669245648,
+      "eval_loss": 0.34417060017585754,
+      "eval_runtime": 1073.2583,
+      "eval_samples_per_second": 0.963,
+      "step": 21500
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 4.749253731343284e-05,
+      "loss": 0.0007,
+      "step": 21504
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 4.749207089552239e-05,
+      "loss": 0.0026,
+      "step": 21508
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 4.7491604477611944e-05,
+      "loss": 0.0012,
+      "step": 21512
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 4.74911380597015e-05,
+      "loss": 0.0002,
+      "step": 21516
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 4.7490671641791047e-05,
+      "loss": 0.0004,
+      "step": 21520
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 4.74902052238806e-05,
+      "loss": 0.009,
+      "step": 21524
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 4.748973880597015e-05,
+      "loss": 0.0002,
+      "step": 21528
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 4.7489272388059705e-05,
+      "loss": 0.0004,
+      "step": 21532
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 4.748880597014926e-05,
+      "loss": 0.0026,
+      "step": 21536
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 4.748833955223881e-05,
+      "loss": 0.0007,
+      "step": 21540
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 4.7487873134328356e-05,
+      "loss": 0.0021,
+      "step": 21544
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 4.748740671641792e-05,
+      "loss": 0.0003,
+      "step": 21548
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 4.7486940298507465e-05,
+      "loss": 0.0002,
+      "step": 21552
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 4.7486473880597014e-05,
+      "loss": 0.0034,
+      "step": 21556
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 4.748600746268657e-05,
+      "loss": 0.0003,
+      "step": 21560
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 4.7485541044776123e-05,
+      "loss": 0.0003,
+      "step": 21564
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 4.748507462686567e-05,
+      "loss": 0.0048,
+      "step": 21568
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 4.7484608208955226e-05,
+      "loss": 0.0008,
+      "step": 21572
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 4.748414179104478e-05,
+      "loss": 0.002,
+      "step": 21576
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 4.748367537313433e-05,
+      "loss": 0.0005,
+      "step": 21580
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 4.7483208955223884e-05,
+      "loss": 0.0014,
+      "step": 21584
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 4.748274253731343e-05,
+      "loss": 0.0035,
+      "step": 21588
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 4.748227611940299e-05,
+      "loss": 0.001,
+      "step": 21592
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 4.748180970149254e-05,
+      "loss": 0.0008,
+      "step": 21596
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 4.748134328358209e-05,
+      "loss": 0.0169,
+      "step": 21600
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 4.748087686567164e-05,
+      "loss": 0.0008,
+      "step": 21604
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 4.74804104477612e-05,
+      "loss": 0.0003,
+      "step": 21608
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 4.747994402985075e-05,
+      "loss": 0.0003,
+      "step": 21612
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 4.7479477611940297e-05,
+      "loss": 0.0005,
+      "step": 21616
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 4.747901119402985e-05,
+      "loss": 0.0,
+      "step": 21620
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 4.7478544776119406e-05,
+      "loss": 0.0018,
+      "step": 21624
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 4.7478078358208955e-05,
+      "loss": 0.0007,
+      "step": 21628
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 4.747761194029851e-05,
+      "loss": 0.0013,
+      "step": 21632
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 4.7477145522388064e-05,
+      "loss": 0.0001,
+      "step": 21636
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 4.747667910447761e-05,
+      "loss": 0.0017,
+      "step": 21640
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 4.747621268656717e-05,
+      "loss": 0.002,
+      "step": 21644
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 4.7475746268656716e-05,
+      "loss": 0.0002,
+      "step": 21648
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 4.747527985074627e-05,
+      "loss": 0.0025,
+      "step": 21652
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 4.7474813432835825e-05,
+      "loss": 0.0003,
+      "step": 21656
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 4.7474347014925373e-05,
+      "loss": 0.0018,
+      "step": 21660
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 4.747388059701493e-05,
+      "loss": 0.0071,
+      "step": 21664
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 4.747341417910448e-05,
+      "loss": 0.0013,
+      "step": 21668
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 4.747294776119403e-05,
+      "loss": 0.0014,
+      "step": 21672
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 4.7472481343283586e-05,
+      "loss": 0.0003,
+      "step": 21676
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 4.7472014925373134e-05,
+      "loss": 0.0003,
+      "step": 21680
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 4.747154850746269e-05,
+      "loss": 0.0034,
+      "step": 21684
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 4.7471082089552244e-05,
+      "loss": 0.0001,
+      "step": 21688
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 4.747061567164179e-05,
+      "loss": 0.0003,
+      "step": 21692
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 4.747014925373135e-05,
+      "loss": 0.0041,
+      "step": 21696
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 4.74696828358209e-05,
+      "loss": 0.0008,
+      "step": 21700
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 4.746921641791045e-05,
+      "loss": 0.0004,
+      "step": 21704
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 4.746875e-05,
+      "loss": 0.0005,
+      "step": 21708
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 4.746828358208955e-05,
+      "loss": 0.0003,
+      "step": 21712
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 4.746781716417911e-05,
+      "loss": 0.0005,
+      "step": 21716
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 4.7467350746268656e-05,
+      "loss": 0.0001,
+      "step": 21720
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 4.746688432835821e-05,
+      "loss": 0.0039,
+      "step": 21724
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 4.7466417910447766e-05,
+      "loss": 0.0005,
+      "step": 21728
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 4.7465951492537314e-05,
+      "loss": 0.0046,
+      "step": 21732
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 4.746548507462687e-05,
+      "loss": 0.0003,
+      "step": 21736
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 4.746501865671642e-05,
+      "loss": 0.0025,
+      "step": 21740
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 4.746455223880597e-05,
+      "loss": 0.0016,
+      "step": 21744
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 4.746408582089553e-05,
+      "loss": 0.0019,
+      "step": 21748
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 4.7463619402985075e-05,
+      "loss": 0.0016,
+      "step": 21752
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 4.746315298507463e-05,
+      "loss": 0.0004,
+      "step": 21756
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 4.7462686567164185e-05,
+      "loss": 0.0004,
+      "step": 21760
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 4.746222014925373e-05,
+      "loss": 0.0007,
+      "step": 21764
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 4.746175373134328e-05,
+      "loss": 0.0002,
+      "step": 21768
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 4.7461287313432836e-05,
+      "loss": 0.0006,
+      "step": 21772
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 4.746082089552239e-05,
+      "loss": 0.0003,
+      "step": 21776
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 4.746035447761194e-05,
+      "loss": 0.0005,
+      "step": 21780
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 4.7459888059701494e-05,
+      "loss": 0.0007,
+      "step": 21784
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 4.745942164179105e-05,
+      "loss": 0.0005,
+      "step": 21788
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 4.74589552238806e-05,
+      "loss": 0.0005,
+      "step": 21792
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 4.745848880597015e-05,
+      "loss": 0.0004,
+      "step": 21796
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 4.74580223880597e-05,
+      "loss": 0.0003,
+      "step": 21800
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 4.7457555970149255e-05,
+      "loss": 0.0032,
+      "step": 21804
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 4.745708955223881e-05,
+      "loss": 0.0004,
+      "step": 21808
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 4.745662313432836e-05,
+      "loss": 0.0013,
+      "step": 21812
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 4.745615671641791e-05,
+      "loss": 0.0008,
+      "step": 21816
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 4.745569029850747e-05,
+      "loss": 0.0047,
+      "step": 21820
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 4.7455223880597016e-05,
+      "loss": 0.0001,
+      "step": 21824
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 4.745475746268657e-05,
+      "loss": 0.0012,
+      "step": 21828
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 4.745429104477612e-05,
+      "loss": 0.0035,
+      "step": 21832
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 4.7453824626865674e-05,
+      "loss": 0.0052,
+      "step": 21836
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 4.745335820895523e-05,
+      "loss": 0.0006,
+      "step": 21840
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 4.745289179104478e-05,
+      "loss": 0.002,
+      "step": 21844
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 4.745242537313433e-05,
+      "loss": 0.0063,
+      "step": 21848
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 4.745195895522389e-05,
+      "loss": 0.0005,
+      "step": 21852
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 4.7451492537313435e-05,
+      "loss": 0.0214,
+      "step": 21856
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 4.745102611940298e-05,
+      "loss": 0.0038,
+      "step": 21860
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 4.7450559701492545e-05,
+      "loss": 0.0043,
+      "step": 21864
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 4.745009328358209e-05,
+      "loss": 0.0023,
+      "step": 21868
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 4.744962686567164e-05,
+      "loss": 0.0001,
+      "step": 21872
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 4.7449160447761196e-05,
+      "loss": 0.0002,
+      "step": 21876
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 4.744869402985075e-05,
+      "loss": 0.0021,
+      "step": 21880
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 4.74482276119403e-05,
+      "loss": 0.0018,
+      "step": 21884
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 4.7447761194029854e-05,
+      "loss": 0.0012,
+      "step": 21888
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 4.74472947761194e-05,
+      "loss": 0.0003,
+      "step": 21892
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 4.744682835820896e-05,
+      "loss": 0.0016,
+      "step": 21896
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 4.744636194029851e-05,
+      "loss": 0.0029,
+      "step": 21900
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 4.744589552238806e-05,
+      "loss": 0.0005,
+      "step": 21904
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 4.7445429104477615e-05,
+      "loss": 0.0028,
+      "step": 21908
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 4.744496268656717e-05,
+      "loss": 0.0006,
+      "step": 21912
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 4.744449626865672e-05,
+      "loss": 0.0002,
+      "step": 21916
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 4.7444029850746266e-05,
+      "loss": 0.0009,
+      "step": 21920
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 4.744356343283583e-05,
+      "loss": 0.0019,
+      "step": 21924
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 4.7443097014925376e-05,
+      "loss": 0.0005,
+      "step": 21928
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 4.7442630597014924e-05,
+      "loss": 0.0018,
+      "step": 21932
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 4.744216417910448e-05,
+      "loss": 0.0024,
+      "step": 21936
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 4.7441697761194034e-05,
+      "loss": 0.0007,
+      "step": 21940
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 4.744123134328358e-05,
+      "loss": 0.0129,
+      "step": 21944
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 4.744076492537314e-05,
+      "loss": 0.005,
+      "step": 21948
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 4.7440298507462685e-05,
+      "loss": 0.0038,
+      "step": 21952
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 4.743983208955224e-05,
+      "loss": 0.0001,
+      "step": 21956
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 4.7439365671641795e-05,
+      "loss": 0.0003,
+      "step": 21960
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 4.743889925373134e-05,
+      "loss": 0.0015,
+      "step": 21964
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 4.74384328358209e-05,
+      "loss": 0.0002,
+      "step": 21968
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 4.743796641791045e-05,
+      "loss": 0.0009,
+      "step": 21972
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 4.74375e-05,
+      "loss": 0.0006,
+      "step": 21976
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 4.7437033582089556e-05,
+      "loss": 0.0001,
+      "step": 21980
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 4.743656716417911e-05,
+      "loss": 0.0005,
+      "step": 21984
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 4.743610074626866e-05,
+      "loss": 0.0021,
+      "step": 21988
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 4.7435634328358214e-05,
+      "loss": 0.0008,
+      "step": 21992
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 4.743516791044776e-05,
+      "loss": 0.0022,
+      "step": 21996
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 4.743470149253732e-05,
+      "loss": 0.009,
+      "step": 22000
+    },
+    {
+      "epoch": 20.52,
+      "eval_exact_match": 0.718568665377176,
+      "eval_exec": 0.746615087040619,
+      "eval_loss": 0.36730751395225525,
+      "eval_runtime": 1232.9452,
+      "eval_samples_per_second": 0.839,
+      "step": 22000
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 4.743423507462687e-05,
+      "loss": 0.0004,
+      "step": 22004
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 4.743376865671642e-05,
+      "loss": 0.0028,
+      "step": 22008
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 4.743330223880597e-05,
+      "loss": 0.0001,
+      "step": 22012
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 4.743283582089553e-05,
+      "loss": 0.0002,
+      "step": 22016
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 4.743236940298508e-05,
+      "loss": 0.0049,
+      "step": 22020
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 4.7431902985074626e-05,
+      "loss": 0.0001,
+      "step": 22024
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 4.743143656716418e-05,
+      "loss": 0.0003,
+      "step": 22028
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 4.7430970149253736e-05,
+      "loss": 0.0003,
+      "step": 22032
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 4.7430503731343284e-05,
+      "loss": 0.0009,
+      "step": 22036
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 4.743003731343284e-05,
+      "loss": 0.0002,
+      "step": 22040
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 4.7429570895522394e-05,
+      "loss": 0.0001,
+      "step": 22044
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 4.742910447761194e-05,
+      "loss": 0.0001,
+      "step": 22048
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 4.74286380597015e-05,
+      "loss": 0.0337,
+      "step": 22052
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 4.7428171641791045e-05,
+      "loss": 0.0002,
+      "step": 22056
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 4.74277052238806e-05,
+      "loss": 0.0001,
+      "step": 22060
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 4.7427238805970155e-05,
+      "loss": 0.0001,
+      "step": 22064
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 4.74267723880597e-05,
+      "loss": 0.0031,
+      "step": 22068
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 4.742630597014925e-05,
+      "loss": 0.0008,
+      "step": 22072
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 4.742583955223881e-05,
+      "loss": 0.0003,
+      "step": 22076
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 4.742537313432836e-05,
+      "loss": 0.0011,
+      "step": 22080
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 4.742490671641791e-05,
+      "loss": 0.0006,
+      "step": 22084
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 4.7424440298507464e-05,
+      "loss": 0.0001,
+      "step": 22088
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 4.742397388059702e-05,
+      "loss": 0.0025,
+      "step": 22092
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 4.742350746268657e-05,
+      "loss": 0.0034,
+      "step": 22096
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 4.742304104477612e-05,
+      "loss": 0.0001,
+      "step": 22100
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 4.742257462686567e-05,
+      "loss": 0.0001,
+      "step": 22104
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 4.7422108208955225e-05,
+      "loss": 0.0015,
+      "step": 22108
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 4.742164179104478e-05,
+      "loss": 0.0001,
+      "step": 22112
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 4.742117537313433e-05,
+      "loss": 0.0002,
+      "step": 22116
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 4.742070895522388e-05,
+      "loss": 0.0075,
+      "step": 22120
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 4.742024253731344e-05,
+      "loss": 0.0001,
+      "step": 22124
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 4.7419776119402986e-05,
+      "loss": 0.0042,
+      "step": 22128
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 4.7419309701492534e-05,
+      "loss": 0.0004,
+      "step": 22132
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 4.7418843283582096e-05,
+      "loss": 0.0016,
+      "step": 22136
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 4.7418376865671644e-05,
+      "loss": 0.0008,
+      "step": 22140
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 4.74179104477612e-05,
+      "loss": 0.0014,
+      "step": 22144
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 4.741744402985075e-05,
+      "loss": 0.0001,
+      "step": 22148
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 4.74169776119403e-05,
+      "loss": 0.0036,
+      "step": 22152
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 4.741651119402986e-05,
+      "loss": 0.0004,
+      "step": 22156
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 4.7416044776119405e-05,
+      "loss": 0.0022,
+      "step": 22160
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 4.741557835820895e-05,
+      "loss": 0.0009,
+      "step": 22164
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 4.7415111940298515e-05,
+      "loss": 0.0004,
+      "step": 22168
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 4.741464552238806e-05,
+      "loss": 0.0057,
+      "step": 22172
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 4.741417910447761e-05,
+      "loss": 0.0005,
+      "step": 22176
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 4.7413712686567166e-05,
+      "loss": 0.0015,
+      "step": 22180
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 4.741324626865672e-05,
+      "loss": 0.0002,
+      "step": 22184
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 4.741277985074627e-05,
+      "loss": 0.0002,
+      "step": 22188
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 4.7412313432835824e-05,
+      "loss": 0.0003,
+      "step": 22192
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 4.741184701492538e-05,
+      "loss": 0.0003,
+      "step": 22196
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 4.741138059701493e-05,
+      "loss": 0.0023,
+      "step": 22200
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 4.741091417910448e-05,
+      "loss": 0.0016,
+      "step": 22204
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 4.741044776119403e-05,
+      "loss": 0.0012,
+      "step": 22208
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 4.7409981343283585e-05,
+      "loss": 0.0025,
+      "step": 22212
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 4.740951492537314e-05,
+      "loss": 0.0003,
+      "step": 22216
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 4.740904850746269e-05,
+      "loss": 0.0009,
+      "step": 22220
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 4.7408582089552236e-05,
+      "loss": 0.0008,
+      "step": 22224
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 4.74081156716418e-05,
+      "loss": 0.004,
+      "step": 22228
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 4.7407649253731346e-05,
+      "loss": 0.0004,
+      "step": 22232
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 4.7407182835820894e-05,
+      "loss": 0.0026,
+      "step": 22236
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 4.740671641791045e-05,
+      "loss": 0.0016,
+      "step": 22240
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 4.7406250000000004e-05,
+      "loss": 0.0011,
+      "step": 22244
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 4.740578358208955e-05,
+      "loss": 0.0045,
+      "step": 22248
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 4.740531716417911e-05,
+      "loss": 0.0007,
+      "step": 22252
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 4.740485074626866e-05,
+      "loss": 0.0013,
+      "step": 22256
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 4.740438432835821e-05,
+      "loss": 0.0009,
+      "step": 22260
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 4.7403917910447765e-05,
+      "loss": 0.0041,
+      "step": 22264
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 4.740345149253731e-05,
+      "loss": 0.0016,
+      "step": 22268
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 4.740298507462687e-05,
+      "loss": 0.0008,
+      "step": 22272
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 4.740251865671642e-05,
+      "loss": 0.0174,
+      "step": 22276
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 4.740205223880597e-05,
+      "loss": 0.0067,
+      "step": 22280
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 4.740158582089552e-05,
+      "loss": 0.0009,
+      "step": 22284
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 4.740111940298508e-05,
+      "loss": 0.0,
+      "step": 22288
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 4.740065298507463e-05,
+      "loss": 0.0009,
+      "step": 22292
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 4.740018656716418e-05,
+      "loss": 0.0014,
+      "step": 22296
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 4.739972014925373e-05,
+      "loss": 0.0009,
+      "step": 22300
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 4.739925373134329e-05,
+      "loss": 0.0001,
+      "step": 22304
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 4.739878731343284e-05,
+      "loss": 0.0046,
+      "step": 22308
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 4.739832089552239e-05,
+      "loss": 0.0004,
+      "step": 22312
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 4.7397854477611945e-05,
+      "loss": 0.0076,
+      "step": 22316
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 4.73973880597015e-05,
+      "loss": 0.0117,
+      "step": 22320
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 4.739692164179105e-05,
+      "loss": 0.0002,
+      "step": 22324
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 4.7396455223880596e-05,
+      "loss": 0.0009,
+      "step": 22328
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 4.739598880597015e-05,
+      "loss": 0.0028,
+      "step": 22332
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 4.7395522388059706e-05,
+      "loss": 0.0009,
+      "step": 22336
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 4.7395055970149254e-05,
+      "loss": 0.0007,
+      "step": 22340
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 4.739458955223881e-05,
+      "loss": 0.001,
+      "step": 22344
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 4.7394123134328364e-05,
+      "loss": 0.0058,
+      "step": 22348
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 4.739365671641791e-05,
+      "loss": 0.0018,
+      "step": 22352
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 4.7393190298507467e-05,
+      "loss": 0.0022,
+      "step": 22356
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 4.7392723880597015e-05,
+      "loss": 0.0006,
+      "step": 22360
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 4.739225746268657e-05,
+      "loss": 0.0006,
+      "step": 22364
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 4.7391791044776125e-05,
+      "loss": 0.0003,
+      "step": 22368
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 4.739132462686567e-05,
+      "loss": 0.0001,
+      "step": 22372
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 4.739085820895523e-05,
+      "loss": 0.0002,
+      "step": 22376
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 4.739039179104478e-05,
+      "loss": 0.0007,
+      "step": 22380
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 4.738992537313433e-05,
+      "loss": 0.0015,
+      "step": 22384
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 4.738945895522388e-05,
+      "loss": 0.0007,
+      "step": 22388
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 4.7388992537313434e-05,
+      "loss": 0.0003,
+      "step": 22392
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 4.738852611940299e-05,
+      "loss": 0.0024,
+      "step": 22396
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 4.738805970149254e-05,
+      "loss": 0.0006,
+      "step": 22400
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 4.738759328358209e-05,
+      "loss": 0.0005,
+      "step": 22404
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 4.7387126865671647e-05,
+      "loss": 0.0026,
+      "step": 22408
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 4.7386660447761195e-05,
+      "loss": 0.0015,
+      "step": 22412
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 4.738619402985075e-05,
+      "loss": 0.0008,
+      "step": 22416
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 4.73857276119403e-05,
+      "loss": 0.0009,
+      "step": 22420
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 4.738526119402985e-05,
+      "loss": 0.0006,
+      "step": 22424
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 4.738479477611941e-05,
+      "loss": 0.0038,
+      "step": 22428
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 4.7384328358208956e-05,
+      "loss": 0.0001,
+      "step": 22432
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 4.738386194029851e-05,
+      "loss": 0.0005,
+      "step": 22436
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 4.7383395522388065e-05,
+      "loss": 0.0003,
+      "step": 22440
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 4.7382929104477614e-05,
+      "loss": 0.0004,
+      "step": 22444
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 4.738246268656716e-05,
+      "loss": 0.0036,
+      "step": 22448
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 4.738199626865672e-05,
+      "loss": 0.0002,
+      "step": 22452
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 4.738152985074627e-05,
+      "loss": 0.0025,
+      "step": 22456
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 4.738106343283582e-05,
+      "loss": 0.0004,
+      "step": 22460
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 4.7380597014925375e-05,
+      "loss": 0.0019,
+      "step": 22464
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 4.738013059701493e-05,
+      "loss": 0.005,
+      "step": 22468
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 4.7379664179104484e-05,
+      "loss": 0.0005,
+      "step": 22472
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 4.737919776119403e-05,
+      "loss": 0.0002,
+      "step": 22476
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 4.737873134328358e-05,
+      "loss": 0.0046,
+      "step": 22480
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 4.737826492537314e-05,
+      "loss": 0.0005,
+      "step": 22484
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 4.737779850746269e-05,
+      "loss": 0.0011,
+      "step": 22488
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 4.737733208955224e-05,
+      "loss": 0.0024,
+      "step": 22492
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 4.7376865671641793e-05,
+      "loss": 0.001,
+      "step": 22496
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 4.737639925373135e-05,
+      "loss": 0.0001,
+      "step": 22500
+    },
+    {
+      "epoch": 20.99,
+      "eval_exact_match": 0.7321083172147002,
+      "eval_exec": 0.7669245647969052,
+      "eval_loss": 0.3726440668106079,
+      "eval_runtime": 1238.1908,
+      "eval_samples_per_second": 0.835,
+      "step": 22500
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 4.7375932835820897e-05,
+      "loss": 0.0001,
+      "step": 22504
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 4.737546641791045e-05,
+      "loss": 0.0035,
+      "step": 22508
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 4.7375e-05,
+      "loss": 0.0003,
+      "step": 22512
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 4.7374533582089554e-05,
+      "loss": 0.0002,
+      "step": 22516
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 4.737406716417911e-05,
+      "loss": 0.0005,
+      "step": 22520
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 4.737360074626866e-05,
+      "loss": 0.006,
+      "step": 22524
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 4.737313432835821e-05,
+      "loss": 0.0019,
+      "step": 22528
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 4.737266791044777e-05,
+      "loss": 0.0008,
+      "step": 22532
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 4.7372201492537315e-05,
+      "loss": 0.002,
+      "step": 22536
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 4.7371735074626864e-05,
+      "loss": 0.0002,
+      "step": 22540
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 4.7371268656716425e-05,
+      "loss": 0.0007,
+      "step": 22544
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 4.7370802238805973e-05,
+      "loss": 0.0004,
+      "step": 22548
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 4.737033582089552e-05,
+      "loss": 0.0002,
+      "step": 22552
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 4.7369869402985076e-05,
+      "loss": 0.0005,
+      "step": 22556
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 4.736940298507463e-05,
+      "loss": 0.0001,
+      "step": 22560
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 4.736893656716418e-05,
+      "loss": 0.0003,
+      "step": 22564
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 4.7368470149253734e-05,
+      "loss": 0.0008,
+      "step": 22568
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 4.736800373134328e-05,
+      "loss": 0.0027,
+      "step": 22572
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 4.736753731343284e-05,
+      "loss": 0.0021,
+      "step": 22576
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 4.736707089552239e-05,
+      "loss": 0.0071,
+      "step": 22580
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 4.736660447761194e-05,
+      "loss": 0.001,
+      "step": 22584
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 4.7366138059701495e-05,
+      "loss": 0.0019,
+      "step": 22588
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 4.736567164179105e-05,
+      "loss": 0.0006,
+      "step": 22592
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 4.73652052238806e-05,
+      "loss": 0.0009,
+      "step": 22596
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 4.7364738805970147e-05,
+      "loss": 0.0016,
+      "step": 22600
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 4.736427238805971e-05,
+      "loss": 0.0001,
+      "step": 22604
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 4.7363805970149256e-05,
+      "loss": 0.0004,
+      "step": 22608
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 4.7363339552238805e-05,
+      "loss": 0.0074,
+      "step": 22612
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 4.736287313432836e-05,
+      "loss": 0.0042,
+      "step": 22616
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 4.7362406716417914e-05,
+      "loss": 0.0001,
+      "step": 22620
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 4.736194029850746e-05,
+      "loss": 0.0036,
+      "step": 22624
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 4.736147388059702e-05,
+      "loss": 0.0009,
+      "step": 22628
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 4.7361007462686565e-05,
+      "loss": 0.0002,
+      "step": 22632
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 4.736054104477613e-05,
+      "loss": 0.0,
+      "step": 22636
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 4.7360074626865675e-05,
+      "loss": 0.0076,
+      "step": 22640
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 4.7359608208955223e-05,
+      "loss": 0.0015,
+      "step": 22644
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 4.735914179104478e-05,
+      "loss": 0.003,
+      "step": 22648
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 4.735867537313433e-05,
+      "loss": 0.0014,
+      "step": 22652
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 4.735820895522388e-05,
+      "loss": 0.0033,
+      "step": 22656
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 4.7357742537313436e-05,
+      "loss": 0.0001,
+      "step": 22660
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 4.735727611940299e-05,
+      "loss": 0.0011,
+      "step": 22664
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 4.735680970149254e-05,
+      "loss": 0.0007,
+      "step": 22668
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 4.7356343283582094e-05,
+      "loss": 0.0007,
+      "step": 22672
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 4.735587686567164e-05,
+      "loss": 0.0041,
+      "step": 22676
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 4.73554104477612e-05,
+      "loss": 0.0011,
+      "step": 22680
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 4.735494402985075e-05,
+      "loss": 0.0036,
+      "step": 22684
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 4.73544776119403e-05,
+      "loss": 0.0049,
+      "step": 22688
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 4.735401119402985e-05,
+      "loss": 0.0011,
+      "step": 22692
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 4.735354477611941e-05,
+      "loss": 0.0011,
+      "step": 22696
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 4.735307835820896e-05,
+      "loss": 0.0031,
+      "step": 22700
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 4.7352611940298506e-05,
+      "loss": 0.0002,
+      "step": 22704
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 4.735214552238806e-05,
+      "loss": 0.0002,
+      "step": 22708
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 4.7351679104477616e-05,
+      "loss": 0.0005,
+      "step": 22712
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 4.7351212686567164e-05,
+      "loss": 0.0044,
+      "step": 22716
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 4.735074626865672e-05,
+      "loss": 0.0009,
+      "step": 22720
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 4.7350279850746274e-05,
+      "loss": 0.0002,
+      "step": 22724
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 4.734981343283582e-05,
+      "loss": 0.0003,
+      "step": 22728
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 4.734934701492538e-05,
+      "loss": 0.0001,
+      "step": 22732
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 4.7348880597014925e-05,
+      "loss": 0.0013,
+      "step": 22736
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 4.734841417910448e-05,
+      "loss": 0.0047,
+      "step": 22740
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 4.7347947761194035e-05,
+      "loss": 0.0031,
+      "step": 22744
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 4.734748134328358e-05,
+      "loss": 0.0044,
+      "step": 22748
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 4.734701492537313e-05,
+      "loss": 0.0001,
+      "step": 22752
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 4.734654850746269e-05,
+      "loss": 0.0006,
+      "step": 22756
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 4.734608208955224e-05,
+      "loss": 0.0004,
+      "step": 22760
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 4.734561567164179e-05,
+      "loss": 0.0019,
+      "step": 22764
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 4.7345149253731344e-05,
+      "loss": 0.0003,
+      "step": 22768
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 4.73446828358209e-05,
+      "loss": 0.0009,
+      "step": 22772
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 4.734421641791045e-05,
+      "loss": 0.0006,
+      "step": 22776
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 4.734375e-05,
+      "loss": 0.0015,
+      "step": 22780
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 4.734328358208955e-05,
+      "loss": 0.0011,
+      "step": 22784
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 4.7342817164179105e-05,
+      "loss": 0.0038,
+      "step": 22788
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 4.734235074626866e-05,
+      "loss": 0.0001,
+      "step": 22792
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 4.734188432835821e-05,
+      "loss": 0.0001,
+      "step": 22796
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 4.734141791044776e-05,
+      "loss": 0.0037,
+      "step": 22800
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 4.734095149253732e-05,
+      "loss": 0.0003,
+      "step": 22804
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 4.7340485074626866e-05,
+      "loss": 0.0006,
+      "step": 22808
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 4.734001865671642e-05,
+      "loss": 0.0003,
+      "step": 22812
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 4.7339552238805976e-05,
+      "loss": 0.0018,
+      "step": 22816
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 4.7339085820895524e-05,
+      "loss": 0.0003,
+      "step": 22820
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 4.733861940298508e-05,
+      "loss": 0.0003,
+      "step": 22824
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 4.733815298507463e-05,
+      "loss": 0.004,
+      "step": 22828
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 4.733768656716418e-05,
+      "loss": 0.0005,
+      "step": 22832
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 4.733722014925374e-05,
+      "loss": 0.0001,
+      "step": 22836
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 4.7336753731343285e-05,
+      "loss": 0.0006,
+      "step": 22840
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 4.733628731343283e-05,
+      "loss": 0.0018,
+      "step": 22844
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 4.7335820895522395e-05,
+      "loss": 0.0,
+      "step": 22848
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 4.733535447761194e-05,
+      "loss": 0.0008,
+      "step": 22852
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 4.733488805970149e-05,
+      "loss": 0.0045,
+      "step": 22856
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 4.7334421641791046e-05,
+      "loss": 0.0006,
+      "step": 22860
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 4.73339552238806e-05,
+      "loss": 0.0014,
+      "step": 22864
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 4.733348880597015e-05,
+      "loss": 0.0005,
+      "step": 22868
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 4.7333022388059704e-05,
+      "loss": 0.0001,
+      "step": 22872
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 4.733255597014926e-05,
+      "loss": 0.0032,
+      "step": 22876
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 4.733208955223881e-05,
+      "loss": 0.0015,
+      "step": 22880
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 4.733162313432836e-05,
+      "loss": 0.0006,
+      "step": 22884
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 4.733115671641791e-05,
+      "loss": 0.0001,
+      "step": 22888
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 4.7330690298507465e-05,
+      "loss": 0.0023,
+      "step": 22892
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 4.733022388059702e-05,
+      "loss": 0.0013,
+      "step": 22896
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 4.732975746268657e-05,
+      "loss": 0.0022,
+      "step": 22900
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 4.7329291044776116e-05,
+      "loss": 0.0071,
+      "step": 22904
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 4.732882462686568e-05,
+      "loss": 0.0038,
+      "step": 22908
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 4.7328358208955226e-05,
+      "loss": 0.0001,
+      "step": 22912
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 4.7327891791044774e-05,
+      "loss": 0.0005,
+      "step": 22916
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 4.732742537313433e-05,
+      "loss": 0.0001,
+      "step": 22920
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 4.7326958955223884e-05,
+      "loss": 0.0085,
+      "step": 22924
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 4.732649253731343e-05,
+      "loss": 0.0053,
+      "step": 22928
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 4.732602611940299e-05,
+      "loss": 0.0007,
+      "step": 22932
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 4.732555970149254e-05,
+      "loss": 0.0013,
+      "step": 22936
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 4.732509328358209e-05,
+      "loss": 0.0002,
+      "step": 22940
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 4.7324626865671645e-05,
+      "loss": 0.0027,
+      "step": 22944
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 4.732416044776119e-05,
+      "loss": 0.0005,
+      "step": 22948
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 4.732369402985075e-05,
+      "loss": 0.0004,
+      "step": 22952
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 4.73232276119403e-05,
+      "loss": 0.0006,
+      "step": 22956
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 4.732276119402985e-05,
+      "loss": 0.0047,
+      "step": 22960
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 4.7322294776119406e-05,
+      "loss": 0.0008,
+      "step": 22964
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 4.732182835820896e-05,
+      "loss": 0.0019,
+      "step": 22968
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 4.732136194029851e-05,
+      "loss": 0.0008,
+      "step": 22972
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 4.7320895522388064e-05,
+      "loss": 0.0001,
+      "step": 22976
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 4.732042910447761e-05,
+      "loss": 0.0002,
+      "step": 22980
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 4.731996268656717e-05,
+      "loss": 0.0008,
+      "step": 22984
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 4.731949626865672e-05,
+      "loss": 0.0007,
+      "step": 22988
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 4.731902985074627e-05,
+      "loss": 0.0002,
+      "step": 22992
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 4.7318563432835825e-05,
+      "loss": 0.0017,
+      "step": 22996
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 4.731809701492538e-05,
+      "loss": 0.0074,
+      "step": 23000
+    },
+    {
+      "epoch": 21.46,
+      "eval_exact_match": 0.7263056092843327,
+      "eval_exec": 0.7630560928433269,
+      "eval_loss": 0.3736368417739868,
+      "eval_runtime": 1044.2285,
+      "eval_samples_per_second": 0.99,
+      "step": 23000
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 4.731763059701493e-05,
+      "loss": 0.0002,
+      "step": 23004
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 4.7317164179104476e-05,
+      "loss": 0.0008,
+      "step": 23008
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 4.731669776119403e-05,
+      "loss": 0.0008,
+      "step": 23012
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 4.7316231343283586e-05,
+      "loss": 0.0004,
+      "step": 23016
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 4.7315764925373134e-05,
+      "loss": 0.0002,
+      "step": 23020
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 4.731529850746269e-05,
+      "loss": 0.0021,
+      "step": 23024
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 4.7314832089552244e-05,
+      "loss": 0.0001,
+      "step": 23028
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 4.731436567164179e-05,
+      "loss": 0.0001,
+      "step": 23032
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 4.731389925373135e-05,
+      "loss": 0.0063,
+      "step": 23036
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 4.7313432835820895e-05,
+      "loss": 0.0068,
+      "step": 23040
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 4.731296641791045e-05,
+      "loss": 0.0021,
+      "step": 23044
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 4.7312500000000005e-05,
+      "loss": 0.0038,
+      "step": 23048
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 4.731203358208955e-05,
+      "loss": 0.0005,
+      "step": 23052
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 4.731156716417911e-05,
+      "loss": 0.0001,
+      "step": 23056
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 4.731110074626866e-05,
+      "loss": 0.0007,
+      "step": 23060
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 4.731063432835821e-05,
+      "loss": 0.0004,
+      "step": 23064
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 4.731016791044776e-05,
+      "loss": 0.0005,
+      "step": 23068
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 4.7309701492537314e-05,
+      "loss": 0.0006,
+      "step": 23072
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 4.730923507462687e-05,
+      "loss": 0.0016,
+      "step": 23076
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 4.730876865671642e-05,
+      "loss": 0.0008,
+      "step": 23080
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 4.730830223880597e-05,
+      "loss": 0.0009,
+      "step": 23084
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 4.730783582089553e-05,
+      "loss": 0.0008,
+      "step": 23088
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 4.7307369402985075e-05,
+      "loss": 0.0016,
+      "step": 23092
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 4.730690298507463e-05,
+      "loss": 0.0001,
+      "step": 23096
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 4.730643656716418e-05,
+      "loss": 0.0047,
+      "step": 23100
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 4.730597014925373e-05,
+      "loss": 0.0004,
+      "step": 23104
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 4.730550373134329e-05,
+      "loss": 0.0008,
+      "step": 23108
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 4.7305037313432836e-05,
+      "loss": 0.0011,
+      "step": 23112
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 4.730457089552239e-05,
+      "loss": 0.0015,
+      "step": 23116
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 4.7304104477611946e-05,
+      "loss": 0.0034,
+      "step": 23120
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 4.7303638059701494e-05,
+      "loss": 0.0004,
+      "step": 23124
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 4.730317164179105e-05,
+      "loss": 0.0009,
+      "step": 23128
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 4.73027052238806e-05,
+      "loss": 0.0052,
+      "step": 23132
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 4.730223880597015e-05,
+      "loss": 0.0007,
+      "step": 23136
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 4.730177238805971e-05,
+      "loss": 0.0061,
+      "step": 23140
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 4.7301305970149255e-05,
+      "loss": 0.0036,
+      "step": 23144
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 4.730083955223881e-05,
+      "loss": 0.0002,
+      "step": 23148
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 4.7300373134328365e-05,
+      "loss": 0.0001,
+      "step": 23152
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 4.729990671641791e-05,
+      "loss": 0.0001,
+      "step": 23156
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 4.729944029850746e-05,
+      "loss": 0.0013,
+      "step": 23160
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 4.729897388059702e-05,
+      "loss": 0.0001,
+      "step": 23164
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 4.729850746268657e-05,
+      "loss": 0.003,
+      "step": 23168
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 4.729804104477612e-05,
+      "loss": 0.0053,
+      "step": 23172
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 4.7297574626865674e-05,
+      "loss": 0.0004,
+      "step": 23176
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 4.729710820895523e-05,
+      "loss": 0.0028,
+      "step": 23180
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 4.729664179104478e-05,
+      "loss": 0.0004,
+      "step": 23184
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 4.729617537313433e-05,
+      "loss": 0.0004,
+      "step": 23188
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 4.729570895522388e-05,
+      "loss": 0.0018,
+      "step": 23192
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 4.7295242537313435e-05,
+      "loss": 0.0004,
+      "step": 23196
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 4.729477611940299e-05,
+      "loss": 0.001,
+      "step": 23200
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 4.729430970149254e-05,
+      "loss": 0.0012,
+      "step": 23204
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 4.729384328358209e-05,
+      "loss": 0.001,
+      "step": 23208
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 4.729337686567165e-05,
+      "loss": 0.005,
+      "step": 23212
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 4.7292910447761196e-05,
+      "loss": 0.0017,
+      "step": 23216
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 4.7292444029850744e-05,
+      "loss": 0.0002,
+      "step": 23220
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 4.7291977611940306e-05,
+      "loss": 0.0004,
+      "step": 23224
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 4.7291511194029854e-05,
+      "loss": 0.0004,
+      "step": 23228
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 4.72910447761194e-05,
+      "loss": 0.001,
+      "step": 23232
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 4.729057835820896e-05,
+      "loss": 0.0002,
+      "step": 23236
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 4.729011194029851e-05,
+      "loss": 0.0063,
+      "step": 23240
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 4.728964552238806e-05,
+      "loss": 0.0015,
+      "step": 23244
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 4.7289179104477615e-05,
+      "loss": 0.0066,
+      "step": 23248
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 4.728871268656716e-05,
+      "loss": 0.0004,
+      "step": 23252
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 4.728824626865672e-05,
+      "loss": 0.0006,
+      "step": 23256
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 4.728777985074627e-05,
+      "loss": 0.0007,
+      "step": 23260
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 4.728731343283582e-05,
+      "loss": 0.0007,
+      "step": 23264
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 4.7286847014925376e-05,
+      "loss": 0.0005,
+      "step": 23268
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 4.728638059701493e-05,
+      "loss": 0.0004,
+      "step": 23272
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 4.728591417910448e-05,
+      "loss": 0.0004,
+      "step": 23276
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 4.728544776119403e-05,
+      "loss": 0.0002,
+      "step": 23280
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 4.728498134328359e-05,
+      "loss": 0.0005,
+      "step": 23284
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 4.728451492537314e-05,
+      "loss": 0.0003,
+      "step": 23288
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 4.728404850746269e-05,
+      "loss": 0.0016,
+      "step": 23292
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 4.728358208955224e-05,
+      "loss": 0.0005,
+      "step": 23296
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 4.7283115671641795e-05,
+      "loss": 0.0026,
+      "step": 23300
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 4.728264925373135e-05,
+      "loss": 0.0004,
+      "step": 23304
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 4.72821828358209e-05,
+      "loss": 0.0041,
+      "step": 23308
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 4.7281716417910446e-05,
+      "loss": 0.0022,
+      "step": 23312
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 4.728125000000001e-05,
+      "loss": 0.0001,
+      "step": 23316
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 4.7280783582089556e-05,
+      "loss": 0.0013,
+      "step": 23320
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 4.7280317164179104e-05,
+      "loss": 0.0005,
+      "step": 23324
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 4.727985074626866e-05,
+      "loss": 0.0039,
+      "step": 23328
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 4.7279384328358214e-05,
+      "loss": 0.0002,
+      "step": 23332
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 4.727891791044776e-05,
+      "loss": 0.0032,
+      "step": 23336
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 4.7278451492537317e-05,
+      "loss": 0.003,
+      "step": 23340
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 4.727798507462687e-05,
+      "loss": 0.0012,
+      "step": 23344
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 4.727751865671642e-05,
+      "loss": 0.0001,
+      "step": 23348
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 4.7277052238805975e-05,
+      "loss": 0.0053,
+      "step": 23352
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 4.727658582089552e-05,
+      "loss": 0.0026,
+      "step": 23356
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 4.727611940298508e-05,
+      "loss": 0.0003,
+      "step": 23360
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 4.727565298507463e-05,
+      "loss": 0.0002,
+      "step": 23364
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 4.727518656716418e-05,
+      "loss": 0.0061,
+      "step": 23368
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 4.727472014925373e-05,
+      "loss": 0.0089,
+      "step": 23372
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 4.727425373134329e-05,
+      "loss": 0.0052,
+      "step": 23376
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 4.727378731343284e-05,
+      "loss": 0.0004,
+      "step": 23380
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 4.727332089552239e-05,
+      "loss": 0.0003,
+      "step": 23384
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 4.727285447761194e-05,
+      "loss": 0.0017,
+      "step": 23388
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 4.7272388059701496e-05,
+      "loss": 0.0068,
+      "step": 23392
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 4.7271921641791045e-05,
+      "loss": 0.0007,
+      "step": 23396
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 4.72714552238806e-05,
+      "loss": 0.0001,
+      "step": 23400
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 4.7270988805970154e-05,
+      "loss": 0.0002,
+      "step": 23404
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 4.72705223880597e-05,
+      "loss": 0.0011,
+      "step": 23408
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 4.727005597014926e-05,
+      "loss": 0.0083,
+      "step": 23412
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 4.7269589552238806e-05,
+      "loss": 0.0012,
+      "step": 23416
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 4.726912313432836e-05,
+      "loss": 0.0002,
+      "step": 23420
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 4.7268656716417915e-05,
+      "loss": 0.0012,
+      "step": 23424
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 4.7268190298507464e-05,
+      "loss": 0.0065,
+      "step": 23428
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 4.726772388059701e-05,
+      "loss": 0.0005,
+      "step": 23432
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 4.726725746268657e-05,
+      "loss": 0.0027,
+      "step": 23436
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 4.726679104477612e-05,
+      "loss": 0.0001,
+      "step": 23440
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 4.726632462686567e-05,
+      "loss": 0.0003,
+      "step": 23444
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 4.7265858208955225e-05,
+      "loss": 0.0003,
+      "step": 23448
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 4.726539179104478e-05,
+      "loss": 0.0001,
+      "step": 23452
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 4.7264925373134334e-05,
+      "loss": 0.0002,
+      "step": 23456
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 4.726445895522388e-05,
+      "loss": 0.0014,
+      "step": 23460
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 4.726399253731344e-05,
+      "loss": 0.0018,
+      "step": 23464
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 4.726352611940299e-05,
+      "loss": 0.0012,
+      "step": 23468
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 4.726305970149254e-05,
+      "loss": 0.0036,
+      "step": 23472
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 4.726259328358209e-05,
+      "loss": 0.0006,
+      "step": 23476
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 4.7262126865671643e-05,
+      "loss": 0.0003,
+      "step": 23480
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 4.72616604477612e-05,
+      "loss": 0.0003,
+      "step": 23484
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 4.7261194029850747e-05,
+      "loss": 0.0009,
+      "step": 23488
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 4.72607276119403e-05,
+      "loss": 0.0003,
+      "step": 23492
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 4.7260261194029856e-05,
+      "loss": 0.0002,
+      "step": 23496
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 4.7259794776119404e-05,
+      "loss": 0.0011,
+      "step": 23500
+    },
+    {
+      "epoch": 21.92,
+      "eval_exact_match": 0.7272727272727273,
+      "eval_exec": 0.7630560928433269,
+      "eval_loss": 0.36809346079826355,
+      "eval_runtime": 1056.1475,
+      "eval_samples_per_second": 0.979,
+      "step": 23500
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 4.725932835820896e-05,
+      "loss": 0.0008,
+      "step": 23504
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 4.725886194029851e-05,
+      "loss": 0.0004,
+      "step": 23508
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 4.725839552238806e-05,
+      "loss": 0.0012,
+      "step": 23512
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 4.725792910447762e-05,
+      "loss": 0.0006,
+      "step": 23516
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 4.7257462686567165e-05,
+      "loss": 0.0019,
+      "step": 23520
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 4.7256996268656714e-05,
+      "loss": 0.0013,
+      "step": 23524
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 4.7256529850746275e-05,
+      "loss": 0.0033,
+      "step": 23528
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 4.7256063432835823e-05,
+      "loss": 0.0003,
+      "step": 23532
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 4.725559701492537e-05,
+      "loss": 0.0004,
+      "step": 23536
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 4.7255130597014926e-05,
+      "loss": 0.001,
+      "step": 23540
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 4.725466417910448e-05,
+      "loss": 0.0015,
+      "step": 23544
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 4.725419776119403e-05,
+      "loss": 0.0011,
+      "step": 23548
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 4.7253731343283584e-05,
+      "loss": 0.0004,
+      "step": 23552
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 4.725326492537314e-05,
+      "loss": 0.0005,
+      "step": 23556
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 4.725279850746269e-05,
+      "loss": 0.0012,
+      "step": 23560
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 4.725233208955224e-05,
+      "loss": 0.003,
+      "step": 23564
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 4.725186567164179e-05,
+      "loss": 0.0087,
+      "step": 23568
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 4.7251399253731345e-05,
+      "loss": 0.0005,
+      "step": 23572
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 4.72509328358209e-05,
+      "loss": 0.0026,
+      "step": 23576
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 4.725046641791045e-05,
+      "loss": 0.0004,
+      "step": 23580
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 4.7249999999999997e-05,
+      "loss": 0.0038,
+      "step": 23584
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 4.724953358208956e-05,
+      "loss": 0.0006,
+      "step": 23588
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 4.7249067164179106e-05,
+      "loss": 0.0012,
+      "step": 23592
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 4.7248600746268654e-05,
+      "loss": 0.0009,
+      "step": 23596
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 4.724813432835821e-05,
+      "loss": 0.0002,
+      "step": 23600
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 4.7247667910447764e-05,
+      "loss": 0.0009,
+      "step": 23604
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 4.724720149253731e-05,
+      "loss": 0.0002,
+      "step": 23608
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 4.724673507462687e-05,
+      "loss": 0.0005,
+      "step": 23612
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 4.724626865671642e-05,
+      "loss": 0.0022,
+      "step": 23616
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 4.724580223880598e-05,
+      "loss": 0.0006,
+      "step": 23620
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 4.7245335820895525e-05,
+      "loss": 0.0001,
+      "step": 23624
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 4.7244869402985073e-05,
+      "loss": 0.0013,
+      "step": 23628
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 4.724440298507463e-05,
+      "loss": 0.0006,
+      "step": 23632
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 4.724393656716418e-05,
+      "loss": 0.0001,
+      "step": 23636
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 4.724347014925373e-05,
+      "loss": 0.0021,
+      "step": 23640
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 4.7243003731343286e-05,
+      "loss": 0.001,
+      "step": 23644
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 4.724253731343284e-05,
+      "loss": 0.002,
+      "step": 23648
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 4.724207089552239e-05,
+      "loss": 0.0002,
+      "step": 23652
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 4.7241604477611944e-05,
+      "loss": 0.0001,
+      "step": 23656
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 4.724113805970149e-05,
+      "loss": 0.0002,
+      "step": 23660
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 4.724067164179105e-05,
+      "loss": 0.0004,
+      "step": 23664
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 4.72402052238806e-05,
+      "loss": 0.0013,
+      "step": 23668
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 4.723973880597015e-05,
+      "loss": 0.0005,
+      "step": 23672
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 4.7239272388059705e-05,
+      "loss": 0.0027,
+      "step": 23676
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 4.723880597014926e-05,
+      "loss": 0.0012,
+      "step": 23680
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 4.723833955223881e-05,
+      "loss": 0.0015,
+      "step": 23684
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 4.7237873134328356e-05,
+      "loss": 0.003,
+      "step": 23688
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 4.723740671641791e-05,
+      "loss": 0.0004,
+      "step": 23692
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 4.7236940298507466e-05,
+      "loss": 0.0002,
+      "step": 23696
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 4.7236473880597014e-05,
+      "loss": 0.0003,
+      "step": 23700
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 4.723600746268657e-05,
+      "loss": 0.0019,
+      "step": 23704
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 4.7235541044776124e-05,
+      "loss": 0.0005,
+      "step": 23708
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 4.723507462686567e-05,
+      "loss": 0.0004,
+      "step": 23712
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 4.723460820895523e-05,
+      "loss": 0.0002,
+      "step": 23716
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 4.7234141791044775e-05,
+      "loss": 0.0002,
+      "step": 23720
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 4.723367537313433e-05,
+      "loss": 0.0002,
+      "step": 23724
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 4.7233208955223885e-05,
+      "loss": 0.0082,
+      "step": 23728
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 4.723274253731343e-05,
+      "loss": 0.0003,
+      "step": 23732
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 4.723227611940299e-05,
+      "loss": 0.0001,
+      "step": 23736
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 4.723180970149254e-05,
+      "loss": 0.0039,
+      "step": 23740
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 4.723134328358209e-05,
+      "loss": 0.0001,
+      "step": 23744
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 4.723087686567164e-05,
+      "loss": 0.0098,
+      "step": 23748
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 4.7230410447761194e-05,
+      "loss": 0.0001,
+      "step": 23752
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 4.722994402985075e-05,
+      "loss": 0.0003,
+      "step": 23756
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 4.72294776119403e-05,
+      "loss": 0.001,
+      "step": 23760
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 4.722901119402985e-05,
+      "loss": 0.0,
+      "step": 23764
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 4.722854477611941e-05,
+      "loss": 0.0005,
+      "step": 23768
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 4.7228078358208955e-05,
+      "loss": 0.0013,
+      "step": 23772
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 4.722761194029851e-05,
+      "loss": 0.0003,
+      "step": 23776
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 4.722714552238806e-05,
+      "loss": 0.0004,
+      "step": 23780
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 4.722667910447762e-05,
+      "loss": 0.0005,
+      "step": 23784
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 4.722621268656717e-05,
+      "loss": 0.0001,
+      "step": 23788
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 4.7225746268656716e-05,
+      "loss": 0.0005,
+      "step": 23792
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 4.722527985074627e-05,
+      "loss": 0.0002,
+      "step": 23796
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 4.7224813432835826e-05,
+      "loss": 0.0021,
+      "step": 23800
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 4.7224347014925374e-05,
+      "loss": 0.0011,
+      "step": 23804
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 4.722388059701493e-05,
+      "loss": 0.001,
+      "step": 23808
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 4.722341417910448e-05,
+      "loss": 0.0004,
+      "step": 23812
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 4.722294776119403e-05,
+      "loss": 0.0082,
+      "step": 23816
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 4.722248134328359e-05,
+      "loss": 0.0006,
+      "step": 23820
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 4.7222014925373135e-05,
+      "loss": 0.0021,
+      "step": 23824
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 4.722154850746269e-05,
+      "loss": 0.0002,
+      "step": 23828
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 4.7221082089552245e-05,
+      "loss": 0.0003,
+      "step": 23832
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 4.722061567164179e-05,
+      "loss": 0.0034,
+      "step": 23836
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 4.722014925373134e-05,
+      "loss": 0.0007,
+      "step": 23840
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 4.72196828358209e-05,
+      "loss": 0.0041,
+      "step": 23844
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 4.721921641791045e-05,
+      "loss": 0.0001,
+      "step": 23848
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 4.721875e-05,
+      "loss": 0.001,
+      "step": 23852
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 4.7218283582089554e-05,
+      "loss": 0.0001,
+      "step": 23856
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 4.721781716417911e-05,
+      "loss": 0.001,
+      "step": 23860
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 4.721735074626866e-05,
+      "loss": 0.0025,
+      "step": 23864
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 4.721688432835821e-05,
+      "loss": 0.0001,
+      "step": 23868
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 4.721641791044776e-05,
+      "loss": 0.0023,
+      "step": 23872
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 4.7215951492537315e-05,
+      "loss": 0.0054,
+      "step": 23876
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 4.721548507462687e-05,
+      "loss": 0.0035,
+      "step": 23880
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 4.721501865671642e-05,
+      "loss": 0.0015,
+      "step": 23884
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 4.721455223880597e-05,
+      "loss": 0.0002,
+      "step": 23888
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 4.721408582089553e-05,
+      "loss": 0.0007,
+      "step": 23892
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 4.7213619402985076e-05,
+      "loss": 0.0007,
+      "step": 23896
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 4.7213152985074624e-05,
+      "loss": 0.0122,
+      "step": 23900
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 4.7212686567164186e-05,
+      "loss": 0.0012,
+      "step": 23904
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 4.7212220149253734e-05,
+      "loss": 0.0023,
+      "step": 23908
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 4.721175373134328e-05,
+      "loss": 0.0001,
+      "step": 23912
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 4.721128731343284e-05,
+      "loss": 0.0003,
+      "step": 23916
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 4.721082089552239e-05,
+      "loss": 0.0008,
+      "step": 23920
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 4.721035447761194e-05,
+      "loss": 0.0004,
+      "step": 23924
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 4.7209888059701495e-05,
+      "loss": 0.0005,
+      "step": 23928
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 4.720942164179104e-05,
+      "loss": 0.0058,
+      "step": 23932
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 4.72089552238806e-05,
+      "loss": 0.0006,
+      "step": 23936
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 4.720848880597015e-05,
+      "loss": 0.0009,
+      "step": 23940
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 4.72080223880597e-05,
+      "loss": 0.0004,
+      "step": 23944
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 4.7207555970149256e-05,
+      "loss": 0.0005,
+      "step": 23948
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 4.720708955223881e-05,
+      "loss": 0.0012,
+      "step": 23952
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 4.720662313432836e-05,
+      "loss": 0.0026,
+      "step": 23956
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 4.7206156716417914e-05,
+      "loss": 0.0011,
+      "step": 23960
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 4.720569029850747e-05,
+      "loss": 0.0041,
+      "step": 23964
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 4.720522388059702e-05,
+      "loss": 0.0008,
+      "step": 23968
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 4.720475746268657e-05,
+      "loss": 0.0008,
+      "step": 23972
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 4.720429104477612e-05,
+      "loss": 0.0001,
+      "step": 23976
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 4.7203824626865675e-05,
+      "loss": 0.007,
+      "step": 23980
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 4.720335820895523e-05,
+      "loss": 0.0003,
+      "step": 23984
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 4.720289179104478e-05,
+      "loss": 0.0006,
+      "step": 23988
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 4.7202425373134326e-05,
+      "loss": 0.0001,
+      "step": 23992
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 4.720195895522389e-05,
+      "loss": 0.0005,
+      "step": 23996
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 4.7201492537313436e-05,
+      "loss": 0.0002,
+      "step": 24000
+    },
+    {
+      "epoch": 22.39,
+      "eval_exact_match": 0.7214700193423598,
+      "eval_exec": 0.7601547388781431,
+      "eval_loss": 0.3911060094833374,
+      "eval_runtime": 1226.8058,
+      "eval_samples_per_second": 0.843,
+      "step": 24000
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 4.7201026119402984e-05,
+      "loss": 0.0042,
+      "step": 24004
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 4.720055970149254e-05,
+      "loss": 0.0012,
+      "step": 24008
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 4.7200093283582094e-05,
+      "loss": 0.0001,
+      "step": 24012
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 4.719962686567164e-05,
+      "loss": 0.0004,
+      "step": 24016
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 4.71991604477612e-05,
+      "loss": 0.0045,
+      "step": 24020
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 4.719869402985075e-05,
+      "loss": 0.0001,
+      "step": 24024
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 4.71982276119403e-05,
+      "loss": 0.0062,
+      "step": 24028
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 4.7197761194029855e-05,
+      "loss": 0.0037,
+      "step": 24032
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 4.71972947761194e-05,
+      "loss": 0.0016,
+      "step": 24036
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 4.719682835820896e-05,
+      "loss": 0.0004,
+      "step": 24040
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 4.719636194029851e-05,
+      "loss": 0.001,
+      "step": 24044
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 4.719589552238806e-05,
+      "loss": 0.0024,
+      "step": 24048
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 4.719542910447761e-05,
+      "loss": 0.0011,
+      "step": 24052
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 4.719496268656717e-05,
+      "loss": 0.0002,
+      "step": 24056
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 4.719449626865672e-05,
+      "loss": 0.0004,
+      "step": 24060
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 4.719402985074627e-05,
+      "loss": 0.0029,
+      "step": 24064
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 4.719356343283582e-05,
+      "loss": 0.0002,
+      "step": 24068
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 4.719309701492538e-05,
+      "loss": 0.0043,
+      "step": 24072
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 4.7192630597014925e-05,
+      "loss": 0.0003,
+      "step": 24076
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 4.719216417910448e-05,
+      "loss": 0.0002,
+      "step": 24080
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 4.7191697761194035e-05,
+      "loss": 0.0014,
+      "step": 24084
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 4.719123134328358e-05,
+      "loss": 0.0008,
+      "step": 24088
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 4.719076492537314e-05,
+      "loss": 0.0002,
+      "step": 24092
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 4.7190298507462686e-05,
+      "loss": 0.002,
+      "step": 24096
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 4.718983208955224e-05,
+      "loss": 0.0002,
+      "step": 24100
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 4.7189365671641796e-05,
+      "loss": 0.0012,
+      "step": 24104
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 4.7188899253731344e-05,
+      "loss": 0.002,
+      "step": 24108
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 4.71884328358209e-05,
+      "loss": 0.0002,
+      "step": 24112
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 4.7187966417910454e-05,
+      "loss": 0.0003,
+      "step": 24116
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 4.71875e-05,
+      "loss": 0.0018,
+      "step": 24120
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 4.718703358208956e-05,
+      "loss": 0.0011,
+      "step": 24124
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 4.7186567164179105e-05,
+      "loss": 0.0014,
+      "step": 24128
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 4.718610074626866e-05,
+      "loss": 0.0001,
+      "step": 24132
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 4.7185634328358215e-05,
+      "loss": 0.0001,
+      "step": 24136
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 4.718516791044776e-05,
+      "loss": 0.0007,
+      "step": 24140
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 4.718470149253732e-05,
+      "loss": 0.002,
+      "step": 24144
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 4.718423507462687e-05,
+      "loss": 0.0022,
+      "step": 24148
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 4.718376865671642e-05,
+      "loss": 0.0014,
+      "step": 24152
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 4.718330223880597e-05,
+      "loss": 0.0003,
+      "step": 24156
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 4.7182835820895524e-05,
+      "loss": 0.0009,
+      "step": 24160
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 4.718236940298508e-05,
+      "loss": 0.0016,
+      "step": 24164
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 4.718190298507463e-05,
+      "loss": 0.0,
+      "step": 24168
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 4.718143656716418e-05,
+      "loss": 0.0005,
+      "step": 24172
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 4.718097014925374e-05,
+      "loss": 0.0004,
+      "step": 24176
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 4.7180503731343285e-05,
+      "loss": 0.0014,
+      "step": 24180
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 4.718003731343284e-05,
+      "loss": 0.0001,
+      "step": 24184
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 4.717957089552239e-05,
+      "loss": 0.001,
+      "step": 24188
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 4.717910447761194e-05,
+      "loss": 0.0002,
+      "step": 24192
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 4.71786380597015e-05,
+      "loss": 0.0005,
+      "step": 24196
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 4.7178171641791046e-05,
+      "loss": 0.0005,
+      "step": 24200
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 4.7177705223880594e-05,
+      "loss": 0.0023,
+      "step": 24204
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 4.7177238805970156e-05,
+      "loss": 0.0051,
+      "step": 24208
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 4.7176772388059704e-05,
+      "loss": 0.0009,
+      "step": 24212
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 4.717630597014925e-05,
+      "loss": 0.0091,
+      "step": 24216
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 4.717583955223881e-05,
+      "loss": 0.0002,
+      "step": 24220
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 4.717537313432836e-05,
+      "loss": 0.0014,
+      "step": 24224
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 4.717490671641791e-05,
+      "loss": 0.0001,
+      "step": 24228
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 4.7174440298507465e-05,
+      "loss": 0.0034,
+      "step": 24232
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 4.717397388059702e-05,
+      "loss": 0.0047,
+      "step": 24236
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 4.717350746268657e-05,
+      "loss": 0.0017,
+      "step": 24240
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 4.717304104477612e-05,
+      "loss": 0.0013,
+      "step": 24244
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 4.717257462686567e-05,
+      "loss": 0.0008,
+      "step": 24248
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 4.7172108208955226e-05,
+      "loss": 0.0029,
+      "step": 24252
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 4.717164179104478e-05,
+      "loss": 0.0023,
+      "step": 24256
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 4.717117537313433e-05,
+      "loss": 0.0011,
+      "step": 24260
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 4.717070895522388e-05,
+      "loss": 0.0121,
+      "step": 24264
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 4.717024253731344e-05,
+      "loss": 0.0003,
+      "step": 24268
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 4.716977611940299e-05,
+      "loss": 0.0011,
+      "step": 24272
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 4.716930970149254e-05,
+      "loss": 0.0001,
+      "step": 24276
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 4.716884328358209e-05,
+      "loss": 0.0002,
+      "step": 24280
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 4.7168376865671645e-05,
+      "loss": 0.0005,
+      "step": 24284
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 4.71679104477612e-05,
+      "loss": 0.0006,
+      "step": 24288
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 4.716744402985075e-05,
+      "loss": 0.0003,
+      "step": 24292
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 4.71669776119403e-05,
+      "loss": 0.0103,
+      "step": 24296
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 4.716651119402986e-05,
+      "loss": 0.0001,
+      "step": 24300
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 4.7166044776119406e-05,
+      "loss": 0.0017,
+      "step": 24304
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 4.7165578358208954e-05,
+      "loss": 0.0011,
+      "step": 24308
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 4.716511194029851e-05,
+      "loss": 0.0001,
+      "step": 24312
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 4.7164645522388064e-05,
+      "loss": 0.0017,
+      "step": 24316
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 4.716417910447761e-05,
+      "loss": 0.0002,
+      "step": 24320
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 4.7163712686567167e-05,
+      "loss": 0.0004,
+      "step": 24324
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 4.716324626865672e-05,
+      "loss": 0.0014,
+      "step": 24328
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 4.716277985074627e-05,
+      "loss": 0.0087,
+      "step": 24332
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 4.7162313432835825e-05,
+      "loss": 0.0003,
+      "step": 24336
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 4.716184701492537e-05,
+      "loss": 0.0002,
+      "step": 24340
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 4.716138059701493e-05,
+      "loss": 0.0077,
+      "step": 24344
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 4.716091417910448e-05,
+      "loss": 0.0058,
+      "step": 24348
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 4.716044776119403e-05,
+      "loss": 0.0052,
+      "step": 24352
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 4.7159981343283585e-05,
+      "loss": 0.0028,
+      "step": 24356
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 4.715951492537314e-05,
+      "loss": 0.0002,
+      "step": 24360
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 4.715904850746269e-05,
+      "loss": 0.0023,
+      "step": 24364
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 4.715858208955224e-05,
+      "loss": 0.0002,
+      "step": 24368
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 4.715811567164179e-05,
+      "loss": 0.0004,
+      "step": 24372
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 4.7157649253731346e-05,
+      "loss": 0.0006,
+      "step": 24376
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 4.7157182835820895e-05,
+      "loss": 0.0004,
+      "step": 24380
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 4.715671641791045e-05,
+      "loss": 0.0004,
+      "step": 24384
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 4.7156250000000004e-05,
+      "loss": 0.0012,
+      "step": 24388
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 4.715578358208955e-05,
+      "loss": 0.0012,
+      "step": 24392
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 4.715531716417911e-05,
+      "loss": 0.0009,
+      "step": 24396
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 4.7154850746268656e-05,
+      "loss": 0.0009,
+      "step": 24400
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 4.715438432835821e-05,
+      "loss": 0.0004,
+      "step": 24404
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 4.7153917910447765e-05,
+      "loss": 0.0004,
+      "step": 24408
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 4.7153451492537314e-05,
+      "loss": 0.001,
+      "step": 24412
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 4.715298507462687e-05,
+      "loss": 0.0005,
+      "step": 24416
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 4.715251865671642e-05,
+      "loss": 0.0003,
+      "step": 24420
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 4.715205223880597e-05,
+      "loss": 0.0019,
+      "step": 24424
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 4.715158582089552e-05,
+      "loss": 0.0008,
+      "step": 24428
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 4.7151119402985075e-05,
+      "loss": 0.0007,
+      "step": 24432
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 4.715065298507463e-05,
+      "loss": 0.0027,
+      "step": 24436
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 4.7150186567164184e-05,
+      "loss": 0.0009,
+      "step": 24440
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 4.714972014925373e-05,
+      "loss": 0.0034,
+      "step": 24444
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 4.714925373134329e-05,
+      "loss": 0.0026,
+      "step": 24448
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 4.714878731343284e-05,
+      "loss": 0.001,
+      "step": 24452
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 4.714832089552239e-05,
+      "loss": 0.0003,
+      "step": 24456
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 4.714785447761194e-05,
+      "loss": 0.0001,
+      "step": 24460
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 4.71473880597015e-05,
+      "loss": 0.0009,
+      "step": 24464
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 4.714692164179105e-05,
+      "loss": 0.0019,
+      "step": 24468
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 4.7146455223880596e-05,
+      "loss": 0.0003,
+      "step": 24472
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 4.714598880597015e-05,
+      "loss": 0.0001,
+      "step": 24476
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 4.7145522388059706e-05,
+      "loss": 0.0003,
+      "step": 24480
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 4.7145055970149254e-05,
+      "loss": 0.0013,
+      "step": 24484
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 4.714458955223881e-05,
+      "loss": 0.0095,
+      "step": 24488
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 4.714412313432836e-05,
+      "loss": 0.0002,
+      "step": 24492
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 4.714365671641791e-05,
+      "loss": 0.0009,
+      "step": 24496
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 4.714319029850747e-05,
+      "loss": 0.0002,
+      "step": 24500
+    },
+    {
+      "epoch": 22.85,
+      "eval_exact_match": 0.7292069632495164,
+      "eval_exec": 0.7572533849129593,
+      "eval_loss": 0.35489827394485474,
+      "eval_runtime": 1237.9317,
+      "eval_samples_per_second": 0.835,
+      "step": 24500
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 4.7142723880597015e-05,
+      "loss": 0.001,
+      "step": 24504
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 4.714225746268657e-05,
+      "loss": 0.0014,
+      "step": 24508
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 4.7141791044776125e-05,
+      "loss": 0.0068,
+      "step": 24512
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 4.714132462686567e-05,
+      "loss": 0.0008,
+      "step": 24516
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 4.714085820895522e-05,
+      "loss": 0.0002,
+      "step": 24520
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 4.714039179104478e-05,
+      "loss": 0.0031,
+      "step": 24524
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 4.713992537313433e-05,
+      "loss": 0.0003,
+      "step": 24528
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 4.713945895522388e-05,
+      "loss": 0.0001,
+      "step": 24532
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 4.7138992537313434e-05,
+      "loss": 0.0041,
+      "step": 24536
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 4.713852611940299e-05,
+      "loss": 0.0003,
+      "step": 24540
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 4.713805970149254e-05,
+      "loss": 0.0001,
+      "step": 24544
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 4.713759328358209e-05,
+      "loss": 0.0058,
+      "step": 24548
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 4.713712686567164e-05,
+      "loss": 0.0006,
+      "step": 24552
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 4.7136660447761195e-05,
+      "loss": 0.0001,
+      "step": 24556
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 4.713619402985075e-05,
+      "loss": 0.0011,
+      "step": 24560
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 4.71357276119403e-05,
+      "loss": 0.0019,
+      "step": 24564
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 4.713526119402985e-05,
+      "loss": 0.0006,
+      "step": 24568
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 4.713479477611941e-05,
+      "loss": 0.0015,
+      "step": 24572
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 4.7134328358208956e-05,
+      "loss": 0.0004,
+      "step": 24576
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 4.7133861940298504e-05,
+      "loss": 0.0049,
+      "step": 24580
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 4.7133395522388066e-05,
+      "loss": 0.0001,
+      "step": 24584
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 4.7132929104477614e-05,
+      "loss": 0.0026,
+      "step": 24588
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 4.713246268656716e-05,
+      "loss": 0.0007,
+      "step": 24592
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 4.713199626865672e-05,
+      "loss": 0.0002,
+      "step": 24596
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 4.713152985074627e-05,
+      "loss": 0.0001,
+      "step": 24600
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 4.713106343283583e-05,
+      "loss": 0.0002,
+      "step": 24604
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 4.7130597014925375e-05,
+      "loss": 0.0204,
+      "step": 24608
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 4.7130130597014923e-05,
+      "loss": 0.0018,
+      "step": 24612
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 4.7129664179104485e-05,
+      "loss": 0.0037,
+      "step": 24616
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 4.712919776119403e-05,
+      "loss": 0.0002,
+      "step": 24620
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 4.712873134328358e-05,
+      "loss": 0.0006,
+      "step": 24624
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 4.7128264925373136e-05,
+      "loss": 0.0034,
+      "step": 24628
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 4.712779850746269e-05,
+      "loss": 0.0018,
+      "step": 24632
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 4.712733208955224e-05,
+      "loss": 0.0031,
+      "step": 24636
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 4.7126865671641794e-05,
+      "loss": 0.0002,
+      "step": 24640
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 4.712639925373135e-05,
+      "loss": 0.002,
+      "step": 24644
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 4.71259328358209e-05,
+      "loss": 0.0027,
+      "step": 24648
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 4.712546641791045e-05,
+      "loss": 0.0018,
+      "step": 24652
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 4.7125e-05,
+      "loss": 0.0015,
+      "step": 24656
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 4.7124533582089555e-05,
+      "loss": 0.0004,
+      "step": 24660
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 4.712406716417911e-05,
+      "loss": 0.0035,
+      "step": 24664
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 4.712360074626866e-05,
+      "loss": 0.0002,
+      "step": 24668
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 4.7123134328358206e-05,
+      "loss": 0.0024,
+      "step": 24672
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 4.712266791044777e-05,
+      "loss": 0.0001,
+      "step": 24676
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 4.7122201492537316e-05,
+      "loss": 0.0006,
+      "step": 24680
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 4.7121735074626864e-05,
+      "loss": 0.0048,
+      "step": 24684
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 4.712126865671642e-05,
+      "loss": 0.0005,
+      "step": 24688
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 4.7120802238805974e-05,
+      "loss": 0.0001,
+      "step": 24692
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 4.712033582089552e-05,
+      "loss": 0.0008,
+      "step": 24696
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 4.711986940298508e-05,
+      "loss": 0.0006,
+      "step": 24700
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 4.711940298507463e-05,
+      "loss": 0.0002,
+      "step": 24704
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 4.711893656716418e-05,
+      "loss": 0.0008,
+      "step": 24708
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 4.7118470149253735e-05,
+      "loss": 0.0002,
+      "step": 24712
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 4.711800373134328e-05,
+      "loss": 0.0002,
+      "step": 24716
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 4.711753731343284e-05,
+      "loss": 0.0002,
+      "step": 24720
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 4.711707089552239e-05,
+      "loss": 0.0005,
+      "step": 24724
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 4.711660447761194e-05,
+      "loss": 0.0009,
+      "step": 24728
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 4.711613805970149e-05,
+      "loss": 0.0001,
+      "step": 24732
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 4.711567164179105e-05,
+      "loss": 0.0002,
+      "step": 24736
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 4.71152052238806e-05,
+      "loss": 0.0012,
+      "step": 24740
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 4.711473880597015e-05,
+      "loss": 0.0002,
+      "step": 24744
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 4.71142723880597e-05,
+      "loss": 0.0017,
+      "step": 24748
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 4.711380597014926e-05,
+      "loss": 0.0016,
+      "step": 24752
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 4.7113339552238805e-05,
+      "loss": 0.0005,
+      "step": 24756
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 4.711287313432836e-05,
+      "loss": 0.0002,
+      "step": 24760
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 4.7112406716417915e-05,
+      "loss": 0.0002,
+      "step": 24764
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 4.711194029850747e-05,
+      "loss": 0.0092,
+      "step": 24768
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 4.711147388059702e-05,
+      "loss": 0.0022,
+      "step": 24772
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 4.7111007462686566e-05,
+      "loss": 0.0004,
+      "step": 24776
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 4.711054104477612e-05,
+      "loss": 0.0006,
+      "step": 24780
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 4.7110074626865676e-05,
+      "loss": 0.0004,
+      "step": 24784
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 4.7109608208955224e-05,
+      "loss": 0.001,
+      "step": 24788
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 4.710914179104478e-05,
+      "loss": 0.0002,
+      "step": 24792
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 4.7108675373134334e-05,
+      "loss": 0.0017,
+      "step": 24796
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 4.710820895522388e-05,
+      "loss": 0.0001,
+      "step": 24800
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 4.710774253731344e-05,
+      "loss": 0.0017,
+      "step": 24804
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 4.7107276119402985e-05,
+      "loss": 0.0002,
+      "step": 24808
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 4.710680970149254e-05,
+      "loss": 0.0002,
+      "step": 24812
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 4.7106343283582095e-05,
+      "loss": 0.0,
+      "step": 24816
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 4.710587686567164e-05,
+      "loss": 0.0,
+      "step": 24820
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 4.71054104477612e-05,
+      "loss": 0.0001,
+      "step": 24824
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 4.710494402985075e-05,
+      "loss": 0.0001,
+      "step": 24828
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 4.71044776119403e-05,
+      "loss": 0.0011,
+      "step": 24832
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 4.710401119402985e-05,
+      "loss": 0.0004,
+      "step": 24836
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 4.7103544776119404e-05,
+      "loss": 0.0014,
+      "step": 24840
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 4.710307835820896e-05,
+      "loss": 0.0002,
+      "step": 24844
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 4.710261194029851e-05,
+      "loss": 0.001,
+      "step": 24848
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 4.710214552238806e-05,
+      "loss": 0.0003,
+      "step": 24852
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 4.710167910447762e-05,
+      "loss": 0.0003,
+      "step": 24856
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 4.7101212686567165e-05,
+      "loss": 0.0049,
+      "step": 24860
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 4.710074626865672e-05,
+      "loss": 0.0021,
+      "step": 24864
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 4.710027985074627e-05,
+      "loss": 0.0002,
+      "step": 24868
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 4.709981343283582e-05,
+      "loss": 0.0063,
+      "step": 24872
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 4.709934701492538e-05,
+      "loss": 0.0003,
+      "step": 24876
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 4.7098880597014926e-05,
+      "loss": 0.0009,
+      "step": 24880
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 4.7098414179104474e-05,
+      "loss": 0.0044,
+      "step": 24884
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 4.7097947761194036e-05,
+      "loss": 0.0003,
+      "step": 24888
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 4.7097481343283584e-05,
+      "loss": 0.0002,
+      "step": 24892
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 4.709701492537313e-05,
+      "loss": 0.0008,
+      "step": 24896
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 4.709654850746269e-05,
+      "loss": 0.0021,
+      "step": 24900
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 4.709608208955224e-05,
+      "loss": 0.0001,
+      "step": 24904
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 4.709561567164179e-05,
+      "loss": 0.0002,
+      "step": 24908
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 4.7095149253731345e-05,
+      "loss": 0.0008,
+      "step": 24912
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 4.70946828358209e-05,
+      "loss": 0.0011,
+      "step": 24916
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 4.709421641791045e-05,
+      "loss": 0.0003,
+      "step": 24920
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 4.709375e-05,
+      "loss": 0.0007,
+      "step": 24924
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 4.709328358208955e-05,
+      "loss": 0.0031,
+      "step": 24928
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 4.709281716417911e-05,
+      "loss": 0.008,
+      "step": 24932
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 4.709235074626866e-05,
+      "loss": 0.0002,
+      "step": 24936
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 4.709188432835821e-05,
+      "loss": 0.0032,
+      "step": 24940
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 4.7091417910447764e-05,
+      "loss": 0.0001,
+      "step": 24944
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 4.709095149253732e-05,
+      "loss": 0.0024,
+      "step": 24948
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 4.709048507462687e-05,
+      "loss": 0.0003,
+      "step": 24952
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 4.709001865671642e-05,
+      "loss": 0.0001,
+      "step": 24956
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 4.708955223880597e-05,
+      "loss": 0.0003,
+      "step": 24960
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 4.7089085820895525e-05,
+      "loss": 0.0003,
+      "step": 24964
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 4.708861940298508e-05,
+      "loss": 0.0076,
+      "step": 24968
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 4.708815298507463e-05,
+      "loss": 0.0044,
+      "step": 24972
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 4.708768656716418e-05,
+      "loss": 0.0026,
+      "step": 24976
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 4.708722014925374e-05,
+      "loss": 0.0001,
+      "step": 24980
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 4.7086753731343286e-05,
+      "loss": 0.0015,
+      "step": 24984
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 4.7086287313432834e-05,
+      "loss": 0.0001,
+      "step": 24988
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 4.708582089552239e-05,
+      "loss": 0.0101,
+      "step": 24992
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 4.7085354477611944e-05,
+      "loss": 0.0002,
+      "step": 24996
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 4.708488805970149e-05,
+      "loss": 0.0001,
+      "step": 25000
+    },
+    {
+      "epoch": 23.32,
+      "eval_exact_match": 0.7263056092843327,
+      "eval_exec": 0.7620889748549323,
+      "eval_loss": 0.3771933615207672,
+      "eval_runtime": 1213.1376,
+      "eval_samples_per_second": 0.852,
+      "step": 25000
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 4.708442164179105e-05,
+      "loss": 0.0006,
+      "step": 25004
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 4.70839552238806e-05,
+      "loss": 0.0035,
+      "step": 25008
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 4.708348880597015e-05,
+      "loss": 0.0,
+      "step": 25012
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 4.7083022388059705e-05,
+      "loss": 0.0009,
+      "step": 25016
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 4.708255597014925e-05,
+      "loss": 0.0004,
+      "step": 25020
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 4.708208955223881e-05,
+      "loss": 0.0027,
+      "step": 25024
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 4.708162313432836e-05,
+      "loss": 0.0013,
+      "step": 25028
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 4.708115671641791e-05,
+      "loss": 0.0016,
+      "step": 25032
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 4.7080690298507466e-05,
+      "loss": 0.0002,
+      "step": 25036
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 4.708022388059702e-05,
+      "loss": 0.0001,
+      "step": 25040
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 4.707975746268657e-05,
+      "loss": 0.0003,
+      "step": 25044
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 4.707929104477612e-05,
+      "loss": 0.0005,
+      "step": 25048
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 4.707882462686567e-05,
+      "loss": 0.0006,
+      "step": 25052
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 4.707835820895523e-05,
+      "loss": 0.008,
+      "step": 25056
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 4.7077891791044775e-05,
+      "loss": 0.0058,
+      "step": 25060
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 4.707742537313433e-05,
+      "loss": 0.0025,
+      "step": 25064
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 4.7076958955223885e-05,
+      "loss": 0.0001,
+      "step": 25068
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 4.707649253731343e-05,
+      "loss": 0.0002,
+      "step": 25072
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 4.707602611940299e-05,
+      "loss": 0.0004,
+      "step": 25076
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 4.7075559701492536e-05,
+      "loss": 0.0101,
+      "step": 25080
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 4.707509328358209e-05,
+      "loss": 0.0012,
+      "step": 25084
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 4.7074626865671646e-05,
+      "loss": 0.0007,
+      "step": 25088
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 4.7074160447761194e-05,
+      "loss": 0.0014,
+      "step": 25092
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 4.707369402985075e-05,
+      "loss": 0.0009,
+      "step": 25096
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 4.7073227611940304e-05,
+      "loss": 0.0004,
+      "step": 25100
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 4.707276119402985e-05,
+      "loss": 0.0006,
+      "step": 25104
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 4.707229477611941e-05,
+      "loss": 0.0011,
+      "step": 25108
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 4.7071828358208955e-05,
+      "loss": 0.0,
+      "step": 25112
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 4.707136194029851e-05,
+      "loss": 0.0004,
+      "step": 25116
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 4.7070895522388065e-05,
+      "loss": 0.0003,
+      "step": 25120
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 4.707042910447761e-05,
+      "loss": 0.0005,
+      "step": 25124
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 4.706996268656717e-05,
+      "loss": 0.0012,
+      "step": 25128
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 4.706949626865672e-05,
+      "loss": 0.0,
+      "step": 25132
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 4.706902985074627e-05,
+      "loss": 0.0012,
+      "step": 25136
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 4.706856343283582e-05,
+      "loss": 0.0002,
+      "step": 25140
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 4.706809701492538e-05,
+      "loss": 0.0078,
+      "step": 25144
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 4.706763059701493e-05,
+      "loss": 0.0002,
+      "step": 25148
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 4.706716417910448e-05,
+      "loss": 0.0062,
+      "step": 25152
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 4.706669776119403e-05,
+      "loss": 0.0002,
+      "step": 25156
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 4.7066231343283587e-05,
+      "loss": 0.0002,
+      "step": 25160
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 4.7065764925373135e-05,
+      "loss": 0.003,
+      "step": 25164
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 4.706529850746269e-05,
+      "loss": 0.0006,
+      "step": 25168
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 4.706483208955224e-05,
+      "loss": 0.0003,
+      "step": 25172
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 4.706436567164179e-05,
+      "loss": 0.0007,
+      "step": 25176
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 4.706389925373135e-05,
+      "loss": 0.0002,
+      "step": 25180
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 4.7063432835820896e-05,
+      "loss": 0.0005,
+      "step": 25184
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 4.706296641791045e-05,
+      "loss": 0.0004,
+      "step": 25188
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 4.7062500000000006e-05,
+      "loss": 0.0025,
+      "step": 25192
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 4.7062033582089554e-05,
+      "loss": 0.0001,
+      "step": 25196
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 4.70615671641791e-05,
+      "loss": 0.0006,
+      "step": 25200
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 4.7061100746268663e-05,
+      "loss": 0.0012,
+      "step": 25204
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 4.706063432835821e-05,
+      "loss": 0.0002,
+      "step": 25208
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 4.706016791044776e-05,
+      "loss": 0.0003,
+      "step": 25212
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 4.7059701492537315e-05,
+      "loss": 0.0005,
+      "step": 25216
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 4.705923507462687e-05,
+      "loss": 0.0005,
+      "step": 25220
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 4.705876865671642e-05,
+      "loss": 0.0001,
+      "step": 25224
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 4.705830223880597e-05,
+      "loss": 0.0005,
+      "step": 25228
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 4.705783582089552e-05,
+      "loss": 0.0128,
+      "step": 25232
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 4.7057369402985076e-05,
+      "loss": 0.0123,
+      "step": 25236
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 4.705690298507463e-05,
+      "loss": 0.0012,
+      "step": 25240
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 4.705643656716418e-05,
+      "loss": 0.0005,
+      "step": 25244
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 4.7055970149253734e-05,
+      "loss": 0.0001,
+      "step": 25248
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 4.705550373134329e-05,
+      "loss": 0.0008,
+      "step": 25252
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 4.705503731343284e-05,
+      "loss": 0.0002,
+      "step": 25256
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 4.705457089552239e-05,
+      "loss": 0.0005,
+      "step": 25260
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 4.7054104477611946e-05,
+      "loss": 0.0,
+      "step": 25264
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 4.7053638059701495e-05,
+      "loss": 0.0,
+      "step": 25268
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 4.705317164179105e-05,
+      "loss": 0.0008,
+      "step": 25272
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 4.70527052238806e-05,
+      "loss": 0.0023,
+      "step": 25276
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 4.705223880597015e-05,
+      "loss": 0.0004,
+      "step": 25280
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 4.705177238805971e-05,
+      "loss": 0.0054,
+      "step": 25284
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 4.7051305970149256e-05,
+      "loss": 0.0,
+      "step": 25288
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 4.7050839552238804e-05,
+      "loss": 0.0004,
+      "step": 25292
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 4.7050373134328365e-05,
+      "loss": 0.0001,
+      "step": 25296
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 4.7049906716417914e-05,
+      "loss": 0.0001,
+      "step": 25300
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 4.704944029850746e-05,
+      "loss": 0.001,
+      "step": 25304
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 4.7048973880597017e-05,
+      "loss": 0.0003,
+      "step": 25308
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 4.704850746268657e-05,
+      "loss": 0.0029,
+      "step": 25312
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 4.704804104477612e-05,
+      "loss": 0.0015,
+      "step": 25316
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 4.7047574626865674e-05,
+      "loss": 0.0019,
+      "step": 25320
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 4.704710820895523e-05,
+      "loss": 0.0018,
+      "step": 25324
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 4.704664179104478e-05,
+      "loss": 0.0,
+      "step": 25328
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 4.704617537313433e-05,
+      "loss": 0.0007,
+      "step": 25332
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 4.704570895522388e-05,
+      "loss": 0.004,
+      "step": 25336
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 4.7045242537313435e-05,
+      "loss": 0.0004,
+      "step": 25340
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 4.704477611940299e-05,
+      "loss": 0.0002,
+      "step": 25344
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 4.704430970149254e-05,
+      "loss": 0.0033,
+      "step": 25348
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 4.704384328358209e-05,
+      "loss": 0.0004,
+      "step": 25352
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 4.704337686567165e-05,
+      "loss": 0.0073,
+      "step": 25356
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 4.7042910447761196e-05,
+      "loss": 0.0004,
+      "step": 25360
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 4.7042444029850745e-05,
+      "loss": 0.0017,
+      "step": 25364
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 4.70419776119403e-05,
+      "loss": 0.0078,
+      "step": 25368
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 4.7041511194029854e-05,
+      "loss": 0.0002,
+      "step": 25372
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 4.70410447761194e-05,
+      "loss": 0.0006,
+      "step": 25376
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 4.704057835820896e-05,
+      "loss": 0.0027,
+      "step": 25380
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 4.704011194029851e-05,
+      "loss": 0.0021,
+      "step": 25384
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 4.703964552238806e-05,
+      "loss": 0.0059,
+      "step": 25388
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 4.7039179104477615e-05,
+      "loss": 0.0002,
+      "step": 25392
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 4.7038712686567164e-05,
+      "loss": 0.0048,
+      "step": 25396
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 4.703824626865672e-05,
+      "loss": 0.0018,
+      "step": 25400
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 4.703777985074627e-05,
+      "loss": 0.0001,
+      "step": 25404
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 4.703731343283582e-05,
+      "loss": 0.0035,
+      "step": 25408
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 4.7036847014925376e-05,
+      "loss": 0.0006,
+      "step": 25412
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 4.703638059701493e-05,
+      "loss": 0.0004,
+      "step": 25416
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 4.703591417910448e-05,
+      "loss": 0.0,
+      "step": 25420
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 4.7035447761194034e-05,
+      "loss": 0.0002,
+      "step": 25424
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 4.703498134328358e-05,
+      "loss": 0.0019,
+      "step": 25428
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 4.703451492537314e-05,
+      "loss": 0.0001,
+      "step": 25432
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 4.703404850746269e-05,
+      "loss": 0.0087,
+      "step": 25436
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 4.703358208955224e-05,
+      "loss": 0.0002,
+      "step": 25440
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 4.7033115671641795e-05,
+      "loss": 0.0046,
+      "step": 25444
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 4.703264925373135e-05,
+      "loss": 0.0002,
+      "step": 25448
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 4.70321828358209e-05,
+      "loss": 0.0007,
+      "step": 25452
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 4.7031716417910446e-05,
+      "loss": 0.0033,
+      "step": 25456
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 4.703125e-05,
+      "loss": 0.0002,
+      "step": 25460
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 4.7030783582089556e-05,
+      "loss": 0.0004,
+      "step": 25464
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 4.7030317164179104e-05,
+      "loss": 0.0019,
+      "step": 25468
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 4.702985074626866e-05,
+      "loss": 0.0056,
+      "step": 25472
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 4.7029384328358214e-05,
+      "loss": 0.0001,
+      "step": 25476
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 4.702891791044776e-05,
+      "loss": 0.0001,
+      "step": 25480
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 4.702845149253732e-05,
+      "loss": 0.0007,
+      "step": 25484
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 4.7027985074626865e-05,
+      "loss": 0.0028,
+      "step": 25488
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 4.702751865671642e-05,
+      "loss": 0.0002,
+      "step": 25492
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 4.7027052238805975e-05,
+      "loss": 0.0005,
+      "step": 25496
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 4.702658582089552e-05,
+      "loss": 0.0029,
+      "step": 25500
+    },
+    {
+      "epoch": 23.79,
+      "eval_exact_match": 0.7369439071566731,
+      "eval_exec": 0.7611218568665378,
+      "eval_loss": 0.3535132110118866,
+      "eval_runtime": 1197.1055,
+      "eval_samples_per_second": 0.864,
+      "step": 25500
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 4.702611940298508e-05,
+      "loss": 0.0041,
+      "step": 25504
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 4.702565298507463e-05,
+      "loss": 0.0005,
+      "step": 25508
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 4.702518656716418e-05,
+      "loss": 0.0014,
+      "step": 25512
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 4.702472014925373e-05,
+      "loss": 0.0017,
+      "step": 25516
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 4.7024253731343284e-05,
+      "loss": 0.0065,
+      "step": 25520
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 4.702378731343284e-05,
+      "loss": 0.0002,
+      "step": 25524
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 4.702332089552239e-05,
+      "loss": 0.0034,
+      "step": 25528
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 4.702285447761194e-05,
+      "loss": 0.0004,
+      "step": 25532
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 4.70223880597015e-05,
+      "loss": 0.0005,
+      "step": 25536
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 4.7021921641791045e-05,
+      "loss": 0.004,
+      "step": 25540
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 4.70214552238806e-05,
+      "loss": 0.002,
+      "step": 25544
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 4.702098880597015e-05,
+      "loss": 0.0006,
+      "step": 25548
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 4.70205223880597e-05,
+      "loss": 0.0009,
+      "step": 25552
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 4.702005597014926e-05,
+      "loss": 0.0043,
+      "step": 25556
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 4.7019589552238806e-05,
+      "loss": 0.0052,
+      "step": 25560
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 4.7019123134328354e-05,
+      "loss": 0.0047,
+      "step": 25564
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 4.7018656716417916e-05,
+      "loss": 0.0028,
+      "step": 25568
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 4.7018190298507464e-05,
+      "loss": 0.0003,
+      "step": 25572
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 4.701772388059702e-05,
+      "loss": 0.0004,
+      "step": 25576
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 4.701725746268657e-05,
+      "loss": 0.0006,
+      "step": 25580
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 4.701679104477612e-05,
+      "loss": 0.0011,
+      "step": 25584
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 4.701632462686568e-05,
+      "loss": 0.0001,
+      "step": 25588
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 4.7015858208955225e-05,
+      "loss": 0.0003,
+      "step": 25592
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 4.701539179104478e-05,
+      "loss": 0.0046,
+      "step": 25596
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 4.7014925373134335e-05,
+      "loss": 0.0064,
+      "step": 25600
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 4.701445895522388e-05,
+      "loss": 0.0011,
+      "step": 25604
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 4.701399253731343e-05,
+      "loss": 0.0002,
+      "step": 25608
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 4.701352611940299e-05,
+      "loss": 0.0006,
+      "step": 25612
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 4.701305970149254e-05,
+      "loss": 0.0022,
+      "step": 25616
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 4.701259328358209e-05,
+      "loss": 0.0024,
+      "step": 25620
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 4.7012126865671644e-05,
+      "loss": 0.0019,
+      "step": 25624
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 4.70116604477612e-05,
+      "loss": 0.0004,
+      "step": 25628
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 4.701119402985075e-05,
+      "loss": 0.0001,
+      "step": 25632
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 4.70107276119403e-05,
+      "loss": 0.0006,
+      "step": 25636
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 4.701026119402985e-05,
+      "loss": 0.0044,
+      "step": 25640
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 4.7009794776119405e-05,
+      "loss": 0.0003,
+      "step": 25644
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 4.700932835820896e-05,
+      "loss": 0.0001,
+      "step": 25648
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 4.700886194029851e-05,
+      "loss": 0.0002,
+      "step": 25652
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 4.700839552238806e-05,
+      "loss": 0.0016,
+      "step": 25656
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 4.700792910447762e-05,
+      "loss": 0.0002,
+      "step": 25660
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 4.7007462686567166e-05,
+      "loss": 0.0004,
+      "step": 25664
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 4.7006996268656714e-05,
+      "loss": 0.0001,
+      "step": 25668
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 4.7006529850746276e-05,
+      "loss": 0.0001,
+      "step": 25672
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 4.7006063432835824e-05,
+      "loss": 0.0003,
+      "step": 25676
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 4.700559701492537e-05,
+      "loss": 0.0012,
+      "step": 25680
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 4.700513059701493e-05,
+      "loss": 0.0006,
+      "step": 25684
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 4.700466417910448e-05,
+      "loss": 0.0028,
+      "step": 25688
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 4.700419776119403e-05,
+      "loss": 0.0014,
+      "step": 25692
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 4.7003731343283585e-05,
+      "loss": 0.0002,
+      "step": 25696
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 4.700326492537313e-05,
+      "loss": 0.0001,
+      "step": 25700
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 4.700279850746269e-05,
+      "loss": 0.0004,
+      "step": 25704
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 4.700233208955224e-05,
+      "loss": 0.0058,
+      "step": 25708
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 4.700186567164179e-05,
+      "loss": 0.0001,
+      "step": 25712
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 4.7001399253731346e-05,
+      "loss": 0.005,
+      "step": 25716
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 4.70009328358209e-05,
+      "loss": 0.0002,
+      "step": 25720
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 4.700046641791045e-05,
+      "loss": 0.0001,
+      "step": 25724
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 4.7e-05,
+      "loss": 0.0021,
+      "step": 25728
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 4.699953358208955e-05,
+      "loss": 0.0013,
+      "step": 25732
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 4.699906716417911e-05,
+      "loss": 0.0007,
+      "step": 25736
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 4.699860074626866e-05,
+      "loss": 0.0076,
+      "step": 25740
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 4.699813432835821e-05,
+      "loss": 0.0079,
+      "step": 25744
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 4.6997667910447765e-05,
+      "loss": 0.0001,
+      "step": 25748
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 4.699720149253732e-05,
+      "loss": 0.0084,
+      "step": 25752
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 4.699673507462687e-05,
+      "loss": 0.0021,
+      "step": 25756
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 4.6996268656716416e-05,
+      "loss": 0.0029,
+      "step": 25760
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 4.699580223880598e-05,
+      "loss": 0.0023,
+      "step": 25764
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 4.6995335820895526e-05,
+      "loss": 0.0001,
+      "step": 25768
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 4.6994869402985074e-05,
+      "loss": 0.0003,
+      "step": 25772
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 4.699440298507463e-05,
+      "loss": 0.0002,
+      "step": 25776
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 4.6993936567164184e-05,
+      "loss": 0.001,
+      "step": 25780
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 4.699347014925373e-05,
+      "loss": 0.0001,
+      "step": 25784
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 4.699300373134329e-05,
+      "loss": 0.0008,
+      "step": 25788
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 4.6992537313432835e-05,
+      "loss": 0.0007,
+      "step": 25792
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 4.699207089552239e-05,
+      "loss": 0.0001,
+      "step": 25796
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 4.6991604477611945e-05,
+      "loss": 0.0021,
+      "step": 25800
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 4.699113805970149e-05,
+      "loss": 0.0002,
+      "step": 25804
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 4.699067164179105e-05,
+      "loss": 0.0013,
+      "step": 25808
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 4.69902052238806e-05,
+      "loss": 0.0077,
+      "step": 25812
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 4.698973880597015e-05,
+      "loss": 0.0002,
+      "step": 25816
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 4.69892723880597e-05,
+      "loss": 0.0005,
+      "step": 25820
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 4.698880597014926e-05,
+      "loss": 0.0043,
+      "step": 25824
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 4.698833955223881e-05,
+      "loss": 0.0001,
+      "step": 25828
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 4.698787313432836e-05,
+      "loss": 0.0002,
+      "step": 25832
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 4.698740671641791e-05,
+      "loss": 0.0002,
+      "step": 25836
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 4.698694029850747e-05,
+      "loss": 0.0008,
+      "step": 25840
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 4.6986473880597015e-05,
+      "loss": 0.0008,
+      "step": 25844
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 4.698600746268657e-05,
+      "loss": 0.0014,
+      "step": 25848
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 4.698554104477612e-05,
+      "loss": 0.0018,
+      "step": 25852
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 4.698507462686567e-05,
+      "loss": 0.0002,
+      "step": 25856
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 4.698460820895523e-05,
+      "loss": 0.0042,
+      "step": 25860
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 4.6984141791044776e-05,
+      "loss": 0.0026,
+      "step": 25864
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 4.698367537313433e-05,
+      "loss": 0.002,
+      "step": 25868
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 4.6983208955223886e-05,
+      "loss": 0.001,
+      "step": 25872
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 4.6982742537313434e-05,
+      "loss": 0.0002,
+      "step": 25876
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 4.698227611940298e-05,
+      "loss": 0.0003,
+      "step": 25880
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 4.6981809701492544e-05,
+      "loss": 0.0003,
+      "step": 25884
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 4.698134328358209e-05,
+      "loss": 0.0002,
+      "step": 25888
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 4.698087686567164e-05,
+      "loss": 0.0003,
+      "step": 25892
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 4.6980410447761195e-05,
+      "loss": 0.0001,
+      "step": 25896
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 4.697994402985075e-05,
+      "loss": 0.0001,
+      "step": 25900
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 4.6979477611940305e-05,
+      "loss": 0.0001,
+      "step": 25904
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 4.697901119402985e-05,
+      "loss": 0.0002,
+      "step": 25908
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 4.69785447761194e-05,
+      "loss": 0.0003,
+      "step": 25912
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 4.697807835820896e-05,
+      "loss": 0.0002,
+      "step": 25916
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 4.697761194029851e-05,
+      "loss": 0.0002,
+      "step": 25920
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 4.697714552238806e-05,
+      "loss": 0.0034,
+      "step": 25924
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 4.6976679104477614e-05,
+      "loss": 0.0027,
+      "step": 25928
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 4.697621268656717e-05,
+      "loss": 0.001,
+      "step": 25932
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 4.697574626865672e-05,
+      "loss": 0.0009,
+      "step": 25936
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 4.697527985074627e-05,
+      "loss": 0.0001,
+      "step": 25940
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 4.697481343283583e-05,
+      "loss": 0.0003,
+      "step": 25944
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 4.6974347014925375e-05,
+      "loss": 0.0002,
+      "step": 25948
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 4.697388059701493e-05,
+      "loss": 0.0006,
+      "step": 25952
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 4.697341417910448e-05,
+      "loss": 0.0006,
+      "step": 25956
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 4.697294776119403e-05,
+      "loss": 0.0022,
+      "step": 25960
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 4.697248134328359e-05,
+      "loss": 0.0067,
+      "step": 25964
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 4.6972014925373136e-05,
+      "loss": 0.0022,
+      "step": 25968
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 4.6971548507462684e-05,
+      "loss": 0.0004,
+      "step": 25972
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 4.6971082089552246e-05,
+      "loss": 0.0001,
+      "step": 25976
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 4.6970615671641794e-05,
+      "loss": 0.0108,
+      "step": 25980
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 4.697014925373134e-05,
+      "loss": 0.0007,
+      "step": 25984
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 4.69696828358209e-05,
+      "loss": 0.0003,
+      "step": 25988
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 4.696921641791045e-05,
+      "loss": 0.0008,
+      "step": 25992
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 4.696875e-05,
+      "loss": 0.0006,
+      "step": 25996
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 4.6968283582089555e-05,
+      "loss": 0.0008,
+      "step": 26000
+    },
+    {
+      "epoch": 24.25,
+      "eval_exact_match": 0.7330754352030948,
+      "eval_exec": 0.7678916827852998,
+      "eval_loss": 0.36795204877853394,
+      "eval_runtime": 1555.2766,
+      "eval_samples_per_second": 0.665,
+      "step": 26000
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 4.696781716417911e-05,
+      "loss": 0.0003,
+      "step": 26004
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 4.696735074626866e-05,
+      "loss": 0.0002,
+      "step": 26008
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 4.696688432835821e-05,
+      "loss": 0.0018,
+      "step": 26012
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 4.696641791044776e-05,
+      "loss": 0.0021,
+      "step": 26016
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 4.6965951492537316e-05,
+      "loss": 0.0,
+      "step": 26020
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 4.696548507462687e-05,
+      "loss": 0.0002,
+      "step": 26024
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 4.696501865671642e-05,
+      "loss": 0.0002,
+      "step": 26028
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 4.696455223880597e-05,
+      "loss": 0.0048,
+      "step": 26032
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 4.696408582089553e-05,
+      "loss": 0.0005,
+      "step": 26036
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 4.696361940298508e-05,
+      "loss": 0.0003,
+      "step": 26040
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 4.6963152985074625e-05,
+      "loss": 0.0004,
+      "step": 26044
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 4.696268656716418e-05,
+      "loss": 0.0064,
+      "step": 26048
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 4.6962220149253735e-05,
+      "loss": 0.0003,
+      "step": 26052
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 4.696175373134328e-05,
+      "loss": 0.0032,
+      "step": 26056
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 4.696128731343284e-05,
+      "loss": 0.001,
+      "step": 26060
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 4.696082089552239e-05,
+      "loss": 0.0002,
+      "step": 26064
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 4.696035447761195e-05,
+      "loss": 0.0021,
+      "step": 26068
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 4.6959888059701496e-05,
+      "loss": 0.0005,
+      "step": 26072
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 4.6959421641791044e-05,
+      "loss": 0.0007,
+      "step": 26076
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 4.69589552238806e-05,
+      "loss": 0.0006,
+      "step": 26080
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 4.6958488805970154e-05,
+      "loss": 0.0029,
+      "step": 26084
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 4.69580223880597e-05,
+      "loss": 0.0002,
+      "step": 26088
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 4.695755597014926e-05,
+      "loss": 0.0014,
+      "step": 26092
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 4.695708955223881e-05,
+      "loss": 0.0,
+      "step": 26096
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 4.695662313432836e-05,
+      "loss": 0.0015,
+      "step": 26100
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 4.6956156716417915e-05,
+      "loss": 0.0001,
+      "step": 26104
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 4.695569029850746e-05,
+      "loss": 0.0002,
+      "step": 26108
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 4.695522388059702e-05,
+      "loss": 0.0002,
+      "step": 26112
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 4.695475746268657e-05,
+      "loss": 0.0002,
+      "step": 26116
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 4.695429104477612e-05,
+      "loss": 0.0002,
+      "step": 26120
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 4.6953824626865676e-05,
+      "loss": 0.0012,
+      "step": 26124
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 4.695335820895523e-05,
+      "loss": 0.0001,
+      "step": 26128
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 4.695289179104478e-05,
+      "loss": 0.0002,
+      "step": 26132
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 4.695242537313433e-05,
+      "loss": 0.0003,
+      "step": 26136
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 4.695195895522388e-05,
+      "loss": 0.0046,
+      "step": 26140
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 4.6951492537313437e-05,
+      "loss": 0.0003,
+      "step": 26144
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 4.6951026119402985e-05,
+      "loss": 0.0003,
+      "step": 26148
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 4.695055970149254e-05,
+      "loss": 0.0003,
+      "step": 26152
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 4.6950093283582095e-05,
+      "loss": 0.0004,
+      "step": 26156
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 4.694962686567164e-05,
+      "loss": 0.0003,
+      "step": 26160
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 4.69491604477612e-05,
+      "loss": 0.0001,
+      "step": 26164
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 4.6948694029850746e-05,
+      "loss": 0.001,
+      "step": 26168
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 4.69482276119403e-05,
+      "loss": 0.0015,
+      "step": 26172
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 4.6947761194029856e-05,
+      "loss": 0.0002,
+      "step": 26176
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 4.6947294776119404e-05,
+      "loss": 0.0003,
+      "step": 26180
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 4.694682835820896e-05,
+      "loss": 0.0017,
+      "step": 26184
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 4.6946361940298513e-05,
+      "loss": 0.0001,
+      "step": 26188
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 4.694589552238806e-05,
+      "loss": 0.0001,
+      "step": 26192
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 4.694542910447761e-05,
+      "loss": 0.001,
+      "step": 26196
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 4.6944962686567165e-05,
+      "loss": 0.0061,
+      "step": 26200
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 4.694449626865672e-05,
+      "loss": 0.0005,
+      "step": 26204
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 4.694402985074627e-05,
+      "loss": 0.0012,
+      "step": 26208
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 4.694356343283582e-05,
+      "loss": 0.0001,
+      "step": 26212
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 4.694309701492538e-05,
+      "loss": 0.0,
+      "step": 26216
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 4.6942630597014926e-05,
+      "loss": 0.0,
+      "step": 26220
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 4.694216417910448e-05,
+      "loss": 0.0006,
+      "step": 26224
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 4.694169776119403e-05,
+      "loss": 0.0,
+      "step": 26228
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 4.694123134328359e-05,
+      "loss": 0.0011,
+      "step": 26232
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 4.694076492537314e-05,
+      "loss": 0.0003,
+      "step": 26236
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 4.6940298507462687e-05,
+      "loss": 0.0007,
+      "step": 26240
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 4.693983208955224e-05,
+      "loss": 0.0036,
+      "step": 26244
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 4.6939365671641796e-05,
+      "loss": 0.0004,
+      "step": 26248
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 4.6938899253731345e-05,
+      "loss": 0.0017,
+      "step": 26252
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 4.69384328358209e-05,
+      "loss": 0.0001,
+      "step": 26256
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 4.693796641791045e-05,
+      "loss": 0.0001,
+      "step": 26260
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 4.69375e-05,
+      "loss": 0.0002,
+      "step": 26264
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 4.693703358208956e-05,
+      "loss": 0.0016,
+      "step": 26268
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 4.6936567164179106e-05,
+      "loss": 0.0008,
+      "step": 26272
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 4.693610074626866e-05,
+      "loss": 0.0031,
+      "step": 26276
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 4.6935634328358215e-05,
+      "loss": 0.0004,
+      "step": 26280
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 4.6935167910447763e-05,
+      "loss": 0.0013,
+      "step": 26284
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 4.693470149253731e-05,
+      "loss": 0.0001,
+      "step": 26288
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 4.693423507462687e-05,
+      "loss": 0.0001,
+      "step": 26292
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 4.693376865671642e-05,
+      "loss": 0.0015,
+      "step": 26296
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 4.693330223880597e-05,
+      "loss": 0.0008,
+      "step": 26300
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 4.6932835820895524e-05,
+      "loss": 0.0001,
+      "step": 26304
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 4.693236940298508e-05,
+      "loss": 0.0003,
+      "step": 26308
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 4.693190298507463e-05,
+      "loss": 0.0002,
+      "step": 26312
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 4.693143656716418e-05,
+      "loss": 0.0005,
+      "step": 26316
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 4.693097014925373e-05,
+      "loss": 0.0002,
+      "step": 26320
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 4.6930503731343285e-05,
+      "loss": 0.0001,
+      "step": 26324
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 4.693003731343284e-05,
+      "loss": 0.0004,
+      "step": 26328
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 4.692957089552239e-05,
+      "loss": 0.002,
+      "step": 26332
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 4.6929104477611943e-05,
+      "loss": 0.0001,
+      "step": 26336
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 4.69286380597015e-05,
+      "loss": 0.0005,
+      "step": 26340
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 4.6928171641791046e-05,
+      "loss": 0.0063,
+      "step": 26344
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 4.6927705223880595e-05,
+      "loss": 0.0003,
+      "step": 26348
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 4.6927238805970156e-05,
+      "loss": 0.0011,
+      "step": 26352
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 4.6926772388059704e-05,
+      "loss": 0.0027,
+      "step": 26356
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 4.692630597014925e-05,
+      "loss": 0.0006,
+      "step": 26360
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 4.692583955223881e-05,
+      "loss": 0.0011,
+      "step": 26364
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 4.692537313432836e-05,
+      "loss": 0.0004,
+      "step": 26368
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 4.692490671641791e-05,
+      "loss": 0.0005,
+      "step": 26372
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 4.6924440298507465e-05,
+      "loss": 0.0002,
+      "step": 26376
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 4.6923973880597014e-05,
+      "loss": 0.0004,
+      "step": 26380
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 4.692350746268657e-05,
+      "loss": 0.0011,
+      "step": 26384
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 4.692304104477612e-05,
+      "loss": 0.0001,
+      "step": 26388
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 4.692257462686567e-05,
+      "loss": 0.0,
+      "step": 26392
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 4.6922108208955226e-05,
+      "loss": 0.0006,
+      "step": 26396
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 4.692164179104478e-05,
+      "loss": 0.0,
+      "step": 26400
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 4.692117537313433e-05,
+      "loss": 0.0031,
+      "step": 26404
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 4.6920708955223884e-05,
+      "loss": 0.0009,
+      "step": 26408
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 4.692024253731343e-05,
+      "loss": 0.0048,
+      "step": 26412
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 4.691977611940299e-05,
+      "loss": 0.0003,
+      "step": 26416
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 4.691930970149254e-05,
+      "loss": 0.0,
+      "step": 26420
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 4.691884328358209e-05,
+      "loss": 0.0005,
+      "step": 26424
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 4.6918376865671645e-05,
+      "loss": 0.0022,
+      "step": 26428
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 4.69179104477612e-05,
+      "loss": 0.0002,
+      "step": 26432
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 4.691744402985075e-05,
+      "loss": 0.0001,
+      "step": 26436
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 4.6916977611940296e-05,
+      "loss": 0.0015,
+      "step": 26440
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 4.691651119402986e-05,
+      "loss": 0.0003,
+      "step": 26444
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 4.6916044776119406e-05,
+      "loss": 0.0035,
+      "step": 26448
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 4.6915578358208954e-05,
+      "loss": 0.0034,
+      "step": 26452
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 4.691511194029851e-05,
+      "loss": 0.0004,
+      "step": 26456
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 4.6914645522388064e-05,
+      "loss": 0.0001,
+      "step": 26460
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 4.691417910447761e-05,
+      "loss": 0.0001,
+      "step": 26464
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 4.691371268656717e-05,
+      "loss": 0.001,
+      "step": 26468
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 4.6913246268656715e-05,
+      "loss": 0.0022,
+      "step": 26472
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 4.691277985074627e-05,
+      "loss": 0.0001,
+      "step": 26476
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 4.6912313432835825e-05,
+      "loss": 0.0,
+      "step": 26480
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 4.691184701492537e-05,
+      "loss": 0.0005,
+      "step": 26484
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 4.691138059701493e-05,
+      "loss": 0.0031,
+      "step": 26488
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 4.691091417910448e-05,
+      "loss": 0.0005,
+      "step": 26492
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 4.691044776119403e-05,
+      "loss": 0.0001,
+      "step": 26496
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 4.690998134328358e-05,
+      "loss": 0.0001,
+      "step": 26500
+    },
+    {
+      "epoch": 24.72,
+      "eval_exact_match": 0.7446808510638298,
+      "eval_exec": 0.7736943907156673,
+      "eval_loss": 0.403182715177536,
+      "eval_runtime": 1851.9773,
+      "eval_samples_per_second": 0.558,
+      "step": 26500
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 4.690951492537314e-05,
+      "loss": 0.0009,
+      "step": 26504
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 4.690904850746269e-05,
+      "loss": 0.0003,
+      "step": 26508
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 4.690858208955224e-05,
+      "loss": 0.0001,
+      "step": 26512
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 4.690811567164179e-05,
+      "loss": 0.0015,
+      "step": 26516
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 4.690764925373135e-05,
+      "loss": 0.0003,
+      "step": 26520
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 4.6907182835820895e-05,
+      "loss": 0.0021,
+      "step": 26524
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 4.690671641791045e-05,
+      "loss": 0.0006,
+      "step": 26528
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 4.690625e-05,
+      "loss": 0.0009,
+      "step": 26532
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 4.690578358208955e-05,
+      "loss": 0.0,
+      "step": 26536
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 4.690531716417911e-05,
+      "loss": 0.0002,
+      "step": 26540
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 4.6904850746268656e-05,
+      "loss": 0.0004,
+      "step": 26544
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 4.690438432835821e-05,
+      "loss": 0.0109,
+      "step": 26548
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 4.6903917910447766e-05,
+      "loss": 0.0008,
+      "step": 26552
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 4.6903451492537314e-05,
+      "loss": 0.001,
+      "step": 26556
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 4.690298507462687e-05,
+      "loss": 0.0027,
+      "step": 26560
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 4.6902518656716424e-05,
+      "loss": 0.0001,
+      "step": 26564
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 4.690205223880597e-05,
+      "loss": 0.0004,
+      "step": 26568
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 4.690158582089553e-05,
+      "loss": 0.0099,
+      "step": 26572
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 4.6901119402985075e-05,
+      "loss": 0.0017,
+      "step": 26576
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 4.690065298507463e-05,
+      "loss": 0.0046,
+      "step": 26580
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 4.6900186567164185e-05,
+      "loss": 0.0015,
+      "step": 26584
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 4.689972014925373e-05,
+      "loss": 0.001,
+      "step": 26588
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 4.689925373134328e-05,
+      "loss": 0.001,
+      "step": 26592
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 4.689878731343284e-05,
+      "loss": 0.0003,
+      "step": 26596
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 4.689832089552239e-05,
+      "loss": 0.0036,
+      "step": 26600
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 4.689785447761194e-05,
+      "loss": 0.0021,
+      "step": 26604
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 4.6897388059701494e-05,
+      "loss": 0.0002,
+      "step": 26608
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 4.689692164179105e-05,
+      "loss": 0.0001,
+      "step": 26612
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 4.68964552238806e-05,
+      "loss": 0.0015,
+      "step": 26616
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 4.689598880597015e-05,
+      "loss": 0.0015,
+      "step": 26620
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 4.689552238805971e-05,
+      "loss": 0.0032,
+      "step": 26624
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 4.6895055970149255e-05,
+      "loss": 0.0009,
+      "step": 26628
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 4.689458955223881e-05,
+      "loss": 0.0013,
+      "step": 26632
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 4.689412313432836e-05,
+      "loss": 0.0004,
+      "step": 26636
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 4.689365671641791e-05,
+      "loss": 0.0029,
+      "step": 26640
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 4.689319029850747e-05,
+      "loss": 0.0001,
+      "step": 26644
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 4.6892723880597016e-05,
+      "loss": 0.0026,
+      "step": 26648
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 4.6892257462686564e-05,
+      "loss": 0.0009,
+      "step": 26652
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 4.6891791044776126e-05,
+      "loss": 0.0006,
+      "step": 26656
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 4.6891324626865674e-05,
+      "loss": 0.0002,
+      "step": 26660
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 4.689085820895522e-05,
+      "loss": 0.0002,
+      "step": 26664
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 4.689039179104478e-05,
+      "loss": 0.0037,
+      "step": 26668
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 4.688992537313433e-05,
+      "loss": 0.0004,
+      "step": 26672
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 4.688945895522388e-05,
+      "loss": 0.0035,
+      "step": 26676
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 4.6888992537313435e-05,
+      "loss": 0.0003,
+      "step": 26680
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 4.688852611940299e-05,
+      "loss": 0.0004,
+      "step": 26684
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 4.688805970149254e-05,
+      "loss": 0.0002,
+      "step": 26688
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 4.688759328358209e-05,
+      "loss": 0.0,
+      "step": 26692
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 4.688712686567164e-05,
+      "loss": 0.003,
+      "step": 26696
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 4.6886660447761196e-05,
+      "loss": 0.0013,
+      "step": 26700
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 4.688619402985075e-05,
+      "loss": 0.0,
+      "step": 26704
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 4.68857276119403e-05,
+      "loss": 0.0002,
+      "step": 26708
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 4.688526119402985e-05,
+      "loss": 0.0008,
+      "step": 26712
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 4.688479477611941e-05,
+      "loss": 0.0005,
+      "step": 26716
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 4.688432835820896e-05,
+      "loss": 0.0007,
+      "step": 26720
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 4.688386194029851e-05,
+      "loss": 0.0,
+      "step": 26724
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 4.688339552238806e-05,
+      "loss": 0.0001,
+      "step": 26728
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 4.6882929104477615e-05,
+      "loss": 0.0001,
+      "step": 26732
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 4.688246268656717e-05,
+      "loss": 0.0008,
+      "step": 26736
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 4.688199626865672e-05,
+      "loss": 0.0001,
+      "step": 26740
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 4.688152985074627e-05,
+      "loss": 0.0001,
+      "step": 26744
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 4.688106343283583e-05,
+      "loss": 0.0013,
+      "step": 26748
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 4.6880597014925376e-05,
+      "loss": 0.0001,
+      "step": 26752
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 4.6880130597014924e-05,
+      "loss": 0.0005,
+      "step": 26756
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 4.687966417910448e-05,
+      "loss": 0.0015,
+      "step": 26760
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 4.6879197761194034e-05,
+      "loss": 0.0004,
+      "step": 26764
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 4.687873134328358e-05,
+      "loss": 0.0023,
+      "step": 26768
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 4.687826492537314e-05,
+      "loss": 0.0001,
+      "step": 26772
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 4.687779850746269e-05,
+      "loss": 0.0001,
+      "step": 26776
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 4.687733208955224e-05,
+      "loss": 0.0015,
+      "step": 26780
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 4.6876865671641795e-05,
+      "loss": 0.0002,
+      "step": 26784
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 4.687639925373134e-05,
+      "loss": 0.0001,
+      "step": 26788
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 4.68759328358209e-05,
+      "loss": 0.0003,
+      "step": 26792
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 4.687546641791045e-05,
+      "loss": 0.0088,
+      "step": 26796
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 4.6875e-05,
+      "loss": 0.0008,
+      "step": 26800
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 4.6874533582089556e-05,
+      "loss": 0.0018,
+      "step": 26804
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 4.687406716417911e-05,
+      "loss": 0.0,
+      "step": 26808
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 4.687360074626866e-05,
+      "loss": 0.0,
+      "step": 26812
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 4.687313432835821e-05,
+      "loss": 0.0,
+      "step": 26816
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 4.687266791044776e-05,
+      "loss": 0.0002,
+      "step": 26820
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 4.687220149253732e-05,
+      "loss": 0.0018,
+      "step": 26824
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 4.6871735074626865e-05,
+      "loss": 0.0002,
+      "step": 26828
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 4.687126865671642e-05,
+      "loss": 0.0037,
+      "step": 26832
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 4.6870802238805975e-05,
+      "loss": 0.0019,
+      "step": 26836
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 4.687033582089552e-05,
+      "loss": 0.0003,
+      "step": 26840
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 4.686986940298508e-05,
+      "loss": 0.0004,
+      "step": 26844
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 4.6869402985074626e-05,
+      "loss": 0.0001,
+      "step": 26848
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 4.686893656716418e-05,
+      "loss": 0.0002,
+      "step": 26852
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 4.6868470149253736e-05,
+      "loss": 0.0001,
+      "step": 26856
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 4.6868003731343284e-05,
+      "loss": 0.0017,
+      "step": 26860
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 4.686753731343284e-05,
+      "loss": 0.0001,
+      "step": 26864
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 4.6867070895522394e-05,
+      "loss": 0.0009,
+      "step": 26868
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 4.686660447761194e-05,
+      "loss": 0.0085,
+      "step": 26872
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 4.686613805970149e-05,
+      "loss": 0.0021,
+      "step": 26876
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 4.6865671641791045e-05,
+      "loss": 0.0007,
+      "step": 26880
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 4.68652052238806e-05,
+      "loss": 0.0021,
+      "step": 26884
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 4.6864738805970155e-05,
+      "loss": 0.0012,
+      "step": 26888
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 4.68642723880597e-05,
+      "loss": 0.0,
+      "step": 26892
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 4.686380597014926e-05,
+      "loss": 0.0011,
+      "step": 26896
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 4.686333955223881e-05,
+      "loss": 0.001,
+      "step": 26900
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 4.686287313432836e-05,
+      "loss": 0.0006,
+      "step": 26904
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 4.686240671641791e-05,
+      "loss": 0.0019,
+      "step": 26908
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 4.686194029850747e-05,
+      "loss": 0.0002,
+      "step": 26912
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 4.686147388059702e-05,
+      "loss": 0.0002,
+      "step": 26916
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 4.686100746268657e-05,
+      "loss": 0.0014,
+      "step": 26920
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 4.686054104477612e-05,
+      "loss": 0.0001,
+      "step": 26924
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 4.686007462686568e-05,
+      "loss": 0.0036,
+      "step": 26928
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 4.6859608208955225e-05,
+      "loss": 0.0018,
+      "step": 26932
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 4.685914179104478e-05,
+      "loss": 0.0045,
+      "step": 26936
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 4.685867537313433e-05,
+      "loss": 0.0017,
+      "step": 26940
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 4.685820895522388e-05,
+      "loss": 0.0011,
+      "step": 26944
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 4.685774253731344e-05,
+      "loss": 0.0015,
+      "step": 26948
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 4.6857276119402986e-05,
+      "loss": 0.0408,
+      "step": 26952
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 4.685680970149254e-05,
+      "loss": 0.008,
+      "step": 26956
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 4.6856343283582096e-05,
+      "loss": 0.0,
+      "step": 26960
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 4.6855876865671644e-05,
+      "loss": 0.0036,
+      "step": 26964
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 4.685541044776119e-05,
+      "loss": 0.0021,
+      "step": 26968
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 4.6854944029850754e-05,
+      "loss": 0.001,
+      "step": 26972
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 4.68544776119403e-05,
+      "loss": 0.0101,
+      "step": 26976
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 4.685401119402985e-05,
+      "loss": 0.0008,
+      "step": 26980
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 4.6853544776119405e-05,
+      "loss": 0.0,
+      "step": 26984
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 4.685307835820896e-05,
+      "loss": 0.001,
+      "step": 26988
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 4.685261194029851e-05,
+      "loss": 0.0004,
+      "step": 26992
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 4.685214552238806e-05,
+      "loss": 0.0001,
+      "step": 26996
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 4.685167910447761e-05,
+      "loss": 0.0002,
+      "step": 27000
+    },
+    {
+      "epoch": 25.19,
+      "eval_exact_match": 0.7427466150870407,
+      "eval_exec": 0.7794970986460348,
+      "eval_loss": 0.3887414336204529,
+      "eval_runtime": 1208.0912,
+      "eval_samples_per_second": 0.856,
+      "step": 27000
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 4.6851212686567166e-05,
+      "loss": 0.0008,
+      "step": 27004
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 4.685074626865672e-05,
+      "loss": 0.0002,
+      "step": 27008
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 4.685027985074627e-05,
+      "loss": 0.0004,
+      "step": 27012
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 4.6849813432835824e-05,
+      "loss": 0.0028,
+      "step": 27016
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 4.684934701492538e-05,
+      "loss": 0.0006,
+      "step": 27020
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 4.684888059701493e-05,
+      "loss": 0.0005,
+      "step": 27024
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 4.6848414179104475e-05,
+      "loss": 0.0001,
+      "step": 27028
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 4.6847947761194037e-05,
+      "loss": 0.0013,
+      "step": 27032
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 4.6847481343283585e-05,
+      "loss": 0.0019,
+      "step": 27036
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 4.684701492537313e-05,
+      "loss": 0.0004,
+      "step": 27040
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 4.684654850746269e-05,
+      "loss": 0.0012,
+      "step": 27044
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 4.684608208955224e-05,
+      "loss": 0.0008,
+      "step": 27048
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 4.68456156716418e-05,
+      "loss": 0.0028,
+      "step": 27052
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 4.6845149253731346e-05,
+      "loss": 0.0005,
+      "step": 27056
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 4.6844682835820894e-05,
+      "loss": 0.0012,
+      "step": 27060
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 4.6844216417910455e-05,
+      "loss": 0.0002,
+      "step": 27064
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 4.6843750000000004e-05,
+      "loss": 0.0007,
+      "step": 27068
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 4.684328358208955e-05,
+      "loss": 0.0023,
+      "step": 27072
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 4.684281716417911e-05,
+      "loss": 0.0008,
+      "step": 27076
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 4.684235074626866e-05,
+      "loss": 0.0001,
+      "step": 27080
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 4.684188432835821e-05,
+      "loss": 0.0008,
+      "step": 27084
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 4.6841417910447765e-05,
+      "loss": 0.004,
+      "step": 27088
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 4.684095149253731e-05,
+      "loss": 0.0075,
+      "step": 27092
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 4.684048507462687e-05,
+      "loss": 0.0006,
+      "step": 27096
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 4.684001865671642e-05,
+      "loss": 0.0002,
+      "step": 27100
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 4.683955223880597e-05,
+      "loss": 0.0002,
+      "step": 27104
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 4.6839085820895526e-05,
+      "loss": 0.0052,
+      "step": 27108
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 4.683861940298508e-05,
+      "loss": 0.0016,
+      "step": 27112
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 4.683815298507463e-05,
+      "loss": 0.0021,
+      "step": 27116
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 4.683768656716418e-05,
+      "loss": 0.0004,
+      "step": 27120
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 4.683722014925374e-05,
+      "loss": 0.0011,
+      "step": 27124
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 4.6836753731343287e-05,
+      "loss": 0.0,
+      "step": 27128
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 4.6836287313432835e-05,
+      "loss": 0.0018,
+      "step": 27132
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 4.683582089552239e-05,
+      "loss": 0.001,
+      "step": 27136
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 4.6835354477611945e-05,
+      "loss": 0.0001,
+      "step": 27140
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 4.683488805970149e-05,
+      "loss": 0.0094,
+      "step": 27144
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 4.683442164179105e-05,
+      "loss": 0.0019,
+      "step": 27148
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 4.6833955223880596e-05,
+      "loss": 0.0067,
+      "step": 27152
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 4.683348880597015e-05,
+      "loss": 0.0001,
+      "step": 27156
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 4.6833022388059705e-05,
+      "loss": 0.0008,
+      "step": 27160
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 4.6832555970149254e-05,
+      "loss": 0.0013,
+      "step": 27164
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 4.683208955223881e-05,
+      "loss": 0.0013,
+      "step": 27168
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 4.6831623134328363e-05,
+      "loss": 0.0001,
+      "step": 27172
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 4.683115671641791e-05,
+      "loss": 0.0001,
+      "step": 27176
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 4.683069029850746e-05,
+      "loss": 0.0005,
+      "step": 27180
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 4.683022388059702e-05,
+      "loss": 0.0006,
+      "step": 27184
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 4.682975746268657e-05,
+      "loss": 0.0001,
+      "step": 27188
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 4.682929104477612e-05,
+      "loss": 0.0001,
+      "step": 27192
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 4.682882462686567e-05,
+      "loss": 0.0012,
+      "step": 27196
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 4.682835820895523e-05,
+      "loss": 0.005,
+      "step": 27200
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 4.6827891791044776e-05,
+      "loss": 0.0013,
+      "step": 27204
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 4.682742537313433e-05,
+      "loss": 0.0078,
+      "step": 27208
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 4.682695895522388e-05,
+      "loss": 0.0002,
+      "step": 27212
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 4.682649253731344e-05,
+      "loss": 0.0006,
+      "step": 27216
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 4.682602611940299e-05,
+      "loss": 0.0002,
+      "step": 27220
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 4.6825559701492537e-05,
+      "loss": 0.0005,
+      "step": 27224
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 4.682509328358209e-05,
+      "loss": 0.004,
+      "step": 27228
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 4.6824626865671646e-05,
+      "loss": 0.0108,
+      "step": 27232
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 4.6824160447761195e-05,
+      "loss": 0.0004,
+      "step": 27236
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 4.682369402985075e-05,
+      "loss": 0.0009,
+      "step": 27240
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 4.6823227611940304e-05,
+      "loss": 0.0001,
+      "step": 27244
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 4.682276119402985e-05,
+      "loss": 0.0001,
+      "step": 27248
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 4.682229477611941e-05,
+      "loss": 0.0002,
+      "step": 27252
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 4.6821828358208956e-05,
+      "loss": 0.0003,
+      "step": 27256
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 4.682136194029851e-05,
+      "loss": 0.0007,
+      "step": 27260
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 4.6820895522388065e-05,
+      "loss": 0.0006,
+      "step": 27264
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 4.6820429104477613e-05,
+      "loss": 0.0002,
+      "step": 27268
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 4.681996268656716e-05,
+      "loss": 0.0004,
+      "step": 27272
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 4.681949626865672e-05,
+      "loss": 0.0001,
+      "step": 27276
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 4.681902985074627e-05,
+      "loss": 0.005,
+      "step": 27280
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 4.681856343283582e-05,
+      "loss": 0.0094,
+      "step": 27284
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 4.6818097014925374e-05,
+      "loss": 0.0005,
+      "step": 27288
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 4.681763059701493e-05,
+      "loss": 0.0038,
+      "step": 27292
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 4.681716417910448e-05,
+      "loss": 0.0005,
+      "step": 27296
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 4.681669776119403e-05,
+      "loss": 0.0001,
+      "step": 27300
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 4.681623134328359e-05,
+      "loss": 0.0018,
+      "step": 27304
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 4.6815764925373135e-05,
+      "loss": 0.0006,
+      "step": 27308
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 4.681529850746269e-05,
+      "loss": 0.0001,
+      "step": 27312
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 4.681483208955224e-05,
+      "loss": 0.0006,
+      "step": 27316
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 4.681436567164179e-05,
+      "loss": 0.0018,
+      "step": 27320
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 4.681389925373135e-05,
+      "loss": 0.0086,
+      "step": 27324
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 4.6813432835820896e-05,
+      "loss": 0.0001,
+      "step": 27328
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 4.6812966417910445e-05,
+      "loss": 0.0012,
+      "step": 27332
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 4.6812500000000006e-05,
+      "loss": 0.0001,
+      "step": 27336
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 4.6812033582089554e-05,
+      "loss": 0.0015,
+      "step": 27340
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 4.68115671641791e-05,
+      "loss": 0.0023,
+      "step": 27344
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 4.681110074626866e-05,
+      "loss": 0.0007,
+      "step": 27348
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 4.681063432835821e-05,
+      "loss": 0.0034,
+      "step": 27352
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 4.681016791044776e-05,
+      "loss": 0.0001,
+      "step": 27356
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 4.6809701492537315e-05,
+      "loss": 0.0001,
+      "step": 27360
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 4.680923507462687e-05,
+      "loss": 0.0019,
+      "step": 27364
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 4.680876865671642e-05,
+      "loss": 0.0001,
+      "step": 27368
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 4.680830223880597e-05,
+      "loss": 0.0042,
+      "step": 27372
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 4.680783582089552e-05,
+      "loss": 0.0001,
+      "step": 27376
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 4.6807369402985076e-05,
+      "loss": 0.0011,
+      "step": 27380
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 4.680690298507463e-05,
+      "loss": 0.0004,
+      "step": 27384
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 4.680643656716418e-05,
+      "loss": 0.0008,
+      "step": 27388
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 4.6805970149253734e-05,
+      "loss": 0.0003,
+      "step": 27392
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 4.680550373134329e-05,
+      "loss": 0.0003,
+      "step": 27396
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 4.680503731343284e-05,
+      "loss": 0.0021,
+      "step": 27400
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 4.680457089552239e-05,
+      "loss": 0.0001,
+      "step": 27404
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 4.680410447761194e-05,
+      "loss": 0.0024,
+      "step": 27408
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 4.6803638059701495e-05,
+      "loss": 0.0041,
+      "step": 27412
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 4.680317164179105e-05,
+      "loss": 0.001,
+      "step": 27416
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 4.68027052238806e-05,
+      "loss": 0.0008,
+      "step": 27420
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 4.680223880597015e-05,
+      "loss": 0.0029,
+      "step": 27424
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 4.680177238805971e-05,
+      "loss": 0.0012,
+      "step": 27428
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 4.6801305970149256e-05,
+      "loss": 0.0008,
+      "step": 27432
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 4.6800839552238804e-05,
+      "loss": 0.0002,
+      "step": 27436
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 4.680037313432836e-05,
+      "loss": 0.0016,
+      "step": 27440
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 4.6799906716417914e-05,
+      "loss": 0.0009,
+      "step": 27444
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 4.679944029850746e-05,
+      "loss": 0.0008,
+      "step": 27448
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 4.679897388059702e-05,
+      "loss": 0.0012,
+      "step": 27452
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 4.679850746268657e-05,
+      "loss": 0.0012,
+      "step": 27456
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 4.679804104477612e-05,
+      "loss": 0.0018,
+      "step": 27460
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 4.6797574626865675e-05,
+      "loss": 0.0002,
+      "step": 27464
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 4.679710820895522e-05,
+      "loss": 0.0001,
+      "step": 27468
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 4.679664179104478e-05,
+      "loss": 0.0017,
+      "step": 27472
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 4.679617537313433e-05,
+      "loss": 0.0011,
+      "step": 27476
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 4.679570895522388e-05,
+      "loss": 0.0003,
+      "step": 27480
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 4.6795242537313436e-05,
+      "loss": 0.01,
+      "step": 27484
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 4.679477611940299e-05,
+      "loss": 0.0016,
+      "step": 27488
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 4.679430970149254e-05,
+      "loss": 0.0048,
+      "step": 27492
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 4.679384328358209e-05,
+      "loss": 0.0038,
+      "step": 27496
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 4.679337686567164e-05,
+      "loss": 0.0004,
+      "step": 27500
+    },
+    {
+      "epoch": 25.65,
+      "eval_exact_match": 0.7427466150870407,
+      "eval_exec": 0.7833655705996132,
+      "eval_loss": 0.3488633632659912,
+      "eval_runtime": 1170.5529,
+      "eval_samples_per_second": 0.883,
+      "step": 27500
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 4.67929104477612e-05,
+      "loss": 0.0005,
+      "step": 27504
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 4.6792444029850745e-05,
+      "loss": 0.0003,
+      "step": 27508
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 4.67919776119403e-05,
+      "loss": 0.0001,
+      "step": 27512
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 4.6791511194029855e-05,
+      "loss": 0.0045,
+      "step": 27516
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 4.67910447761194e-05,
+      "loss": 0.0001,
+      "step": 27520
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 4.679057835820896e-05,
+      "loss": 0.0014,
+      "step": 27524
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 4.6790111940298506e-05,
+      "loss": 0.0004,
+      "step": 27528
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 4.678964552238806e-05,
+      "loss": 0.0047,
+      "step": 27532
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 4.6789179104477616e-05,
+      "loss": 0.0002,
+      "step": 27536
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 4.6788712686567164e-05,
+      "loss": 0.0001,
+      "step": 27540
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 4.678824626865672e-05,
+      "loss": 0.0021,
+      "step": 27544
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 4.6787779850746274e-05,
+      "loss": 0.0018,
+      "step": 27548
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 4.678731343283582e-05,
+      "loss": 0.0036,
+      "step": 27552
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 4.678684701492538e-05,
+      "loss": 0.0001,
+      "step": 27556
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 4.6786380597014925e-05,
+      "loss": 0.0041,
+      "step": 27560
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 4.678591417910448e-05,
+      "loss": 0.0001,
+      "step": 27564
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 4.6785447761194035e-05,
+      "loss": 0.0009,
+      "step": 27568
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 4.678498134328358e-05,
+      "loss": 0.0035,
+      "step": 27572
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 4.678451492537314e-05,
+      "loss": 0.002,
+      "step": 27576
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 4.678404850746269e-05,
+      "loss": 0.0001,
+      "step": 27580
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 4.678358208955224e-05,
+      "loss": 0.0027,
+      "step": 27584
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 4.678311567164179e-05,
+      "loss": 0.0073,
+      "step": 27588
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 4.678264925373135e-05,
+      "loss": 0.0004,
+      "step": 27592
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 4.67821828358209e-05,
+      "loss": 0.0008,
+      "step": 27596
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 4.678171641791045e-05,
+      "loss": 0.0004,
+      "step": 27600
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 4.678125e-05,
+      "loss": 0.0095,
+      "step": 27604
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 4.678078358208956e-05,
+      "loss": 0.0004,
+      "step": 27608
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 4.6780317164179105e-05,
+      "loss": 0.0162,
+      "step": 27612
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 4.677985074626866e-05,
+      "loss": 0.0008,
+      "step": 27616
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 4.677938432835821e-05,
+      "loss": 0.0013,
+      "step": 27620
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 4.677891791044776e-05,
+      "loss": 0.0006,
+      "step": 27624
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 4.677845149253732e-05,
+      "loss": 0.0004,
+      "step": 27628
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 4.6777985074626866e-05,
+      "loss": 0.0016,
+      "step": 27632
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 4.677751865671642e-05,
+      "loss": 0.0001,
+      "step": 27636
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 4.6777052238805976e-05,
+      "loss": 0.0001,
+      "step": 27640
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 4.6776585820895524e-05,
+      "loss": 0.0002,
+      "step": 27644
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 4.677611940298507e-05,
+      "loss": 0.0003,
+      "step": 27648
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 4.6775652985074634e-05,
+      "loss": 0.0002,
+      "step": 27652
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 4.677518656716418e-05,
+      "loss": 0.0,
+      "step": 27656
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 4.677472014925373e-05,
+      "loss": 0.0011,
+      "step": 27660
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 4.6774253731343285e-05,
+      "loss": 0.0002,
+      "step": 27664
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 4.677378731343284e-05,
+      "loss": 0.0009,
+      "step": 27668
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 4.677332089552239e-05,
+      "loss": 0.0003,
+      "step": 27672
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 4.677285447761194e-05,
+      "loss": 0.0011,
+      "step": 27676
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 4.677238805970149e-05,
+      "loss": 0.0007,
+      "step": 27680
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 4.6771921641791046e-05,
+      "loss": 0.0129,
+      "step": 27684
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 4.67714552238806e-05,
+      "loss": 0.0053,
+      "step": 27688
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 4.677098880597015e-05,
+      "loss": 0.0096,
+      "step": 27692
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 4.6770522388059704e-05,
+      "loss": 0.0001,
+      "step": 27696
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 4.677005597014926e-05,
+      "loss": 0.0019,
+      "step": 27700
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 4.676958955223881e-05,
+      "loss": 0.0001,
+      "step": 27704
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 4.676912313432836e-05,
+      "loss": 0.0017,
+      "step": 27708
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 4.676865671641792e-05,
+      "loss": 0.0008,
+      "step": 27712
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 4.6768190298507465e-05,
+      "loss": 0.0004,
+      "step": 27716
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 4.676772388059702e-05,
+      "loss": 0.0052,
+      "step": 27720
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 4.676725746268657e-05,
+      "loss": 0.0012,
+      "step": 27724
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 4.676679104477612e-05,
+      "loss": 0.001,
+      "step": 27728
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 4.676632462686568e-05,
+      "loss": 0.0001,
+      "step": 27732
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 4.6765858208955226e-05,
+      "loss": 0.0006,
+      "step": 27736
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 4.6765391791044774e-05,
+      "loss": 0.0004,
+      "step": 27740
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 4.6764925373134336e-05,
+      "loss": 0.0002,
+      "step": 27744
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 4.6764458955223884e-05,
+      "loss": 0.0022,
+      "step": 27748
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 4.676399253731343e-05,
+      "loss": 0.001,
+      "step": 27752
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 4.676352611940299e-05,
+      "loss": 0.0005,
+      "step": 27756
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 4.676305970149254e-05,
+      "loss": 0.0013,
+      "step": 27760
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 4.676259328358209e-05,
+      "loss": 0.0015,
+      "step": 27764
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 4.6762126865671645e-05,
+      "loss": 0.0012,
+      "step": 27768
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 4.676166044776119e-05,
+      "loss": 0.0005,
+      "step": 27772
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 4.676119402985075e-05,
+      "loss": 0.0005,
+      "step": 27776
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 4.67607276119403e-05,
+      "loss": 0.0037,
+      "step": 27780
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 4.676026119402985e-05,
+      "loss": 0.0003,
+      "step": 27784
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 4.6759794776119406e-05,
+      "loss": 0.0002,
+      "step": 27788
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 4.675932835820896e-05,
+      "loss": 0.0004,
+      "step": 27792
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 4.675886194029851e-05,
+      "loss": 0.0013,
+      "step": 27796
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 4.675839552238806e-05,
+      "loss": 0.0001,
+      "step": 27800
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 4.675792910447762e-05,
+      "loss": 0.0,
+      "step": 27804
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 4.675746268656717e-05,
+      "loss": 0.0001,
+      "step": 27808
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 4.6756996268656715e-05,
+      "loss": 0.0005,
+      "step": 27812
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 4.675652985074627e-05,
+      "loss": 0.0002,
+      "step": 27816
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 4.6756063432835825e-05,
+      "loss": 0.0018,
+      "step": 27820
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 4.675559701492537e-05,
+      "loss": 0.0099,
+      "step": 27824
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 4.675513059701493e-05,
+      "loss": 0.004,
+      "step": 27828
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 4.6754664179104476e-05,
+      "loss": 0.0025,
+      "step": 27832
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 4.675419776119403e-05,
+      "loss": 0.0001,
+      "step": 27836
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 4.6753731343283586e-05,
+      "loss": 0.0002,
+      "step": 27840
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 4.6753264925373134e-05,
+      "loss": 0.0,
+      "step": 27844
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 4.675279850746269e-05,
+      "loss": 0.0001,
+      "step": 27848
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 4.6752332089552244e-05,
+      "loss": 0.0013,
+      "step": 27852
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 4.675186567164179e-05,
+      "loss": 0.0002,
+      "step": 27856
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 4.675139925373134e-05,
+      "loss": 0.0009,
+      "step": 27860
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 4.67509328358209e-05,
+      "loss": 0.001,
+      "step": 27864
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 4.675046641791045e-05,
+      "loss": 0.0001,
+      "step": 27868
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 4.6750000000000005e-05,
+      "loss": 0.005,
+      "step": 27872
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 4.674953358208955e-05,
+      "loss": 0.001,
+      "step": 27876
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 4.674906716417911e-05,
+      "loss": 0.0177,
+      "step": 27880
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 4.674860074626866e-05,
+      "loss": 0.0028,
+      "step": 27884
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 4.674813432835821e-05,
+      "loss": 0.001,
+      "step": 27888
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 4.674766791044776e-05,
+      "loss": 0.0001,
+      "step": 27892
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 4.674720149253732e-05,
+      "loss": 0.001,
+      "step": 27896
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 4.674673507462687e-05,
+      "loss": 0.0004,
+      "step": 27900
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 4.674626865671642e-05,
+      "loss": 0.0003,
+      "step": 27904
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 4.674580223880597e-05,
+      "loss": 0.0005,
+      "step": 27908
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 4.674533582089553e-05,
+      "loss": 0.0001,
+      "step": 27912
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 4.6744869402985075e-05,
+      "loss": 0.0012,
+      "step": 27916
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 4.674440298507463e-05,
+      "loss": 0.0008,
+      "step": 27920
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 4.6743936567164185e-05,
+      "loss": 0.0008,
+      "step": 27924
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 4.674347014925373e-05,
+      "loss": 0.0003,
+      "step": 27928
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 4.674300373134329e-05,
+      "loss": 0.0002,
+      "step": 27932
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 4.6742537313432836e-05,
+      "loss": 0.0047,
+      "step": 27936
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 4.674207089552239e-05,
+      "loss": 0.0001,
+      "step": 27940
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 4.6741604477611946e-05,
+      "loss": 0.0004,
+      "step": 27944
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 4.6741138059701494e-05,
+      "loss": 0.0001,
+      "step": 27948
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 4.674067164179104e-05,
+      "loss": 0.0002,
+      "step": 27952
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 4.6740205223880604e-05,
+      "loss": 0.0034,
+      "step": 27956
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 4.673973880597015e-05,
+      "loss": 0.0018,
+      "step": 27960
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 4.67392723880597e-05,
+      "loss": 0.0017,
+      "step": 27964
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 4.6738805970149255e-05,
+      "loss": 0.0006,
+      "step": 27968
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 4.673833955223881e-05,
+      "loss": 0.0081,
+      "step": 27972
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 4.673787313432836e-05,
+      "loss": 0.0001,
+      "step": 27976
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 4.673740671641791e-05,
+      "loss": 0.0002,
+      "step": 27980
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 4.673694029850747e-05,
+      "loss": 0.0002,
+      "step": 27984
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 4.6736473880597016e-05,
+      "loss": 0.0027,
+      "step": 27988
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 4.673600746268657e-05,
+      "loss": 0.0018,
+      "step": 27992
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 4.673554104477612e-05,
+      "loss": 0.0001,
+      "step": 27996
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 4.6735074626865674e-05,
+      "loss": 0.0,
+      "step": 28000
+    },
+    {
+      "epoch": 26.12,
+      "eval_exact_match": 0.7437137330754352,
+      "eval_exec": 0.7785299806576402,
+      "eval_loss": 0.38033804297447205,
+      "eval_runtime": 1184.2263,
+      "eval_samples_per_second": 0.873,
+      "step": 28000
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 4.673460820895523e-05,
+      "loss": 0.0016,
+      "step": 28004
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 4.673414179104478e-05,
+      "loss": 0.0084,
+      "step": 28008
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 4.6733675373134325e-05,
+      "loss": 0.0003,
+      "step": 28012
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 4.6733208955223887e-05,
+      "loss": 0.0001,
+      "step": 28016
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 4.6732742537313435e-05,
+      "loss": 0.0006,
+      "step": 28020
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 4.673227611940298e-05,
+      "loss": 0.0018,
+      "step": 28024
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 4.673180970149254e-05,
+      "loss": 0.0001,
+      "step": 28028
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 4.673134328358209e-05,
+      "loss": 0.0002,
+      "step": 28032
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 4.673087686567165e-05,
+      "loss": 0.0003,
+      "step": 28036
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 4.6730410447761196e-05,
+      "loss": 0.0005,
+      "step": 28040
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 4.672994402985075e-05,
+      "loss": 0.0001,
+      "step": 28044
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 4.6729477611940305e-05,
+      "loss": 0.0034,
+      "step": 28048
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 4.6729011194029854e-05,
+      "loss": 0.0013,
+      "step": 28052
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 4.67285447761194e-05,
+      "loss": 0.0001,
+      "step": 28056
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 4.672807835820896e-05,
+      "loss": 0.0014,
+      "step": 28060
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 4.672761194029851e-05,
+      "loss": 0.0001,
+      "step": 28064
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 4.672714552238806e-05,
+      "loss": 0.0106,
+      "step": 28068
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 4.6726679104477615e-05,
+      "loss": 0.0022,
+      "step": 28072
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 4.672621268656717e-05,
+      "loss": 0.0007,
+      "step": 28076
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 4.672574626865672e-05,
+      "loss": 0.0005,
+      "step": 28080
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 4.672527985074627e-05,
+      "loss": 0.0035,
+      "step": 28084
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 4.672481343283582e-05,
+      "loss": 0.0045,
+      "step": 28088
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 4.6724347014925376e-05,
+      "loss": 0.0002,
+      "step": 28092
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 4.672388059701493e-05,
+      "loss": 0.0033,
+      "step": 28096
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 4.672341417910448e-05,
+      "loss": 0.001,
+      "step": 28100
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 4.6722947761194034e-05,
+      "loss": 0.0002,
+      "step": 28104
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 4.672248134328359e-05,
+      "loss": 0.0014,
+      "step": 28108
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 4.6722014925373137e-05,
+      "loss": 0.0095,
+      "step": 28112
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 4.6721548507462685e-05,
+      "loss": 0.0016,
+      "step": 28116
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 4.672108208955224e-05,
+      "loss": 0.0088,
+      "step": 28120
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 4.6720615671641794e-05,
+      "loss": 0.0013,
+      "step": 28124
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 4.672014925373134e-05,
+      "loss": 0.0004,
+      "step": 28128
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 4.67196828358209e-05,
+      "loss": 0.0004,
+      "step": 28132
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 4.671921641791045e-05,
+      "loss": 0.0002,
+      "step": 28136
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 4.671875e-05,
+      "loss": 0.0006,
+      "step": 28140
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 4.6718283582089555e-05,
+      "loss": 0.0003,
+      "step": 28144
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 4.6717817164179104e-05,
+      "loss": 0.0004,
+      "step": 28148
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 4.671735074626866e-05,
+      "loss": 0.0003,
+      "step": 28152
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 4.6716884328358213e-05,
+      "loss": 0.0008,
+      "step": 28156
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 4.671641791044776e-05,
+      "loss": 0.0,
+      "step": 28160
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 4.6715951492537316e-05,
+      "loss": 0.0065,
+      "step": 28164
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 4.671548507462687e-05,
+      "loss": 0.0005,
+      "step": 28168
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 4.671501865671642e-05,
+      "loss": 0.0013,
+      "step": 28172
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 4.671455223880597e-05,
+      "loss": 0.001,
+      "step": 28176
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 4.671408582089552e-05,
+      "loss": 0.0159,
+      "step": 28180
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 4.671361940298508e-05,
+      "loss": 0.0004,
+      "step": 28184
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 4.6713152985074626e-05,
+      "loss": 0.0047,
+      "step": 28188
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 4.671268656716418e-05,
+      "loss": 0.0001,
+      "step": 28192
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 4.6712220149253735e-05,
+      "loss": 0.0007,
+      "step": 28196
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 4.671175373134329e-05,
+      "loss": 0.0003,
+      "step": 28200
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 4.671128731343284e-05,
+      "loss": 0.0004,
+      "step": 28204
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 4.6710820895522387e-05,
+      "loss": 0.0002,
+      "step": 28208
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 4.671035447761195e-05,
+      "loss": 0.0001,
+      "step": 28212
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 4.6709888059701496e-05,
+      "loss": 0.0028,
+      "step": 28216
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 4.6709421641791045e-05,
+      "loss": 0.0002,
+      "step": 28220
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 4.67089552238806e-05,
+      "loss": 0.0002,
+      "step": 28224
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 4.6708488805970154e-05,
+      "loss": 0.001,
+      "step": 28228
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 4.67080223880597e-05,
+      "loss": 0.0003,
+      "step": 28232
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 4.670755597014926e-05,
+      "loss": 0.0005,
+      "step": 28236
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 4.6707089552238805e-05,
+      "loss": 0.0007,
+      "step": 28240
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 4.670662313432836e-05,
+      "loss": 0.0002,
+      "step": 28244
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 4.6706156716417915e-05,
+      "loss": 0.0001,
+      "step": 28248
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 4.6705690298507463e-05,
+      "loss": 0.0002,
+      "step": 28252
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 4.670522388059702e-05,
+      "loss": 0.0017,
+      "step": 28256
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 4.670475746268657e-05,
+      "loss": 0.0005,
+      "step": 28260
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 4.670429104477612e-05,
+      "loss": 0.0003,
+      "step": 28264
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 4.670382462686567e-05,
+      "loss": 0.0,
+      "step": 28268
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 4.670335820895523e-05,
+      "loss": 0.0008,
+      "step": 28272
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 4.670289179104478e-05,
+      "loss": 0.0004,
+      "step": 28276
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 4.670242537313433e-05,
+      "loss": 0.0003,
+      "step": 28280
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 4.670195895522388e-05,
+      "loss": 0.0003,
+      "step": 28284
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 4.670149253731344e-05,
+      "loss": 0.0025,
+      "step": 28288
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 4.6701026119402985e-05,
+      "loss": 0.0014,
+      "step": 28292
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 4.670055970149254e-05,
+      "loss": 0.0002,
+      "step": 28296
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 4.670009328358209e-05,
+      "loss": 0.0002,
+      "step": 28300
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 4.669962686567164e-05,
+      "loss": 0.0004,
+      "step": 28304
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 4.66991604477612e-05,
+      "loss": 0.0002,
+      "step": 28308
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 4.6698694029850746e-05,
+      "loss": 0.0004,
+      "step": 28312
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 4.66982276119403e-05,
+      "loss": 0.0001,
+      "step": 28316
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 4.6697761194029856e-05,
+      "loss": 0.0004,
+      "step": 28320
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 4.6697294776119404e-05,
+      "loss": 0.0008,
+      "step": 28324
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 4.669682835820895e-05,
+      "loss": 0.0024,
+      "step": 28328
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 4.6696361940298514e-05,
+      "loss": 0.0005,
+      "step": 28332
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 4.669589552238806e-05,
+      "loss": 0.006,
+      "step": 28336
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 4.669542910447761e-05,
+      "loss": 0.0046,
+      "step": 28340
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 4.6694962686567165e-05,
+      "loss": 0.0001,
+      "step": 28344
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 4.669449626865672e-05,
+      "loss": 0.0009,
+      "step": 28348
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 4.669402985074627e-05,
+      "loss": 0.0006,
+      "step": 28352
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 4.669356343283582e-05,
+      "loss": 0.0001,
+      "step": 28356
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 4.669309701492537e-05,
+      "loss": 0.0001,
+      "step": 28360
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 4.669263059701493e-05,
+      "loss": 0.0004,
+      "step": 28364
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 4.669216417910448e-05,
+      "loss": 0.0003,
+      "step": 28368
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 4.669169776119403e-05,
+      "loss": 0.0002,
+      "step": 28372
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 4.6691231343283584e-05,
+      "loss": 0.0001,
+      "step": 28376
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 4.669076492537314e-05,
+      "loss": 0.0043,
+      "step": 28380
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 4.669029850746269e-05,
+      "loss": 0.0052,
+      "step": 28384
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 4.668983208955224e-05,
+      "loss": 0.0012,
+      "step": 28388
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 4.66893656716418e-05,
+      "loss": 0.0005,
+      "step": 28392
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 4.6688899253731345e-05,
+      "loss": 0.001,
+      "step": 28396
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 4.66884328358209e-05,
+      "loss": 0.0001,
+      "step": 28400
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 4.668796641791045e-05,
+      "loss": 0.0012,
+      "step": 28404
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 4.66875e-05,
+      "loss": 0.0002,
+      "step": 28408
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 4.668703358208956e-05,
+      "loss": 0.0002,
+      "step": 28412
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 4.6686567164179106e-05,
+      "loss": 0.0001,
+      "step": 28416
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 4.6686100746268654e-05,
+      "loss": 0.0001,
+      "step": 28420
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 4.6685634328358216e-05,
+      "loss": 0.0009,
+      "step": 28424
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 4.6685167910447764e-05,
+      "loss": 0.0003,
+      "step": 28428
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 4.668470149253731e-05,
+      "loss": 0.0001,
+      "step": 28432
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 4.668423507462687e-05,
+      "loss": 0.0021,
+      "step": 28436
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 4.668376865671642e-05,
+      "loss": 0.0024,
+      "step": 28440
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 4.668330223880597e-05,
+      "loss": 0.001,
+      "step": 28444
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 4.6682835820895525e-05,
+      "loss": 0.0006,
+      "step": 28448
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 4.668236940298507e-05,
+      "loss": 0.0001,
+      "step": 28452
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 4.668190298507463e-05,
+      "loss": 0.0005,
+      "step": 28456
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 4.668143656716418e-05,
+      "loss": 0.0041,
+      "step": 28460
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 4.668097014925373e-05,
+      "loss": 0.0049,
+      "step": 28464
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 4.6680503731343286e-05,
+      "loss": 0.0023,
+      "step": 28468
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 4.668003731343284e-05,
+      "loss": 0.0006,
+      "step": 28472
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 4.667957089552239e-05,
+      "loss": 0.0001,
+      "step": 28476
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 4.667910447761194e-05,
+      "loss": 0.0009,
+      "step": 28480
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 4.66786380597015e-05,
+      "loss": 0.0008,
+      "step": 28484
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 4.667817164179105e-05,
+      "loss": 0.0014,
+      "step": 28488
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 4.6677705223880595e-05,
+      "loss": 0.0002,
+      "step": 28492
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 4.667723880597015e-05,
+      "loss": 0.001,
+      "step": 28496
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 4.6676772388059705e-05,
+      "loss": 0.0014,
+      "step": 28500
+    },
+    {
+      "epoch": 26.59,
+      "eval_exact_match": 0.7321083172147002,
+      "eval_exec": 0.7688588007736944,
+      "eval_loss": 0.3977905809879303,
+      "eval_runtime": 1155.6229,
+      "eval_samples_per_second": 0.895,
+      "step": 28500
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 4.667630597014925e-05,
+      "loss": 0.0002,
+      "step": 28504
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 4.667583955223881e-05,
+      "loss": 0.0001,
+      "step": 28508
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 4.6675373134328356e-05,
+      "loss": 0.0006,
+      "step": 28512
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 4.667490671641792e-05,
+      "loss": 0.0018,
+      "step": 28516
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 4.6674440298507466e-05,
+      "loss": 0.0003,
+      "step": 28520
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 4.6673973880597014e-05,
+      "loss": 0.0005,
+      "step": 28524
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 4.667350746268657e-05,
+      "loss": 0.0001,
+      "step": 28528
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 4.6673041044776124e-05,
+      "loss": 0.0022,
+      "step": 28532
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 4.667257462686567e-05,
+      "loss": 0.0006,
+      "step": 28536
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 4.667210820895523e-05,
+      "loss": 0.0002,
+      "step": 28540
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 4.667164179104478e-05,
+      "loss": 0.0001,
+      "step": 28544
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 4.667117537313433e-05,
+      "loss": 0.0001,
+      "step": 28548
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 4.6670708955223885e-05,
+      "loss": 0.0036,
+      "step": 28552
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 4.667024253731343e-05,
+      "loss": 0.0002,
+      "step": 28556
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 4.666977611940299e-05,
+      "loss": 0.0001,
+      "step": 28560
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 4.666930970149254e-05,
+      "loss": 0.0002,
+      "step": 28564
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 4.666884328358209e-05,
+      "loss": 0.001,
+      "step": 28568
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 4.666837686567164e-05,
+      "loss": 0.0032,
+      "step": 28572
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 4.66679104477612e-05,
+      "loss": 0.0002,
+      "step": 28576
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 4.666744402985075e-05,
+      "loss": 0.0015,
+      "step": 28580
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 4.66669776119403e-05,
+      "loss": 0.0034,
+      "step": 28584
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 4.666651119402985e-05,
+      "loss": 0.0009,
+      "step": 28588
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 4.666604477611941e-05,
+      "loss": 0.0019,
+      "step": 28592
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 4.6665578358208955e-05,
+      "loss": 0.0029,
+      "step": 28596
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 4.666511194029851e-05,
+      "loss": 0.0,
+      "step": 28600
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 4.6664645522388065e-05,
+      "loss": 0.0006,
+      "step": 28604
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 4.666417910447761e-05,
+      "loss": 0.001,
+      "step": 28608
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 4.666371268656717e-05,
+      "loss": 0.0,
+      "step": 28612
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 4.6663246268656716e-05,
+      "loss": 0.0001,
+      "step": 28616
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 4.666277985074627e-05,
+      "loss": 0.003,
+      "step": 28620
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 4.6662313432835826e-05,
+      "loss": 0.0009,
+      "step": 28624
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 4.6661847014925374e-05,
+      "loss": 0.0001,
+      "step": 28628
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 4.666138059701492e-05,
+      "loss": 0.0007,
+      "step": 28632
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 4.6660914179104484e-05,
+      "loss": 0.0061,
+      "step": 28636
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 4.666044776119403e-05,
+      "loss": 0.0041,
+      "step": 28640
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 4.665998134328358e-05,
+      "loss": 0.0067,
+      "step": 28644
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 4.6659514925373135e-05,
+      "loss": 0.0009,
+      "step": 28648
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 4.665904850746269e-05,
+      "loss": 0.0003,
+      "step": 28652
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 4.665858208955224e-05,
+      "loss": 0.0005,
+      "step": 28656
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 4.665811567164179e-05,
+      "loss": 0.0002,
+      "step": 28660
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 4.665764925373135e-05,
+      "loss": 0.0002,
+      "step": 28664
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 4.6657182835820896e-05,
+      "loss": 0.0076,
+      "step": 28668
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 4.665671641791045e-05,
+      "loss": 0.0014,
+      "step": 28672
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 4.665625e-05,
+      "loss": 0.0002,
+      "step": 28676
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 4.6655783582089554e-05,
+      "loss": 0.0002,
+      "step": 28680
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 4.665531716417911e-05,
+      "loss": 0.0006,
+      "step": 28684
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 4.665485074626866e-05,
+      "loss": 0.0048,
+      "step": 28688
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 4.665438432835821e-05,
+      "loss": 0.0002,
+      "step": 28692
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 4.665391791044777e-05,
+      "loss": 0.0012,
+      "step": 28696
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 4.6653451492537315e-05,
+      "loss": 0.0002,
+      "step": 28700
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 4.665298507462687e-05,
+      "loss": 0.0004,
+      "step": 28704
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 4.665251865671642e-05,
+      "loss": 0.0001,
+      "step": 28708
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 4.665205223880597e-05,
+      "loss": 0.0001,
+      "step": 28712
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 4.665158582089553e-05,
+      "loss": 0.0023,
+      "step": 28716
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 4.6651119402985076e-05,
+      "loss": 0.0012,
+      "step": 28720
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 4.665065298507463e-05,
+      "loss": 0.0002,
+      "step": 28724
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 4.6650186567164186e-05,
+      "loss": 0.0015,
+      "step": 28728
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 4.6649720149253734e-05,
+      "loss": 0.0008,
+      "step": 28732
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 4.664925373134328e-05,
+      "loss": 0.0019,
+      "step": 28736
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 4.664878731343284e-05,
+      "loss": 0.0001,
+      "step": 28740
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 4.664832089552239e-05,
+      "loss": 0.0007,
+      "step": 28744
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 4.664785447761194e-05,
+      "loss": 0.0013,
+      "step": 28748
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 4.6647388059701495e-05,
+      "loss": 0.0018,
+      "step": 28752
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 4.664692164179105e-05,
+      "loss": 0.0005,
+      "step": 28756
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 4.66464552238806e-05,
+      "loss": 0.0042,
+      "step": 28760
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 4.664598880597015e-05,
+      "loss": 0.0186,
+      "step": 28764
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 4.66455223880597e-05,
+      "loss": 0.0004,
+      "step": 28768
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 4.6645055970149256e-05,
+      "loss": 0.0026,
+      "step": 28772
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 4.664458955223881e-05,
+      "loss": 0.0029,
+      "step": 28776
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 4.664412313432836e-05,
+      "loss": 0.0073,
+      "step": 28780
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 4.6643656716417914e-05,
+      "loss": 0.0005,
+      "step": 28784
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 4.664319029850747e-05,
+      "loss": 0.0006,
+      "step": 28788
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 4.664272388059702e-05,
+      "loss": 0.0002,
+      "step": 28792
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 4.6642257462686565e-05,
+      "loss": 0.0004,
+      "step": 28796
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 4.664179104477612e-05,
+      "loss": 0.0002,
+      "step": 28800
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 4.6641324626865675e-05,
+      "loss": 0.0021,
+      "step": 28804
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 4.664085820895522e-05,
+      "loss": 0.0002,
+      "step": 28808
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 4.664039179104478e-05,
+      "loss": 0.0009,
+      "step": 28812
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 4.663992537313433e-05,
+      "loss": 0.0011,
+      "step": 28816
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 4.663945895522388e-05,
+      "loss": 0.0003,
+      "step": 28820
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 4.6638992537313436e-05,
+      "loss": 0.0001,
+      "step": 28824
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 4.6638526119402984e-05,
+      "loss": 0.0001,
+      "step": 28828
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 4.663805970149254e-05,
+      "loss": 0.0011,
+      "step": 28832
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 4.6637593283582094e-05,
+      "loss": 0.0003,
+      "step": 28836
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 4.663712686567164e-05,
+      "loss": 0.0001,
+      "step": 28840
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 4.66366604477612e-05,
+      "loss": 0.0015,
+      "step": 28844
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 4.663619402985075e-05,
+      "loss": 0.0002,
+      "step": 28848
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 4.66357276119403e-05,
+      "loss": 0.0007,
+      "step": 28852
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 4.6635261194029855e-05,
+      "loss": 0.001,
+      "step": 28856
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 4.66347947761194e-05,
+      "loss": 0.0003,
+      "step": 28860
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 4.663432835820896e-05,
+      "loss": 0.0026,
+      "step": 28864
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 4.663386194029851e-05,
+      "loss": 0.0001,
+      "step": 28868
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 4.663339552238806e-05,
+      "loss": 0.0007,
+      "step": 28872
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 4.6632929104477616e-05,
+      "loss": 0.0001,
+      "step": 28876
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 4.663246268656717e-05,
+      "loss": 0.0009,
+      "step": 28880
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 4.663199626865672e-05,
+      "loss": 0.0006,
+      "step": 28884
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 4.663152985074627e-05,
+      "loss": 0.0013,
+      "step": 28888
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 4.663106343283583e-05,
+      "loss": 0.001,
+      "step": 28892
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 4.663059701492538e-05,
+      "loss": 0.0058,
+      "step": 28896
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 4.6630130597014925e-05,
+      "loss": 0.0191,
+      "step": 28900
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 4.662966417910448e-05,
+      "loss": 0.0006,
+      "step": 28904
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 4.6629197761194035e-05,
+      "loss": 0.0017,
+      "step": 28908
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 4.662873134328358e-05,
+      "loss": 0.0013,
+      "step": 28912
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 4.662826492537314e-05,
+      "loss": 0.0019,
+      "step": 28916
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 4.6627798507462686e-05,
+      "loss": 0.0083,
+      "step": 28920
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 4.662733208955224e-05,
+      "loss": 0.0022,
+      "step": 28924
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 4.6626865671641796e-05,
+      "loss": 0.0003,
+      "step": 28928
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 4.6626399253731344e-05,
+      "loss": 0.0031,
+      "step": 28932
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 4.66259328358209e-05,
+      "loss": 0.0046,
+      "step": 28936
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 4.6625466417910454e-05,
+      "loss": 0.0008,
+      "step": 28940
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 4.6625e-05,
+      "loss": 0.0012,
+      "step": 28944
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 4.662453358208955e-05,
+      "loss": 0.0001,
+      "step": 28948
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 4.662406716417911e-05,
+      "loss": 0.0021,
+      "step": 28952
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 4.662360074626866e-05,
+      "loss": 0.0017,
+      "step": 28956
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 4.662313432835821e-05,
+      "loss": 0.0023,
+      "step": 28960
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 4.662266791044776e-05,
+      "loss": 0.0001,
+      "step": 28964
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 4.662220149253732e-05,
+      "loss": 0.001,
+      "step": 28968
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 4.6621735074626866e-05,
+      "loss": 0.0003,
+      "step": 28972
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 4.662126865671642e-05,
+      "loss": 0.0002,
+      "step": 28976
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 4.662080223880597e-05,
+      "loss": 0.0003,
+      "step": 28980
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 4.6620335820895524e-05,
+      "loss": 0.0007,
+      "step": 28984
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 4.661986940298508e-05,
+      "loss": 0.0008,
+      "step": 28988
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 4.661940298507463e-05,
+      "loss": 0.0025,
+      "step": 28992
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 4.661893656716418e-05,
+      "loss": 0.0004,
+      "step": 28996
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 4.6618470149253737e-05,
+      "loss": 0.0005,
+      "step": 29000
+    },
+    {
+      "epoch": 27.05,
+      "eval_exact_match": 0.7330754352030948,
+      "eval_exec": 0.7736943907156673,
+      "eval_loss": 0.374284952878952,
+      "eval_runtime": 1172.5719,
+      "eval_samples_per_second": 0.882,
+      "step": 29000
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 4.6618003731343285e-05,
+      "loss": 0.0004,
+      "step": 29004
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 4.661753731343284e-05,
+      "loss": 0.0003,
+      "step": 29008
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 4.6617070895522394e-05,
+      "loss": 0.0004,
+      "step": 29012
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 4.661660447761194e-05,
+      "loss": 0.0001,
+      "step": 29016
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 4.66161380597015e-05,
+      "loss": 0.0004,
+      "step": 29020
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 4.6615671641791046e-05,
+      "loss": 0.0005,
+      "step": 29024
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 4.66152052238806e-05,
+      "loss": 0.0149,
+      "step": 29028
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 4.6614738805970155e-05,
+      "loss": 0.0012,
+      "step": 29032
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 4.6614272388059704e-05,
+      "loss": 0.0,
+      "step": 29036
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 4.661380597014925e-05,
+      "loss": 0.0001,
+      "step": 29040
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 4.661333955223881e-05,
+      "loss": 0.0014,
+      "step": 29044
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 4.661287313432836e-05,
+      "loss": 0.0009,
+      "step": 29048
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 4.661240671641791e-05,
+      "loss": 0.003,
+      "step": 29052
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 4.6611940298507465e-05,
+      "loss": 0.0019,
+      "step": 29056
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 4.661147388059702e-05,
+      "loss": 0.0001,
+      "step": 29060
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 4.661100746268657e-05,
+      "loss": 0.0001,
+      "step": 29064
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 4.661054104477612e-05,
+      "loss": 0.0005,
+      "step": 29068
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 4.661007462686568e-05,
+      "loss": 0.0,
+      "step": 29072
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 4.6609608208955226e-05,
+      "loss": 0.0,
+      "step": 29076
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 4.660914179104478e-05,
+      "loss": 0.0008,
+      "step": 29080
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 4.660867537313433e-05,
+      "loss": 0.0003,
+      "step": 29084
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 4.6608208955223883e-05,
+      "loss": 0.0001,
+      "step": 29088
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 4.660774253731344e-05,
+      "loss": 0.001,
+      "step": 29092
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 4.6607276119402987e-05,
+      "loss": 0.0016,
+      "step": 29096
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 4.6606809701492535e-05,
+      "loss": 0.0016,
+      "step": 29100
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 4.6606343283582096e-05,
+      "loss": 0.0003,
+      "step": 29104
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 4.6605876865671644e-05,
+      "loss": 0.0002,
+      "step": 29108
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 4.660541044776119e-05,
+      "loss": 0.0002,
+      "step": 29112
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 4.660494402985075e-05,
+      "loss": 0.0001,
+      "step": 29116
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 4.66044776119403e-05,
+      "loss": 0.0,
+      "step": 29120
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 4.660401119402985e-05,
+      "loss": 0.0002,
+      "step": 29124
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 4.6603544776119405e-05,
+      "loss": 0.0005,
+      "step": 29128
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 4.6603078358208954e-05,
+      "loss": 0.0001,
+      "step": 29132
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 4.660261194029851e-05,
+      "loss": 0.0,
+      "step": 29136
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 4.6602145522388063e-05,
+      "loss": 0.0,
+      "step": 29140
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 4.660167910447761e-05,
+      "loss": 0.0002,
+      "step": 29144
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 4.6601212686567166e-05,
+      "loss": 0.0005,
+      "step": 29148
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 4.660074626865672e-05,
+      "loss": 0.0007,
+      "step": 29152
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 4.660027985074627e-05,
+      "loss": 0.0002,
+      "step": 29156
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 4.659981343283582e-05,
+      "loss": 0.0,
+      "step": 29160
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 4.659934701492538e-05,
+      "loss": 0.0026,
+      "step": 29164
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 4.659888059701493e-05,
+      "loss": 0.0001,
+      "step": 29168
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 4.659841417910448e-05,
+      "loss": 0.0012,
+      "step": 29172
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 4.659794776119403e-05,
+      "loss": 0.0158,
+      "step": 29176
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 4.6597481343283585e-05,
+      "loss": 0.0003,
+      "step": 29180
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 4.659701492537314e-05,
+      "loss": 0.0022,
+      "step": 29184
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 4.659654850746269e-05,
+      "loss": 0.0002,
+      "step": 29188
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 4.6596082089552237e-05,
+      "loss": 0.0001,
+      "step": 29192
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 4.65956156716418e-05,
+      "loss": 0.0028,
+      "step": 29196
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 4.6595149253731346e-05,
+      "loss": 0.0001,
+      "step": 29200
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 4.6594682835820894e-05,
+      "loss": 0.0009,
+      "step": 29204
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 4.659421641791045e-05,
+      "loss": 0.0,
+      "step": 29208
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 4.6593750000000004e-05,
+      "loss": 0.0012,
+      "step": 29212
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 4.659328358208955e-05,
+      "loss": 0.0034,
+      "step": 29216
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 4.659281716417911e-05,
+      "loss": 0.0001,
+      "step": 29220
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 4.659235074626866e-05,
+      "loss": 0.0,
+      "step": 29224
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 4.659188432835821e-05,
+      "loss": 0.0002,
+      "step": 29228
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 4.6591417910447765e-05,
+      "loss": 0.0052,
+      "step": 29232
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 4.6590951492537313e-05,
+      "loss": 0.0002,
+      "step": 29236
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 4.659048507462687e-05,
+      "loss": 0.0,
+      "step": 29240
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 4.659001865671642e-05,
+      "loss": 0.0001,
+      "step": 29244
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 4.658955223880597e-05,
+      "loss": 0.0018,
+      "step": 29248
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 4.658908582089552e-05,
+      "loss": 0.0002,
+      "step": 29252
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 4.658861940298508e-05,
+      "loss": 0.0019,
+      "step": 29256
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 4.658815298507463e-05,
+      "loss": 0.0004,
+      "step": 29260
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 4.658768656716418e-05,
+      "loss": 0.0008,
+      "step": 29264
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 4.658722014925373e-05,
+      "loss": 0.0001,
+      "step": 29268
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 4.658675373134329e-05,
+      "loss": 0.0013,
+      "step": 29272
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 4.6586287313432835e-05,
+      "loss": 0.0,
+      "step": 29276
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 4.658582089552239e-05,
+      "loss": 0.0001,
+      "step": 29280
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 4.6585354477611945e-05,
+      "loss": 0.0021,
+      "step": 29284
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 4.658488805970149e-05,
+      "loss": 0.0001,
+      "step": 29288
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 4.658442164179105e-05,
+      "loss": 0.0001,
+      "step": 29292
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 4.6583955223880596e-05,
+      "loss": 0.0003,
+      "step": 29296
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 4.658348880597015e-05,
+      "loss": 0.0001,
+      "step": 29300
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 4.6583022388059706e-05,
+      "loss": 0.0001,
+      "step": 29304
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 4.6582555970149254e-05,
+      "loss": 0.0,
+      "step": 29308
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 4.65820895522388e-05,
+      "loss": 0.0001,
+      "step": 29312
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 4.6581623134328364e-05,
+      "loss": 0.0003,
+      "step": 29316
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 4.658115671641791e-05,
+      "loss": 0.0,
+      "step": 29320
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 4.658069029850746e-05,
+      "loss": 0.0,
+      "step": 29324
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 4.6580223880597015e-05,
+      "loss": 0.0042,
+      "step": 29328
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 4.657975746268657e-05,
+      "loss": 0.0003,
+      "step": 29332
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 4.6579291044776125e-05,
+      "loss": 0.0,
+      "step": 29336
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 4.657882462686567e-05,
+      "loss": 0.0093,
+      "step": 29340
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 4.657835820895523e-05,
+      "loss": 0.0035,
+      "step": 29344
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 4.657789179104478e-05,
+      "loss": 0.004,
+      "step": 29348
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 4.657742537313433e-05,
+      "loss": 0.0006,
+      "step": 29352
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 4.657695895522388e-05,
+      "loss": 0.0009,
+      "step": 29356
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 4.6576492537313434e-05,
+      "loss": 0.0006,
+      "step": 29360
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 4.657602611940299e-05,
+      "loss": 0.0001,
+      "step": 29364
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 4.657555970149254e-05,
+      "loss": 0.0001,
+      "step": 29368
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 4.657509328358209e-05,
+      "loss": 0.0066,
+      "step": 29372
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 4.657462686567165e-05,
+      "loss": 0.0003,
+      "step": 29376
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 4.6574160447761195e-05,
+      "loss": 0.0004,
+      "step": 29380
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 4.657369402985075e-05,
+      "loss": 0.003,
+      "step": 29384
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 4.65732276119403e-05,
+      "loss": 0.0017,
+      "step": 29388
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 4.657276119402985e-05,
+      "loss": 0.002,
+      "step": 29392
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 4.657229477611941e-05,
+      "loss": 0.0001,
+      "step": 29396
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 4.6571828358208956e-05,
+      "loss": 0.0082,
+      "step": 29400
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 4.657136194029851e-05,
+      "loss": 0.001,
+      "step": 29404
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 4.6570895522388066e-05,
+      "loss": 0.0002,
+      "step": 29408
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 4.6570429104477614e-05,
+      "loss": 0.0013,
+      "step": 29412
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 4.656996268656716e-05,
+      "loss": 0.0003,
+      "step": 29416
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 4.656949626865672e-05,
+      "loss": 0.0004,
+      "step": 29420
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 4.656902985074627e-05,
+      "loss": 0.0001,
+      "step": 29424
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 4.656856343283582e-05,
+      "loss": 0.0029,
+      "step": 29428
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 4.6568097014925375e-05,
+      "loss": 0.0001,
+      "step": 29432
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 4.656763059701493e-05,
+      "loss": 0.0055,
+      "step": 29436
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 4.656716417910448e-05,
+      "loss": 0.0005,
+      "step": 29440
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 4.656669776119403e-05,
+      "loss": 0.0002,
+      "step": 29444
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 4.656623134328358e-05,
+      "loss": 0.0033,
+      "step": 29448
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 4.6565764925373136e-05,
+      "loss": 0.0005,
+      "step": 29452
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 4.656529850746269e-05,
+      "loss": 0.0002,
+      "step": 29456
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 4.656483208955224e-05,
+      "loss": 0.0003,
+      "step": 29460
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 4.6564365671641794e-05,
+      "loss": 0.0001,
+      "step": 29464
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 4.656389925373135e-05,
+      "loss": 0.001,
+      "step": 29468
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 4.65634328358209e-05,
+      "loss": 0.0001,
+      "step": 29472
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 4.6562966417910445e-05,
+      "loss": 0.0062,
+      "step": 29476
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 4.65625e-05,
+      "loss": 0.0055,
+      "step": 29480
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 4.6562033582089555e-05,
+      "loss": 0.0001,
+      "step": 29484
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 4.65615671641791e-05,
+      "loss": 0.0002,
+      "step": 29488
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 4.656110074626866e-05,
+      "loss": 0.0049,
+      "step": 29492
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 4.656063432835821e-05,
+      "loss": 0.0002,
+      "step": 29496
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 4.656016791044777e-05,
+      "loss": 0.0001,
+      "step": 29500
+    },
+    {
+      "epoch": 27.52,
+      "eval_exact_match": 0.741779497098646,
+      "eval_exec": 0.769825918762089,
+      "eval_loss": 0.3933384418487549,
+      "eval_runtime": 1168.2038,
+      "eval_samples_per_second": 0.885,
+      "step": 29500
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 4.6559701492537316e-05,
+      "loss": 0.0001,
+      "step": 29504
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 4.6559235074626864e-05,
+      "loss": 0.0025,
+      "step": 29508
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 4.6558768656716426e-05,
+      "loss": 0.0003,
+      "step": 29512
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 4.6558302238805974e-05,
+      "loss": 0.0003,
+      "step": 29516
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 4.655783582089552e-05,
+      "loss": 0.0016,
+      "step": 29520
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 4.655736940298508e-05,
+      "loss": 0.0001,
+      "step": 29524
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 4.655690298507463e-05,
+      "loss": 0.0007,
+      "step": 29528
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 4.655643656716418e-05,
+      "loss": 0.0013,
+      "step": 29532
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 4.6555970149253735e-05,
+      "loss": 0.0001,
+      "step": 29536
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 4.655550373134328e-05,
+      "loss": 0.0009,
+      "step": 29540
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 4.655503731343284e-05,
+      "loss": 0.0015,
+      "step": 29544
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 4.655457089552239e-05,
+      "loss": 0.0002,
+      "step": 29548
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 4.655410447761194e-05,
+      "loss": 0.0001,
+      "step": 29552
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 4.6553638059701496e-05,
+      "loss": 0.001,
+      "step": 29556
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 4.655317164179105e-05,
+      "loss": 0.0001,
+      "step": 29560
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 4.65527052238806e-05,
+      "loss": 0.0001,
+      "step": 29564
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 4.655223880597015e-05,
+      "loss": 0.0067,
+      "step": 29568
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 4.655177238805971e-05,
+      "loss": 0.0029,
+      "step": 29572
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 4.655130597014926e-05,
+      "loss": 0.0007,
+      "step": 29576
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 4.6550839552238805e-05,
+      "loss": 0.0004,
+      "step": 29580
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 4.655037313432836e-05,
+      "loss": 0.0001,
+      "step": 29584
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 4.6549906716417915e-05,
+      "loss": 0.0002,
+      "step": 29588
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 4.654944029850746e-05,
+      "loss": 0.0001,
+      "step": 29592
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 4.654897388059702e-05,
+      "loss": 0.0003,
+      "step": 29596
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 4.6548507462686566e-05,
+      "loss": 0.0017,
+      "step": 29600
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 4.654804104477612e-05,
+      "loss": 0.0031,
+      "step": 29604
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 4.6547574626865676e-05,
+      "loss": 0.0004,
+      "step": 29608
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 4.6547108208955224e-05,
+      "loss": 0.0003,
+      "step": 29612
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 4.654664179104478e-05,
+      "loss": 0.0002,
+      "step": 29616
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 4.6546175373134334e-05,
+      "loss": 0.0034,
+      "step": 29620
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 4.654570895522388e-05,
+      "loss": 0.0008,
+      "step": 29624
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 4.654524253731343e-05,
+      "loss": 0.0001,
+      "step": 29628
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 4.654477611940299e-05,
+      "loss": 0.0001,
+      "step": 29632
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 4.654430970149254e-05,
+      "loss": 0.0002,
+      "step": 29636
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 4.654384328358209e-05,
+      "loss": 0.0035,
+      "step": 29640
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 4.654337686567164e-05,
+      "loss": 0.0001,
+      "step": 29644
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 4.65429104477612e-05,
+      "loss": 0.0001,
+      "step": 29648
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 4.6542444029850746e-05,
+      "loss": 0.0001,
+      "step": 29652
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 4.65419776119403e-05,
+      "loss": 0.0007,
+      "step": 29656
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 4.654151119402985e-05,
+      "loss": 0.0005,
+      "step": 29660
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 4.654104477611941e-05,
+      "loss": 0.0015,
+      "step": 29664
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 4.654057835820896e-05,
+      "loss": 0.0003,
+      "step": 29668
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 4.654011194029851e-05,
+      "loss": 0.0001,
+      "step": 29672
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 4.653964552238806e-05,
+      "loss": 0.0027,
+      "step": 29676
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 4.653917910447762e-05,
+      "loss": 0.0002,
+      "step": 29680
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 4.6538712686567165e-05,
+      "loss": 0.0016,
+      "step": 29684
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 4.653824626865672e-05,
+      "loss": 0.0001,
+      "step": 29688
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 4.6537779850746275e-05,
+      "loss": 0.0004,
+      "step": 29692
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 4.653731343283582e-05,
+      "loss": 0.0001,
+      "step": 29696
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 4.653684701492538e-05,
+      "loss": 0.0019,
+      "step": 29700
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 4.6536380597014926e-05,
+      "loss": 0.0015,
+      "step": 29704
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 4.653591417910448e-05,
+      "loss": 0.0001,
+      "step": 29708
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 4.6535447761194036e-05,
+      "loss": 0.0005,
+      "step": 29712
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 4.6534981343283584e-05,
+      "loss": 0.0011,
+      "step": 29716
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 4.653451492537313e-05,
+      "loss": 0.0086,
+      "step": 29720
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 4.6534048507462694e-05,
+      "loss": 0.0006,
+      "step": 29724
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 4.653358208955224e-05,
+      "loss": 0.0001,
+      "step": 29728
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 4.653311567164179e-05,
+      "loss": 0.001,
+      "step": 29732
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 4.6532649253731345e-05,
+      "loss": 0.0023,
+      "step": 29736
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 4.65321828358209e-05,
+      "loss": 0.0094,
+      "step": 29740
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 4.653171641791045e-05,
+      "loss": 0.0033,
+      "step": 29744
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 4.653125e-05,
+      "loss": 0.0016,
+      "step": 29748
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 4.653078358208956e-05,
+      "loss": 0.0007,
+      "step": 29752
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 4.6530317164179106e-05,
+      "loss": 0.0053,
+      "step": 29756
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 4.652985074626866e-05,
+      "loss": 0.0008,
+      "step": 29760
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 4.652938432835821e-05,
+      "loss": 0.0001,
+      "step": 29764
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 4.6528917910447764e-05,
+      "loss": 0.0009,
+      "step": 29768
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 4.652845149253732e-05,
+      "loss": 0.0,
+      "step": 29772
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 4.652798507462687e-05,
+      "loss": 0.0078,
+      "step": 29776
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 4.6527518656716415e-05,
+      "loss": 0.0015,
+      "step": 29780
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 4.652705223880598e-05,
+      "loss": 0.0003,
+      "step": 29784
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 4.6526585820895525e-05,
+      "loss": 0.0029,
+      "step": 29788
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 4.652611940298507e-05,
+      "loss": 0.0012,
+      "step": 29792
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 4.652565298507463e-05,
+      "loss": 0.0003,
+      "step": 29796
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 4.652518656716418e-05,
+      "loss": 0.0,
+      "step": 29800
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 4.652472014925373e-05,
+      "loss": 0.0024,
+      "step": 29804
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 4.6524253731343286e-05,
+      "loss": 0.0003,
+      "step": 29808
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 4.6523787313432834e-05,
+      "loss": 0.0006,
+      "step": 29812
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 4.652332089552239e-05,
+      "loss": 0.0004,
+      "step": 29816
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 4.6522854477611944e-05,
+      "loss": 0.0001,
+      "step": 29820
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 4.652238805970149e-05,
+      "loss": 0.0,
+      "step": 29824
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 4.652192164179105e-05,
+      "loss": 0.0001,
+      "step": 29828
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 4.65214552238806e-05,
+      "loss": 0.0023,
+      "step": 29832
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 4.652098880597015e-05,
+      "loss": 0.0027,
+      "step": 29836
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 4.6520522388059705e-05,
+      "loss": 0.0025,
+      "step": 29840
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 4.652005597014926e-05,
+      "loss": 0.0049,
+      "step": 29844
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 4.651958955223881e-05,
+      "loss": 0.0001,
+      "step": 29848
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 4.651912313432836e-05,
+      "loss": 0.0002,
+      "step": 29852
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 4.651865671641791e-05,
+      "loss": 0.0004,
+      "step": 29856
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 4.6518190298507466e-05,
+      "loss": 0.0039,
+      "step": 29860
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 4.651772388059702e-05,
+      "loss": 0.0001,
+      "step": 29864
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 4.651725746268657e-05,
+      "loss": 0.0002,
+      "step": 29868
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 4.651679104477612e-05,
+      "loss": 0.0003,
+      "step": 29872
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 4.651632462686568e-05,
+      "loss": 0.0078,
+      "step": 29876
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 4.651585820895523e-05,
+      "loss": 0.0001,
+      "step": 29880
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 4.6515391791044775e-05,
+      "loss": 0.0015,
+      "step": 29884
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 4.651492537313433e-05,
+      "loss": 0.0007,
+      "step": 29888
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 4.6514458955223885e-05,
+      "loss": 0.0002,
+      "step": 29892
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 4.651399253731343e-05,
+      "loss": 0.0013,
+      "step": 29896
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 4.651352611940299e-05,
+      "loss": 0.0008,
+      "step": 29900
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 4.651305970149254e-05,
+      "loss": 0.0056,
+      "step": 29904
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 4.651259328358209e-05,
+      "loss": 0.0073,
+      "step": 29908
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 4.6512126865671646e-05,
+      "loss": 0.0015,
+      "step": 29912
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 4.6511660447761194e-05,
+      "loss": 0.0005,
+      "step": 29916
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 4.651119402985075e-05,
+      "loss": 0.0001,
+      "step": 29920
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 4.6510727611940304e-05,
+      "loss": 0.0006,
+      "step": 29924
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 4.651026119402985e-05,
+      "loss": 0.0004,
+      "step": 29928
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 4.65097947761194e-05,
+      "loss": 0.0015,
+      "step": 29932
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 4.650932835820896e-05,
+      "loss": 0.002,
+      "step": 29936
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 4.650886194029851e-05,
+      "loss": 0.0002,
+      "step": 29940
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 4.650839552238806e-05,
+      "loss": 0.0006,
+      "step": 29944
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 4.650792910447761e-05,
+      "loss": 0.0062,
+      "step": 29948
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 4.650746268656717e-05,
+      "loss": 0.0003,
+      "step": 29952
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 4.6506996268656716e-05,
+      "loss": 0.0018,
+      "step": 29956
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 4.650652985074627e-05,
+      "loss": 0.0003,
+      "step": 29960
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 4.6506063432835825e-05,
+      "loss": 0.0001,
+      "step": 29964
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 4.6505597014925374e-05,
+      "loss": 0.0017,
+      "step": 29968
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 4.650513059701493e-05,
+      "loss": 0.0019,
+      "step": 29972
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 4.650466417910448e-05,
+      "loss": 0.0003,
+      "step": 29976
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 4.650419776119403e-05,
+      "loss": 0.0001,
+      "step": 29980
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 4.6503731343283586e-05,
+      "loss": 0.0011,
+      "step": 29984
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 4.6503264925373135e-05,
+      "loss": 0.009,
+      "step": 29988
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 4.650279850746269e-05,
+      "loss": 0.0013,
+      "step": 29992
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 4.6502332089552244e-05,
+      "loss": 0.0,
+      "step": 29996
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 4.650186567164179e-05,
+      "loss": 0.0007,
+      "step": 30000
+    },
+    {
+      "epoch": 27.98,
+      "eval_exact_match": 0.7350096711798839,
+      "eval_exec": 0.769825918762089,
+      "eval_loss": 0.3727231025695801,
+      "eval_runtime": 1201.9878,
+      "eval_samples_per_second": 0.86,
+      "step": 30000
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 4.650139925373135e-05,
+      "loss": 0.0001,
+      "step": 30004
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 4.6500932835820896e-05,
+      "loss": 0.0002,
+      "step": 30008
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 4.650046641791045e-05,
+      "loss": 0.0004,
+      "step": 30012
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 4.6500000000000005e-05,
+      "loss": 0.0001,
+      "step": 30016
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 4.6499533582089554e-05,
+      "loss": 0.0026,
+      "step": 30020
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 4.649906716417911e-05,
+      "loss": 0.0003,
+      "step": 30024
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 4.649860074626866e-05,
+      "loss": 0.0002,
+      "step": 30028
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 4.649813432835821e-05,
+      "loss": 0.0043,
+      "step": 30032
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 4.649766791044776e-05,
+      "loss": 0.0,
+      "step": 30036
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 4.6497201492537315e-05,
+      "loss": 0.0001,
+      "step": 30040
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 4.649673507462687e-05,
+      "loss": 0.0041,
+      "step": 30044
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 4.649626865671642e-05,
+      "loss": 0.0003,
+      "step": 30048
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 4.649580223880597e-05,
+      "loss": 0.0,
+      "step": 30052
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 4.649533582089553e-05,
+      "loss": 0.0002,
+      "step": 30056
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 4.6494869402985076e-05,
+      "loss": 0.0005,
+      "step": 30060
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 4.649440298507463e-05,
+      "loss": 0.002,
+      "step": 30064
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 4.649393656716418e-05,
+      "loss": 0.0003,
+      "step": 30068
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 4.6493470149253733e-05,
+      "loss": 0.0008,
+      "step": 30072
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 4.649300373134329e-05,
+      "loss": 0.0001,
+      "step": 30076
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 4.6492537313432837e-05,
+      "loss": 0.0001,
+      "step": 30080
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 4.649207089552239e-05,
+      "loss": 0.0001,
+      "step": 30084
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 4.6491604477611946e-05,
+      "loss": 0.0005,
+      "step": 30088
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 4.6491138059701494e-05,
+      "loss": 0.0009,
+      "step": 30092
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 4.649067164179104e-05,
+      "loss": 0.0015,
+      "step": 30096
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 4.64902052238806e-05,
+      "loss": 0.0009,
+      "step": 30100
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 4.648973880597015e-05,
+      "loss": 0.0013,
+      "step": 30104
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 4.64892723880597e-05,
+      "loss": 0.0003,
+      "step": 30108
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 4.6488805970149255e-05,
+      "loss": 0.0005,
+      "step": 30112
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 4.648833955223881e-05,
+      "loss": 0.0006,
+      "step": 30116
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 4.648787313432836e-05,
+      "loss": 0.0001,
+      "step": 30120
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 4.648740671641791e-05,
+      "loss": 0.0002,
+      "step": 30124
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 4.648694029850746e-05,
+      "loss": 0.0013,
+      "step": 30128
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 4.6486473880597016e-05,
+      "loss": 0.0001,
+      "step": 30132
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 4.648600746268657e-05,
+      "loss": 0.0031,
+      "step": 30136
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 4.648554104477612e-05,
+      "loss": 0.0004,
+      "step": 30140
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 4.6485074626865674e-05,
+      "loss": 0.0002,
+      "step": 30144
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 4.648460820895523e-05,
+      "loss": 0.0003,
+      "step": 30148
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 4.648414179104478e-05,
+      "loss": 0.0041,
+      "step": 30152
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 4.648367537313433e-05,
+      "loss": 0.0002,
+      "step": 30156
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 4.648320895522388e-05,
+      "loss": 0.0002,
+      "step": 30160
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 4.6482742537313435e-05,
+      "loss": 0.0001,
+      "step": 30164
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 4.648227611940299e-05,
+      "loss": 0.0004,
+      "step": 30168
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 4.648180970149254e-05,
+      "loss": 0.0033,
+      "step": 30172
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 4.648134328358209e-05,
+      "loss": 0.0013,
+      "step": 30176
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 4.648087686567165e-05,
+      "loss": 0.0021,
+      "step": 30180
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 4.6480410447761196e-05,
+      "loss": 0.0003,
+      "step": 30184
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 4.6479944029850744e-05,
+      "loss": 0.0004,
+      "step": 30188
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 4.6479477611940306e-05,
+      "loss": 0.0002,
+      "step": 30192
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 4.6479011194029854e-05,
+      "loss": 0.004,
+      "step": 30196
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 4.64785447761194e-05,
+      "loss": 0.0082,
+      "step": 30200
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 4.647807835820896e-05,
+      "loss": 0.0004,
+      "step": 30204
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 4.647761194029851e-05,
+      "loss": 0.0002,
+      "step": 30208
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 4.647714552238806e-05,
+      "loss": 0.0001,
+      "step": 30212
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 4.6476679104477615e-05,
+      "loss": 0.0004,
+      "step": 30216
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 4.6476212686567163e-05,
+      "loss": 0.0022,
+      "step": 30220
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 4.647574626865672e-05,
+      "loss": 0.0004,
+      "step": 30224
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 4.647527985074627e-05,
+      "loss": 0.0001,
+      "step": 30228
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 4.647481343283582e-05,
+      "loss": 0.0005,
+      "step": 30232
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 4.6474347014925376e-05,
+      "loss": 0.0001,
+      "step": 30236
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 4.647388059701493e-05,
+      "loss": 0.003,
+      "step": 30240
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 4.647341417910448e-05,
+      "loss": 0.0,
+      "step": 30244
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 4.647294776119403e-05,
+      "loss": 0.0043,
+      "step": 30248
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 4.647248134328359e-05,
+      "loss": 0.0021,
+      "step": 30252
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 4.647201492537314e-05,
+      "loss": 0.0038,
+      "step": 30256
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 4.6471548507462685e-05,
+      "loss": 0.0008,
+      "step": 30260
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 4.647108208955224e-05,
+      "loss": 0.0005,
+      "step": 30264
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 4.6470615671641795e-05,
+      "loss": 0.0001,
+      "step": 30268
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 4.647014925373134e-05,
+      "loss": 0.0038,
+      "step": 30272
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 4.64696828358209e-05,
+      "loss": 0.0013,
+      "step": 30276
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 4.6469216417910446e-05,
+      "loss": 0.0002,
+      "step": 30280
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 4.646875e-05,
+      "loss": 0.0029,
+      "step": 30284
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 4.6468283582089556e-05,
+      "loss": 0.0001,
+      "step": 30288
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 4.6467817164179104e-05,
+      "loss": 0.0001,
+      "step": 30292
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 4.646735074626866e-05,
+      "loss": 0.0012,
+      "step": 30296
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 4.6466884328358214e-05,
+      "loss": 0.0002,
+      "step": 30300
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 4.646641791044776e-05,
+      "loss": 0.0018,
+      "step": 30304
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 4.646595149253731e-05,
+      "loss": 0.0,
+      "step": 30308
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 4.646548507462687e-05,
+      "loss": 0.0003,
+      "step": 30312
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 4.646501865671642e-05,
+      "loss": 0.0001,
+      "step": 30316
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 4.6464552238805975e-05,
+      "loss": 0.0003,
+      "step": 30320
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 4.646408582089552e-05,
+      "loss": 0.0025,
+      "step": 30324
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 4.646361940298508e-05,
+      "loss": 0.0001,
+      "step": 30328
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 4.646315298507463e-05,
+      "loss": 0.0018,
+      "step": 30332
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 4.646268656716418e-05,
+      "loss": 0.0011,
+      "step": 30336
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 4.646222014925373e-05,
+      "loss": 0.0006,
+      "step": 30340
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 4.646175373134329e-05,
+      "loss": 0.0001,
+      "step": 30344
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 4.646128731343284e-05,
+      "loss": 0.0004,
+      "step": 30348
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 4.646082089552239e-05,
+      "loss": 0.0002,
+      "step": 30352
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 4.646035447761194e-05,
+      "loss": 0.0016,
+      "step": 30356
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 4.64598880597015e-05,
+      "loss": 0.0006,
+      "step": 30360
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 4.6459421641791045e-05,
+      "loss": 0.004,
+      "step": 30364
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 4.64589552238806e-05,
+      "loss": 0.0001,
+      "step": 30368
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 4.6458488805970155e-05,
+      "loss": 0.0001,
+      "step": 30372
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 4.64580223880597e-05,
+      "loss": 0.0029,
+      "step": 30376
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 4.645755597014926e-05,
+      "loss": 0.0078,
+      "step": 30380
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 4.6457089552238806e-05,
+      "loss": 0.0001,
+      "step": 30384
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 4.645662313432836e-05,
+      "loss": 0.001,
+      "step": 30388
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 4.6456156716417916e-05,
+      "loss": 0.0031,
+      "step": 30392
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 4.6455690298507464e-05,
+      "loss": 0.0004,
+      "step": 30396
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 4.645522388059701e-05,
+      "loss": 0.0001,
+      "step": 30400
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 4.6454757462686574e-05,
+      "loss": 0.0002,
+      "step": 30404
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 4.645429104477612e-05,
+      "loss": 0.0005,
+      "step": 30408
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 4.645382462686567e-05,
+      "loss": 0.0001,
+      "step": 30412
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 4.6453358208955225e-05,
+      "loss": 0.0001,
+      "step": 30416
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 4.645289179104478e-05,
+      "loss": 0.0012,
+      "step": 30420
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 4.645242537313433e-05,
+      "loss": 0.0074,
+      "step": 30424
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 4.645195895522388e-05,
+      "loss": 0.002,
+      "step": 30428
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 4.645149253731344e-05,
+      "loss": 0.0008,
+      "step": 30432
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 4.6451026119402986e-05,
+      "loss": 0.0001,
+      "step": 30436
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 4.645055970149254e-05,
+      "loss": 0.004,
+      "step": 30440
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 4.645009328358209e-05,
+      "loss": 0.0011,
+      "step": 30444
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 4.6449626865671644e-05,
+      "loss": 0.0064,
+      "step": 30448
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 4.64491604477612e-05,
+      "loss": 0.0002,
+      "step": 30452
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 4.644869402985075e-05,
+      "loss": 0.0027,
+      "step": 30456
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 4.6448227611940295e-05,
+      "loss": 0.0004,
+      "step": 30460
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 4.644776119402986e-05,
+      "loss": 0.0001,
+      "step": 30464
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 4.6447294776119405e-05,
+      "loss": 0.0006,
+      "step": 30468
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 4.644682835820895e-05,
+      "loss": 0.0003,
+      "step": 30472
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 4.644636194029851e-05,
+      "loss": 0.0026,
+      "step": 30476
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 4.644589552238806e-05,
+      "loss": 0.0056,
+      "step": 30480
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 4.644542910447762e-05,
+      "loss": 0.0002,
+      "step": 30484
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 4.6444962686567166e-05,
+      "loss": 0.0002,
+      "step": 30488
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 4.6444496268656714e-05,
+      "loss": 0.0001,
+      "step": 30492
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 4.6444029850746276e-05,
+      "loss": 0.0002,
+      "step": 30496
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 4.6443563432835824e-05,
+      "loss": 0.0003,
+      "step": 30500
+    },
+    {
+      "epoch": 28.45,
+      "eval_exact_match": 0.718568665377176,
+      "eval_exec": 0.7514506769825918,
+      "eval_loss": 0.3722231984138489,
+      "eval_runtime": 1165.7612,
+      "eval_samples_per_second": 0.887,
+      "step": 30500
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 4.644309701492537e-05,
+      "loss": 0.0001,
+      "step": 30504
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 4.644263059701493e-05,
+      "loss": 0.0008,
+      "step": 30508
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 4.644216417910448e-05,
+      "loss": 0.0006,
+      "step": 30512
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 4.644169776119403e-05,
+      "loss": 0.0001,
+      "step": 30516
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 4.6441231343283585e-05,
+      "loss": 0.0001,
+      "step": 30520
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 4.644076492537314e-05,
+      "loss": 0.0,
+      "step": 30524
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 4.644029850746269e-05,
+      "loss": 0.0001,
+      "step": 30528
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 4.643983208955224e-05,
+      "loss": 0.0025,
+      "step": 30532
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 4.643936567164179e-05,
+      "loss": 0.0004,
+      "step": 30536
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 4.6438899253731346e-05,
+      "loss": 0.0,
+      "step": 30540
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 4.64384328358209e-05,
+      "loss": 0.0002,
+      "step": 30544
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 4.643796641791045e-05,
+      "loss": 0.0001,
+      "step": 30548
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 4.64375e-05,
+      "loss": 0.0001,
+      "step": 30552
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 4.643703358208956e-05,
+      "loss": 0.0002,
+      "step": 30556
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 4.643656716417911e-05,
+      "loss": 0.0001,
+      "step": 30560
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 4.6436100746268655e-05,
+      "loss": 0.0009,
+      "step": 30564
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 4.643563432835821e-05,
+      "loss": 0.0066,
+      "step": 30568
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 4.6435167910447765e-05,
+      "loss": 0.0007,
+      "step": 30572
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 4.643470149253731e-05,
+      "loss": 0.0004,
+      "step": 30576
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 4.643423507462687e-05,
+      "loss": 0.0007,
+      "step": 30580
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 4.643376865671642e-05,
+      "loss": 0.0012,
+      "step": 30584
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 4.643330223880597e-05,
+      "loss": 0.0001,
+      "step": 30588
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 4.6432835820895526e-05,
+      "loss": 0.0001,
+      "step": 30592
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 4.6432369402985074e-05,
+      "loss": 0.004,
+      "step": 30596
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 4.643190298507463e-05,
+      "loss": 0.0001,
+      "step": 30600
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 4.6431436567164184e-05,
+      "loss": 0.0005,
+      "step": 30604
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 4.643097014925373e-05,
+      "loss": 0.0011,
+      "step": 30608
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 4.643050373134328e-05,
+      "loss": 0.0001,
+      "step": 30612
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 4.643003731343284e-05,
+      "loss": 0.0005,
+      "step": 30616
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 4.642957089552239e-05,
+      "loss": 0.0004,
+      "step": 30620
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 4.642910447761194e-05,
+      "loss": 0.0,
+      "step": 30624
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 4.642863805970149e-05,
+      "loss": 0.0001,
+      "step": 30628
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 4.642817164179105e-05,
+      "loss": 0.0033,
+      "step": 30632
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 4.6427705223880596e-05,
+      "loss": 0.0002,
+      "step": 30636
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 4.642723880597015e-05,
+      "loss": 0.0046,
+      "step": 30640
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 4.6426772388059706e-05,
+      "loss": 0.0022,
+      "step": 30644
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 4.642630597014926e-05,
+      "loss": 0.0011,
+      "step": 30648
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 4.642583955223881e-05,
+      "loss": 0.0047,
+      "step": 30652
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 4.642537313432836e-05,
+      "loss": 0.0001,
+      "step": 30656
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 4.642490671641791e-05,
+      "loss": 0.0002,
+      "step": 30660
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 4.642444029850747e-05,
+      "loss": 0.0005,
+      "step": 30664
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 4.6423973880597015e-05,
+      "loss": 0.0007,
+      "step": 30668
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 4.642350746268657e-05,
+      "loss": 0.0027,
+      "step": 30672
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 4.6423041044776125e-05,
+      "loss": 0.0001,
+      "step": 30676
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 4.642257462686567e-05,
+      "loss": 0.0003,
+      "step": 30680
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 4.642210820895523e-05,
+      "loss": 0.0053,
+      "step": 30684
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 4.6421641791044776e-05,
+      "loss": 0.0001,
+      "step": 30688
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 4.642117537313433e-05,
+      "loss": 0.0002,
+      "step": 30692
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 4.6420708955223886e-05,
+      "loss": 0.0011,
+      "step": 30696
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 4.6420242537313434e-05,
+      "loss": 0.0009,
+      "step": 30700
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 4.641977611940299e-05,
+      "loss": 0.0007,
+      "step": 30704
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 4.6419309701492544e-05,
+      "loss": 0.0011,
+      "step": 30708
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 4.641884328358209e-05,
+      "loss": 0.0003,
+      "step": 30712
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 4.641837686567164e-05,
+      "loss": 0.0016,
+      "step": 30716
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 4.6417910447761195e-05,
+      "loss": 0.0002,
+      "step": 30720
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 4.641744402985075e-05,
+      "loss": 0.0001,
+      "step": 30724
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 4.64169776119403e-05,
+      "loss": 0.0003,
+      "step": 30728
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 4.641651119402985e-05,
+      "loss": 0.0002,
+      "step": 30732
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 4.641604477611941e-05,
+      "loss": 0.0003,
+      "step": 30736
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 4.6415578358208956e-05,
+      "loss": 0.0001,
+      "step": 30740
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 4.641511194029851e-05,
+      "loss": 0.0001,
+      "step": 30744
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 4.641464552238806e-05,
+      "loss": 0.0,
+      "step": 30748
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 4.6414179104477614e-05,
+      "loss": 0.0014,
+      "step": 30752
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 4.641371268656717e-05,
+      "loss": 0.0002,
+      "step": 30756
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 4.641324626865672e-05,
+      "loss": 0.0,
+      "step": 30760
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 4.641277985074627e-05,
+      "loss": 0.0006,
+      "step": 30764
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 4.6412313432835827e-05,
+      "loss": 0.0004,
+      "step": 30768
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 4.6411847014925375e-05,
+      "loss": 0.0,
+      "step": 30772
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 4.641138059701492e-05,
+      "loss": 0.0,
+      "step": 30776
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 4.641091417910448e-05,
+      "loss": 0.0002,
+      "step": 30780
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 4.641044776119403e-05,
+      "loss": 0.0029,
+      "step": 30784
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 4.640998134328358e-05,
+      "loss": 0.0002,
+      "step": 30788
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 4.6409514925373136e-05,
+      "loss": 0.0094,
+      "step": 30792
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 4.640904850746269e-05,
+      "loss": 0.0011,
+      "step": 30796
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 4.640858208955224e-05,
+      "loss": 0.0027,
+      "step": 30800
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 4.6408115671641794e-05,
+      "loss": 0.0009,
+      "step": 30804
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 4.640764925373134e-05,
+      "loss": 0.0007,
+      "step": 30808
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 4.6407182835820903e-05,
+      "loss": 0.0001,
+      "step": 30812
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 4.640671641791045e-05,
+      "loss": 0.0018,
+      "step": 30816
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 4.640625e-05,
+      "loss": 0.0005,
+      "step": 30820
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 4.6405783582089555e-05,
+      "loss": 0.0003,
+      "step": 30824
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 4.640531716417911e-05,
+      "loss": 0.0003,
+      "step": 30828
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 4.640485074626866e-05,
+      "loss": 0.0011,
+      "step": 30832
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 4.640438432835821e-05,
+      "loss": 0.0047,
+      "step": 30836
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 4.640391791044776e-05,
+      "loss": 0.0002,
+      "step": 30840
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 4.6403451492537316e-05,
+      "loss": 0.0002,
+      "step": 30844
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 4.640298507462687e-05,
+      "loss": 0.0,
+      "step": 30848
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 4.640251865671642e-05,
+      "loss": 0.0,
+      "step": 30852
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 4.6402052238805974e-05,
+      "loss": 0.0,
+      "step": 30856
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 4.640158582089553e-05,
+      "loss": 0.0002,
+      "step": 30860
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 4.640111940298508e-05,
+      "loss": 0.0031,
+      "step": 30864
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 4.6400652985074625e-05,
+      "loss": 0.0001,
+      "step": 30868
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 4.6400186567164186e-05,
+      "loss": 0.0001,
+      "step": 30872
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 4.6399720149253735e-05,
+      "loss": 0.0007,
+      "step": 30876
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 4.639925373134328e-05,
+      "loss": 0.0001,
+      "step": 30880
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 4.639878731343284e-05,
+      "loss": 0.0004,
+      "step": 30884
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 4.639832089552239e-05,
+      "loss": 0.0003,
+      "step": 30888
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 4.639785447761194e-05,
+      "loss": 0.0042,
+      "step": 30892
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 4.6397388059701496e-05,
+      "loss": 0.0001,
+      "step": 30896
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 4.6396921641791044e-05,
+      "loss": 0.0,
+      "step": 30900
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 4.63964552238806e-05,
+      "loss": 0.0001,
+      "step": 30904
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 4.6395988805970154e-05,
+      "loss": 0.0,
+      "step": 30908
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 4.63955223880597e-05,
+      "loss": 0.0007,
+      "step": 30912
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 4.6395055970149257e-05,
+      "loss": 0.0009,
+      "step": 30916
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 4.639458955223881e-05,
+      "loss": 0.0066,
+      "step": 30920
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 4.639412313432836e-05,
+      "loss": 0.0005,
+      "step": 30924
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 4.639365671641791e-05,
+      "loss": 0.0012,
+      "step": 30928
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 4.639319029850747e-05,
+      "loss": 0.0002,
+      "step": 30932
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 4.639272388059702e-05,
+      "loss": 0.0,
+      "step": 30936
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 4.6392257462686566e-05,
+      "loss": 0.0003,
+      "step": 30940
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 4.639179104477612e-05,
+      "loss": 0.0007,
+      "step": 30944
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 4.6391324626865675e-05,
+      "loss": 0.0001,
+      "step": 30948
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 4.6390858208955224e-05,
+      "loss": 0.0001,
+      "step": 30952
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 4.639039179104478e-05,
+      "loss": 0.0002,
+      "step": 30956
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 4.638992537313433e-05,
+      "loss": 0.0007,
+      "step": 30960
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 4.638945895522388e-05,
+      "loss": 0.0002,
+      "step": 30964
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 4.6388992537313436e-05,
+      "loss": 0.0039,
+      "step": 30968
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 4.6388526119402985e-05,
+      "loss": 0.0,
+      "step": 30972
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 4.638805970149254e-05,
+      "loss": 0.0006,
+      "step": 30976
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 4.6387593283582094e-05,
+      "loss": 0.0,
+      "step": 30980
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 4.638712686567164e-05,
+      "loss": 0.0018,
+      "step": 30984
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 4.63866604477612e-05,
+      "loss": 0.0024,
+      "step": 30988
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 4.638619402985075e-05,
+      "loss": 0.0023,
+      "step": 30992
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 4.63857276119403e-05,
+      "loss": 0.001,
+      "step": 30996
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 4.6385261194029855e-05,
+      "loss": 0.0003,
+      "step": 31000
+    },
+    {
+      "epoch": 28.92,
+      "eval_exact_match": 0.7263056092843327,
+      "eval_exec": 0.7649903288201161,
+      "eval_loss": 0.41026368737220764,
+      "eval_runtime": 1150.5617,
+      "eval_samples_per_second": 0.899,
+      "step": 31000
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 4.6384794776119404e-05,
+      "loss": 0.0009,
+      "step": 31004
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 4.638432835820896e-05,
+      "loss": 0.0017,
+      "step": 31008
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 4.638386194029851e-05,
+      "loss": 0.0015,
+      "step": 31012
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 4.638339552238806e-05,
+      "loss": 0.0,
+      "step": 31016
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 4.638292910447761e-05,
+      "loss": 0.0016,
+      "step": 31020
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 4.638246268656717e-05,
+      "loss": 0.0142,
+      "step": 31024
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 4.638199626865672e-05,
+      "loss": 0.0027,
+      "step": 31028
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 4.638152985074627e-05,
+      "loss": 0.0,
+      "step": 31032
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 4.638106343283582e-05,
+      "loss": 0.0025,
+      "step": 31036
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 4.638059701492538e-05,
+      "loss": 0.0004,
+      "step": 31040
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 4.6380130597014925e-05,
+      "loss": 0.0007,
+      "step": 31044
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 4.637966417910448e-05,
+      "loss": 0.0001,
+      "step": 31048
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 4.6379197761194035e-05,
+      "loss": 0.0002,
+      "step": 31052
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 4.6378731343283583e-05,
+      "loss": 0.0045,
+      "step": 31056
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 4.637826492537314e-05,
+      "loss": 0.0011,
+      "step": 31060
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 4.6377798507462686e-05,
+      "loss": 0.0001,
+      "step": 31064
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 4.637733208955224e-05,
+      "loss": 0.0005,
+      "step": 31068
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 4.6376865671641796e-05,
+      "loss": 0.0009,
+      "step": 31072
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 4.6376399253731344e-05,
+      "loss": 0.0003,
+      "step": 31076
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 4.637593283582089e-05,
+      "loss": 0.0017,
+      "step": 31080
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 4.6375466417910454e-05,
+      "loss": 0.0004,
+      "step": 31084
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 4.6375e-05,
+      "loss": 0.0047,
+      "step": 31088
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 4.637453358208955e-05,
+      "loss": 0.0004,
+      "step": 31092
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 4.6374067164179105e-05,
+      "loss": 0.0001,
+      "step": 31096
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 4.637360074626866e-05,
+      "loss": 0.0001,
+      "step": 31100
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 4.637313432835821e-05,
+      "loss": 0.0003,
+      "step": 31104
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 4.637266791044776e-05,
+      "loss": 0.0004,
+      "step": 31108
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 4.637220149253732e-05,
+      "loss": 0.0036,
+      "step": 31112
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 4.6371735074626866e-05,
+      "loss": 0.0001,
+      "step": 31116
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 4.637126865671642e-05,
+      "loss": 0.0001,
+      "step": 31120
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 4.637080223880597e-05,
+      "loss": 0.0019,
+      "step": 31124
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 4.6370335820895524e-05,
+      "loss": 0.0003,
+      "step": 31128
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 4.636986940298508e-05,
+      "loss": 0.0018,
+      "step": 31132
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 4.636940298507463e-05,
+      "loss": 0.0001,
+      "step": 31136
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 4.636893656716418e-05,
+      "loss": 0.0005,
+      "step": 31140
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 4.636847014925374e-05,
+      "loss": 0.0001,
+      "step": 31144
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 4.6368003731343285e-05,
+      "loss": 0.0001,
+      "step": 31148
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 4.636753731343284e-05,
+      "loss": 0.0,
+      "step": 31152
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 4.636707089552239e-05,
+      "loss": 0.0,
+      "step": 31156
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 4.636660447761194e-05,
+      "loss": 0.0001,
+      "step": 31160
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 4.63661380597015e-05,
+      "loss": 0.0001,
+      "step": 31164
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 4.6365671641791046e-05,
+      "loss": 0.0,
+      "step": 31168
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 4.63652052238806e-05,
+      "loss": 0.0006,
+      "step": 31172
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 4.6364738805970156e-05,
+      "loss": 0.0001,
+      "step": 31176
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 4.6364272388059704e-05,
+      "loss": 0.0007,
+      "step": 31180
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 4.636380597014925e-05,
+      "loss": 0.0053,
+      "step": 31184
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 4.636333955223881e-05,
+      "loss": 0.0003,
+      "step": 31188
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 4.636287313432836e-05,
+      "loss": 0.0001,
+      "step": 31192
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 4.636240671641791e-05,
+      "loss": 0.0003,
+      "step": 31196
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 4.6361940298507465e-05,
+      "loss": 0.0001,
+      "step": 31200
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 4.636147388059702e-05,
+      "loss": 0.0007,
+      "step": 31204
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 4.636100746268657e-05,
+      "loss": 0.0002,
+      "step": 31208
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 4.636054104477612e-05,
+      "loss": 0.0013,
+      "step": 31212
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 4.636007462686567e-05,
+      "loss": 0.0014,
+      "step": 31216
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 4.6359608208955226e-05,
+      "loss": 0.0,
+      "step": 31220
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 4.635914179104478e-05,
+      "loss": 0.0011,
+      "step": 31224
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 4.635867537313433e-05,
+      "loss": 0.0001,
+      "step": 31228
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 4.635820895522388e-05,
+      "loss": 0.0006,
+      "step": 31232
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 4.635774253731344e-05,
+      "loss": 0.0071,
+      "step": 31236
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 4.635727611940299e-05,
+      "loss": 0.0001,
+      "step": 31240
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 4.6356809701492535e-05,
+      "loss": 0.0003,
+      "step": 31244
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 4.635634328358209e-05,
+      "loss": 0.0029,
+      "step": 31248
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 4.6355876865671645e-05,
+      "loss": 0.0001,
+      "step": 31252
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 4.635541044776119e-05,
+      "loss": 0.0001,
+      "step": 31256
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 4.635494402985075e-05,
+      "loss": 0.0004,
+      "step": 31260
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 4.63544776119403e-05,
+      "loss": 0.0,
+      "step": 31264
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 4.635401119402985e-05,
+      "loss": 0.0007,
+      "step": 31268
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 4.6353544776119406e-05,
+      "loss": 0.0059,
+      "step": 31272
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 4.6353078358208954e-05,
+      "loss": 0.0023,
+      "step": 31276
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 4.635261194029851e-05,
+      "loss": 0.0004,
+      "step": 31280
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 4.6352145522388064e-05,
+      "loss": 0.0004,
+      "step": 31284
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 4.635167910447761e-05,
+      "loss": 0.0006,
+      "step": 31288
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 4.635121268656716e-05,
+      "loss": 0.0001,
+      "step": 31292
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 4.635074626865672e-05,
+      "loss": 0.0001,
+      "step": 31296
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 4.635027985074627e-05,
+      "loss": 0.0003,
+      "step": 31300
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 4.6349813432835825e-05,
+      "loss": 0.0,
+      "step": 31304
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 4.634934701492537e-05,
+      "loss": 0.0007,
+      "step": 31308
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 4.634888059701493e-05,
+      "loss": 0.0015,
+      "step": 31312
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 4.634841417910448e-05,
+      "loss": 0.0005,
+      "step": 31316
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 4.634794776119403e-05,
+      "loss": 0.0006,
+      "step": 31320
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 4.6347481343283586e-05,
+      "loss": 0.0005,
+      "step": 31324
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 4.634701492537314e-05,
+      "loss": 0.0001,
+      "step": 31328
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 4.634654850746269e-05,
+      "loss": 0.0018,
+      "step": 31332
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 4.634608208955224e-05,
+      "loss": 0.0002,
+      "step": 31336
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 4.634561567164179e-05,
+      "loss": 0.0,
+      "step": 31340
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 4.634514925373135e-05,
+      "loss": 0.0001,
+      "step": 31344
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 4.6344682835820895e-05,
+      "loss": 0.0,
+      "step": 31348
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 4.634421641791045e-05,
+      "loss": 0.0,
+      "step": 31352
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 4.6343750000000005e-05,
+      "loss": 0.0006,
+      "step": 31356
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 4.634328358208955e-05,
+      "loss": 0.0002,
+      "step": 31360
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 4.634281716417911e-05,
+      "loss": 0.0002,
+      "step": 31364
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 4.6342350746268656e-05,
+      "loss": 0.0097,
+      "step": 31368
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 4.634188432835821e-05,
+      "loss": 0.0,
+      "step": 31372
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 4.6341417910447766e-05,
+      "loss": 0.0008,
+      "step": 31376
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 4.6340951492537314e-05,
+      "loss": 0.0001,
+      "step": 31380
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 4.634048507462687e-05,
+      "loss": 0.0015,
+      "step": 31384
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 4.6340018656716424e-05,
+      "loss": 0.0018,
+      "step": 31388
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 4.633955223880597e-05,
+      "loss": 0.0008,
+      "step": 31392
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 4.633908582089552e-05,
+      "loss": 0.0001,
+      "step": 31396
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 4.6338619402985075e-05,
+      "loss": 0.0008,
+      "step": 31400
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 4.633815298507463e-05,
+      "loss": 0.0,
+      "step": 31404
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 4.633768656716418e-05,
+      "loss": 0.0063,
+      "step": 31408
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 4.633722014925373e-05,
+      "loss": 0.0001,
+      "step": 31412
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 4.633675373134329e-05,
+      "loss": 0.0002,
+      "step": 31416
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 4.6336287313432836e-05,
+      "loss": 0.0,
+      "step": 31420
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 4.633582089552239e-05,
+      "loss": 0.0003,
+      "step": 31424
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 4.633535447761194e-05,
+      "loss": 0.0007,
+      "step": 31428
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 4.6334888059701494e-05,
+      "loss": 0.0002,
+      "step": 31432
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 4.633442164179105e-05,
+      "loss": 0.0011,
+      "step": 31436
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 4.63339552238806e-05,
+      "loss": 0.0022,
+      "step": 31440
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 4.633348880597015e-05,
+      "loss": 0.0,
+      "step": 31444
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 4.633302238805971e-05,
+      "loss": 0.0002,
+      "step": 31448
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 4.6332555970149255e-05,
+      "loss": 0.0003,
+      "step": 31452
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 4.63320895522388e-05,
+      "loss": 0.0002,
+      "step": 31456
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 4.633162313432836e-05,
+      "loss": 0.0002,
+      "step": 31460
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 4.633115671641791e-05,
+      "loss": 0.0003,
+      "step": 31464
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 4.633069029850747e-05,
+      "loss": 0.0003,
+      "step": 31468
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 4.6330223880597016e-05,
+      "loss": 0.0035,
+      "step": 31472
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 4.632975746268657e-05,
+      "loss": 0.0011,
+      "step": 31476
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 4.6329291044776126e-05,
+      "loss": 0.0001,
+      "step": 31480
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 4.6328824626865674e-05,
+      "loss": 0.0002,
+      "step": 31484
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 4.632835820895522e-05,
+      "loss": 0.0005,
+      "step": 31488
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 4.6327891791044784e-05,
+      "loss": 0.0004,
+      "step": 31492
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 4.632742537313433e-05,
+      "loss": 0.0008,
+      "step": 31496
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 4.632695895522388e-05,
+      "loss": 0.0002,
+      "step": 31500
+    },
+    {
+      "epoch": 29.38,
+      "eval_exact_match": 0.741779497098646,
+      "eval_exec": 0.7746615087040619,
+      "eval_loss": 0.42285799980163574,
+      "eval_runtime": 1077.7296,
+      "eval_samples_per_second": 0.959,
+      "step": 31500
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 4.6326492537313435e-05,
+      "loss": 0.0001,
+      "step": 31504
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 4.632602611940299e-05,
+      "loss": 0.0,
+      "step": 31508
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 4.632555970149254e-05,
+      "loss": 0.0004,
+      "step": 31512
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 4.632509328358209e-05,
+      "loss": 0.0,
+      "step": 31516
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 4.632462686567164e-05,
+      "loss": 0.0014,
+      "step": 31520
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 4.6324160447761196e-05,
+      "loss": 0.037,
+      "step": 31524
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 4.632369402985075e-05,
+      "loss": 0.004,
+      "step": 31528
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 4.63232276119403e-05,
+      "loss": 0.0,
+      "step": 31532
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 4.6322761194029854e-05,
+      "loss": 0.0001,
+      "step": 31536
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 4.632229477611941e-05,
+      "loss": 0.0005,
+      "step": 31540
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 4.632182835820896e-05,
+      "loss": 0.0022,
+      "step": 31544
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 4.6321361940298505e-05,
+      "loss": 0.0001,
+      "step": 31548
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 4.632089552238807e-05,
+      "loss": 0.001,
+      "step": 31552
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 4.6320429104477615e-05,
+      "loss": 0.001,
+      "step": 31556
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 4.631996268656716e-05,
+      "loss": 0.0,
+      "step": 31560
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 4.631949626865672e-05,
+      "loss": 0.0012,
+      "step": 31564
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 4.631902985074627e-05,
+      "loss": 0.0011,
+      "step": 31568
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 4.631856343283582e-05,
+      "loss": 0.0005,
+      "step": 31572
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 4.6318097014925376e-05,
+      "loss": 0.0018,
+      "step": 31576
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 4.6317630597014924e-05,
+      "loss": 0.0,
+      "step": 31580
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 4.631716417910448e-05,
+      "loss": 0.0001,
+      "step": 31584
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 4.6316697761194034e-05,
+      "loss": 0.002,
+      "step": 31588
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 4.631623134328358e-05,
+      "loss": 0.0004,
+      "step": 31592
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 4.631576492537314e-05,
+      "loss": 0.0172,
+      "step": 31596
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 4.631529850746269e-05,
+      "loss": 0.0001,
+      "step": 31600
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 4.631483208955224e-05,
+      "loss": 0.0021,
+      "step": 31604
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 4.631436567164179e-05,
+      "loss": 0.0003,
+      "step": 31608
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 4.631389925373135e-05,
+      "loss": 0.0005,
+      "step": 31612
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 4.63134328358209e-05,
+      "loss": 0.0004,
+      "step": 31616
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 4.6312966417910446e-05,
+      "loss": 0.0002,
+      "step": 31620
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 4.63125e-05,
+      "loss": 0.0,
+      "step": 31624
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 4.6312033582089556e-05,
+      "loss": 0.0007,
+      "step": 31628
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 4.631156716417911e-05,
+      "loss": 0.0001,
+      "step": 31632
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 4.631110074626866e-05,
+      "loss": 0.0015,
+      "step": 31636
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 4.631063432835821e-05,
+      "loss": 0.0001,
+      "step": 31640
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 4.631016791044777e-05,
+      "loss": 0.0019,
+      "step": 31644
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 4.630970149253732e-05,
+      "loss": 0.0003,
+      "step": 31648
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 4.6309235074626865e-05,
+      "loss": 0.0,
+      "step": 31652
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 4.630876865671642e-05,
+      "loss": 0.0235,
+      "step": 31656
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 4.6308302238805975e-05,
+      "loss": 0.0001,
+      "step": 31660
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 4.630783582089552e-05,
+      "loss": 0.0015,
+      "step": 31664
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 4.630736940298508e-05,
+      "loss": 0.0002,
+      "step": 31668
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 4.630690298507463e-05,
+      "loss": 0.0001,
+      "step": 31672
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 4.630643656716418e-05,
+      "loss": 0.0002,
+      "step": 31676
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 4.6305970149253736e-05,
+      "loss": 0.0,
+      "step": 31680
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 4.6305503731343284e-05,
+      "loss": 0.0001,
+      "step": 31684
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 4.630503731343284e-05,
+      "loss": 0.0003,
+      "step": 31688
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 4.6304570895522394e-05,
+      "loss": 0.0,
+      "step": 31692
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 4.630410447761194e-05,
+      "loss": 0.0001,
+      "step": 31696
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 4.630363805970149e-05,
+      "loss": 0.0002,
+      "step": 31700
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 4.630317164179105e-05,
+      "loss": 0.0009,
+      "step": 31704
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 4.63027052238806e-05,
+      "loss": 0.0001,
+      "step": 31708
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 4.630223880597015e-05,
+      "loss": 0.0001,
+      "step": 31712
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 4.63017723880597e-05,
+      "loss": 0.0,
+      "step": 31716
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 4.630130597014926e-05,
+      "loss": 0.0001,
+      "step": 31720
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 4.6300839552238806e-05,
+      "loss": 0.0023,
+      "step": 31724
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 4.630037313432836e-05,
+      "loss": 0.001,
+      "step": 31728
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 4.6299906716417916e-05,
+      "loss": 0.0001,
+      "step": 31732
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 4.6299440298507464e-05,
+      "loss": 0.0034,
+      "step": 31736
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 4.629897388059702e-05,
+      "loss": 0.0007,
+      "step": 31740
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 4.629850746268657e-05,
+      "loss": 0.0002,
+      "step": 31744
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 4.629804104477612e-05,
+      "loss": 0.0001,
+      "step": 31748
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 4.6297574626865677e-05,
+      "loss": 0.0007,
+      "step": 31752
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 4.6297108208955225e-05,
+      "loss": 0.0001,
+      "step": 31756
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 4.629664179104477e-05,
+      "loss": 0.0028,
+      "step": 31760
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 4.6296175373134335e-05,
+      "loss": 0.0001,
+      "step": 31764
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 4.629570895522388e-05,
+      "loss": 0.0111,
+      "step": 31768
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 4.629524253731343e-05,
+      "loss": 0.0001,
+      "step": 31772
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 4.6294776119402986e-05,
+      "loss": 0.0001,
+      "step": 31776
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 4.629430970149254e-05,
+      "loss": 0.0013,
+      "step": 31780
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 4.629384328358209e-05,
+      "loss": 0.0004,
+      "step": 31784
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 4.6293376865671644e-05,
+      "loss": 0.0011,
+      "step": 31788
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 4.62929104477612e-05,
+      "loss": 0.0001,
+      "step": 31792
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 4.6292444029850753e-05,
+      "loss": 0.0015,
+      "step": 31796
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 4.62919776119403e-05,
+      "loss": 0.001,
+      "step": 31800
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 4.629151119402985e-05,
+      "loss": 0.0008,
+      "step": 31804
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 4.6291044776119405e-05,
+      "loss": 0.0002,
+      "step": 31808
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 4.629057835820896e-05,
+      "loss": 0.0002,
+      "step": 31812
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 4.629011194029851e-05,
+      "loss": 0.0117,
+      "step": 31816
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 4.628964552238806e-05,
+      "loss": 0.0005,
+      "step": 31820
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 4.628917910447762e-05,
+      "loss": 0.0025,
+      "step": 31824
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 4.6288712686567166e-05,
+      "loss": 0.0005,
+      "step": 31828
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 4.628824626865672e-05,
+      "loss": 0.003,
+      "step": 31832
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 4.628777985074627e-05,
+      "loss": 0.0,
+      "step": 31836
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 4.6287313432835824e-05,
+      "loss": 0.0002,
+      "step": 31840
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 4.628684701492538e-05,
+      "loss": 0.0016,
+      "step": 31844
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 4.6286380597014927e-05,
+      "loss": 0.0002,
+      "step": 31848
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 4.628591417910448e-05,
+      "loss": 0.0002,
+      "step": 31852
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 4.6285447761194036e-05,
+      "loss": 0.0002,
+      "step": 31856
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 4.6284981343283585e-05,
+      "loss": 0.0003,
+      "step": 31860
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 4.628451492537313e-05,
+      "loss": 0.002,
+      "step": 31864
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 4.628404850746269e-05,
+      "loss": 0.0025,
+      "step": 31868
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 4.628358208955224e-05,
+      "loss": 0.0001,
+      "step": 31872
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 4.628311567164179e-05,
+      "loss": 0.0001,
+      "step": 31876
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 4.6282649253731346e-05,
+      "loss": 0.0001,
+      "step": 31880
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 4.62821828358209e-05,
+      "loss": 0.0002,
+      "step": 31884
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 4.628171641791045e-05,
+      "loss": 0.0001,
+      "step": 31888
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 4.6281250000000003e-05,
+      "loss": 0.0004,
+      "step": 31892
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 4.628078358208955e-05,
+      "loss": 0.0013,
+      "step": 31896
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 4.6280317164179107e-05,
+      "loss": 0.0084,
+      "step": 31900
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 4.627985074626866e-05,
+      "loss": 0.002,
+      "step": 31904
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 4.627938432835821e-05,
+      "loss": 0.0001,
+      "step": 31908
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 4.627891791044776e-05,
+      "loss": 0.0001,
+      "step": 31912
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 4.627845149253732e-05,
+      "loss": 0.0007,
+      "step": 31916
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 4.627798507462687e-05,
+      "loss": 0.0002,
+      "step": 31920
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 4.6277518656716416e-05,
+      "loss": 0.0034,
+      "step": 31924
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 4.627705223880597e-05,
+      "loss": 0.0002,
+      "step": 31928
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 4.6276585820895525e-05,
+      "loss": 0.0013,
+      "step": 31932
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 4.6276119402985074e-05,
+      "loss": 0.0002,
+      "step": 31936
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 4.627565298507463e-05,
+      "loss": 0.0,
+      "step": 31940
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 4.6275186567164183e-05,
+      "loss": 0.0003,
+      "step": 31944
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 4.627472014925374e-05,
+      "loss": 0.0001,
+      "step": 31948
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 4.6274253731343286e-05,
+      "loss": 0.0002,
+      "step": 31952
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 4.6273787313432835e-05,
+      "loss": 0.0003,
+      "step": 31956
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 4.6273320895522396e-05,
+      "loss": 0.001,
+      "step": 31960
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 4.6272854477611944e-05,
+      "loss": 0.0,
+      "step": 31964
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 4.627238805970149e-05,
+      "loss": 0.0015,
+      "step": 31968
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 4.627192164179105e-05,
+      "loss": 0.001,
+      "step": 31972
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 4.62714552238806e-05,
+      "loss": 0.0001,
+      "step": 31976
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 4.627098880597015e-05,
+      "loss": 0.0011,
+      "step": 31980
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 4.6270522388059705e-05,
+      "loss": 0.0006,
+      "step": 31984
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 4.6270055970149254e-05,
+      "loss": 0.0002,
+      "step": 31988
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 4.626958955223881e-05,
+      "loss": 0.0014,
+      "step": 31992
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 4.626912313432836e-05,
+      "loss": 0.0007,
+      "step": 31996
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 4.626865671641791e-05,
+      "loss": 0.0037,
+      "step": 32000
+    },
+    {
+      "epoch": 29.85,
+      "eval_exact_match": 0.7437137330754352,
+      "eval_exec": 0.7717601547388782,
+      "eval_loss": 0.40884315967559814,
+      "eval_runtime": 1094.761,
+      "eval_samples_per_second": 0.944,
+      "step": 32000
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 4.6268190298507466e-05,
+      "loss": 0.0,
+      "step": 32004
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 4.626772388059702e-05,
+      "loss": 0.0149,
+      "step": 32008
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 4.626725746268657e-05,
+      "loss": 0.0004,
+      "step": 32012
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 4.626679104477612e-05,
+      "loss": 0.0004,
+      "step": 32016
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 4.626632462686567e-05,
+      "loss": 0.0001,
+      "step": 32020
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 4.626585820895523e-05,
+      "loss": 0.0001,
+      "step": 32024
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 4.6265391791044775e-05,
+      "loss": 0.0,
+      "step": 32028
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 4.626492537313433e-05,
+      "loss": 0.0046,
+      "step": 32032
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 4.6264458955223885e-05,
+      "loss": 0.0023,
+      "step": 32036
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 4.6263992537313433e-05,
+      "loss": 0.0013,
+      "step": 32040
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 4.626352611940299e-05,
+      "loss": 0.0003,
+      "step": 32044
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 4.6263059701492536e-05,
+      "loss": 0.0,
+      "step": 32048
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 4.626259328358209e-05,
+      "loss": 0.0001,
+      "step": 32052
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 4.6262126865671646e-05,
+      "loss": 0.0001,
+      "step": 32056
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 4.6261660447761194e-05,
+      "loss": 0.0001,
+      "step": 32060
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 4.626119402985075e-05,
+      "loss": 0.0023,
+      "step": 32064
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 4.6260727611940304e-05,
+      "loss": 0.0001,
+      "step": 32068
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 4.626026119402985e-05,
+      "loss": 0.0004,
+      "step": 32072
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 4.62597947761194e-05,
+      "loss": 0.0009,
+      "step": 32076
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 4.6259328358208955e-05,
+      "loss": 0.0062,
+      "step": 32080
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 4.625886194029851e-05,
+      "loss": 0.002,
+      "step": 32084
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 4.625839552238806e-05,
+      "loss": 0.0001,
+      "step": 32088
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 4.625792910447761e-05,
+      "loss": 0.0055,
+      "step": 32092
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 4.625746268656717e-05,
+      "loss": 0.0002,
+      "step": 32096
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 4.6256996268656716e-05,
+      "loss": 0.0003,
+      "step": 32100
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 4.625652985074627e-05,
+      "loss": 0.0006,
+      "step": 32104
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 4.625606343283582e-05,
+      "loss": 0.0005,
+      "step": 32108
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 4.625559701492538e-05,
+      "loss": 0.0001,
+      "step": 32112
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 4.625513059701493e-05,
+      "loss": 0.0001,
+      "step": 32116
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 4.625466417910448e-05,
+      "loss": 0.0003,
+      "step": 32120
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 4.625419776119403e-05,
+      "loss": 0.0014,
+      "step": 32124
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 4.625373134328359e-05,
+      "loss": 0.0,
+      "step": 32128
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 4.6253264925373135e-05,
+      "loss": 0.0021,
+      "step": 32132
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 4.625279850746269e-05,
+      "loss": 0.0006,
+      "step": 32136
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 4.625233208955224e-05,
+      "loss": 0.0003,
+      "step": 32140
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 4.625186567164179e-05,
+      "loss": 0.0011,
+      "step": 32144
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 4.625139925373135e-05,
+      "loss": 0.0005,
+      "step": 32148
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 4.6250932835820896e-05,
+      "loss": 0.0001,
+      "step": 32152
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 4.625046641791045e-05,
+      "loss": 0.0002,
+      "step": 32156
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 4.6250000000000006e-05,
+      "loss": 0.0002,
+      "step": 32160
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 4.6249533582089554e-05,
+      "loss": 0.0002,
+      "step": 32164
+    },
+    {
+      "epoch": 30.01,
+      "learning_rate": 4.62490671641791e-05,
+      "loss": 0.0005,
+      "step": 32168
+    },
+    {
+      "epoch": 30.01,
+      "learning_rate": 4.6248600746268664e-05,
+      "loss": 0.0006,
+      "step": 32172
+    },
+    {
+      "epoch": 30.01,
+      "learning_rate": 4.624813432835821e-05,
+      "loss": 0.0003,
+      "step": 32176
+    },
+    {
+      "epoch": 30.02,
+      "learning_rate": 4.624766791044776e-05,
+      "loss": 0.0139,
+      "step": 32180
+    },
+    {
+      "epoch": 30.02,
+      "learning_rate": 4.6247201492537315e-05,
+      "loss": 0.0009,
+      "step": 32184
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 4.624673507462687e-05,
+      "loss": 0.0002,
+      "step": 32188
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 4.624626865671642e-05,
+      "loss": 0.0001,
+      "step": 32192
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 4.624580223880597e-05,
+      "loss": 0.0001,
+      "step": 32196
+    },
+    {
+      "epoch": 30.04,
+      "learning_rate": 4.624533582089552e-05,
+      "loss": 0.0001,
+      "step": 32200
+    },
+    {
+      "epoch": 30.04,
+      "learning_rate": 4.6244869402985076e-05,
+      "loss": 0.0008,
+      "step": 32204
+    },
+    {
+      "epoch": 30.04,
+      "learning_rate": 4.624440298507463e-05,
+      "loss": 0.0071,
+      "step": 32208
+    },
+    {
+      "epoch": 30.05,
+      "learning_rate": 4.624393656716418e-05,
+      "loss": 0.0001,
+      "step": 32212
+    },
+    {
+      "epoch": 30.05,
+      "learning_rate": 4.6243470149253734e-05,
+      "loss": 0.0001,
+      "step": 32216
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 4.624300373134329e-05,
+      "loss": 0.0002,
+      "step": 32220
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 4.624253731343284e-05,
+      "loss": 0.0,
+      "step": 32224
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 4.6242070895522385e-05,
+      "loss": 0.0001,
+      "step": 32228
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 4.624160447761195e-05,
+      "loss": 0.0008,
+      "step": 32232
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 4.6241138059701495e-05,
+      "loss": 0.0003,
+      "step": 32236
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 4.624067164179104e-05,
+      "loss": 0.0002,
+      "step": 32240
+    },
+    {
+      "epoch": 30.08,
+      "learning_rate": 4.62402052238806e-05,
+      "loss": 0.0013,
+      "step": 32244
+    },
+    {
+      "epoch": 30.08,
+      "learning_rate": 4.623973880597015e-05,
+      "loss": 0.0013,
+      "step": 32248
+    },
+    {
+      "epoch": 30.09,
+      "learning_rate": 4.62392723880597e-05,
+      "loss": 0.0003,
+      "step": 32252
+    },
+    {
+      "epoch": 30.09,
+      "learning_rate": 4.6238805970149256e-05,
+      "loss": 0.0002,
+      "step": 32256
+    },
+    {
+      "epoch": 30.09,
+      "learning_rate": 4.6238339552238804e-05,
+      "loss": 0.0004,
+      "step": 32260
+    },
+    {
+      "epoch": 30.1,
+      "learning_rate": 4.623787313432836e-05,
+      "loss": 0.0005,
+      "step": 32264
+    },
+    {
+      "epoch": 30.1,
+      "learning_rate": 4.6237406716417914e-05,
+      "loss": 0.0001,
+      "step": 32268
+    },
+    {
+      "epoch": 30.1,
+      "learning_rate": 4.623694029850746e-05,
+      "loss": 0.0003,
+      "step": 32272
+    },
+    {
+      "epoch": 30.11,
+      "learning_rate": 4.623647388059702e-05,
+      "loss": 0.0003,
+      "step": 32276
+    },
+    {
+      "epoch": 30.11,
+      "learning_rate": 4.623600746268657e-05,
+      "loss": 0.0002,
+      "step": 32280
+    },
+    {
+      "epoch": 30.12,
+      "learning_rate": 4.623554104477612e-05,
+      "loss": 0.0001,
+      "step": 32284
+    },
+    {
+      "epoch": 30.12,
+      "learning_rate": 4.6235074626865675e-05,
+      "loss": 0.0,
+      "step": 32288
+    },
+    {
+      "epoch": 30.12,
+      "learning_rate": 4.623460820895523e-05,
+      "loss": 0.0004,
+      "step": 32292
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 4.623414179104478e-05,
+      "loss": 0.0,
+      "step": 32296
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 4.623367537313433e-05,
+      "loss": 0.0006,
+      "step": 32300
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 4.623320895522388e-05,
+      "loss": 0.001,
+      "step": 32304
+    },
+    {
+      "epoch": 30.14,
+      "learning_rate": 4.6232742537313436e-05,
+      "loss": 0.005,
+      "step": 32308
+    },
+    {
+      "epoch": 30.14,
+      "learning_rate": 4.623227611940299e-05,
+      "loss": 0.0001,
+      "step": 32312
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 4.623180970149254e-05,
+      "loss": 0.001,
+      "step": 32316
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 4.623134328358209e-05,
+      "loss": 0.0,
+      "step": 32320
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 4.623087686567165e-05,
+      "loss": 0.0001,
+      "step": 32324
+    },
+    {
+      "epoch": 30.16,
+      "learning_rate": 4.62304104477612e-05,
+      "loss": 0.0008,
+      "step": 32328
+    },
+    {
+      "epoch": 30.16,
+      "learning_rate": 4.6229944029850745e-05,
+      "loss": 0.0002,
+      "step": 32332
+    },
+    {
+      "epoch": 30.16,
+      "learning_rate": 4.62294776119403e-05,
+      "loss": 0.0001,
+      "step": 32336
+    },
+    {
+      "epoch": 30.17,
+      "learning_rate": 4.6229011194029855e-05,
+      "loss": 0.0008,
+      "step": 32340
+    },
+    {
+      "epoch": 30.17,
+      "learning_rate": 4.62285447761194e-05,
+      "loss": 0.003,
+      "step": 32344
+    },
+    {
+      "epoch": 30.18,
+      "learning_rate": 4.622807835820896e-05,
+      "loss": 0.0002,
+      "step": 32348
+    },
+    {
+      "epoch": 30.18,
+      "learning_rate": 4.622761194029851e-05,
+      "loss": 0.0005,
+      "step": 32352
+    },
+    {
+      "epoch": 30.18,
+      "learning_rate": 4.622714552238806e-05,
+      "loss": 0.0007,
+      "step": 32356
+    },
+    {
+      "epoch": 30.19,
+      "learning_rate": 4.6226679104477616e-05,
+      "loss": 0.0001,
+      "step": 32360
+    },
+    {
+      "epoch": 30.19,
+      "learning_rate": 4.6226212686567164e-05,
+      "loss": 0.0001,
+      "step": 32364
+    },
+    {
+      "epoch": 30.19,
+      "learning_rate": 4.622574626865672e-05,
+      "loss": 0.0001,
+      "step": 32368
+    },
+    {
+      "epoch": 30.2,
+      "learning_rate": 4.6225279850746274e-05,
+      "loss": 0.0004,
+      "step": 32372
+    },
+    {
+      "epoch": 30.2,
+      "learning_rate": 4.622481343283582e-05,
+      "loss": 0.0002,
+      "step": 32376
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 4.622434701492537e-05,
+      "loss": 0.0,
+      "step": 32380
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 4.622388059701493e-05,
+      "loss": 0.0017,
+      "step": 32384
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 4.622341417910448e-05,
+      "loss": 0.0007,
+      "step": 32388
+    },
+    {
+      "epoch": 30.22,
+      "learning_rate": 4.622294776119403e-05,
+      "loss": 0.0014,
+      "step": 32392
+    },
+    {
+      "epoch": 30.22,
+      "learning_rate": 4.622248134328358e-05,
+      "loss": 0.0003,
+      "step": 32396
+    },
+    {
+      "epoch": 30.22,
+      "learning_rate": 4.622201492537314e-05,
+      "loss": 0.0131,
+      "step": 32400
+    },
+    {
+      "epoch": 30.23,
+      "learning_rate": 4.6221548507462686e-05,
+      "loss": 0.0001,
+      "step": 32404
+    },
+    {
+      "epoch": 30.23,
+      "learning_rate": 4.622108208955224e-05,
+      "loss": 0.0023,
+      "step": 32408
+    },
+    {
+      "epoch": 30.24,
+      "learning_rate": 4.6220615671641796e-05,
+      "loss": 0.0002,
+      "step": 32412
+    },
+    {
+      "epoch": 30.24,
+      "learning_rate": 4.6220149253731344e-05,
+      "loss": 0.0004,
+      "step": 32416
+    },
+    {
+      "epoch": 30.24,
+      "learning_rate": 4.62196828358209e-05,
+      "loss": 0.0001,
+      "step": 32420
+    },
+    {
+      "epoch": 30.25,
+      "learning_rate": 4.621921641791045e-05,
+      "loss": 0.001,
+      "step": 32424
+    },
+    {
+      "epoch": 30.25,
+      "learning_rate": 4.621875e-05,
+      "loss": 0.0001,
+      "step": 32428
+    },
+    {
+      "epoch": 30.25,
+      "learning_rate": 4.621828358208956e-05,
+      "loss": 0.0006,
+      "step": 32432
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 4.6217817164179105e-05,
+      "loss": 0.0003,
+      "step": 32436
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 4.621735074626866e-05,
+      "loss": 0.0002,
+      "step": 32440
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 4.6216884328358215e-05,
+      "loss": 0.0003,
+      "step": 32444
+    },
+    {
+      "epoch": 30.27,
+      "learning_rate": 4.621641791044776e-05,
+      "loss": 0.0001,
+      "step": 32448
+    },
+    {
+      "epoch": 30.27,
+      "learning_rate": 4.621595149253732e-05,
+      "loss": 0.0006,
+      "step": 32452
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 4.6215485074626866e-05,
+      "loss": 0.0,
+      "step": 32456
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 4.621501865671642e-05,
+      "loss": 0.0002,
+      "step": 32460
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 4.6214552238805976e-05,
+      "loss": 0.0003,
+      "step": 32464
+    },
+    {
+      "epoch": 30.29,
+      "learning_rate": 4.6214085820895524e-05,
+      "loss": 0.0016,
+      "step": 32468
+    },
+    {
+      "epoch": 30.29,
+      "learning_rate": 4.621361940298508e-05,
+      "loss": 0.0013,
+      "step": 32472
+    },
+    {
+      "epoch": 30.29,
+      "learning_rate": 4.6213152985074634e-05,
+      "loss": 0.0,
+      "step": 32476
+    },
+    {
+      "epoch": 30.3,
+      "learning_rate": 4.621268656716418e-05,
+      "loss": 0.0,
+      "step": 32480
+    },
+    {
+      "epoch": 30.3,
+      "learning_rate": 4.621222014925373e-05,
+      "loss": 0.0001,
+      "step": 32484
+    },
+    {
+      "epoch": 30.31,
+      "learning_rate": 4.6211753731343285e-05,
+      "loss": 0.0008,
+      "step": 32488
+    },
+    {
+      "epoch": 30.31,
+      "learning_rate": 4.621128731343284e-05,
+      "loss": 0.0003,
+      "step": 32492
+    },
+    {
+      "epoch": 30.31,
+      "learning_rate": 4.621082089552239e-05,
+      "loss": 0.0001,
+      "step": 32496
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 4.621035447761194e-05,
+      "loss": 0.0,
+      "step": 32500
+    },
+    {
+      "epoch": 30.32,
+      "eval_exact_match": 0.7388781431334622,
+      "eval_exec": 0.7678916827852998,
+      "eval_loss": 0.4083205461502075,
+      "eval_runtime": 1184.6901,
+      "eval_samples_per_second": 0.873,
+      "step": 32500
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 4.62098880597015e-05,
+      "loss": 0.0,
+      "step": 32504
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 4.6209421641791046e-05,
+      "loss": 0.0046,
+      "step": 32508
+    },
+    {
+      "epoch": 30.33,
+      "learning_rate": 4.62089552238806e-05,
+      "loss": 0.0001,
+      "step": 32512
+    },
+    {
+      "epoch": 30.33,
+      "learning_rate": 4.620848880597015e-05,
+      "loss": 0.0001,
+      "step": 32516
+    },
+    {
+      "epoch": 30.34,
+      "learning_rate": 4.6208022388059704e-05,
+      "loss": 0.0014,
+      "step": 32520
+    },
+    {
+      "epoch": 30.34,
+      "learning_rate": 4.620755597014926e-05,
+      "loss": 0.0004,
+      "step": 32524
+    },
+    {
+      "epoch": 30.34,
+      "learning_rate": 4.620708955223881e-05,
+      "loss": 0.0001,
+      "step": 32528
+    },
+    {
+      "epoch": 30.35,
+      "learning_rate": 4.620662313432836e-05,
+      "loss": 0.0028,
+      "step": 32532
+    },
+    {
+      "epoch": 30.35,
+      "learning_rate": 4.620615671641792e-05,
+      "loss": 0.0,
+      "step": 32536
+    },
+    {
+      "epoch": 30.35,
+      "learning_rate": 4.6205690298507465e-05,
+      "loss": 0.0001,
+      "step": 32540
+    },
+    {
+      "epoch": 30.36,
+      "learning_rate": 4.620522388059701e-05,
+      "loss": 0.0002,
+      "step": 32544
+    },
+    {
+      "epoch": 30.36,
+      "learning_rate": 4.620475746268657e-05,
+      "loss": 0.0001,
+      "step": 32548
+    },
+    {
+      "epoch": 30.37,
+      "learning_rate": 4.620429104477612e-05,
+      "loss": 0.001,
+      "step": 32552
+    },
+    {
+      "epoch": 30.37,
+      "learning_rate": 4.620382462686567e-05,
+      "loss": 0.0002,
+      "step": 32556
+    },
+    {
+      "epoch": 30.37,
+      "learning_rate": 4.6203358208955226e-05,
+      "loss": 0.0167,
+      "step": 32560
+    },
+    {
+      "epoch": 30.38,
+      "learning_rate": 4.620289179104478e-05,
+      "loss": 0.0051,
+      "step": 32564
+    },
+    {
+      "epoch": 30.38,
+      "learning_rate": 4.620242537313433e-05,
+      "loss": 0.0002,
+      "step": 32568
+    },
+    {
+      "epoch": 30.38,
+      "learning_rate": 4.6201958955223884e-05,
+      "loss": 0.0,
+      "step": 32572
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 4.620149253731343e-05,
+      "loss": 0.0027,
+      "step": 32576
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 4.620102611940299e-05,
+      "loss": 0.007,
+      "step": 32580
+    },
+    {
+      "epoch": 30.4,
+      "learning_rate": 4.620055970149254e-05,
+      "loss": 0.0001,
+      "step": 32584
+    },
+    {
+      "epoch": 30.4,
+      "learning_rate": 4.620009328358209e-05,
+      "loss": 0.0037,
+      "step": 32588
+    },
+    {
+      "epoch": 30.4,
+      "learning_rate": 4.619962686567164e-05,
+      "loss": 0.0001,
+      "step": 32592
+    },
+    {
+      "epoch": 30.41,
+      "learning_rate": 4.61991604477612e-05,
+      "loss": 0.0018,
+      "step": 32596
+    },
+    {
+      "epoch": 30.41,
+      "learning_rate": 4.619869402985075e-05,
+      "loss": 0.0007,
+      "step": 32600
+    },
+    {
+      "epoch": 30.41,
+      "learning_rate": 4.61982276119403e-05,
+      "loss": 0.0001,
+      "step": 32604
+    },
+    {
+      "epoch": 30.42,
+      "learning_rate": 4.619776119402985e-05,
+      "loss": 0.0002,
+      "step": 32608
+    },
+    {
+      "epoch": 30.42,
+      "learning_rate": 4.6197294776119406e-05,
+      "loss": 0.0001,
+      "step": 32612
+    },
+    {
+      "epoch": 30.43,
+      "learning_rate": 4.619682835820896e-05,
+      "loss": 0.0016,
+      "step": 32616
+    },
+    {
+      "epoch": 30.43,
+      "learning_rate": 4.619636194029851e-05,
+      "loss": 0.0,
+      "step": 32620
+    },
+    {
+      "epoch": 30.43,
+      "learning_rate": 4.6195895522388064e-05,
+      "loss": 0.0002,
+      "step": 32624
+    },
+    {
+      "epoch": 30.44,
+      "learning_rate": 4.619542910447762e-05,
+      "loss": 0.0003,
+      "step": 32628
+    },
+    {
+      "epoch": 30.44,
+      "learning_rate": 4.619496268656717e-05,
+      "loss": 0.0002,
+      "step": 32632
+    },
+    {
+      "epoch": 30.44,
+      "learning_rate": 4.6194496268656715e-05,
+      "loss": 0.0001,
+      "step": 32636
+    },
+    {
+      "epoch": 30.45,
+      "learning_rate": 4.6194029850746277e-05,
+      "loss": 0.0002,
+      "step": 32640
+    },
+    {
+      "epoch": 30.45,
+      "learning_rate": 4.6193563432835825e-05,
+      "loss": 0.0001,
+      "step": 32644
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 4.619309701492537e-05,
+      "loss": 0.0003,
+      "step": 32648
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 4.619263059701493e-05,
+      "loss": 0.0014,
+      "step": 32652
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 4.619216417910448e-05,
+      "loss": 0.0,
+      "step": 32656
+    },
+    {
+      "epoch": 30.47,
+      "learning_rate": 4.619169776119403e-05,
+      "loss": 0.0085,
+      "step": 32660
+    },
+    {
+      "epoch": 30.47,
+      "learning_rate": 4.6191231343283586e-05,
+      "loss": 0.0002,
+      "step": 32664
+    },
+    {
+      "epoch": 30.47,
+      "learning_rate": 4.6190764925373134e-05,
+      "loss": 0.0008,
+      "step": 32668
+    },
+    {
+      "epoch": 30.48,
+      "learning_rate": 4.619029850746269e-05,
+      "loss": 0.0007,
+      "step": 32672
+    },
+    {
+      "epoch": 30.48,
+      "learning_rate": 4.6189832089552244e-05,
+      "loss": 0.0025,
+      "step": 32676
+    },
+    {
+      "epoch": 30.49,
+      "learning_rate": 4.618936567164179e-05,
+      "loss": 0.0002,
+      "step": 32680
+    },
+    {
+      "epoch": 30.49,
+      "learning_rate": 4.618889925373135e-05,
+      "loss": 0.0009,
+      "step": 32684
+    },
+    {
+      "epoch": 30.49,
+      "learning_rate": 4.61884328358209e-05,
+      "loss": 0.0041,
+      "step": 32688
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 4.618796641791045e-05,
+      "loss": 0.0002,
+      "step": 32692
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 4.61875e-05,
+      "loss": 0.0012,
+      "step": 32696
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 4.618703358208955e-05,
+      "loss": 0.0005,
+      "step": 32700
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 4.618656716417911e-05,
+      "loss": 0.0004,
+      "step": 32704
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 4.6186100746268656e-05,
+      "loss": 0.0019,
+      "step": 32708
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 4.618563432835821e-05,
+      "loss": 0.001,
+      "step": 32712
+    },
+    {
+      "epoch": 30.52,
+      "learning_rate": 4.6185167910447766e-05,
+      "loss": 0.0,
+      "step": 32716
+    },
+    {
+      "epoch": 30.52,
+      "learning_rate": 4.6184701492537314e-05,
+      "loss": 0.001,
+      "step": 32720
+    },
+    {
+      "epoch": 30.53,
+      "learning_rate": 4.618423507462687e-05,
+      "loss": 0.0001,
+      "step": 32724
+    },
+    {
+      "epoch": 30.53,
+      "learning_rate": 4.618376865671642e-05,
+      "loss": 0.0002,
+      "step": 32728
+    },
+    {
+      "epoch": 30.53,
+      "learning_rate": 4.618330223880597e-05,
+      "loss": 0.0001,
+      "step": 32732
+    },
+    {
+      "epoch": 30.54,
+      "learning_rate": 4.6182835820895527e-05,
+      "loss": 0.0002,
+      "step": 32736
+    },
+    {
+      "epoch": 30.54,
+      "learning_rate": 4.6182369402985075e-05,
+      "loss": 0.0002,
+      "step": 32740
+    },
+    {
+      "epoch": 30.54,
+      "learning_rate": 4.618190298507463e-05,
+      "loss": 0.0,
+      "step": 32744
+    },
+    {
+      "epoch": 30.55,
+      "learning_rate": 4.6181436567164185e-05,
+      "loss": 0.0012,
+      "step": 32748
+    },
+    {
+      "epoch": 30.55,
+      "learning_rate": 4.618097014925373e-05,
+      "loss": 0.0002,
+      "step": 32752
+    },
+    {
+      "epoch": 30.56,
+      "learning_rate": 4.618050373134328e-05,
+      "loss": 0.0014,
+      "step": 32756
+    },
+    {
+      "epoch": 30.56,
+      "learning_rate": 4.6180037313432836e-05,
+      "loss": 0.0005,
+      "step": 32760
+    },
+    {
+      "epoch": 30.56,
+      "learning_rate": 4.617957089552239e-05,
+      "loss": 0.0001,
+      "step": 32764
+    },
+    {
+      "epoch": 30.57,
+      "learning_rate": 4.6179104477611945e-05,
+      "loss": 0.0012,
+      "step": 32768
+    },
+    {
+      "epoch": 30.57,
+      "learning_rate": 4.6178638059701494e-05,
+      "loss": 0.0006,
+      "step": 32772
+    },
+    {
+      "epoch": 30.57,
+      "learning_rate": 4.617817164179105e-05,
+      "loss": 0.0002,
+      "step": 32776
+    },
+    {
+      "epoch": 30.58,
+      "learning_rate": 4.6177705223880603e-05,
+      "loss": 0.0023,
+      "step": 32780
+    },
+    {
+      "epoch": 30.58,
+      "learning_rate": 4.617723880597015e-05,
+      "loss": 0.0011,
+      "step": 32784
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 4.61767723880597e-05,
+      "loss": 0.0008,
+      "step": 32788
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 4.617630597014926e-05,
+      "loss": 0.0011,
+      "step": 32792
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 4.617583955223881e-05,
+      "loss": 0.0,
+      "step": 32796
+    },
+    {
+      "epoch": 30.6,
+      "learning_rate": 4.617537313432836e-05,
+      "loss": 0.0021,
+      "step": 32800
+    },
+    {
+      "epoch": 30.6,
+      "learning_rate": 4.617490671641791e-05,
+      "loss": 0.001,
+      "step": 32804
+    },
+    {
+      "epoch": 30.6,
+      "learning_rate": 4.617444029850747e-05,
+      "loss": 0.0003,
+      "step": 32808
+    },
+    {
+      "epoch": 30.61,
+      "learning_rate": 4.6173973880597016e-05,
+      "loss": 0.0001,
+      "step": 32812
+    },
+    {
+      "epoch": 30.61,
+      "learning_rate": 4.617350746268657e-05,
+      "loss": 0.0076,
+      "step": 32816
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 4.617304104477612e-05,
+      "loss": 0.0001,
+      "step": 32820
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 4.6172574626865674e-05,
+      "loss": 0.0002,
+      "step": 32824
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 4.617210820895523e-05,
+      "loss": 0.0001,
+      "step": 32828
+    },
+    {
+      "epoch": 30.63,
+      "learning_rate": 4.6171641791044777e-05,
+      "loss": 0.002,
+      "step": 32832
+    },
+    {
+      "epoch": 30.63,
+      "learning_rate": 4.617117537313433e-05,
+      "loss": 0.0037,
+      "step": 32836
+    },
+    {
+      "epoch": 30.63,
+      "learning_rate": 4.6170708955223886e-05,
+      "loss": 0.0004,
+      "step": 32840
+    },
+    {
+      "epoch": 30.64,
+      "learning_rate": 4.6170242537313435e-05,
+      "loss": 0.001,
+      "step": 32844
+    },
+    {
+      "epoch": 30.64,
+      "learning_rate": 4.616977611940298e-05,
+      "loss": 0.0001,
+      "step": 32848
+    },
+    {
+      "epoch": 30.65,
+      "learning_rate": 4.6169309701492544e-05,
+      "loss": 0.0001,
+      "step": 32852
+    },
+    {
+      "epoch": 30.65,
+      "learning_rate": 4.616884328358209e-05,
+      "loss": 0.0,
+      "step": 32856
+    },
+    {
+      "epoch": 30.65,
+      "learning_rate": 4.616837686567164e-05,
+      "loss": 0.0011,
+      "step": 32860
+    },
+    {
+      "epoch": 30.66,
+      "learning_rate": 4.6167910447761196e-05,
+      "loss": 0.0023,
+      "step": 32864
+    },
+    {
+      "epoch": 30.66,
+      "learning_rate": 4.616744402985075e-05,
+      "loss": 0.0,
+      "step": 32868
+    },
+    {
+      "epoch": 30.66,
+      "learning_rate": 4.61669776119403e-05,
+      "loss": 0.0001,
+      "step": 32872
+    },
+    {
+      "epoch": 30.67,
+      "learning_rate": 4.6166511194029853e-05,
+      "loss": 0.0013,
+      "step": 32876
+    },
+    {
+      "epoch": 30.67,
+      "learning_rate": 4.61660447761194e-05,
+      "loss": 0.0,
+      "step": 32880
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 4.6165578358208957e-05,
+      "loss": 0.0006,
+      "step": 32884
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 4.616511194029851e-05,
+      "loss": 0.0001,
+      "step": 32888
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 4.616464552238806e-05,
+      "loss": 0.0001,
+      "step": 32892
+    },
+    {
+      "epoch": 30.69,
+      "learning_rate": 4.6164179104477614e-05,
+      "loss": 0.0009,
+      "step": 32896
+    },
+    {
+      "epoch": 30.69,
+      "learning_rate": 4.616371268656717e-05,
+      "loss": 0.0,
+      "step": 32900
+    },
+    {
+      "epoch": 30.69,
+      "learning_rate": 4.616324626865672e-05,
+      "loss": 0.002,
+      "step": 32904
+    },
+    {
+      "epoch": 30.7,
+      "learning_rate": 4.6162779850746266e-05,
+      "loss": 0.0001,
+      "step": 32908
+    },
+    {
+      "epoch": 30.7,
+      "learning_rate": 4.616231343283583e-05,
+      "loss": 0.0,
+      "step": 32912
+    },
+    {
+      "epoch": 30.71,
+      "learning_rate": 4.6161847014925375e-05,
+      "loss": 0.0041,
+      "step": 32916
+    },
+    {
+      "epoch": 30.71,
+      "learning_rate": 4.6161380597014924e-05,
+      "loss": 0.0001,
+      "step": 32920
+    },
+    {
+      "epoch": 30.71,
+      "learning_rate": 4.616091417910448e-05,
+      "loss": 0.0002,
+      "step": 32924
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 4.616044776119403e-05,
+      "loss": 0.0001,
+      "step": 32928
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 4.615998134328359e-05,
+      "loss": 0.0005,
+      "step": 32932
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 4.6159514925373136e-05,
+      "loss": 0.0006,
+      "step": 32936
+    },
+    {
+      "epoch": 30.73,
+      "learning_rate": 4.6159048507462685e-05,
+      "loss": 0.0002,
+      "step": 32940
+    },
+    {
+      "epoch": 30.73,
+      "learning_rate": 4.6158582089552246e-05,
+      "loss": 0.0004,
+      "step": 32944
+    },
+    {
+      "epoch": 30.73,
+      "learning_rate": 4.6158115671641794e-05,
+      "loss": 0.0068,
+      "step": 32948
+    },
+    {
+      "epoch": 30.74,
+      "learning_rate": 4.615764925373134e-05,
+      "loss": 0.0001,
+      "step": 32952
+    },
+    {
+      "epoch": 30.74,
+      "learning_rate": 4.61571828358209e-05,
+      "loss": 0.0001,
+      "step": 32956
+    },
+    {
+      "epoch": 30.75,
+      "learning_rate": 4.615671641791045e-05,
+      "loss": 0.0022,
+      "step": 32960
+    },
+    {
+      "epoch": 30.75,
+      "learning_rate": 4.615625e-05,
+      "loss": 0.0028,
+      "step": 32964
+    },
+    {
+      "epoch": 30.75,
+      "learning_rate": 4.6155783582089555e-05,
+      "loss": 0.0011,
+      "step": 32968
+    },
+    {
+      "epoch": 30.76,
+      "learning_rate": 4.615531716417911e-05,
+      "loss": 0.0004,
+      "step": 32972
+    },
+    {
+      "epoch": 30.76,
+      "learning_rate": 4.615485074626866e-05,
+      "loss": 0.0,
+      "step": 32976
+    },
+    {
+      "epoch": 30.76,
+      "learning_rate": 4.615438432835821e-05,
+      "loss": 0.0038,
+      "step": 32980
+    },
+    {
+      "epoch": 30.77,
+      "learning_rate": 4.615391791044776e-05,
+      "loss": 0.0007,
+      "step": 32984
+    },
+    {
+      "epoch": 30.77,
+      "learning_rate": 4.6153451492537316e-05,
+      "loss": 0.0,
+      "step": 32988
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 4.615298507462687e-05,
+      "loss": 0.0004,
+      "step": 32992
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 4.615251865671642e-05,
+      "loss": 0.0,
+      "step": 32996
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 4.615205223880597e-05,
+      "loss": 0.0,
+      "step": 33000
+    },
+    {
+      "epoch": 30.78,
+      "eval_exact_match": 0.7427466150870407,
+      "eval_exec": 0.7707930367504836,
+      "eval_loss": 0.3826291561126709,
+      "eval_runtime": 1590.6112,
+      "eval_samples_per_second": 0.65,
+      "step": 33000
+    },
+    {
+      "epoch": 30.79,
+      "learning_rate": 4.615158582089553e-05,
+      "loss": 0.0001,
+      "step": 33004
+    },
+    {
+      "epoch": 30.79,
+      "learning_rate": 4.615111940298508e-05,
+      "loss": 0.0,
+      "step": 33008
+    },
+    {
+      "epoch": 30.79,
+      "learning_rate": 4.6150652985074625e-05,
+      "loss": 0.0005,
+      "step": 33012
+    },
+    {
+      "epoch": 30.8,
+      "learning_rate": 4.615018656716418e-05,
+      "loss": 0.0007,
+      "step": 33016
+    },
+    {
+      "epoch": 30.8,
+      "learning_rate": 4.6149720149253735e-05,
+      "loss": 0.002,
+      "step": 33020
+    },
+    {
+      "epoch": 30.81,
+      "learning_rate": 4.6149253731343283e-05,
+      "loss": 0.0002,
+      "step": 33024
+    },
+    {
+      "epoch": 30.81,
+      "learning_rate": 4.614878731343284e-05,
+      "loss": 0.0003,
+      "step": 33028
+    },
+    {
+      "epoch": 30.81,
+      "learning_rate": 4.614832089552239e-05,
+      "loss": 0.0012,
+      "step": 33032
+    },
+    {
+      "epoch": 30.82,
+      "learning_rate": 4.614785447761194e-05,
+      "loss": 0.0001,
+      "step": 33036
+    },
+    {
+      "epoch": 30.82,
+      "learning_rate": 4.6147388059701496e-05,
+      "loss": 0.0003,
+      "step": 33040
+    },
+    {
+      "epoch": 30.82,
+      "learning_rate": 4.6146921641791044e-05,
+      "loss": 0.0001,
+      "step": 33044
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 4.61464552238806e-05,
+      "loss": 0.0,
+      "step": 33048
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 4.6145988805970154e-05,
+      "loss": 0.0001,
+      "step": 33052
+    },
+    {
+      "epoch": 30.84,
+      "learning_rate": 4.61455223880597e-05,
+      "loss": 0.0002,
+      "step": 33056
+    },
+    {
+      "epoch": 30.84,
+      "learning_rate": 4.614505597014925e-05,
+      "loss": 0.0001,
+      "step": 33060
+    },
+    {
+      "epoch": 30.84,
+      "learning_rate": 4.614458955223881e-05,
+      "loss": 0.0002,
+      "step": 33064
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 4.614412313432836e-05,
+      "loss": 0.0001,
+      "step": 33068
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 4.614365671641791e-05,
+      "loss": 0.001,
+      "step": 33072
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 4.614319029850746e-05,
+      "loss": 0.0001,
+      "step": 33076
+    },
+    {
+      "epoch": 30.86,
+      "learning_rate": 4.614272388059702e-05,
+      "loss": 0.0031,
+      "step": 33080
+    },
+    {
+      "epoch": 30.86,
+      "learning_rate": 4.6142257462686566e-05,
+      "loss": 0.0001,
+      "step": 33084
+    },
+    {
+      "epoch": 30.87,
+      "learning_rate": 4.614179104477612e-05,
+      "loss": 0.001,
+      "step": 33088
+    },
+    {
+      "epoch": 30.87,
+      "learning_rate": 4.6141324626865676e-05,
+      "loss": 0.0007,
+      "step": 33092
+    },
+    {
+      "epoch": 30.87,
+      "learning_rate": 4.614085820895523e-05,
+      "loss": 0.0015,
+      "step": 33096
+    },
+    {
+      "epoch": 30.88,
+      "learning_rate": 4.614039179104478e-05,
+      "loss": 0.0001,
+      "step": 33100
+    },
+    {
+      "epoch": 30.88,
+      "learning_rate": 4.613992537313433e-05,
+      "loss": 0.0,
+      "step": 33104
+    },
+    {
+      "epoch": 30.88,
+      "learning_rate": 4.613945895522388e-05,
+      "loss": 0.0024,
+      "step": 33108
+    },
+    {
+      "epoch": 30.89,
+      "learning_rate": 4.613899253731344e-05,
+      "loss": 0.0012,
+      "step": 33112
+    },
+    {
+      "epoch": 30.89,
+      "learning_rate": 4.6138526119402985e-05,
+      "loss": 0.0003,
+      "step": 33116
+    },
+    {
+      "epoch": 30.9,
+      "learning_rate": 4.613805970149254e-05,
+      "loss": 0.0001,
+      "step": 33120
+    },
+    {
+      "epoch": 30.9,
+      "learning_rate": 4.6137593283582095e-05,
+      "loss": 0.0001,
+      "step": 33124
+    },
+    {
+      "epoch": 30.9,
+      "learning_rate": 4.613712686567164e-05,
+      "loss": 0.0004,
+      "step": 33128
+    },
+    {
+      "epoch": 30.91,
+      "learning_rate": 4.61366604477612e-05,
+      "loss": 0.0007,
+      "step": 33132
+    },
+    {
+      "epoch": 30.91,
+      "learning_rate": 4.6136194029850746e-05,
+      "loss": 0.0001,
+      "step": 33136
+    },
+    {
+      "epoch": 30.91,
+      "learning_rate": 4.61357276119403e-05,
+      "loss": 0.0002,
+      "step": 33140
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 4.6135261194029856e-05,
+      "loss": 0.0002,
+      "step": 33144
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 4.6134794776119404e-05,
+      "loss": 0.0,
+      "step": 33148
+    },
+    {
+      "epoch": 30.93,
+      "learning_rate": 4.613432835820896e-05,
+      "loss": 0.0001,
+      "step": 33152
+    },
+    {
+      "epoch": 30.93,
+      "learning_rate": 4.6133861940298514e-05,
+      "loss": 0.0001,
+      "step": 33156
+    },
+    {
+      "epoch": 30.93,
+      "learning_rate": 4.613339552238806e-05,
+      "loss": 0.0002,
+      "step": 33160
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 4.613292910447761e-05,
+      "loss": 0.0013,
+      "step": 33164
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 4.6132462686567165e-05,
+      "loss": 0.0011,
+      "step": 33168
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 4.613199626865672e-05,
+      "loss": 0.0,
+      "step": 33172
+    },
+    {
+      "epoch": 30.95,
+      "learning_rate": 4.613152985074627e-05,
+      "loss": 0.0,
+      "step": 33176
+    },
+    {
+      "epoch": 30.95,
+      "learning_rate": 4.613106343283582e-05,
+      "loss": 0.0003,
+      "step": 33180
+    },
+    {
+      "epoch": 30.96,
+      "learning_rate": 4.613059701492538e-05,
+      "loss": 0.0002,
+      "step": 33184
+    },
+    {
+      "epoch": 30.96,
+      "learning_rate": 4.6130130597014926e-05,
+      "loss": 0.005,
+      "step": 33188
+    },
+    {
+      "epoch": 30.96,
+      "learning_rate": 4.612966417910448e-05,
+      "loss": 0.0004,
+      "step": 33192
+    },
+    {
+      "epoch": 30.97,
+      "learning_rate": 4.612919776119403e-05,
+      "loss": 0.0026,
+      "step": 33196
+    },
+    {
+      "epoch": 30.97,
+      "learning_rate": 4.6128731343283584e-05,
+      "loss": 0.0001,
+      "step": 33200
+    },
+    {
+      "epoch": 30.97,
+      "learning_rate": 4.612826492537314e-05,
+      "loss": 0.0016,
+      "step": 33204
+    },
+    {
+      "epoch": 30.98,
+      "learning_rate": 4.612779850746269e-05,
+      "loss": 0.0008,
+      "step": 33208
+    },
+    {
+      "epoch": 30.98,
+      "learning_rate": 4.612733208955224e-05,
+      "loss": 0.0031,
+      "step": 33212
+    },
+    {
+      "epoch": 30.98,
+      "learning_rate": 4.61268656716418e-05,
+      "loss": 0.0001,
+      "step": 33216
+    },
+    {
+      "epoch": 30.99,
+      "learning_rate": 4.6126399253731345e-05,
+      "loss": 0.0002,
+      "step": 33220
+    },
+    {
+      "epoch": 30.99,
+      "learning_rate": 4.612593283582089e-05,
+      "loss": 0.0014,
+      "step": 33224
+    },
+    {
+      "epoch": 31.0,
+      "learning_rate": 4.612546641791045e-05,
+      "loss": 0.0001,
+      "step": 33228
+    },
+    {
+      "epoch": 31.0,
+      "learning_rate": 4.6125e-05,
+      "loss": 0.003,
+      "step": 33232
+    },
+    {
+      "epoch": 31.0,
+      "learning_rate": 4.612453358208955e-05,
+      "loss": 0.0009,
+      "step": 33236
+    },
+    {
+      "epoch": 31.01,
+      "learning_rate": 4.6124067164179106e-05,
+      "loss": 0.001,
+      "step": 33240
+    },
+    {
+      "epoch": 31.01,
+      "learning_rate": 4.612360074626866e-05,
+      "loss": 0.003,
+      "step": 33244
+    },
+    {
+      "epoch": 31.01,
+      "learning_rate": 4.612313432835821e-05,
+      "loss": 0.0005,
+      "step": 33248
+    },
+    {
+      "epoch": 31.02,
+      "learning_rate": 4.6122667910447764e-05,
+      "loss": 0.0011,
+      "step": 33252
+    },
+    {
+      "epoch": 31.02,
+      "learning_rate": 4.612220149253731e-05,
+      "loss": 0.0022,
+      "step": 33256
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 4.6121735074626874e-05,
+      "loss": 0.0002,
+      "step": 33260
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 4.612126865671642e-05,
+      "loss": 0.0028,
+      "step": 33264
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 4.612080223880597e-05,
+      "loss": 0.0013,
+      "step": 33268
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 4.6120335820895525e-05,
+      "loss": 0.001,
+      "step": 33272
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 4.611986940298508e-05,
+      "loss": 0.0004,
+      "step": 33276
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 4.611940298507463e-05,
+      "loss": 0.0002,
+      "step": 33280
+    },
+    {
+      "epoch": 31.05,
+      "learning_rate": 4.611893656716418e-05,
+      "loss": 0.0001,
+      "step": 33284
+    },
+    {
+      "epoch": 31.05,
+      "learning_rate": 4.611847014925373e-05,
+      "loss": 0.005,
+      "step": 33288
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 4.6118003731343286e-05,
+      "loss": 0.0001,
+      "step": 33292
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 4.611753731343284e-05,
+      "loss": 0.0,
+      "step": 33296
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 4.611707089552239e-05,
+      "loss": 0.0003,
+      "step": 33300
+    },
+    {
+      "epoch": 31.07,
+      "learning_rate": 4.6116604477611944e-05,
+      "loss": 0.0015,
+      "step": 33304
+    },
+    {
+      "epoch": 31.07,
+      "learning_rate": 4.61161380597015e-05,
+      "loss": 0.0002,
+      "step": 33308
+    },
+    {
+      "epoch": 31.07,
+      "learning_rate": 4.611567164179105e-05,
+      "loss": 0.0001,
+      "step": 33312
+    },
+    {
+      "epoch": 31.08,
+      "learning_rate": 4.6115205223880595e-05,
+      "loss": 0.0001,
+      "step": 33316
+    },
+    {
+      "epoch": 31.08,
+      "learning_rate": 4.611473880597016e-05,
+      "loss": 0.0,
+      "step": 33320
+    },
+    {
+      "epoch": 31.09,
+      "learning_rate": 4.6114272388059705e-05,
+      "loss": 0.0,
+      "step": 33324
+    },
+    {
+      "epoch": 31.09,
+      "learning_rate": 4.611380597014925e-05,
+      "loss": 0.0006,
+      "step": 33328
+    },
+    {
+      "epoch": 31.09,
+      "learning_rate": 4.611333955223881e-05,
+      "loss": 0.0002,
+      "step": 33332
+    },
+    {
+      "epoch": 31.1,
+      "learning_rate": 4.611287313432836e-05,
+      "loss": 0.0004,
+      "step": 33336
+    },
+    {
+      "epoch": 31.1,
+      "learning_rate": 4.611240671641791e-05,
+      "loss": 0.0,
+      "step": 33340
+    },
+    {
+      "epoch": 31.1,
+      "learning_rate": 4.6111940298507466e-05,
+      "loss": 0.0001,
+      "step": 33344
+    },
+    {
+      "epoch": 31.11,
+      "learning_rate": 4.6111473880597014e-05,
+      "loss": 0.0,
+      "step": 33348
+    },
+    {
+      "epoch": 31.11,
+      "learning_rate": 4.611100746268657e-05,
+      "loss": 0.0006,
+      "step": 33352
+    },
+    {
+      "epoch": 31.12,
+      "learning_rate": 4.6110541044776124e-05,
+      "loss": 0.0013,
+      "step": 33356
+    },
+    {
+      "epoch": 31.12,
+      "learning_rate": 4.611007462686567e-05,
+      "loss": 0.0001,
+      "step": 33360
+    },
+    {
+      "epoch": 31.12,
+      "learning_rate": 4.610960820895523e-05,
+      "loss": 0.0038,
+      "step": 33364
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 4.610914179104478e-05,
+      "loss": 0.0016,
+      "step": 33368
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 4.610867537313433e-05,
+      "loss": 0.0002,
+      "step": 33372
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 4.610820895522388e-05,
+      "loss": 0.0001,
+      "step": 33376
+    },
+    {
+      "epoch": 31.14,
+      "learning_rate": 4.610774253731344e-05,
+      "loss": 0.0003,
+      "step": 33380
+    },
+    {
+      "epoch": 31.14,
+      "learning_rate": 4.610727611940299e-05,
+      "loss": 0.0107,
+      "step": 33384
+    },
+    {
+      "epoch": 31.15,
+      "learning_rate": 4.6106809701492536e-05,
+      "loss": 0.0001,
+      "step": 33388
+    },
+    {
+      "epoch": 31.15,
+      "learning_rate": 4.610634328358209e-05,
+      "loss": 0.0005,
+      "step": 33392
+    },
+    {
+      "epoch": 31.15,
+      "learning_rate": 4.6105876865671646e-05,
+      "loss": 0.0004,
+      "step": 33396
+    },
+    {
+      "epoch": 31.16,
+      "learning_rate": 4.6105410447761194e-05,
+      "loss": 0.0072,
+      "step": 33400
+    },
+    {
+      "epoch": 31.16,
+      "learning_rate": 4.610494402985075e-05,
+      "loss": 0.0,
+      "step": 33404
+    },
+    {
+      "epoch": 31.16,
+      "learning_rate": 4.61044776119403e-05,
+      "loss": 0.0003,
+      "step": 33408
+    },
+    {
+      "epoch": 31.17,
+      "learning_rate": 4.610401119402985e-05,
+      "loss": 0.0001,
+      "step": 33412
+    },
+    {
+      "epoch": 31.17,
+      "learning_rate": 4.610354477611941e-05,
+      "loss": 0.0004,
+      "step": 33416
+    },
+    {
+      "epoch": 31.18,
+      "learning_rate": 4.6103078358208955e-05,
+      "loss": 0.0,
+      "step": 33420
+    },
+    {
+      "epoch": 31.18,
+      "learning_rate": 4.610261194029851e-05,
+      "loss": 0.0001,
+      "step": 33424
+    },
+    {
+      "epoch": 31.18,
+      "learning_rate": 4.6102145522388065e-05,
+      "loss": 0.0001,
+      "step": 33428
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 4.610167910447761e-05,
+      "loss": 0.0002,
+      "step": 33432
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 4.610121268656717e-05,
+      "loss": 0.0032,
+      "step": 33436
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 4.6100746268656716e-05,
+      "loss": 0.0002,
+      "step": 33440
+    },
+    {
+      "epoch": 31.2,
+      "learning_rate": 4.610027985074627e-05,
+      "loss": 0.0001,
+      "step": 33444
+    },
+    {
+      "epoch": 31.2,
+      "learning_rate": 4.6099813432835826e-05,
+      "loss": 0.0005,
+      "step": 33448
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 4.6099347014925374e-05,
+      "loss": 0.0001,
+      "step": 33452
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 4.609888059701493e-05,
+      "loss": 0.0001,
+      "step": 33456
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 4.6098414179104484e-05,
+      "loss": 0.0,
+      "step": 33460
+    },
+    {
+      "epoch": 31.22,
+      "learning_rate": 4.609794776119403e-05,
+      "loss": 0.0001,
+      "step": 33464
+    },
+    {
+      "epoch": 31.22,
+      "learning_rate": 4.609748134328358e-05,
+      "loss": 0.0014,
+      "step": 33468
+    },
+    {
+      "epoch": 31.22,
+      "learning_rate": 4.609701492537314e-05,
+      "loss": 0.0008,
+      "step": 33472
+    },
+    {
+      "epoch": 31.23,
+      "learning_rate": 4.609654850746269e-05,
+      "loss": 0.0004,
+      "step": 33476
+    },
+    {
+      "epoch": 31.23,
+      "learning_rate": 4.609608208955224e-05,
+      "loss": 0.0002,
+      "step": 33480
+    },
+    {
+      "epoch": 31.24,
+      "learning_rate": 4.609561567164179e-05,
+      "loss": 0.0001,
+      "step": 33484
+    },
+    {
+      "epoch": 31.24,
+      "learning_rate": 4.609514925373135e-05,
+      "loss": 0.0001,
+      "step": 33488
+    },
+    {
+      "epoch": 31.24,
+      "learning_rate": 4.6094682835820896e-05,
+      "loss": 0.0036,
+      "step": 33492
+    },
+    {
+      "epoch": 31.25,
+      "learning_rate": 4.609421641791045e-05,
+      "loss": 0.0031,
+      "step": 33496
+    },
+    {
+      "epoch": 31.25,
+      "learning_rate": 4.609375e-05,
+      "loss": 0.005,
+      "step": 33500
+    },
+    {
+      "epoch": 31.25,
+      "eval_exact_match": 0.7456479690522244,
+      "eval_exec": 0.7659574468085106,
+      "eval_loss": 0.4037356376647949,
+      "eval_runtime": 1057.7845,
+      "eval_samples_per_second": 0.978,
+      "step": 33500
+    },
+    {
+      "epoch": 31.25,
+      "learning_rate": 4.6093283582089554e-05,
+      "loss": 0.0006,
+      "step": 33504
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 4.609281716417911e-05,
+      "loss": 0.0002,
+      "step": 33508
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 4.609235074626866e-05,
+      "loss": 0.0019,
+      "step": 33512
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 4.609188432835821e-05,
+      "loss": 0.0002,
+      "step": 33516
+    },
+    {
+      "epoch": 31.27,
+      "learning_rate": 4.609141791044777e-05,
+      "loss": 0.0003,
+      "step": 33520
+    },
+    {
+      "epoch": 31.27,
+      "learning_rate": 4.6090951492537315e-05,
+      "loss": 0.0,
+      "step": 33524
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 4.609048507462686e-05,
+      "loss": 0.0033,
+      "step": 33528
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 4.6090018656716425e-05,
+      "loss": 0.0,
+      "step": 33532
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 4.608955223880597e-05,
+      "loss": 0.0001,
+      "step": 33536
+    },
+    {
+      "epoch": 31.29,
+      "learning_rate": 4.608908582089552e-05,
+      "loss": 0.0003,
+      "step": 33540
+    },
+    {
+      "epoch": 31.29,
+      "learning_rate": 4.6088619402985076e-05,
+      "loss": 0.0002,
+      "step": 33544
+    },
+    {
+      "epoch": 31.29,
+      "learning_rate": 4.608815298507463e-05,
+      "loss": 0.0035,
+      "step": 33548
+    },
+    {
+      "epoch": 31.3,
+      "learning_rate": 4.608768656716418e-05,
+      "loss": 0.0,
+      "step": 33552
+    },
+    {
+      "epoch": 31.3,
+      "learning_rate": 4.6087220149253734e-05,
+      "loss": 0.0004,
+      "step": 33556
+    },
+    {
+      "epoch": 31.31,
+      "learning_rate": 4.608675373134328e-05,
+      "loss": 0.0001,
+      "step": 33560
+    },
+    {
+      "epoch": 31.31,
+      "learning_rate": 4.608628731343284e-05,
+      "loss": 0.0001,
+      "step": 33564
+    },
+    {
+      "epoch": 31.31,
+      "learning_rate": 4.608582089552239e-05,
+      "loss": 0.0002,
+      "step": 33568
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 4.608535447761194e-05,
+      "loss": 0.0005,
+      "step": 33572
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 4.6084888059701495e-05,
+      "loss": 0.0007,
+      "step": 33576
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 4.608442164179105e-05,
+      "loss": 0.0002,
+      "step": 33580
+    },
+    {
+      "epoch": 31.33,
+      "learning_rate": 4.60839552238806e-05,
+      "loss": 0.0003,
+      "step": 33584
+    },
+    {
+      "epoch": 31.33,
+      "learning_rate": 4.608348880597015e-05,
+      "loss": 0.0001,
+      "step": 33588
+    },
+    {
+      "epoch": 31.34,
+      "learning_rate": 4.608302238805971e-05,
+      "loss": 0.0,
+      "step": 33592
+    },
+    {
+      "epoch": 31.34,
+      "learning_rate": 4.6082555970149256e-05,
+      "loss": 0.0001,
+      "step": 33596
+    },
+    {
+      "epoch": 31.34,
+      "learning_rate": 4.608208955223881e-05,
+      "loss": 0.0,
+      "step": 33600
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 4.608162313432836e-05,
+      "loss": 0.0013,
+      "step": 33604
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 4.6081156716417914e-05,
+      "loss": 0.0,
+      "step": 33608
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 4.608069029850747e-05,
+      "loss": 0.0002,
+      "step": 33612
+    },
+    {
+      "epoch": 31.36,
+      "learning_rate": 4.608022388059702e-05,
+      "loss": 0.002,
+      "step": 33616
+    },
+    {
+      "epoch": 31.36,
+      "learning_rate": 4.6079757462686565e-05,
+      "loss": 0.0,
+      "step": 33620
+    },
+    {
+      "epoch": 31.37,
+      "learning_rate": 4.6079291044776127e-05,
+      "loss": 0.0001,
+      "step": 33624
+    },
+    {
+      "epoch": 31.37,
+      "learning_rate": 4.6078824626865675e-05,
+      "loss": 0.0053,
+      "step": 33628
+    },
+    {
+      "epoch": 31.37,
+      "learning_rate": 4.607835820895522e-05,
+      "loss": 0.0003,
+      "step": 33632
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 4.607789179104478e-05,
+      "loss": 0.0008,
+      "step": 33636
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 4.607742537313433e-05,
+      "loss": 0.0,
+      "step": 33640
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 4.607695895522388e-05,
+      "loss": 0.0028,
+      "step": 33644
+    },
+    {
+      "epoch": 31.39,
+      "learning_rate": 4.6076492537313436e-05,
+      "loss": 0.0001,
+      "step": 33648
+    },
+    {
+      "epoch": 31.39,
+      "learning_rate": 4.607602611940299e-05,
+      "loss": 0.001,
+      "step": 33652
+    },
+    {
+      "epoch": 31.4,
+      "learning_rate": 4.607555970149254e-05,
+      "loss": 0.0003,
+      "step": 33656
+    },
+    {
+      "epoch": 31.4,
+      "learning_rate": 4.6075093283582094e-05,
+      "loss": 0.0077,
+      "step": 33660
+    },
+    {
+      "epoch": 31.4,
+      "learning_rate": 4.607462686567164e-05,
+      "loss": 0.0003,
+      "step": 33664
+    },
+    {
+      "epoch": 31.41,
+      "learning_rate": 4.60741604477612e-05,
+      "loss": 0.0001,
+      "step": 33668
+    },
+    {
+      "epoch": 31.41,
+      "learning_rate": 4.607369402985075e-05,
+      "loss": 0.0001,
+      "step": 33672
+    },
+    {
+      "epoch": 31.41,
+      "learning_rate": 4.60732276119403e-05,
+      "loss": 0.0019,
+      "step": 33676
+    },
+    {
+      "epoch": 31.42,
+      "learning_rate": 4.607276119402985e-05,
+      "loss": 0.0003,
+      "step": 33680
+    },
+    {
+      "epoch": 31.42,
+      "learning_rate": 4.607229477611941e-05,
+      "loss": 0.0001,
+      "step": 33684
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 4.607182835820896e-05,
+      "loss": 0.0036,
+      "step": 33688
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 4.6071361940298506e-05,
+      "loss": 0.0,
+      "step": 33692
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 4.607089552238806e-05,
+      "loss": 0.0002,
+      "step": 33696
+    },
+    {
+      "epoch": 31.44,
+      "learning_rate": 4.6070429104477616e-05,
+      "loss": 0.0004,
+      "step": 33700
+    },
+    {
+      "epoch": 31.44,
+      "learning_rate": 4.6069962686567164e-05,
+      "loss": 0.0,
+      "step": 33704
+    },
+    {
+      "epoch": 31.44,
+      "learning_rate": 4.606949626865672e-05,
+      "loss": 0.0001,
+      "step": 33708
+    },
+    {
+      "epoch": 31.45,
+      "learning_rate": 4.6069029850746274e-05,
+      "loss": 0.0001,
+      "step": 33712
+    },
+    {
+      "epoch": 31.45,
+      "learning_rate": 4.606856343283582e-05,
+      "loss": 0.0005,
+      "step": 33716
+    },
+    {
+      "epoch": 31.46,
+      "learning_rate": 4.6068097014925377e-05,
+      "loss": 0.0005,
+      "step": 33720
+    },
+    {
+      "epoch": 31.46,
+      "learning_rate": 4.6067630597014925e-05,
+      "loss": 0.001,
+      "step": 33724
+    },
+    {
+      "epoch": 31.46,
+      "learning_rate": 4.606716417910448e-05,
+      "loss": 0.0002,
+      "step": 33728
+    },
+    {
+      "epoch": 31.47,
+      "learning_rate": 4.6066697761194034e-05,
+      "loss": 0.0002,
+      "step": 33732
+    },
+    {
+      "epoch": 31.47,
+      "learning_rate": 4.606623134328358e-05,
+      "loss": 0.0006,
+      "step": 33736
+    },
+    {
+      "epoch": 31.47,
+      "learning_rate": 4.606576492537313e-05,
+      "loss": 0.0002,
+      "step": 33740
+    },
+    {
+      "epoch": 31.48,
+      "learning_rate": 4.606529850746269e-05,
+      "loss": 0.0,
+      "step": 33744
+    },
+    {
+      "epoch": 31.48,
+      "learning_rate": 4.606483208955224e-05,
+      "loss": 0.0001,
+      "step": 33748
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 4.6064365671641795e-05,
+      "loss": 0.0,
+      "step": 33752
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 4.6063899253731344e-05,
+      "loss": 0.0014,
+      "step": 33756
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 4.60634328358209e-05,
+      "loss": 0.0001,
+      "step": 33760
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 4.6062966417910453e-05,
+      "loss": 0.0001,
+      "step": 33764
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 4.60625e-05,
+      "loss": 0.0038,
+      "step": 33768
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 4.6062033582089556e-05,
+      "loss": 0.0001,
+      "step": 33772
+    },
+    {
+      "epoch": 31.51,
+      "learning_rate": 4.606156716417911e-05,
+      "loss": 0.0,
+      "step": 33776
+    },
+    {
+      "epoch": 31.51,
+      "learning_rate": 4.606110074626866e-05,
+      "loss": 0.0001,
+      "step": 33780
+    },
+    {
+      "epoch": 31.51,
+      "learning_rate": 4.606063432835821e-05,
+      "loss": 0.0028,
+      "step": 33784
+    },
+    {
+      "epoch": 31.52,
+      "learning_rate": 4.606016791044776e-05,
+      "loss": 0.0003,
+      "step": 33788
+    },
+    {
+      "epoch": 31.52,
+      "learning_rate": 4.605970149253732e-05,
+      "loss": 0.0,
+      "step": 33792
+    },
+    {
+      "epoch": 31.53,
+      "learning_rate": 4.6059235074626866e-05,
+      "loss": 0.0,
+      "step": 33796
+    },
+    {
+      "epoch": 31.53,
+      "learning_rate": 4.605876865671642e-05,
+      "loss": 0.0,
+      "step": 33800
+    },
+    {
+      "epoch": 31.53,
+      "learning_rate": 4.6058302238805975e-05,
+      "loss": 0.0001,
+      "step": 33804
+    },
+    {
+      "epoch": 31.54,
+      "learning_rate": 4.6057835820895524e-05,
+      "loss": 0.0001,
+      "step": 33808
+    },
+    {
+      "epoch": 31.54,
+      "learning_rate": 4.605736940298508e-05,
+      "loss": 0.0105,
+      "step": 33812
+    },
+    {
+      "epoch": 31.54,
+      "learning_rate": 4.6056902985074627e-05,
+      "loss": 0.0003,
+      "step": 33816
+    },
+    {
+      "epoch": 31.55,
+      "learning_rate": 4.605643656716418e-05,
+      "loss": 0.0014,
+      "step": 33820
+    },
+    {
+      "epoch": 31.55,
+      "learning_rate": 4.6055970149253736e-05,
+      "loss": 0.0003,
+      "step": 33824
+    },
+    {
+      "epoch": 31.56,
+      "learning_rate": 4.6055503731343285e-05,
+      "loss": 0.0152,
+      "step": 33828
+    },
+    {
+      "epoch": 31.56,
+      "learning_rate": 4.605503731343284e-05,
+      "loss": 0.0,
+      "step": 33832
+    },
+    {
+      "epoch": 31.56,
+      "learning_rate": 4.6054570895522394e-05,
+      "loss": 0.0018,
+      "step": 33836
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 4.605410447761194e-05,
+      "loss": 0.0004,
+      "step": 33840
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 4.605363805970149e-05,
+      "loss": 0.0031,
+      "step": 33844
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 4.6053171641791045e-05,
+      "loss": 0.0,
+      "step": 33848
+    },
+    {
+      "epoch": 31.58,
+      "learning_rate": 4.60527052238806e-05,
+      "loss": 0.0001,
+      "step": 33852
+    },
+    {
+      "epoch": 31.58,
+      "learning_rate": 4.605223880597015e-05,
+      "loss": 0.0002,
+      "step": 33856
+    },
+    {
+      "epoch": 31.59,
+      "learning_rate": 4.6051772388059703e-05,
+      "loss": 0.0066,
+      "step": 33860
+    },
+    {
+      "epoch": 31.59,
+      "learning_rate": 4.605130597014926e-05,
+      "loss": 0.0006,
+      "step": 33864
+    },
+    {
+      "epoch": 31.59,
+      "learning_rate": 4.6050839552238806e-05,
+      "loss": 0.0001,
+      "step": 33868
+    },
+    {
+      "epoch": 31.6,
+      "learning_rate": 4.605037313432836e-05,
+      "loss": 0.0,
+      "step": 33872
+    },
+    {
+      "epoch": 31.6,
+      "learning_rate": 4.604990671641791e-05,
+      "loss": 0.0017,
+      "step": 33876
+    },
+    {
+      "epoch": 31.6,
+      "learning_rate": 4.6049440298507464e-05,
+      "loss": 0.0001,
+      "step": 33880
+    },
+    {
+      "epoch": 31.61,
+      "learning_rate": 4.604897388059702e-05,
+      "loss": 0.0003,
+      "step": 33884
+    },
+    {
+      "epoch": 31.61,
+      "learning_rate": 4.604850746268657e-05,
+      "loss": 0.0007,
+      "step": 33888
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 4.604804104477612e-05,
+      "loss": 0.0003,
+      "step": 33892
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 4.604757462686568e-05,
+      "loss": 0.001,
+      "step": 33896
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 4.6047108208955225e-05,
+      "loss": 0.0001,
+      "step": 33900
+    },
+    {
+      "epoch": 31.63,
+      "learning_rate": 4.6046641791044774e-05,
+      "loss": 0.0033,
+      "step": 33904
+    },
+    {
+      "epoch": 31.63,
+      "learning_rate": 4.604617537313433e-05,
+      "loss": 0.0002,
+      "step": 33908
+    },
+    {
+      "epoch": 31.63,
+      "learning_rate": 4.604570895522388e-05,
+      "loss": 0.0012,
+      "step": 33912
+    },
+    {
+      "epoch": 31.64,
+      "learning_rate": 4.604524253731344e-05,
+      "loss": 0.0001,
+      "step": 33916
+    },
+    {
+      "epoch": 31.64,
+      "learning_rate": 4.6044776119402986e-05,
+      "loss": 0.0001,
+      "step": 33920
+    },
+    {
+      "epoch": 31.65,
+      "learning_rate": 4.604430970149254e-05,
+      "loss": 0.0001,
+      "step": 33924
+    },
+    {
+      "epoch": 31.65,
+      "learning_rate": 4.6043843283582096e-05,
+      "loss": 0.0002,
+      "step": 33928
+    },
+    {
+      "epoch": 31.65,
+      "learning_rate": 4.6043376865671644e-05,
+      "loss": 0.0003,
+      "step": 33932
+    },
+    {
+      "epoch": 31.66,
+      "learning_rate": 4.604291044776119e-05,
+      "loss": 0.0003,
+      "step": 33936
+    },
+    {
+      "epoch": 31.66,
+      "learning_rate": 4.6042444029850754e-05,
+      "loss": 0.0039,
+      "step": 33940
+    },
+    {
+      "epoch": 31.66,
+      "learning_rate": 4.60419776119403e-05,
+      "loss": 0.0004,
+      "step": 33944
+    },
+    {
+      "epoch": 31.67,
+      "learning_rate": 4.604151119402985e-05,
+      "loss": 0.0002,
+      "step": 33948
+    },
+    {
+      "epoch": 31.67,
+      "learning_rate": 4.6041044776119405e-05,
+      "loss": 0.0,
+      "step": 33952
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 4.604057835820896e-05,
+      "loss": 0.0001,
+      "step": 33956
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 4.604011194029851e-05,
+      "loss": 0.0003,
+      "step": 33960
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 4.603964552238806e-05,
+      "loss": 0.0001,
+      "step": 33964
+    },
+    {
+      "epoch": 31.69,
+      "learning_rate": 4.603917910447761e-05,
+      "loss": 0.0018,
+      "step": 33968
+    },
+    {
+      "epoch": 31.69,
+      "learning_rate": 4.6038712686567166e-05,
+      "loss": 0.0001,
+      "step": 33972
+    },
+    {
+      "epoch": 31.69,
+      "learning_rate": 4.603824626865672e-05,
+      "loss": 0.0002,
+      "step": 33976
+    },
+    {
+      "epoch": 31.7,
+      "learning_rate": 4.603777985074627e-05,
+      "loss": 0.0014,
+      "step": 33980
+    },
+    {
+      "epoch": 31.7,
+      "learning_rate": 4.6037313432835824e-05,
+      "loss": 0.0003,
+      "step": 33984
+    },
+    {
+      "epoch": 31.71,
+      "learning_rate": 4.603684701492538e-05,
+      "loss": 0.0001,
+      "step": 33988
+    },
+    {
+      "epoch": 31.71,
+      "learning_rate": 4.603638059701493e-05,
+      "loss": 0.0005,
+      "step": 33992
+    },
+    {
+      "epoch": 31.71,
+      "learning_rate": 4.6035914179104475e-05,
+      "loss": 0.0028,
+      "step": 33996
+    },
+    {
+      "epoch": 31.72,
+      "learning_rate": 4.603544776119404e-05,
+      "loss": 0.0,
+      "step": 34000
+    },
+    {
+      "epoch": 31.72,
+      "eval_exact_match": 0.7379110251450677,
+      "eval_exec": 0.7572533849129593,
+      "eval_loss": 0.42316770553588867,
+      "eval_runtime": 1045.9822,
+      "eval_samples_per_second": 0.989,
+      "step": 34000
+    },
+    {
+      "epoch": 31.72,
+      "learning_rate": 4.6034981343283585e-05,
+      "loss": 0.0001,
+      "step": 34004
+    },
+    {
+      "epoch": 31.72,
+      "learning_rate": 4.603451492537313e-05,
+      "loss": 0.0,
+      "step": 34008
+    },
+    {
+      "epoch": 31.73,
+      "learning_rate": 4.603404850746269e-05,
+      "loss": 0.0,
+      "step": 34012
+    },
+    {
+      "epoch": 31.73,
+      "learning_rate": 4.603358208955224e-05,
+      "loss": 0.0002,
+      "step": 34016
+    },
+    {
+      "epoch": 31.73,
+      "learning_rate": 4.603311567164179e-05,
+      "loss": 0.0004,
+      "step": 34020
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 4.6032649253731346e-05,
+      "loss": 0.0029,
+      "step": 34024
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 4.6032182835820894e-05,
+      "loss": 0.0001,
+      "step": 34028
+    },
+    {
+      "epoch": 31.75,
+      "learning_rate": 4.603171641791045e-05,
+      "loss": 0.0004,
+      "step": 34032
+    },
+    {
+      "epoch": 31.75,
+      "learning_rate": 4.6031250000000004e-05,
+      "loss": 0.0001,
+      "step": 34036
+    },
+    {
+      "epoch": 31.75,
+      "learning_rate": 4.603078358208955e-05,
+      "loss": 0.0,
+      "step": 34040
+    },
+    {
+      "epoch": 31.76,
+      "learning_rate": 4.603031716417911e-05,
+      "loss": 0.0021,
+      "step": 34044
+    },
+    {
+      "epoch": 31.76,
+      "learning_rate": 4.602985074626866e-05,
+      "loss": 0.0,
+      "step": 34048
+    },
+    {
+      "epoch": 31.76,
+      "learning_rate": 4.602938432835821e-05,
+      "loss": 0.001,
+      "step": 34052
+    },
+    {
+      "epoch": 31.77,
+      "learning_rate": 4.602891791044776e-05,
+      "loss": 0.0001,
+      "step": 34056
+    },
+    {
+      "epoch": 31.77,
+      "learning_rate": 4.602845149253732e-05,
+      "loss": 0.0001,
+      "step": 34060
+    },
+    {
+      "epoch": 31.78,
+      "learning_rate": 4.602798507462687e-05,
+      "loss": 0.0004,
+      "step": 34064
+    },
+    {
+      "epoch": 31.78,
+      "learning_rate": 4.6027518656716416e-05,
+      "loss": 0.0,
+      "step": 34068
+    },
+    {
+      "epoch": 31.78,
+      "learning_rate": 4.602705223880597e-05,
+      "loss": 0.0,
+      "step": 34072
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 4.6026585820895526e-05,
+      "loss": 0.0002,
+      "step": 34076
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 4.602611940298508e-05,
+      "loss": 0.0005,
+      "step": 34080
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 4.602565298507463e-05,
+      "loss": 0.0015,
+      "step": 34084
+    },
+    {
+      "epoch": 31.8,
+      "learning_rate": 4.602518656716418e-05,
+      "loss": 0.0002,
+      "step": 34088
+    },
+    {
+      "epoch": 31.8,
+      "learning_rate": 4.602472014925374e-05,
+      "loss": 0.0008,
+      "step": 34092
+    },
+    {
+      "epoch": 31.81,
+      "learning_rate": 4.602425373134329e-05,
+      "loss": 0.0009,
+      "step": 34096
+    },
+    {
+      "epoch": 31.81,
+      "learning_rate": 4.6023787313432835e-05,
+      "loss": 0.0005,
+      "step": 34100
+    },
+    {
+      "epoch": 31.81,
+      "learning_rate": 4.602332089552239e-05,
+      "loss": 0.0059,
+      "step": 34104
+    },
+    {
+      "epoch": 31.82,
+      "learning_rate": 4.6022854477611945e-05,
+      "loss": 0.0012,
+      "step": 34108
+    },
+    {
+      "epoch": 31.82,
+      "learning_rate": 4.602238805970149e-05,
+      "loss": 0.0017,
+      "step": 34112
+    },
+    {
+      "epoch": 31.82,
+      "learning_rate": 4.602192164179105e-05,
+      "loss": 0.0003,
+      "step": 34116
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 4.6021455223880596e-05,
+      "loss": 0.0021,
+      "step": 34120
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 4.602098880597015e-05,
+      "loss": 0.0008,
+      "step": 34124
+    },
+    {
+      "epoch": 31.84,
+      "learning_rate": 4.6020522388059706e-05,
+      "loss": 0.0001,
+      "step": 34128
+    },
+    {
+      "epoch": 31.84,
+      "learning_rate": 4.6020055970149254e-05,
+      "loss": 0.0001,
+      "step": 34132
+    },
+    {
+      "epoch": 31.84,
+      "learning_rate": 4.601958955223881e-05,
+      "loss": 0.0,
+      "step": 34136
+    },
+    {
+      "epoch": 31.85,
+      "learning_rate": 4.6019123134328364e-05,
+      "loss": 0.0001,
+      "step": 34140
+    },
+    {
+      "epoch": 31.85,
+      "learning_rate": 4.601865671641791e-05,
+      "loss": 0.0032,
+      "step": 34144
+    },
+    {
+      "epoch": 31.85,
+      "learning_rate": 4.601819029850746e-05,
+      "loss": 0.0001,
+      "step": 34148
+    },
+    {
+      "epoch": 31.86,
+      "learning_rate": 4.601772388059702e-05,
+      "loss": 0.0007,
+      "step": 34152
+    },
+    {
+      "epoch": 31.86,
+      "learning_rate": 4.601725746268657e-05,
+      "loss": 0.0001,
+      "step": 34156
+    },
+    {
+      "epoch": 31.87,
+      "learning_rate": 4.601679104477612e-05,
+      "loss": 0.003,
+      "step": 34160
+    },
+    {
+      "epoch": 31.87,
+      "learning_rate": 4.601632462686567e-05,
+      "loss": 0.0003,
+      "step": 34164
+    },
+    {
+      "epoch": 31.87,
+      "learning_rate": 4.601585820895523e-05,
+      "loss": 0.0003,
+      "step": 34168
+    },
+    {
+      "epoch": 31.88,
+      "learning_rate": 4.6015391791044776e-05,
+      "loss": 0.0027,
+      "step": 34172
+    },
+    {
+      "epoch": 31.88,
+      "learning_rate": 4.601492537313433e-05,
+      "loss": 0.0,
+      "step": 34176
+    },
+    {
+      "epoch": 31.88,
+      "learning_rate": 4.601445895522388e-05,
+      "loss": 0.0005,
+      "step": 34180
+    },
+    {
+      "epoch": 31.89,
+      "learning_rate": 4.6013992537313434e-05,
+      "loss": 0.0002,
+      "step": 34184
+    },
+    {
+      "epoch": 31.89,
+      "learning_rate": 4.601352611940299e-05,
+      "loss": 0.0001,
+      "step": 34188
+    },
+    {
+      "epoch": 31.9,
+      "learning_rate": 4.601305970149254e-05,
+      "loss": 0.0,
+      "step": 34192
+    },
+    {
+      "epoch": 31.9,
+      "learning_rate": 4.601259328358209e-05,
+      "loss": 0.0004,
+      "step": 34196
+    },
+    {
+      "epoch": 31.9,
+      "learning_rate": 4.601212686567165e-05,
+      "loss": 0.0,
+      "step": 34200
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 4.6011660447761195e-05,
+      "loss": 0.0008,
+      "step": 34204
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 4.601119402985074e-05,
+      "loss": 0.0017,
+      "step": 34208
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 4.6010727611940305e-05,
+      "loss": 0.0002,
+      "step": 34212
+    },
+    {
+      "epoch": 31.92,
+      "learning_rate": 4.601026119402985e-05,
+      "loss": 0.0001,
+      "step": 34216
+    },
+    {
+      "epoch": 31.92,
+      "learning_rate": 4.60097947761194e-05,
+      "loss": 0.0001,
+      "step": 34220
+    },
+    {
+      "epoch": 31.93,
+      "learning_rate": 4.6009328358208956e-05,
+      "loss": 0.0009,
+      "step": 34224
+    },
+    {
+      "epoch": 31.93,
+      "learning_rate": 4.600886194029851e-05,
+      "loss": 0.001,
+      "step": 34228
+    },
+    {
+      "epoch": 31.93,
+      "learning_rate": 4.600839552238806e-05,
+      "loss": 0.0001,
+      "step": 34232
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 4.6007929104477614e-05,
+      "loss": 0.001,
+      "step": 34236
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 4.600746268656716e-05,
+      "loss": 0.0001,
+      "step": 34240
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 4.6006996268656724e-05,
+      "loss": 0.0,
+      "step": 34244
+    },
+    {
+      "epoch": 31.95,
+      "learning_rate": 4.600652985074627e-05,
+      "loss": 0.0044,
+      "step": 34248
+    },
+    {
+      "epoch": 31.95,
+      "learning_rate": 4.600606343283582e-05,
+      "loss": 0.0002,
+      "step": 34252
+    },
+    {
+      "epoch": 31.96,
+      "learning_rate": 4.6005597014925375e-05,
+      "loss": 0.0004,
+      "step": 34256
+    },
+    {
+      "epoch": 31.96,
+      "learning_rate": 4.600513059701493e-05,
+      "loss": 0.0011,
+      "step": 34260
+    },
+    {
+      "epoch": 31.96,
+      "learning_rate": 4.600466417910448e-05,
+      "loss": 0.0005,
+      "step": 34264
+    },
+    {
+      "epoch": 31.97,
+      "learning_rate": 4.600419776119403e-05,
+      "loss": 0.0033,
+      "step": 34268
+    },
+    {
+      "epoch": 31.97,
+      "learning_rate": 4.600373134328359e-05,
+      "loss": 0.0001,
+      "step": 34272
+    },
+    {
+      "epoch": 31.97,
+      "learning_rate": 4.6003264925373136e-05,
+      "loss": 0.0001,
+      "step": 34276
+    },
+    {
+      "epoch": 31.98,
+      "learning_rate": 4.600279850746269e-05,
+      "loss": 0.0001,
+      "step": 34280
+    },
+    {
+      "epoch": 31.98,
+      "learning_rate": 4.600233208955224e-05,
+      "loss": 0.0,
+      "step": 34284
+    },
+    {
+      "epoch": 31.98,
+      "learning_rate": 4.6001865671641794e-05,
+      "loss": 0.0001,
+      "step": 34288
+    },
+    {
+      "epoch": 31.99,
+      "learning_rate": 4.600139925373135e-05,
+      "loss": 0.0015,
+      "step": 34292
+    },
+    {
+      "epoch": 31.99,
+      "learning_rate": 4.60009328358209e-05,
+      "loss": 0.0002,
+      "step": 34296
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 4.6000466417910445e-05,
+      "loss": 0.0002,
+      "step": 34300
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 4.600000000000001e-05,
+      "loss": 0.001,
+      "step": 34304
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 4.5999533582089555e-05,
+      "loss": 0.0001,
+      "step": 34308
+    },
+    {
+      "epoch": 32.01,
+      "learning_rate": 4.59990671641791e-05,
+      "loss": 0.0013,
+      "step": 34312
+    },
+    {
+      "epoch": 32.01,
+      "learning_rate": 4.599860074626866e-05,
+      "loss": 0.0017,
+      "step": 34316
+    },
+    {
+      "epoch": 32.01,
+      "learning_rate": 4.599813432835821e-05,
+      "loss": 0.0014,
+      "step": 34320
+    },
+    {
+      "epoch": 32.02,
+      "learning_rate": 4.599766791044776e-05,
+      "loss": 0.0001,
+      "step": 34324
+    },
+    {
+      "epoch": 32.02,
+      "learning_rate": 4.5997201492537316e-05,
+      "loss": 0.0007,
+      "step": 34328
+    },
+    {
+      "epoch": 32.03,
+      "learning_rate": 4.599673507462687e-05,
+      "loss": 0.0001,
+      "step": 34332
+    },
+    {
+      "epoch": 32.03,
+      "learning_rate": 4.599626865671642e-05,
+      "loss": 0.0001,
+      "step": 34336
+    },
+    {
+      "epoch": 32.03,
+      "learning_rate": 4.5995802238805974e-05,
+      "loss": 0.0001,
+      "step": 34340
+    },
+    {
+      "epoch": 32.04,
+      "learning_rate": 4.599533582089552e-05,
+      "loss": 0.0001,
+      "step": 34344
+    },
+    {
+      "epoch": 32.04,
+      "learning_rate": 4.599486940298508e-05,
+      "loss": 0.0,
+      "step": 34348
+    },
+    {
+      "epoch": 32.04,
+      "learning_rate": 4.599440298507463e-05,
+      "loss": 0.0,
+      "step": 34352
+    },
+    {
+      "epoch": 32.05,
+      "learning_rate": 4.599393656716418e-05,
+      "loss": 0.0,
+      "step": 34356
+    },
+    {
+      "epoch": 32.05,
+      "learning_rate": 4.599347014925373e-05,
+      "loss": 0.0003,
+      "step": 34360
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 4.599300373134329e-05,
+      "loss": 0.0,
+      "step": 34364
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 4.599253731343284e-05,
+      "loss": 0.0004,
+      "step": 34368
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 4.5992070895522386e-05,
+      "loss": 0.0,
+      "step": 34372
+    },
+    {
+      "epoch": 32.07,
+      "learning_rate": 4.599160447761194e-05,
+      "loss": 0.0001,
+      "step": 34376
+    },
+    {
+      "epoch": 32.07,
+      "learning_rate": 4.5991138059701496e-05,
+      "loss": 0.0001,
+      "step": 34380
+    },
+    {
+      "epoch": 32.07,
+      "learning_rate": 4.5990671641791044e-05,
+      "loss": 0.0128,
+      "step": 34384
+    },
+    {
+      "epoch": 32.08,
+      "learning_rate": 4.59902052238806e-05,
+      "loss": 0.0002,
+      "step": 34388
+    },
+    {
+      "epoch": 32.08,
+      "learning_rate": 4.5989738805970154e-05,
+      "loss": 0.0038,
+      "step": 34392
+    },
+    {
+      "epoch": 32.09,
+      "learning_rate": 4.59892723880597e-05,
+      "loss": 0.0015,
+      "step": 34396
+    },
+    {
+      "epoch": 32.09,
+      "learning_rate": 4.598880597014926e-05,
+      "loss": 0.0,
+      "step": 34400
+    },
+    {
+      "epoch": 32.09,
+      "learning_rate": 4.5988339552238805e-05,
+      "loss": 0.0067,
+      "step": 34404
+    },
+    {
+      "epoch": 32.1,
+      "learning_rate": 4.598787313432836e-05,
+      "loss": 0.0001,
+      "step": 34408
+    },
+    {
+      "epoch": 32.1,
+      "learning_rate": 4.5987406716417915e-05,
+      "loss": 0.0,
+      "step": 34412
+    },
+    {
+      "epoch": 32.1,
+      "learning_rate": 4.598694029850746e-05,
+      "loss": 0.0002,
+      "step": 34416
+    },
+    {
+      "epoch": 32.11,
+      "learning_rate": 4.598647388059702e-05,
+      "loss": 0.0,
+      "step": 34420
+    },
+    {
+      "epoch": 32.11,
+      "learning_rate": 4.598600746268657e-05,
+      "loss": 0.0002,
+      "step": 34424
+    },
+    {
+      "epoch": 32.12,
+      "learning_rate": 4.598554104477612e-05,
+      "loss": 0.0001,
+      "step": 34428
+    },
+    {
+      "epoch": 32.12,
+      "learning_rate": 4.5985074626865676e-05,
+      "loss": 0.0,
+      "step": 34432
+    },
+    {
+      "epoch": 32.12,
+      "learning_rate": 4.5984608208955224e-05,
+      "loss": 0.0,
+      "step": 34436
+    },
+    {
+      "epoch": 32.13,
+      "learning_rate": 4.598414179104478e-05,
+      "loss": 0.0003,
+      "step": 34440
+    },
+    {
+      "epoch": 32.13,
+      "learning_rate": 4.5983675373134334e-05,
+      "loss": 0.0,
+      "step": 34444
+    },
+    {
+      "epoch": 32.13,
+      "learning_rate": 4.598320895522388e-05,
+      "loss": 0.0,
+      "step": 34448
+    },
+    {
+      "epoch": 32.14,
+      "learning_rate": 4.598274253731344e-05,
+      "loss": 0.0001,
+      "step": 34452
+    },
+    {
+      "epoch": 32.14,
+      "learning_rate": 4.598227611940299e-05,
+      "loss": 0.0005,
+      "step": 34456
+    },
+    {
+      "epoch": 32.15,
+      "learning_rate": 4.598180970149254e-05,
+      "loss": 0.0066,
+      "step": 34460
+    },
+    {
+      "epoch": 32.15,
+      "learning_rate": 4.598134328358209e-05,
+      "loss": 0.0007,
+      "step": 34464
+    },
+    {
+      "epoch": 32.15,
+      "learning_rate": 4.598087686567164e-05,
+      "loss": 0.0022,
+      "step": 34468
+    },
+    {
+      "epoch": 32.16,
+      "learning_rate": 4.59804104477612e-05,
+      "loss": 0.0,
+      "step": 34472
+    },
+    {
+      "epoch": 32.16,
+      "learning_rate": 4.5979944029850746e-05,
+      "loss": 0.001,
+      "step": 34476
+    },
+    {
+      "epoch": 32.16,
+      "learning_rate": 4.59794776119403e-05,
+      "loss": 0.0,
+      "step": 34480
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 4.5979011194029856e-05,
+      "loss": 0.0001,
+      "step": 34484
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 4.5978544776119404e-05,
+      "loss": 0.0052,
+      "step": 34488
+    },
+    {
+      "epoch": 32.18,
+      "learning_rate": 4.597807835820896e-05,
+      "loss": 0.0,
+      "step": 34492
+    },
+    {
+      "epoch": 32.18,
+      "learning_rate": 4.597761194029851e-05,
+      "loss": 0.0,
+      "step": 34496
+    },
+    {
+      "epoch": 32.18,
+      "learning_rate": 4.597714552238806e-05,
+      "loss": 0.0005,
+      "step": 34500
+    },
+    {
+      "epoch": 32.18,
+      "eval_exact_match": 0.7282398452611218,
+      "eval_exec": 0.7495164410058027,
+      "eval_loss": 0.42082998156547546,
+      "eval_runtime": 1063.8322,
+      "eval_samples_per_second": 0.972,
+      "step": 34500
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 4.597667910447762e-05,
+      "loss": 0.0003,
+      "step": 34504
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 4.5976212686567165e-05,
+      "loss": 0.0007,
+      "step": 34508
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 4.597574626865672e-05,
+      "loss": 0.0003,
+      "step": 34512
+    },
+    {
+      "epoch": 32.2,
+      "learning_rate": 4.5975279850746275e-05,
+      "loss": 0.0014,
+      "step": 34516
+    },
+    {
+      "epoch": 32.2,
+      "learning_rate": 4.597481343283582e-05,
+      "loss": 0.0015,
+      "step": 34520
+    },
+    {
+      "epoch": 32.21,
+      "learning_rate": 4.597434701492537e-05,
+      "loss": 0.0,
+      "step": 34524
+    },
+    {
+      "epoch": 32.21,
+      "learning_rate": 4.5973880597014926e-05,
+      "loss": 0.0003,
+      "step": 34528
+    },
+    {
+      "epoch": 32.21,
+      "learning_rate": 4.597341417910448e-05,
+      "loss": 0.0005,
+      "step": 34532
+    },
+    {
+      "epoch": 32.22,
+      "learning_rate": 4.597294776119403e-05,
+      "loss": 0.0,
+      "step": 34536
+    },
+    {
+      "epoch": 32.22,
+      "learning_rate": 4.5972481343283584e-05,
+      "loss": 0.0002,
+      "step": 34540
+    },
+    {
+      "epoch": 32.22,
+      "learning_rate": 4.597201492537314e-05,
+      "loss": 0.0,
+      "step": 34544
+    },
+    {
+      "epoch": 32.23,
+      "learning_rate": 4.597154850746269e-05,
+      "loss": 0.0001,
+      "step": 34548
+    },
+    {
+      "epoch": 32.23,
+      "learning_rate": 4.597108208955224e-05,
+      "loss": 0.0026,
+      "step": 34552
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 4.597061567164179e-05,
+      "loss": 0.0001,
+      "step": 34556
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 4.5970149253731345e-05,
+      "loss": 0.0009,
+      "step": 34560
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 4.59696828358209e-05,
+      "loss": 0.0001,
+      "step": 34564
+    },
+    {
+      "epoch": 32.25,
+      "learning_rate": 4.596921641791045e-05,
+      "loss": 0.008,
+      "step": 34568
+    },
+    {
+      "epoch": 32.25,
+      "learning_rate": 4.596875e-05,
+      "loss": 0.0003,
+      "step": 34572
+    },
+    {
+      "epoch": 32.25,
+      "learning_rate": 4.596828358208956e-05,
+      "loss": 0.0008,
+      "step": 34576
+    },
+    {
+      "epoch": 32.26,
+      "learning_rate": 4.5967817164179106e-05,
+      "loss": 0.0063,
+      "step": 34580
+    },
+    {
+      "epoch": 32.26,
+      "learning_rate": 4.596735074626866e-05,
+      "loss": 0.0002,
+      "step": 34584
+    },
+    {
+      "epoch": 32.26,
+      "learning_rate": 4.596688432835821e-05,
+      "loss": 0.0002,
+      "step": 34588
+    },
+    {
+      "epoch": 32.27,
+      "learning_rate": 4.5966417910447764e-05,
+      "loss": 0.0001,
+      "step": 34592
+    },
+    {
+      "epoch": 32.27,
+      "learning_rate": 4.596595149253732e-05,
+      "loss": 0.0008,
+      "step": 34596
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 4.596548507462687e-05,
+      "loss": 0.0,
+      "step": 34600
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 4.596501865671642e-05,
+      "loss": 0.0031,
+      "step": 34604
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 4.5964552238805977e-05,
+      "loss": 0.0001,
+      "step": 34608
+    },
+    {
+      "epoch": 32.29,
+      "learning_rate": 4.5964085820895525e-05,
+      "loss": 0.0,
+      "step": 34612
+    },
+    {
+      "epoch": 32.29,
+      "learning_rate": 4.596361940298507e-05,
+      "loss": 0.0001,
+      "step": 34616
+    },
+    {
+      "epoch": 32.29,
+      "learning_rate": 4.5963152985074634e-05,
+      "loss": 0.0011,
+      "step": 34620
+    },
+    {
+      "epoch": 32.3,
+      "learning_rate": 4.596268656716418e-05,
+      "loss": 0.0011,
+      "step": 34624
+    },
+    {
+      "epoch": 32.3,
+      "learning_rate": 4.596222014925373e-05,
+      "loss": 0.0004,
+      "step": 34628
+    },
+    {
+      "epoch": 32.31,
+      "learning_rate": 4.5961753731343286e-05,
+      "loss": 0.0001,
+      "step": 34632
+    },
+    {
+      "epoch": 32.31,
+      "learning_rate": 4.596128731343284e-05,
+      "loss": 0.0006,
+      "step": 34636
+    },
+    {
+      "epoch": 32.31,
+      "learning_rate": 4.596082089552239e-05,
+      "loss": 0.0007,
+      "step": 34640
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 4.5960354477611944e-05,
+      "loss": 0.0005,
+      "step": 34644
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 4.595988805970149e-05,
+      "loss": 0.0015,
+      "step": 34648
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 4.5959421641791047e-05,
+      "loss": 0.0017,
+      "step": 34652
+    },
+    {
+      "epoch": 32.33,
+      "learning_rate": 4.59589552238806e-05,
+      "loss": 0.0,
+      "step": 34656
+    },
+    {
+      "epoch": 32.33,
+      "learning_rate": 4.595848880597015e-05,
+      "loss": 0.0,
+      "step": 34660
+    },
+    {
+      "epoch": 32.34,
+      "learning_rate": 4.5958022388059705e-05,
+      "loss": 0.0,
+      "step": 34664
+    },
+    {
+      "epoch": 32.34,
+      "learning_rate": 4.595755597014926e-05,
+      "loss": 0.0001,
+      "step": 34668
+    },
+    {
+      "epoch": 32.34,
+      "learning_rate": 4.595708955223881e-05,
+      "loss": 0.0005,
+      "step": 34672
+    },
+    {
+      "epoch": 32.35,
+      "learning_rate": 4.5956623134328356e-05,
+      "loss": 0.0001,
+      "step": 34676
+    },
+    {
+      "epoch": 32.35,
+      "learning_rate": 4.595615671641792e-05,
+      "loss": 0.0001,
+      "step": 34680
+    },
+    {
+      "epoch": 32.35,
+      "learning_rate": 4.5955690298507466e-05,
+      "loss": 0.0005,
+      "step": 34684
+    },
+    {
+      "epoch": 32.36,
+      "learning_rate": 4.5955223880597014e-05,
+      "loss": 0.0002,
+      "step": 34688
+    },
+    {
+      "epoch": 32.36,
+      "learning_rate": 4.595475746268657e-05,
+      "loss": 0.0002,
+      "step": 34692
+    },
+    {
+      "epoch": 32.37,
+      "learning_rate": 4.5954291044776123e-05,
+      "loss": 0.0003,
+      "step": 34696
+    },
+    {
+      "epoch": 32.37,
+      "learning_rate": 4.595382462686567e-05,
+      "loss": 0.0009,
+      "step": 34700
+    },
+    {
+      "epoch": 32.37,
+      "learning_rate": 4.5953358208955227e-05,
+      "loss": 0.0001,
+      "step": 34704
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 4.5952891791044775e-05,
+      "loss": 0.0001,
+      "step": 34708
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 4.595242537313433e-05,
+      "loss": 0.0015,
+      "step": 34712
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 4.5951958955223884e-05,
+      "loss": 0.0001,
+      "step": 34716
+    },
+    {
+      "epoch": 32.39,
+      "learning_rate": 4.595149253731343e-05,
+      "loss": 0.0001,
+      "step": 34720
+    },
+    {
+      "epoch": 32.39,
+      "learning_rate": 4.595102611940299e-05,
+      "loss": 0.0005,
+      "step": 34724
+    },
+    {
+      "epoch": 32.4,
+      "learning_rate": 4.595055970149254e-05,
+      "loss": 0.0,
+      "step": 34728
+    },
+    {
+      "epoch": 32.4,
+      "learning_rate": 4.595009328358209e-05,
+      "loss": 0.0002,
+      "step": 34732
+    },
+    {
+      "epoch": 32.4,
+      "learning_rate": 4.5949626865671645e-05,
+      "loss": 0.0,
+      "step": 34736
+    },
+    {
+      "epoch": 32.41,
+      "learning_rate": 4.59491604477612e-05,
+      "loss": 0.0,
+      "step": 34740
+    },
+    {
+      "epoch": 32.41,
+      "learning_rate": 4.594869402985075e-05,
+      "loss": 0.0,
+      "step": 34744
+    },
+    {
+      "epoch": 32.41,
+      "learning_rate": 4.5948227611940303e-05,
+      "loss": 0.0006,
+      "step": 34748
+    },
+    {
+      "epoch": 32.42,
+      "learning_rate": 4.594776119402985e-05,
+      "loss": 0.0009,
+      "step": 34752
+    },
+    {
+      "epoch": 32.42,
+      "learning_rate": 4.5947294776119406e-05,
+      "loss": 0.0001,
+      "step": 34756
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 4.594682835820896e-05,
+      "loss": 0.0004,
+      "step": 34760
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 4.594636194029851e-05,
+      "loss": 0.002,
+      "step": 34764
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 4.594589552238806e-05,
+      "loss": 0.0014,
+      "step": 34768
+    },
+    {
+      "epoch": 32.44,
+      "learning_rate": 4.594542910447762e-05,
+      "loss": 0.0021,
+      "step": 34772
+    },
+    {
+      "epoch": 32.44,
+      "learning_rate": 4.594496268656717e-05,
+      "loss": 0.0005,
+      "step": 34776
+    },
+    {
+      "epoch": 32.44,
+      "learning_rate": 4.5944496268656716e-05,
+      "loss": 0.0,
+      "step": 34780
+    },
+    {
+      "epoch": 32.45,
+      "learning_rate": 4.594402985074627e-05,
+      "loss": 0.0006,
+      "step": 34784
+    },
+    {
+      "epoch": 32.45,
+      "learning_rate": 4.5943563432835825e-05,
+      "loss": 0.0,
+      "step": 34788
+    },
+    {
+      "epoch": 32.46,
+      "learning_rate": 4.5943097014925374e-05,
+      "loss": 0.0002,
+      "step": 34792
+    },
+    {
+      "epoch": 32.46,
+      "learning_rate": 4.594263059701493e-05,
+      "loss": 0.0047,
+      "step": 34796
+    },
+    {
+      "epoch": 32.46,
+      "learning_rate": 4.5942164179104477e-05,
+      "loss": 0.0005,
+      "step": 34800
+    },
+    {
+      "epoch": 32.47,
+      "learning_rate": 4.594169776119403e-05,
+      "loss": 0.0003,
+      "step": 34804
+    },
+    {
+      "epoch": 32.47,
+      "learning_rate": 4.5941231343283586e-05,
+      "loss": 0.0183,
+      "step": 34808
+    },
+    {
+      "epoch": 32.47,
+      "learning_rate": 4.5940764925373134e-05,
+      "loss": 0.0,
+      "step": 34812
+    },
+    {
+      "epoch": 32.48,
+      "learning_rate": 4.594029850746269e-05,
+      "loss": 0.0005,
+      "step": 34816
+    },
+    {
+      "epoch": 32.48,
+      "learning_rate": 4.5939832089552244e-05,
+      "loss": 0.0013,
+      "step": 34820
+    },
+    {
+      "epoch": 32.49,
+      "learning_rate": 4.593936567164179e-05,
+      "loss": 0.0042,
+      "step": 34824
+    },
+    {
+      "epoch": 32.49,
+      "learning_rate": 4.593889925373134e-05,
+      "loss": 0.0003,
+      "step": 34828
+    },
+    {
+      "epoch": 32.49,
+      "learning_rate": 4.59384328358209e-05,
+      "loss": 0.0,
+      "step": 34832
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 4.593796641791045e-05,
+      "loss": 0.0001,
+      "step": 34836
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 4.59375e-05,
+      "loss": 0.0001,
+      "step": 34840
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 4.5937033582089553e-05,
+      "loss": 0.0001,
+      "step": 34844
+    },
+    {
+      "epoch": 32.51,
+      "learning_rate": 4.593656716417911e-05,
+      "loss": 0.0002,
+      "step": 34848
+    },
+    {
+      "epoch": 32.51,
+      "learning_rate": 4.5936100746268656e-05,
+      "loss": 0.0001,
+      "step": 34852
+    },
+    {
+      "epoch": 32.51,
+      "learning_rate": 4.593563432835821e-05,
+      "loss": 0.0,
+      "step": 34856
+    },
+    {
+      "epoch": 32.52,
+      "learning_rate": 4.593516791044776e-05,
+      "loss": 0.0035,
+      "step": 34860
+    },
+    {
+      "epoch": 32.52,
+      "learning_rate": 4.5934701492537314e-05,
+      "loss": 0.0043,
+      "step": 34864
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 4.593423507462687e-05,
+      "loss": 0.0,
+      "step": 34868
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 4.593376865671642e-05,
+      "loss": 0.0001,
+      "step": 34872
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 4.593330223880597e-05,
+      "loss": 0.0007,
+      "step": 34876
+    },
+    {
+      "epoch": 32.54,
+      "learning_rate": 4.593283582089553e-05,
+      "loss": 0.0002,
+      "step": 34880
+    },
+    {
+      "epoch": 32.54,
+      "learning_rate": 4.5932369402985075e-05,
+      "loss": 0.0004,
+      "step": 34884
+    },
+    {
+      "epoch": 32.54,
+      "learning_rate": 4.5931902985074624e-05,
+      "loss": 0.0,
+      "step": 34888
+    },
+    {
+      "epoch": 32.55,
+      "learning_rate": 4.5931436567164185e-05,
+      "loss": 0.0021,
+      "step": 34892
+    },
+    {
+      "epoch": 32.55,
+      "learning_rate": 4.593097014925373e-05,
+      "loss": 0.0,
+      "step": 34896
+    },
+    {
+      "epoch": 32.56,
+      "learning_rate": 4.593050373134329e-05,
+      "loss": 0.0002,
+      "step": 34900
+    },
+    {
+      "epoch": 32.56,
+      "learning_rate": 4.5930037313432836e-05,
+      "loss": 0.0002,
+      "step": 34904
+    },
+    {
+      "epoch": 32.56,
+      "learning_rate": 4.592957089552239e-05,
+      "loss": 0.0001,
+      "step": 34908
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 4.5929104477611946e-05,
+      "loss": 0.0036,
+      "step": 34912
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 4.5928638059701494e-05,
+      "loss": 0.0019,
+      "step": 34916
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 4.592817164179104e-05,
+      "loss": 0.0009,
+      "step": 34920
+    },
+    {
+      "epoch": 32.58,
+      "learning_rate": 4.5927705223880604e-05,
+      "loss": 0.0015,
+      "step": 34924
+    },
+    {
+      "epoch": 32.58,
+      "learning_rate": 4.592723880597015e-05,
+      "loss": 0.0002,
+      "step": 34928
+    },
+    {
+      "epoch": 32.59,
+      "learning_rate": 4.59267723880597e-05,
+      "loss": 0.0,
+      "step": 34932
+    },
+    {
+      "epoch": 32.59,
+      "learning_rate": 4.5926305970149255e-05,
+      "loss": 0.0001,
+      "step": 34936
+    },
+    {
+      "epoch": 32.59,
+      "learning_rate": 4.592583955223881e-05,
+      "loss": 0.0014,
+      "step": 34940
+    },
+    {
+      "epoch": 32.6,
+      "learning_rate": 4.592537313432836e-05,
+      "loss": 0.0001,
+      "step": 34944
+    },
+    {
+      "epoch": 32.6,
+      "learning_rate": 4.592490671641791e-05,
+      "loss": 0.0001,
+      "step": 34948
+    },
+    {
+      "epoch": 32.6,
+      "learning_rate": 4.592444029850747e-05,
+      "loss": 0.0002,
+      "step": 34952
+    },
+    {
+      "epoch": 32.61,
+      "learning_rate": 4.5923973880597016e-05,
+      "loss": 0.0001,
+      "step": 34956
+    },
+    {
+      "epoch": 32.61,
+      "learning_rate": 4.592350746268657e-05,
+      "loss": 0.001,
+      "step": 34960
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 4.592304104477612e-05,
+      "loss": 0.0002,
+      "step": 34964
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 4.5922574626865674e-05,
+      "loss": 0.0014,
+      "step": 34968
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 4.592210820895523e-05,
+      "loss": 0.0,
+      "step": 34972
+    },
+    {
+      "epoch": 32.63,
+      "learning_rate": 4.592164179104478e-05,
+      "loss": 0.0013,
+      "step": 34976
+    },
+    {
+      "epoch": 32.63,
+      "learning_rate": 4.5921175373134325e-05,
+      "loss": 0.0001,
+      "step": 34980
+    },
+    {
+      "epoch": 32.63,
+      "learning_rate": 4.592070895522389e-05,
+      "loss": 0.0029,
+      "step": 34984
+    },
+    {
+      "epoch": 32.64,
+      "learning_rate": 4.5920242537313435e-05,
+      "loss": 0.0,
+      "step": 34988
+    },
+    {
+      "epoch": 32.64,
+      "learning_rate": 4.591977611940298e-05,
+      "loss": 0.0009,
+      "step": 34992
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 4.591930970149254e-05,
+      "loss": 0.0,
+      "step": 34996
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 4.591884328358209e-05,
+      "loss": 0.0,
+      "step": 35000
+    },
+    {
+      "epoch": 32.65,
+      "eval_exact_match": 0.7350096711798839,
+      "eval_exec": 0.7543520309477756,
+      "eval_loss": 0.41416093707084656,
+      "eval_runtime": 1061.3828,
+      "eval_samples_per_second": 0.974,
+      "step": 35000
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 4.591837686567164e-05,
+      "loss": 0.0063,
+      "step": 35004
+    },
+    {
+      "epoch": 32.66,
+      "learning_rate": 4.5917910447761196e-05,
+      "loss": 0.0001,
+      "step": 35008
+    },
+    {
+      "epoch": 32.66,
+      "learning_rate": 4.591744402985075e-05,
+      "loss": 0.0001,
+      "step": 35012
+    },
+    {
+      "epoch": 32.66,
+      "learning_rate": 4.59169776119403e-05,
+      "loss": 0.0014,
+      "step": 35016
+    },
+    {
+      "epoch": 32.67,
+      "learning_rate": 4.5916511194029854e-05,
+      "loss": 0.0,
+      "step": 35020
+    },
+    {
+      "epoch": 32.67,
+      "learning_rate": 4.59160447761194e-05,
+      "loss": 0.0001,
+      "step": 35024
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 4.591557835820896e-05,
+      "loss": 0.0005,
+      "step": 35028
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 4.591511194029851e-05,
+      "loss": 0.0,
+      "step": 35032
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 4.591464552238806e-05,
+      "loss": 0.0,
+      "step": 35036
+    },
+    {
+      "epoch": 32.69,
+      "learning_rate": 4.591417910447761e-05,
+      "loss": 0.0001,
+      "step": 35040
+    },
+    {
+      "epoch": 32.69,
+      "learning_rate": 4.591371268656717e-05,
+      "loss": 0.0002,
+      "step": 35044
+    },
+    {
+      "epoch": 32.69,
+      "learning_rate": 4.591324626865672e-05,
+      "loss": 0.0001,
+      "step": 35048
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 4.5912779850746266e-05,
+      "loss": 0.0016,
+      "step": 35052
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 4.591231343283582e-05,
+      "loss": 0.0025,
+      "step": 35056
+    },
+    {
+      "epoch": 32.71,
+      "learning_rate": 4.5911847014925376e-05,
+      "loss": 0.0,
+      "step": 35060
+    },
+    {
+      "epoch": 32.71,
+      "learning_rate": 4.591138059701493e-05,
+      "loss": 0.0001,
+      "step": 35064
+    },
+    {
+      "epoch": 32.71,
+      "learning_rate": 4.591091417910448e-05,
+      "loss": 0.0001,
+      "step": 35068
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 4.5910447761194034e-05,
+      "loss": 0.0002,
+      "step": 35072
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 4.590998134328359e-05,
+      "loss": 0.0001,
+      "step": 35076
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 4.590951492537314e-05,
+      "loss": 0.0014,
+      "step": 35080
+    },
+    {
+      "epoch": 32.73,
+      "learning_rate": 4.5909048507462685e-05,
+      "loss": 0.0007,
+      "step": 35084
+    },
+    {
+      "epoch": 32.73,
+      "learning_rate": 4.590858208955224e-05,
+      "loss": 0.0001,
+      "step": 35088
+    },
+    {
+      "epoch": 32.73,
+      "learning_rate": 4.5908115671641795e-05,
+      "loss": 0.0001,
+      "step": 35092
+    },
+    {
+      "epoch": 32.74,
+      "learning_rate": 4.590764925373134e-05,
+      "loss": 0.0001,
+      "step": 35096
+    },
+    {
+      "epoch": 32.74,
+      "learning_rate": 4.59071828358209e-05,
+      "loss": 0.0,
+      "step": 35100
+    },
+    {
+      "epoch": 32.75,
+      "learning_rate": 4.590671641791045e-05,
+      "loss": 0.0,
+      "step": 35104
+    },
+    {
+      "epoch": 32.75,
+      "learning_rate": 4.590625e-05,
+      "loss": 0.0,
+      "step": 35108
+    },
+    {
+      "epoch": 32.75,
+      "learning_rate": 4.5905783582089556e-05,
+      "loss": 0.0004,
+      "step": 35112
+    },
+    {
+      "epoch": 32.76,
+      "learning_rate": 4.5905317164179104e-05,
+      "loss": 0.0,
+      "step": 35116
+    },
+    {
+      "epoch": 32.76,
+      "learning_rate": 4.590485074626866e-05,
+      "loss": 0.0,
+      "step": 35120
+    },
+    {
+      "epoch": 32.76,
+      "learning_rate": 4.5904384328358214e-05,
+      "loss": 0.0003,
+      "step": 35124
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 4.590391791044776e-05,
+      "loss": 0.0003,
+      "step": 35128
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 4.590345149253732e-05,
+      "loss": 0.0002,
+      "step": 35132
+    },
+    {
+      "epoch": 32.78,
+      "learning_rate": 4.590298507462687e-05,
+      "loss": 0.0001,
+      "step": 35136
+    },
+    {
+      "epoch": 32.78,
+      "learning_rate": 4.590251865671642e-05,
+      "loss": 0.0014,
+      "step": 35140
+    },
+    {
+      "epoch": 32.78,
+      "learning_rate": 4.590205223880597e-05,
+      "loss": 0.0,
+      "step": 35144
+    },
+    {
+      "epoch": 32.79,
+      "learning_rate": 4.590158582089552e-05,
+      "loss": 0.0,
+      "step": 35148
+    },
+    {
+      "epoch": 32.79,
+      "learning_rate": 4.590111940298508e-05,
+      "loss": 0.0005,
+      "step": 35152
+    },
+    {
+      "epoch": 32.79,
+      "learning_rate": 4.5900652985074626e-05,
+      "loss": 0.0022,
+      "step": 35156
+    },
+    {
+      "epoch": 32.8,
+      "learning_rate": 4.590018656716418e-05,
+      "loss": 0.0002,
+      "step": 35160
+    },
+    {
+      "epoch": 32.8,
+      "learning_rate": 4.5899720149253736e-05,
+      "loss": 0.0002,
+      "step": 35164
+    },
+    {
+      "epoch": 32.81,
+      "learning_rate": 4.5899253731343284e-05,
+      "loss": 0.0079,
+      "step": 35168
+    },
+    {
+      "epoch": 32.81,
+      "learning_rate": 4.589878731343284e-05,
+      "loss": 0.0002,
+      "step": 35172
+    },
+    {
+      "epoch": 32.81,
+      "learning_rate": 4.589832089552239e-05,
+      "loss": 0.0001,
+      "step": 35176
+    },
+    {
+      "epoch": 32.82,
+      "learning_rate": 4.589785447761194e-05,
+      "loss": 0.0002,
+      "step": 35180
+    },
+    {
+      "epoch": 32.82,
+      "learning_rate": 4.58973880597015e-05,
+      "loss": 0.0001,
+      "step": 35184
+    },
+    {
+      "epoch": 32.82,
+      "learning_rate": 4.5896921641791045e-05,
+      "loss": 0.0001,
+      "step": 35188
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 4.58964552238806e-05,
+      "loss": 0.0,
+      "step": 35192
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 4.5895988805970155e-05,
+      "loss": 0.0001,
+      "step": 35196
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 4.58955223880597e-05,
+      "loss": 0.0,
+      "step": 35200
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 4.589505597014925e-05,
+      "loss": 0.0008,
+      "step": 35204
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 4.5894589552238806e-05,
+      "loss": 0.0003,
+      "step": 35208
+    },
+    {
+      "epoch": 32.85,
+      "learning_rate": 4.589412313432836e-05,
+      "loss": 0.0003,
+      "step": 35212
+    },
+    {
+      "epoch": 32.85,
+      "learning_rate": 4.5893656716417916e-05,
+      "loss": 0.0001,
+      "step": 35216
+    },
+    {
+      "epoch": 32.85,
+      "learning_rate": 4.5893190298507464e-05,
+      "loss": 0.0,
+      "step": 35220
+    },
+    {
+      "epoch": 32.86,
+      "learning_rate": 4.589272388059702e-05,
+      "loss": 0.0,
+      "step": 35224
+    },
+    {
+      "epoch": 32.86,
+      "learning_rate": 4.5892257462686574e-05,
+      "loss": 0.0002,
+      "step": 35228
+    },
+    {
+      "epoch": 32.87,
+      "learning_rate": 4.589179104477612e-05,
+      "loss": 0.0002,
+      "step": 35232
+    },
+    {
+      "epoch": 32.87,
+      "learning_rate": 4.589132462686567e-05,
+      "loss": 0.0,
+      "step": 35236
+    },
+    {
+      "epoch": 32.87,
+      "learning_rate": 4.589085820895523e-05,
+      "loss": 0.0014,
+      "step": 35240
+    },
+    {
+      "epoch": 32.88,
+      "learning_rate": 4.589039179104478e-05,
+      "loss": 0.0002,
+      "step": 35244
+    },
+    {
+      "epoch": 32.88,
+      "learning_rate": 4.588992537313433e-05,
+      "loss": 0.0006,
+      "step": 35248
+    },
+    {
+      "epoch": 32.88,
+      "learning_rate": 4.588945895522388e-05,
+      "loss": 0.0001,
+      "step": 35252
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 4.588899253731344e-05,
+      "loss": 0.0019,
+      "step": 35256
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 4.5888526119402986e-05,
+      "loss": 0.0007,
+      "step": 35260
+    },
+    {
+      "epoch": 32.9,
+      "learning_rate": 4.588805970149254e-05,
+      "loss": 0.0,
+      "step": 35264
+    },
+    {
+      "epoch": 32.9,
+      "learning_rate": 4.588759328358209e-05,
+      "loss": 0.0004,
+      "step": 35268
+    },
+    {
+      "epoch": 32.9,
+      "learning_rate": 4.5887126865671644e-05,
+      "loss": 0.0,
+      "step": 35272
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 4.58866604477612e-05,
+      "loss": 0.0009,
+      "step": 35276
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 4.588619402985075e-05,
+      "loss": 0.0001,
+      "step": 35280
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 4.58857276119403e-05,
+      "loss": 0.0014,
+      "step": 35284
+    },
+    {
+      "epoch": 32.92,
+      "learning_rate": 4.588526119402986e-05,
+      "loss": 0.0025,
+      "step": 35288
+    },
+    {
+      "epoch": 32.92,
+      "learning_rate": 4.5884794776119405e-05,
+      "loss": 0.0001,
+      "step": 35292
+    },
+    {
+      "epoch": 32.93,
+      "learning_rate": 4.588432835820895e-05,
+      "loss": 0.0064,
+      "step": 35296
+    },
+    {
+      "epoch": 32.93,
+      "learning_rate": 4.5883861940298515e-05,
+      "loss": 0.0001,
+      "step": 35300
+    },
+    {
+      "epoch": 32.93,
+      "learning_rate": 4.588339552238806e-05,
+      "loss": 0.0006,
+      "step": 35304
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 4.588292910447761e-05,
+      "loss": 0.0004,
+      "step": 35308
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 4.5882462686567166e-05,
+      "loss": 0.0076,
+      "step": 35312
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 4.588199626865672e-05,
+      "loss": 0.0001,
+      "step": 35316
+    },
+    {
+      "epoch": 32.95,
+      "learning_rate": 4.588152985074627e-05,
+      "loss": 0.0001,
+      "step": 35320
+    },
+    {
+      "epoch": 32.95,
+      "learning_rate": 4.5881063432835824e-05,
+      "loss": 0.0001,
+      "step": 35324
+    },
+    {
+      "epoch": 32.96,
+      "learning_rate": 4.588059701492537e-05,
+      "loss": 0.0,
+      "step": 35328
+    },
+    {
+      "epoch": 32.96,
+      "learning_rate": 4.588013059701493e-05,
+      "loss": 0.001,
+      "step": 35332
+    },
+    {
+      "epoch": 32.96,
+      "learning_rate": 4.587966417910448e-05,
+      "loss": 0.0001,
+      "step": 35336
+    },
+    {
+      "epoch": 32.97,
+      "learning_rate": 4.587919776119403e-05,
+      "loss": 0.0001,
+      "step": 35340
+    },
+    {
+      "epoch": 32.97,
+      "learning_rate": 4.5878731343283585e-05,
+      "loss": 0.0001,
+      "step": 35344
+    },
+    {
+      "epoch": 32.97,
+      "learning_rate": 4.587826492537314e-05,
+      "loss": 0.0001,
+      "step": 35348
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 4.587779850746269e-05,
+      "loss": 0.0,
+      "step": 35352
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 4.5877332089552236e-05,
+      "loss": 0.0001,
+      "step": 35356
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 4.58768656716418e-05,
+      "loss": 0.0001,
+      "step": 35360
+    },
+    {
+      "epoch": 32.99,
+      "learning_rate": 4.5876399253731346e-05,
+      "loss": 0.0,
+      "step": 35364
+    },
+    {
+      "epoch": 32.99,
+      "learning_rate": 4.5875932835820894e-05,
+      "loss": 0.0003,
+      "step": 35368
+    },
+    {
+      "epoch": 33.0,
+      "learning_rate": 4.587546641791045e-05,
+      "loss": 0.0008,
+      "step": 35372
+    },
+    {
+      "epoch": 33.0,
+      "learning_rate": 4.5875000000000004e-05,
+      "loss": 0.0001,
+      "step": 35376
+    },
+    {
+      "epoch": 33.0,
+      "learning_rate": 4.587453358208956e-05,
+      "loss": 0.0008,
+      "step": 35380
+    },
+    {
+      "epoch": 33.01,
+      "learning_rate": 4.587406716417911e-05,
+      "loss": 0.0006,
+      "step": 35384
+    },
+    {
+      "epoch": 33.01,
+      "learning_rate": 4.5873600746268655e-05,
+      "loss": 0.0,
+      "step": 35388
+    },
+    {
+      "epoch": 33.01,
+      "learning_rate": 4.587313432835822e-05,
+      "loss": 0.0007,
+      "step": 35392
+    },
+    {
+      "epoch": 33.02,
+      "learning_rate": 4.5872667910447765e-05,
+      "loss": 0.0001,
+      "step": 35396
+    },
+    {
+      "epoch": 33.02,
+      "learning_rate": 4.587220149253731e-05,
+      "loss": 0.0,
+      "step": 35400
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 4.587173507462687e-05,
+      "loss": 0.0001,
+      "step": 35404
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 4.587126865671642e-05,
+      "loss": 0.001,
+      "step": 35408
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 4.587080223880597e-05,
+      "loss": 0.0012,
+      "step": 35412
+    },
+    {
+      "epoch": 33.04,
+      "learning_rate": 4.5870335820895526e-05,
+      "loss": 0.0008,
+      "step": 35416
+    },
+    {
+      "epoch": 33.04,
+      "learning_rate": 4.586986940298508e-05,
+      "loss": 0.0001,
+      "step": 35420
+    },
+    {
+      "epoch": 33.04,
+      "learning_rate": 4.586940298507463e-05,
+      "loss": 0.0,
+      "step": 35424
+    },
+    {
+      "epoch": 33.05,
+      "learning_rate": 4.5868936567164184e-05,
+      "loss": 0.0005,
+      "step": 35428
+    },
+    {
+      "epoch": 33.05,
+      "learning_rate": 4.586847014925373e-05,
+      "loss": 0.0,
+      "step": 35432
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 4.586800373134329e-05,
+      "loss": 0.0001,
+      "step": 35436
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 4.586753731343284e-05,
+      "loss": 0.0009,
+      "step": 35440
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 4.586707089552239e-05,
+      "loss": 0.0002,
+      "step": 35444
+    },
+    {
+      "epoch": 33.07,
+      "learning_rate": 4.586660447761194e-05,
+      "loss": 0.0005,
+      "step": 35448
+    },
+    {
+      "epoch": 33.07,
+      "learning_rate": 4.58661380597015e-05,
+      "loss": 0.0005,
+      "step": 35452
+    },
+    {
+      "epoch": 33.07,
+      "learning_rate": 4.586567164179105e-05,
+      "loss": 0.0042,
+      "step": 35456
+    },
+    {
+      "epoch": 33.08,
+      "learning_rate": 4.5865205223880596e-05,
+      "loss": 0.0,
+      "step": 35460
+    },
+    {
+      "epoch": 33.08,
+      "learning_rate": 4.586473880597015e-05,
+      "loss": 0.0001,
+      "step": 35464
+    },
+    {
+      "epoch": 33.09,
+      "learning_rate": 4.5864272388059706e-05,
+      "loss": 0.0064,
+      "step": 35468
+    },
+    {
+      "epoch": 33.09,
+      "learning_rate": 4.5863805970149254e-05,
+      "loss": 0.0004,
+      "step": 35472
+    },
+    {
+      "epoch": 33.09,
+      "learning_rate": 4.586333955223881e-05,
+      "loss": 0.0001,
+      "step": 35476
+    },
+    {
+      "epoch": 33.1,
+      "learning_rate": 4.586287313432836e-05,
+      "loss": 0.0,
+      "step": 35480
+    },
+    {
+      "epoch": 33.1,
+      "learning_rate": 4.586240671641791e-05,
+      "loss": 0.0005,
+      "step": 35484
+    },
+    {
+      "epoch": 33.1,
+      "learning_rate": 4.586194029850747e-05,
+      "loss": 0.0002,
+      "step": 35488
+    },
+    {
+      "epoch": 33.11,
+      "learning_rate": 4.5861473880597015e-05,
+      "loss": 0.0002,
+      "step": 35492
+    },
+    {
+      "epoch": 33.11,
+      "learning_rate": 4.586100746268657e-05,
+      "loss": 0.0002,
+      "step": 35496
+    },
+    {
+      "epoch": 33.12,
+      "learning_rate": 4.5860541044776125e-05,
+      "loss": 0.0068,
+      "step": 35500
+    },
+    {
+      "epoch": 33.12,
+      "eval_exact_match": 0.7379110251450677,
+      "eval_exec": 0.758220502901354,
+      "eval_loss": 0.4318758547306061,
+      "eval_runtime": 1370.5932,
+      "eval_samples_per_second": 0.754,
+      "step": 35500
+    },
+    {
+      "epoch": 33.12,
+      "learning_rate": 4.586007462686567e-05,
+      "loss": 0.0,
+      "step": 35504
+    },
+    {
+      "epoch": 33.12,
+      "learning_rate": 4.585960820895522e-05,
+      "loss": 0.0004,
+      "step": 35508
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 4.585914179104478e-05,
+      "loss": 0.0012,
+      "step": 35512
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 4.585867537313433e-05,
+      "loss": 0.0001,
+      "step": 35516
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 4.585820895522388e-05,
+      "loss": 0.0006,
+      "step": 35520
+    },
+    {
+      "epoch": 33.14,
+      "learning_rate": 4.5857742537313434e-05,
+      "loss": 0.0001,
+      "step": 35524
+    },
+    {
+      "epoch": 33.14,
+      "learning_rate": 4.585727611940299e-05,
+      "loss": 0.0,
+      "step": 35528
+    },
+    {
+      "epoch": 33.15,
+      "learning_rate": 4.585680970149254e-05,
+      "loss": 0.0002,
+      "step": 35532
+    },
+    {
+      "epoch": 33.15,
+      "learning_rate": 4.585634328358209e-05,
+      "loss": 0.0001,
+      "step": 35536
+    },
+    {
+      "epoch": 33.15,
+      "learning_rate": 4.585587686567164e-05,
+      "loss": 0.0,
+      "step": 35540
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 4.58554104477612e-05,
+      "loss": 0.0011,
+      "step": 35544
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 4.585494402985075e-05,
+      "loss": 0.0001,
+      "step": 35548
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 4.58544776119403e-05,
+      "loss": 0.0,
+      "step": 35552
+    },
+    {
+      "epoch": 33.17,
+      "learning_rate": 4.585401119402985e-05,
+      "loss": 0.0001,
+      "step": 35556
+    },
+    {
+      "epoch": 33.17,
+      "learning_rate": 4.585354477611941e-05,
+      "loss": 0.0,
+      "step": 35560
+    },
+    {
+      "epoch": 33.18,
+      "learning_rate": 4.5853078358208956e-05,
+      "loss": 0.0,
+      "step": 35564
+    },
+    {
+      "epoch": 33.18,
+      "learning_rate": 4.585261194029851e-05,
+      "loss": 0.0001,
+      "step": 35568
+    },
+    {
+      "epoch": 33.18,
+      "learning_rate": 4.5852145522388065e-05,
+      "loss": 0.0003,
+      "step": 35572
+    },
+    {
+      "epoch": 33.19,
+      "learning_rate": 4.5851679104477614e-05,
+      "loss": 0.0001,
+      "step": 35576
+    },
+    {
+      "epoch": 33.19,
+      "learning_rate": 4.585121268656717e-05,
+      "loss": 0.0,
+      "step": 35580
+    },
+    {
+      "epoch": 33.19,
+      "learning_rate": 4.585074626865672e-05,
+      "loss": 0.0004,
+      "step": 35584
+    },
+    {
+      "epoch": 33.2,
+      "learning_rate": 4.585027985074627e-05,
+      "loss": 0.0002,
+      "step": 35588
+    },
+    {
+      "epoch": 33.2,
+      "learning_rate": 4.5849813432835826e-05,
+      "loss": 0.0006,
+      "step": 35592
+    },
+    {
+      "epoch": 33.21,
+      "learning_rate": 4.5849347014925375e-05,
+      "loss": 0.0,
+      "step": 35596
+    },
+    {
+      "epoch": 33.21,
+      "learning_rate": 4.584888059701492e-05,
+      "loss": 0.0001,
+      "step": 35600
+    },
+    {
+      "epoch": 33.21,
+      "learning_rate": 4.5848414179104484e-05,
+      "loss": 0.0003,
+      "step": 35604
+    },
+    {
+      "epoch": 33.22,
+      "learning_rate": 4.584794776119403e-05,
+      "loss": 0.0001,
+      "step": 35608
+    },
+    {
+      "epoch": 33.22,
+      "learning_rate": 4.584748134328358e-05,
+      "loss": 0.001,
+      "step": 35612
+    },
+    {
+      "epoch": 33.22,
+      "learning_rate": 4.5847014925373136e-05,
+      "loss": 0.0,
+      "step": 35616
+    },
+    {
+      "epoch": 33.23,
+      "learning_rate": 4.584654850746269e-05,
+      "loss": 0.0016,
+      "step": 35620
+    },
+    {
+      "epoch": 33.23,
+      "learning_rate": 4.584608208955224e-05,
+      "loss": 0.0,
+      "step": 35624
+    },
+    {
+      "epoch": 33.24,
+      "learning_rate": 4.5845615671641794e-05,
+      "loss": 0.0,
+      "step": 35628
+    },
+    {
+      "epoch": 33.24,
+      "learning_rate": 4.584514925373135e-05,
+      "loss": 0.0001,
+      "step": 35632
+    },
+    {
+      "epoch": 33.24,
+      "learning_rate": 4.5844682835820897e-05,
+      "loss": 0.0,
+      "step": 35636
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 4.584421641791045e-05,
+      "loss": 0.0,
+      "step": 35640
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 4.584375e-05,
+      "loss": 0.0,
+      "step": 35644
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 4.5843283582089555e-05,
+      "loss": 0.0017,
+      "step": 35648
+    },
+    {
+      "epoch": 33.26,
+      "learning_rate": 4.584281716417911e-05,
+      "loss": 0.0,
+      "step": 35652
+    },
+    {
+      "epoch": 33.26,
+      "learning_rate": 4.584235074626866e-05,
+      "loss": 0.0001,
+      "step": 35656
+    },
+    {
+      "epoch": 33.26,
+      "learning_rate": 4.5841884328358206e-05,
+      "loss": 0.0,
+      "step": 35660
+    },
+    {
+      "epoch": 33.27,
+      "learning_rate": 4.584141791044777e-05,
+      "loss": 0.0,
+      "step": 35664
+    },
+    {
+      "epoch": 33.27,
+      "learning_rate": 4.5840951492537316e-05,
+      "loss": 0.0,
+      "step": 35668
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 4.5840485074626864e-05,
+      "loss": 0.0,
+      "step": 35672
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 4.584001865671642e-05,
+      "loss": 0.0001,
+      "step": 35676
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 4.5839552238805973e-05,
+      "loss": 0.0001,
+      "step": 35680
+    },
+    {
+      "epoch": 33.29,
+      "learning_rate": 4.583908582089552e-05,
+      "loss": 0.0,
+      "step": 35684
+    },
+    {
+      "epoch": 33.29,
+      "learning_rate": 4.5838619402985077e-05,
+      "loss": 0.0,
+      "step": 35688
+    },
+    {
+      "epoch": 33.29,
+      "learning_rate": 4.583815298507463e-05,
+      "loss": 0.0027,
+      "step": 35692
+    },
+    {
+      "epoch": 33.3,
+      "learning_rate": 4.583768656716418e-05,
+      "loss": 0.0003,
+      "step": 35696
+    },
+    {
+      "epoch": 33.3,
+      "learning_rate": 4.5837220149253734e-05,
+      "loss": 0.0032,
+      "step": 35700
+    },
+    {
+      "epoch": 33.31,
+      "learning_rate": 4.583675373134328e-05,
+      "loss": 0.0007,
+      "step": 35704
+    },
+    {
+      "epoch": 33.31,
+      "learning_rate": 4.583628731343284e-05,
+      "loss": 0.0005,
+      "step": 35708
+    },
+    {
+      "epoch": 33.31,
+      "learning_rate": 4.583582089552239e-05,
+      "loss": 0.0032,
+      "step": 35712
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 4.583535447761194e-05,
+      "loss": 0.0002,
+      "step": 35716
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 4.5834888059701495e-05,
+      "loss": 0.0027,
+      "step": 35720
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 4.583442164179105e-05,
+      "loss": 0.0004,
+      "step": 35724
+    },
+    {
+      "epoch": 33.33,
+      "learning_rate": 4.58339552238806e-05,
+      "loss": 0.0,
+      "step": 35728
+    },
+    {
+      "epoch": 33.33,
+      "learning_rate": 4.583348880597015e-05,
+      "loss": 0.0,
+      "step": 35732
+    },
+    {
+      "epoch": 33.34,
+      "learning_rate": 4.58330223880597e-05,
+      "loss": 0.0001,
+      "step": 35736
+    },
+    {
+      "epoch": 33.34,
+      "learning_rate": 4.5832555970149256e-05,
+      "loss": 0.0,
+      "step": 35740
+    },
+    {
+      "epoch": 33.34,
+      "learning_rate": 4.583208955223881e-05,
+      "loss": 0.0023,
+      "step": 35744
+    },
+    {
+      "epoch": 33.35,
+      "learning_rate": 4.583162313432836e-05,
+      "loss": 0.0001,
+      "step": 35748
+    },
+    {
+      "epoch": 33.35,
+      "learning_rate": 4.5831156716417914e-05,
+      "loss": 0.0145,
+      "step": 35752
+    },
+    {
+      "epoch": 33.35,
+      "learning_rate": 4.583069029850747e-05,
+      "loss": 0.0028,
+      "step": 35756
+    },
+    {
+      "epoch": 33.36,
+      "learning_rate": 4.583022388059702e-05,
+      "loss": 0.0007,
+      "step": 35760
+    },
+    {
+      "epoch": 33.36,
+      "learning_rate": 4.5829757462686566e-05,
+      "loss": 0.0,
+      "step": 35764
+    },
+    {
+      "epoch": 33.37,
+      "learning_rate": 4.582929104477612e-05,
+      "loss": 0.0007,
+      "step": 35768
+    },
+    {
+      "epoch": 33.37,
+      "learning_rate": 4.5828824626865675e-05,
+      "loss": 0.0005,
+      "step": 35772
+    },
+    {
+      "epoch": 33.37,
+      "learning_rate": 4.5828358208955223e-05,
+      "loss": 0.0,
+      "step": 35776
+    },
+    {
+      "epoch": 33.38,
+      "learning_rate": 4.582789179104478e-05,
+      "loss": 0.001,
+      "step": 35780
+    },
+    {
+      "epoch": 33.38,
+      "learning_rate": 4.582742537313433e-05,
+      "loss": 0.0002,
+      "step": 35784
+    },
+    {
+      "epoch": 33.38,
+      "learning_rate": 4.582695895522388e-05,
+      "loss": 0.0001,
+      "step": 35788
+    },
+    {
+      "epoch": 33.39,
+      "learning_rate": 4.5826492537313436e-05,
+      "loss": 0.0011,
+      "step": 35792
+    },
+    {
+      "epoch": 33.39,
+      "learning_rate": 4.5826026119402984e-05,
+      "loss": 0.0091,
+      "step": 35796
+    },
+    {
+      "epoch": 33.4,
+      "learning_rate": 4.582555970149254e-05,
+      "loss": 0.0017,
+      "step": 35800
+    },
+    {
+      "epoch": 33.4,
+      "learning_rate": 4.5825093283582094e-05,
+      "loss": 0.0003,
+      "step": 35804
+    },
+    {
+      "epoch": 33.4,
+      "learning_rate": 4.582462686567164e-05,
+      "loss": 0.002,
+      "step": 35808
+    },
+    {
+      "epoch": 33.41,
+      "learning_rate": 4.58241604477612e-05,
+      "loss": 0.0002,
+      "step": 35812
+    },
+    {
+      "epoch": 33.41,
+      "learning_rate": 4.582369402985075e-05,
+      "loss": 0.0001,
+      "step": 35816
+    },
+    {
+      "epoch": 33.41,
+      "learning_rate": 4.58232276119403e-05,
+      "loss": 0.0098,
+      "step": 35820
+    },
+    {
+      "epoch": 33.42,
+      "learning_rate": 4.582276119402985e-05,
+      "loss": 0.0001,
+      "step": 35824
+    },
+    {
+      "epoch": 33.42,
+      "learning_rate": 4.5822294776119403e-05,
+      "loss": 0.0004,
+      "step": 35828
+    },
+    {
+      "epoch": 33.43,
+      "learning_rate": 4.582182835820896e-05,
+      "loss": 0.0001,
+      "step": 35832
+    },
+    {
+      "epoch": 33.43,
+      "learning_rate": 4.5821361940298506e-05,
+      "loss": 0.0001,
+      "step": 35836
+    },
+    {
+      "epoch": 33.43,
+      "learning_rate": 4.582089552238806e-05,
+      "loss": 0.0001,
+      "step": 35840
+    },
+    {
+      "epoch": 33.44,
+      "learning_rate": 4.5820429104477616e-05,
+      "loss": 0.0,
+      "step": 35844
+    },
+    {
+      "epoch": 33.44,
+      "learning_rate": 4.5819962686567164e-05,
+      "loss": 0.0018,
+      "step": 35848
+    },
+    {
+      "epoch": 33.44,
+      "learning_rate": 4.581949626865672e-05,
+      "loss": 0.0005,
+      "step": 35852
+    },
+    {
+      "epoch": 33.45,
+      "learning_rate": 4.581902985074627e-05,
+      "loss": 0.0002,
+      "step": 35856
+    },
+    {
+      "epoch": 33.45,
+      "learning_rate": 4.581856343283582e-05,
+      "loss": 0.0006,
+      "step": 35860
+    },
+    {
+      "epoch": 33.46,
+      "learning_rate": 4.581809701492538e-05,
+      "loss": 0.0001,
+      "step": 35864
+    },
+    {
+      "epoch": 33.46,
+      "learning_rate": 4.5817630597014925e-05,
+      "loss": 0.0001,
+      "step": 35868
+    },
+    {
+      "epoch": 33.46,
+      "learning_rate": 4.581716417910448e-05,
+      "loss": 0.0001,
+      "step": 35872
+    },
+    {
+      "epoch": 33.47,
+      "learning_rate": 4.5816697761194035e-05,
+      "loss": 0.0001,
+      "step": 35876
+    },
+    {
+      "epoch": 33.47,
+      "learning_rate": 4.581623134328358e-05,
+      "loss": 0.0007,
+      "step": 35880
+    },
+    {
+      "epoch": 33.47,
+      "learning_rate": 4.581576492537314e-05,
+      "loss": 0.0047,
+      "step": 35884
+    },
+    {
+      "epoch": 33.48,
+      "learning_rate": 4.5815298507462686e-05,
+      "loss": 0.0004,
+      "step": 35888
+    },
+    {
+      "epoch": 33.48,
+      "learning_rate": 4.581483208955224e-05,
+      "loss": 0.0001,
+      "step": 35892
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 4.5814365671641796e-05,
+      "loss": 0.0003,
+      "step": 35896
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 4.5813899253731344e-05,
+      "loss": 0.0047,
+      "step": 35900
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 4.58134328358209e-05,
+      "loss": 0.0012,
+      "step": 35904
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 4.5812966417910454e-05,
+      "loss": 0.0042,
+      "step": 35908
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 4.58125e-05,
+      "loss": 0.0,
+      "step": 35912
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 4.581203358208955e-05,
+      "loss": 0.0256,
+      "step": 35916
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 4.581156716417911e-05,
+      "loss": 0.0033,
+      "step": 35920
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 4.581110074626866e-05,
+      "loss": 0.0001,
+      "step": 35924
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 4.581063432835821e-05,
+      "loss": 0.0015,
+      "step": 35928
+    },
+    {
+      "epoch": 33.52,
+      "learning_rate": 4.581016791044776e-05,
+      "loss": 0.0003,
+      "step": 35932
+    },
+    {
+      "epoch": 33.52,
+      "learning_rate": 4.580970149253732e-05,
+      "loss": 0.0001,
+      "step": 35936
+    },
+    {
+      "epoch": 33.53,
+      "learning_rate": 4.5809235074626866e-05,
+      "loss": 0.0003,
+      "step": 35940
+    },
+    {
+      "epoch": 33.53,
+      "learning_rate": 4.580876865671642e-05,
+      "loss": 0.0,
+      "step": 35944
+    },
+    {
+      "epoch": 33.53,
+      "learning_rate": 4.580830223880597e-05,
+      "loss": 0.0015,
+      "step": 35948
+    },
+    {
+      "epoch": 33.54,
+      "learning_rate": 4.5807835820895524e-05,
+      "loss": 0.0002,
+      "step": 35952
+    },
+    {
+      "epoch": 33.54,
+      "learning_rate": 4.580736940298508e-05,
+      "loss": 0.0005,
+      "step": 35956
+    },
+    {
+      "epoch": 33.54,
+      "learning_rate": 4.580690298507463e-05,
+      "loss": 0.0001,
+      "step": 35960
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 4.580643656716418e-05,
+      "loss": 0.0033,
+      "step": 35964
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 4.580597014925374e-05,
+      "loss": 0.0027,
+      "step": 35968
+    },
+    {
+      "epoch": 33.56,
+      "learning_rate": 4.5805503731343285e-05,
+      "loss": 0.0024,
+      "step": 35972
+    },
+    {
+      "epoch": 33.56,
+      "learning_rate": 4.580503731343283e-05,
+      "loss": 0.0041,
+      "step": 35976
+    },
+    {
+      "epoch": 33.56,
+      "learning_rate": 4.5804570895522395e-05,
+      "loss": 0.0017,
+      "step": 35980
+    },
+    {
+      "epoch": 33.57,
+      "learning_rate": 4.580410447761194e-05,
+      "loss": 0.0017,
+      "step": 35984
+    },
+    {
+      "epoch": 33.57,
+      "learning_rate": 4.580363805970149e-05,
+      "loss": 0.0001,
+      "step": 35988
+    },
+    {
+      "epoch": 33.57,
+      "learning_rate": 4.5803171641791046e-05,
+      "loss": 0.0015,
+      "step": 35992
+    },
+    {
+      "epoch": 33.58,
+      "learning_rate": 4.58027052238806e-05,
+      "loss": 0.0001,
+      "step": 35996
+    },
+    {
+      "epoch": 33.58,
+      "learning_rate": 4.580223880597015e-05,
+      "loss": 0.0006,
+      "step": 36000
+    },
+    {
+      "epoch": 33.58,
+      "eval_exact_match": 0.723404255319149,
+      "eval_exec": 0.7630560928433269,
+      "eval_loss": 0.4088174104690552,
+      "eval_runtime": 1192.806,
+      "eval_samples_per_second": 0.867,
+      "step": 36000
+    },
+    {
+      "epoch": 33.59,
+      "learning_rate": 4.5801772388059704e-05,
+      "loss": 0.0,
+      "step": 36004
+    },
+    {
+      "epoch": 33.59,
+      "learning_rate": 4.580130597014925e-05,
+      "loss": 0.0005,
+      "step": 36008
+    },
+    {
+      "epoch": 33.59,
+      "learning_rate": 4.580083955223881e-05,
+      "loss": 0.0001,
+      "step": 36012
+    },
+    {
+      "epoch": 33.6,
+      "learning_rate": 4.580037313432836e-05,
+      "loss": 0.0003,
+      "step": 36016
+    },
+    {
+      "epoch": 33.6,
+      "learning_rate": 4.579990671641791e-05,
+      "loss": 0.0,
+      "step": 36020
+    },
+    {
+      "epoch": 33.6,
+      "learning_rate": 4.5799440298507465e-05,
+      "loss": 0.0,
+      "step": 36024
+    },
+    {
+      "epoch": 33.61,
+      "learning_rate": 4.579897388059702e-05,
+      "loss": 0.0002,
+      "step": 36028
+    },
+    {
+      "epoch": 33.61,
+      "learning_rate": 4.579850746268657e-05,
+      "loss": 0.0064,
+      "step": 36032
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 4.579804104477612e-05,
+      "loss": 0.0014,
+      "step": 36036
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 4.579757462686568e-05,
+      "loss": 0.0011,
+      "step": 36040
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 4.5797108208955226e-05,
+      "loss": 0.001,
+      "step": 36044
+    },
+    {
+      "epoch": 33.63,
+      "learning_rate": 4.579664179104478e-05,
+      "loss": 0.0271,
+      "step": 36048
+    },
+    {
+      "epoch": 33.63,
+      "learning_rate": 4.579617537313433e-05,
+      "loss": 0.0015,
+      "step": 36052
+    },
+    {
+      "epoch": 33.63,
+      "learning_rate": 4.5795708955223884e-05,
+      "loss": 0.0014,
+      "step": 36056
+    },
+    {
+      "epoch": 33.64,
+      "learning_rate": 4.579524253731344e-05,
+      "loss": 0.0008,
+      "step": 36060
+    },
+    {
+      "epoch": 33.64,
+      "learning_rate": 4.579477611940299e-05,
+      "loss": 0.0,
+      "step": 36064
+    },
+    {
+      "epoch": 33.65,
+      "learning_rate": 4.5794309701492535e-05,
+      "loss": 0.0001,
+      "step": 36068
+    },
+    {
+      "epoch": 33.65,
+      "learning_rate": 4.57938432835821e-05,
+      "loss": 0.0007,
+      "step": 36072
+    },
+    {
+      "epoch": 33.65,
+      "learning_rate": 4.5793376865671645e-05,
+      "loss": 0.0004,
+      "step": 36076
+    },
+    {
+      "epoch": 33.66,
+      "learning_rate": 4.579291044776119e-05,
+      "loss": 0.0001,
+      "step": 36080
+    },
+    {
+      "epoch": 33.66,
+      "learning_rate": 4.579244402985075e-05,
+      "loss": 0.0001,
+      "step": 36084
+    },
+    {
+      "epoch": 33.66,
+      "learning_rate": 4.57919776119403e-05,
+      "loss": 0.0002,
+      "step": 36088
+    },
+    {
+      "epoch": 33.67,
+      "learning_rate": 4.579151119402985e-05,
+      "loss": 0.0001,
+      "step": 36092
+    },
+    {
+      "epoch": 33.67,
+      "learning_rate": 4.5791044776119406e-05,
+      "loss": 0.0001,
+      "step": 36096
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 4.579057835820896e-05,
+      "loss": 0.0032,
+      "step": 36100
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 4.579011194029851e-05,
+      "loss": 0.0002,
+      "step": 36104
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 4.5789645522388064e-05,
+      "loss": 0.0025,
+      "step": 36108
+    },
+    {
+      "epoch": 33.69,
+      "learning_rate": 4.578917910447761e-05,
+      "loss": 0.0007,
+      "step": 36112
+    },
+    {
+      "epoch": 33.69,
+      "learning_rate": 4.578871268656717e-05,
+      "loss": 0.0001,
+      "step": 36116
+    },
+    {
+      "epoch": 33.69,
+      "learning_rate": 4.578824626865672e-05,
+      "loss": 0.0003,
+      "step": 36120
+    },
+    {
+      "epoch": 33.7,
+      "learning_rate": 4.578777985074627e-05,
+      "loss": 0.001,
+      "step": 36124
+    },
+    {
+      "epoch": 33.7,
+      "learning_rate": 4.578731343283582e-05,
+      "loss": 0.0001,
+      "step": 36128
+    },
+    {
+      "epoch": 33.71,
+      "learning_rate": 4.578684701492538e-05,
+      "loss": 0.0,
+      "step": 36132
+    },
+    {
+      "epoch": 33.71,
+      "learning_rate": 4.578638059701493e-05,
+      "loss": 0.0002,
+      "step": 36136
+    },
+    {
+      "epoch": 33.71,
+      "learning_rate": 4.5785914179104476e-05,
+      "loss": 0.0029,
+      "step": 36140
+    },
+    {
+      "epoch": 33.72,
+      "learning_rate": 4.578544776119403e-05,
+      "loss": 0.0002,
+      "step": 36144
+    },
+    {
+      "epoch": 33.72,
+      "learning_rate": 4.5784981343283586e-05,
+      "loss": 0.0,
+      "step": 36148
+    },
+    {
+      "epoch": 33.72,
+      "learning_rate": 4.5784514925373134e-05,
+      "loss": 0.0,
+      "step": 36152
+    },
+    {
+      "epoch": 33.73,
+      "learning_rate": 4.578404850746269e-05,
+      "loss": 0.0066,
+      "step": 36156
+    },
+    {
+      "epoch": 33.73,
+      "learning_rate": 4.578358208955224e-05,
+      "loss": 0.0017,
+      "step": 36160
+    },
+    {
+      "epoch": 33.73,
+      "learning_rate": 4.578311567164179e-05,
+      "loss": 0.0028,
+      "step": 36164
+    },
+    {
+      "epoch": 33.74,
+      "learning_rate": 4.578264925373135e-05,
+      "loss": 0.0022,
+      "step": 36168
+    },
+    {
+      "epoch": 33.74,
+      "learning_rate": 4.5782182835820895e-05,
+      "loss": 0.0001,
+      "step": 36172
+    },
+    {
+      "epoch": 33.75,
+      "learning_rate": 4.578171641791045e-05,
+      "loss": 0.0001,
+      "step": 36176
+    },
+    {
+      "epoch": 33.75,
+      "learning_rate": 4.5781250000000005e-05,
+      "loss": 0.0001,
+      "step": 36180
+    },
+    {
+      "epoch": 33.75,
+      "learning_rate": 4.578078358208955e-05,
+      "loss": 0.0003,
+      "step": 36184
+    },
+    {
+      "epoch": 33.76,
+      "learning_rate": 4.57803171641791e-05,
+      "loss": 0.0,
+      "step": 36188
+    },
+    {
+      "epoch": 33.76,
+      "learning_rate": 4.577985074626866e-05,
+      "loss": 0.0001,
+      "step": 36192
+    },
+    {
+      "epoch": 33.76,
+      "learning_rate": 4.577938432835821e-05,
+      "loss": 0.0005,
+      "step": 36196
+    },
+    {
+      "epoch": 33.77,
+      "learning_rate": 4.5778917910447766e-05,
+      "loss": 0.0019,
+      "step": 36200
+    },
+    {
+      "epoch": 33.77,
+      "learning_rate": 4.5778451492537314e-05,
+      "loss": 0.0,
+      "step": 36204
+    },
+    {
+      "epoch": 33.78,
+      "learning_rate": 4.577798507462687e-05,
+      "loss": 0.0004,
+      "step": 36208
+    },
+    {
+      "epoch": 33.78,
+      "learning_rate": 4.5777518656716424e-05,
+      "loss": 0.0003,
+      "step": 36212
+    },
+    {
+      "epoch": 33.78,
+      "learning_rate": 4.577705223880597e-05,
+      "loss": 0.0003,
+      "step": 36216
+    },
+    {
+      "epoch": 33.79,
+      "learning_rate": 4.577658582089552e-05,
+      "loss": 0.0013,
+      "step": 36220
+    },
+    {
+      "epoch": 33.79,
+      "learning_rate": 4.577611940298508e-05,
+      "loss": 0.0005,
+      "step": 36224
+    },
+    {
+      "epoch": 33.79,
+      "learning_rate": 4.577565298507463e-05,
+      "loss": 0.0002,
+      "step": 36228
+    },
+    {
+      "epoch": 33.8,
+      "learning_rate": 4.577518656716418e-05,
+      "loss": 0.0006,
+      "step": 36232
+    },
+    {
+      "epoch": 33.8,
+      "learning_rate": 4.577472014925373e-05,
+      "loss": 0.0065,
+      "step": 36236
+    },
+    {
+      "epoch": 33.81,
+      "learning_rate": 4.577425373134329e-05,
+      "loss": 0.0,
+      "step": 36240
+    },
+    {
+      "epoch": 33.81,
+      "learning_rate": 4.5773787313432836e-05,
+      "loss": 0.0014,
+      "step": 36244
+    },
+    {
+      "epoch": 33.81,
+      "learning_rate": 4.577332089552239e-05,
+      "loss": 0.0,
+      "step": 36248
+    },
+    {
+      "epoch": 33.82,
+      "learning_rate": 4.5772854477611946e-05,
+      "loss": 0.0003,
+      "step": 36252
+    },
+    {
+      "epoch": 33.82,
+      "learning_rate": 4.5772388059701494e-05,
+      "loss": 0.0,
+      "step": 36256
+    },
+    {
+      "epoch": 33.82,
+      "learning_rate": 4.577192164179105e-05,
+      "loss": 0.0006,
+      "step": 36260
+    },
+    {
+      "epoch": 33.83,
+      "learning_rate": 4.57714552238806e-05,
+      "loss": 0.0001,
+      "step": 36264
+    },
+    {
+      "epoch": 33.83,
+      "learning_rate": 4.577098880597015e-05,
+      "loss": 0.0007,
+      "step": 36268
+    },
+    {
+      "epoch": 33.84,
+      "learning_rate": 4.577052238805971e-05,
+      "loss": 0.0001,
+      "step": 36272
+    },
+    {
+      "epoch": 33.84,
+      "learning_rate": 4.5770055970149255e-05,
+      "loss": 0.0029,
+      "step": 36276
+    },
+    {
+      "epoch": 33.84,
+      "learning_rate": 4.57695895522388e-05,
+      "loss": 0.0002,
+      "step": 36280
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 4.5769123134328365e-05,
+      "loss": 0.0001,
+      "step": 36284
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 4.576865671641791e-05,
+      "loss": 0.0002,
+      "step": 36288
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 4.576819029850746e-05,
+      "loss": 0.0014,
+      "step": 36292
+    },
+    {
+      "epoch": 33.86,
+      "learning_rate": 4.5767723880597016e-05,
+      "loss": 0.0,
+      "step": 36296
+    },
+    {
+      "epoch": 33.86,
+      "learning_rate": 4.576725746268657e-05,
+      "loss": 0.0003,
+      "step": 36300
+    },
+    {
+      "epoch": 33.87,
+      "learning_rate": 4.576679104477612e-05,
+      "loss": 0.0002,
+      "step": 36304
+    },
+    {
+      "epoch": 33.87,
+      "learning_rate": 4.5766324626865674e-05,
+      "loss": 0.0005,
+      "step": 36308
+    },
+    {
+      "epoch": 33.87,
+      "learning_rate": 4.576585820895523e-05,
+      "loss": 0.0,
+      "step": 36312
+    },
+    {
+      "epoch": 33.88,
+      "learning_rate": 4.576539179104478e-05,
+      "loss": 0.0,
+      "step": 36316
+    },
+    {
+      "epoch": 33.88,
+      "learning_rate": 4.576492537313433e-05,
+      "loss": 0.0026,
+      "step": 36320
+    },
+    {
+      "epoch": 33.88,
+      "learning_rate": 4.576445895522388e-05,
+      "loss": 0.0037,
+      "step": 36324
+    },
+    {
+      "epoch": 33.89,
+      "learning_rate": 4.5763992537313435e-05,
+      "loss": 0.0001,
+      "step": 36328
+    },
+    {
+      "epoch": 33.89,
+      "learning_rate": 4.576352611940299e-05,
+      "loss": 0.0001,
+      "step": 36332
+    },
+    {
+      "epoch": 33.9,
+      "learning_rate": 4.576305970149254e-05,
+      "loss": 0.0008,
+      "step": 36336
+    },
+    {
+      "epoch": 33.9,
+      "learning_rate": 4.5762593283582086e-05,
+      "loss": 0.0,
+      "step": 36340
+    },
+    {
+      "epoch": 33.9,
+      "learning_rate": 4.576212686567165e-05,
+      "loss": 0.0,
+      "step": 36344
+    },
+    {
+      "epoch": 33.91,
+      "learning_rate": 4.5761660447761196e-05,
+      "loss": 0.0,
+      "step": 36348
+    },
+    {
+      "epoch": 33.91,
+      "learning_rate": 4.5761194029850744e-05,
+      "loss": 0.0001,
+      "step": 36352
+    },
+    {
+      "epoch": 33.91,
+      "learning_rate": 4.57607276119403e-05,
+      "loss": 0.0002,
+      "step": 36356
+    },
+    {
+      "epoch": 33.92,
+      "learning_rate": 4.5760261194029854e-05,
+      "loss": 0.0018,
+      "step": 36360
+    },
+    {
+      "epoch": 33.92,
+      "learning_rate": 4.575979477611941e-05,
+      "loss": 0.0018,
+      "step": 36364
+    },
+    {
+      "epoch": 33.93,
+      "learning_rate": 4.575932835820896e-05,
+      "loss": 0.0017,
+      "step": 36368
+    },
+    {
+      "epoch": 33.93,
+      "learning_rate": 4.575886194029851e-05,
+      "loss": 0.0,
+      "step": 36372
+    },
+    {
+      "epoch": 33.93,
+      "learning_rate": 4.5758395522388067e-05,
+      "loss": 0.0001,
+      "step": 36376
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 4.5757929104477615e-05,
+      "loss": 0.0009,
+      "step": 36380
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 4.575746268656716e-05,
+      "loss": 0.0002,
+      "step": 36384
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 4.575699626865672e-05,
+      "loss": 0.0,
+      "step": 36388
+    },
+    {
+      "epoch": 33.95,
+      "learning_rate": 4.575652985074627e-05,
+      "loss": 0.0055,
+      "step": 36392
+    },
+    {
+      "epoch": 33.95,
+      "learning_rate": 4.575606343283582e-05,
+      "loss": 0.0003,
+      "step": 36396
+    },
+    {
+      "epoch": 33.96,
+      "learning_rate": 4.5755597014925376e-05,
+      "loss": 0.0001,
+      "step": 36400
+    },
+    {
+      "epoch": 33.96,
+      "learning_rate": 4.575513059701493e-05,
+      "loss": 0.0001,
+      "step": 36404
+    },
+    {
+      "epoch": 33.96,
+      "learning_rate": 4.575466417910448e-05,
+      "loss": 0.001,
+      "step": 36408
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 4.5754197761194034e-05,
+      "loss": 0.0001,
+      "step": 36412
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 4.575373134328358e-05,
+      "loss": 0.0,
+      "step": 36416
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 4.575326492537314e-05,
+      "loss": 0.0,
+      "step": 36420
+    },
+    {
+      "epoch": 33.98,
+      "learning_rate": 4.575279850746269e-05,
+      "loss": 0.0009,
+      "step": 36424
+    },
+    {
+      "epoch": 33.98,
+      "learning_rate": 4.575233208955224e-05,
+      "loss": 0.0045,
+      "step": 36428
+    },
+    {
+      "epoch": 33.98,
+      "learning_rate": 4.5751865671641795e-05,
+      "loss": 0.0003,
+      "step": 36432
+    },
+    {
+      "epoch": 33.99,
+      "learning_rate": 4.575139925373135e-05,
+      "loss": 0.0,
+      "step": 36436
+    },
+    {
+      "epoch": 33.99,
+      "learning_rate": 4.57509328358209e-05,
+      "loss": 0.0,
+      "step": 36440
+    },
+    {
+      "epoch": 34.0,
+      "learning_rate": 4.5750466417910446e-05,
+      "loss": 0.0079,
+      "step": 36444
+    },
+    {
+      "epoch": 34.0,
+      "learning_rate": 4.575e-05,
+      "loss": 0.0003,
+      "step": 36448
+    },
+    {
+      "epoch": 34.0,
+      "learning_rate": 4.5749533582089556e-05,
+      "loss": 0.0001,
+      "step": 36452
+    },
+    {
+      "epoch": 34.01,
+      "learning_rate": 4.5749067164179104e-05,
+      "loss": 0.0001,
+      "step": 36456
+    },
+    {
+      "epoch": 34.01,
+      "learning_rate": 4.574860074626866e-05,
+      "loss": 0.0002,
+      "step": 36460
+    },
+    {
+      "epoch": 34.01,
+      "learning_rate": 4.5748134328358214e-05,
+      "loss": 0.0031,
+      "step": 36464
+    },
+    {
+      "epoch": 34.02,
+      "learning_rate": 4.574766791044776e-05,
+      "loss": 0.0002,
+      "step": 36468
+    },
+    {
+      "epoch": 34.02,
+      "learning_rate": 4.574720149253732e-05,
+      "loss": 0.0013,
+      "step": 36472
+    },
+    {
+      "epoch": 34.03,
+      "learning_rate": 4.5746735074626865e-05,
+      "loss": 0.0006,
+      "step": 36476
+    },
+    {
+      "epoch": 34.03,
+      "learning_rate": 4.574626865671642e-05,
+      "loss": 0.0052,
+      "step": 36480
+    },
+    {
+      "epoch": 34.03,
+      "learning_rate": 4.5745802238805975e-05,
+      "loss": 0.001,
+      "step": 36484
+    },
+    {
+      "epoch": 34.04,
+      "learning_rate": 4.574533582089552e-05,
+      "loss": 0.0001,
+      "step": 36488
+    },
+    {
+      "epoch": 34.04,
+      "learning_rate": 4.574486940298508e-05,
+      "loss": 0.0011,
+      "step": 36492
+    },
+    {
+      "epoch": 34.04,
+      "learning_rate": 4.574440298507463e-05,
+      "loss": 0.0001,
+      "step": 36496
+    },
+    {
+      "epoch": 34.05,
+      "learning_rate": 4.574393656716418e-05,
+      "loss": 0.0,
+      "step": 36500
+    },
+    {
+      "epoch": 34.05,
+      "eval_exact_match": 0.730174081237911,
+      "eval_exec": 0.7659574468085106,
+      "eval_loss": 0.4040814936161041,
+      "eval_runtime": 1222.2246,
+      "eval_samples_per_second": 0.846,
+      "step": 36500
+    },
+    {
+      "epoch": 34.05,
+      "learning_rate": 4.574347014925373e-05,
+      "loss": 0.0004,
+      "step": 36504
+    },
+    {
+      "epoch": 34.06,
+      "learning_rate": 4.5743003731343284e-05,
+      "loss": 0.0036,
+      "step": 36508
+    },
+    {
+      "epoch": 34.06,
+      "learning_rate": 4.574253731343284e-05,
+      "loss": 0.0008,
+      "step": 36512
+    },
+    {
+      "epoch": 34.06,
+      "learning_rate": 4.574207089552239e-05,
+      "loss": 0.0002,
+      "step": 36516
+    },
+    {
+      "epoch": 34.07,
+      "learning_rate": 4.574160447761194e-05,
+      "loss": 0.0006,
+      "step": 36520
+    },
+    {
+      "epoch": 34.07,
+      "learning_rate": 4.5741138059701497e-05,
+      "loss": 0.0,
+      "step": 36524
+    },
+    {
+      "epoch": 34.07,
+      "learning_rate": 4.574067164179105e-05,
+      "loss": 0.0017,
+      "step": 36528
+    },
+    {
+      "epoch": 34.08,
+      "learning_rate": 4.57402052238806e-05,
+      "loss": 0.0001,
+      "step": 36532
+    },
+    {
+      "epoch": 34.08,
+      "learning_rate": 4.573973880597015e-05,
+      "loss": 0.0001,
+      "step": 36536
+    },
+    {
+      "epoch": 34.09,
+      "learning_rate": 4.573927238805971e-05,
+      "loss": 0.0003,
+      "step": 36540
+    },
+    {
+      "epoch": 34.09,
+      "learning_rate": 4.573880597014926e-05,
+      "loss": 0.0013,
+      "step": 36544
+    },
+    {
+      "epoch": 34.09,
+      "learning_rate": 4.5738339552238806e-05,
+      "loss": 0.0003,
+      "step": 36548
+    },
+    {
+      "epoch": 34.1,
+      "learning_rate": 4.573787313432836e-05,
+      "loss": 0.0004,
+      "step": 36552
+    },
+    {
+      "epoch": 34.1,
+      "learning_rate": 4.5737406716417915e-05,
+      "loss": 0.0003,
+      "step": 36556
+    },
+    {
+      "epoch": 34.1,
+      "learning_rate": 4.5736940298507464e-05,
+      "loss": 0.0016,
+      "step": 36560
+    },
+    {
+      "epoch": 34.11,
+      "learning_rate": 4.573647388059702e-05,
+      "loss": 0.0024,
+      "step": 36564
+    },
+    {
+      "epoch": 34.11,
+      "learning_rate": 4.573600746268657e-05,
+      "loss": 0.0003,
+      "step": 36568
+    },
+    {
+      "epoch": 34.12,
+      "learning_rate": 4.573554104477612e-05,
+      "loss": 0.0001,
+      "step": 36572
+    },
+    {
+      "epoch": 34.12,
+      "learning_rate": 4.5735074626865676e-05,
+      "loss": 0.0005,
+      "step": 36576
+    },
+    {
+      "epoch": 34.12,
+      "learning_rate": 4.5734608208955225e-05,
+      "loss": 0.0008,
+      "step": 36580
+    },
+    {
+      "epoch": 34.13,
+      "learning_rate": 4.573414179104478e-05,
+      "loss": 0.0001,
+      "step": 36584
+    },
+    {
+      "epoch": 34.13,
+      "learning_rate": 4.5733675373134334e-05,
+      "loss": 0.013,
+      "step": 36588
+    },
+    {
+      "epoch": 34.13,
+      "learning_rate": 4.573320895522388e-05,
+      "loss": 0.0001,
+      "step": 36592
+    },
+    {
+      "epoch": 34.14,
+      "learning_rate": 4.573274253731343e-05,
+      "loss": 0.0001,
+      "step": 36596
+    },
+    {
+      "epoch": 34.14,
+      "learning_rate": 4.573227611940299e-05,
+      "loss": 0.0003,
+      "step": 36600
+    },
+    {
+      "epoch": 34.15,
+      "learning_rate": 4.573180970149254e-05,
+      "loss": 0.0001,
+      "step": 36604
+    },
+    {
+      "epoch": 34.15,
+      "learning_rate": 4.573134328358209e-05,
+      "loss": 0.0,
+      "step": 36608
+    },
+    {
+      "epoch": 34.15,
+      "learning_rate": 4.5730876865671644e-05,
+      "loss": 0.0001,
+      "step": 36612
+    },
+    {
+      "epoch": 34.16,
+      "learning_rate": 4.57304104477612e-05,
+      "loss": 0.002,
+      "step": 36616
+    },
+    {
+      "epoch": 34.16,
+      "learning_rate": 4.5729944029850747e-05,
+      "loss": 0.0002,
+      "step": 36620
+    },
+    {
+      "epoch": 34.16,
+      "learning_rate": 4.57294776119403e-05,
+      "loss": 0.0003,
+      "step": 36624
+    },
+    {
+      "epoch": 34.17,
+      "learning_rate": 4.572901119402985e-05,
+      "loss": 0.0003,
+      "step": 36628
+    },
+    {
+      "epoch": 34.17,
+      "learning_rate": 4.5728544776119405e-05,
+      "loss": 0.0001,
+      "step": 36632
+    },
+    {
+      "epoch": 34.18,
+      "learning_rate": 4.572807835820896e-05,
+      "loss": 0.0002,
+      "step": 36636
+    },
+    {
+      "epoch": 34.18,
+      "learning_rate": 4.572761194029851e-05,
+      "loss": 0.0001,
+      "step": 36640
+    },
+    {
+      "epoch": 34.18,
+      "learning_rate": 4.572714552238806e-05,
+      "loss": 0.0015,
+      "step": 36644
+    },
+    {
+      "epoch": 34.19,
+      "learning_rate": 4.572667910447762e-05,
+      "loss": 0.0003,
+      "step": 36648
+    },
+    {
+      "epoch": 34.19,
+      "learning_rate": 4.5726212686567165e-05,
+      "loss": 0.0016,
+      "step": 36652
+    },
+    {
+      "epoch": 34.19,
+      "learning_rate": 4.5725746268656714e-05,
+      "loss": 0.0003,
+      "step": 36656
+    },
+    {
+      "epoch": 34.2,
+      "learning_rate": 4.5725279850746275e-05,
+      "loss": 0.0002,
+      "step": 36660
+    },
+    {
+      "epoch": 34.2,
+      "learning_rate": 4.5724813432835823e-05,
+      "loss": 0.0,
+      "step": 36664
+    },
+    {
+      "epoch": 34.21,
+      "learning_rate": 4.572434701492537e-05,
+      "loss": 0.0001,
+      "step": 36668
+    },
+    {
+      "epoch": 34.21,
+      "learning_rate": 4.5723880597014926e-05,
+      "loss": 0.0001,
+      "step": 36672
+    },
+    {
+      "epoch": 34.21,
+      "learning_rate": 4.572341417910448e-05,
+      "loss": 0.0003,
+      "step": 36676
+    },
+    {
+      "epoch": 34.22,
+      "learning_rate": 4.572294776119403e-05,
+      "loss": 0.0001,
+      "step": 36680
+    },
+    {
+      "epoch": 34.22,
+      "learning_rate": 4.5722481343283584e-05,
+      "loss": 0.0002,
+      "step": 36684
+    },
+    {
+      "epoch": 34.22,
+      "learning_rate": 4.572201492537313e-05,
+      "loss": 0.0016,
+      "step": 36688
+    },
+    {
+      "epoch": 34.23,
+      "learning_rate": 4.5721548507462694e-05,
+      "loss": 0.0011,
+      "step": 36692
+    },
+    {
+      "epoch": 34.23,
+      "learning_rate": 4.572108208955224e-05,
+      "loss": 0.0,
+      "step": 36696
+    },
+    {
+      "epoch": 34.24,
+      "learning_rate": 4.572061567164179e-05,
+      "loss": 0.0001,
+      "step": 36700
+    },
+    {
+      "epoch": 34.24,
+      "learning_rate": 4.5720149253731345e-05,
+      "loss": 0.0,
+      "step": 36704
+    },
+    {
+      "epoch": 34.24,
+      "learning_rate": 4.57196828358209e-05,
+      "loss": 0.0,
+      "step": 36708
+    },
+    {
+      "epoch": 34.25,
+      "learning_rate": 4.571921641791045e-05,
+      "loss": 0.0,
+      "step": 36712
+    },
+    {
+      "epoch": 34.25,
+      "learning_rate": 4.571875e-05,
+      "loss": 0.0022,
+      "step": 36716
+    },
+    {
+      "epoch": 34.25,
+      "learning_rate": 4.571828358208956e-05,
+      "loss": 0.0002,
+      "step": 36720
+    },
+    {
+      "epoch": 34.26,
+      "learning_rate": 4.5717817164179106e-05,
+      "loss": 0.0001,
+      "step": 36724
+    },
+    {
+      "epoch": 34.26,
+      "learning_rate": 4.571735074626866e-05,
+      "loss": 0.0001,
+      "step": 36728
+    },
+    {
+      "epoch": 34.26,
+      "learning_rate": 4.571688432835821e-05,
+      "loss": 0.0064,
+      "step": 36732
+    },
+    {
+      "epoch": 34.27,
+      "learning_rate": 4.5716417910447764e-05,
+      "loss": 0.0001,
+      "step": 36736
+    },
+    {
+      "epoch": 34.27,
+      "learning_rate": 4.571595149253732e-05,
+      "loss": 0.0001,
+      "step": 36740
+    },
+    {
+      "epoch": 34.28,
+      "learning_rate": 4.571548507462687e-05,
+      "loss": 0.0001,
+      "step": 36744
+    },
+    {
+      "epoch": 34.28,
+      "learning_rate": 4.5715018656716416e-05,
+      "loss": 0.0001,
+      "step": 36748
+    },
+    {
+      "epoch": 34.28,
+      "learning_rate": 4.571455223880598e-05,
+      "loss": 0.0,
+      "step": 36752
+    },
+    {
+      "epoch": 34.29,
+      "learning_rate": 4.5714085820895525e-05,
+      "loss": 0.012,
+      "step": 36756
+    },
+    {
+      "epoch": 34.29,
+      "learning_rate": 4.5713619402985073e-05,
+      "loss": 0.0025,
+      "step": 36760
+    },
+    {
+      "epoch": 34.29,
+      "learning_rate": 4.571315298507463e-05,
+      "loss": 0.0001,
+      "step": 36764
+    },
+    {
+      "epoch": 34.3,
+      "learning_rate": 4.571268656716418e-05,
+      "loss": 0.0016,
+      "step": 36768
+    },
+    {
+      "epoch": 34.3,
+      "learning_rate": 4.571222014925373e-05,
+      "loss": 0.0001,
+      "step": 36772
+    },
+    {
+      "epoch": 34.31,
+      "learning_rate": 4.5711753731343286e-05,
+      "loss": 0.0007,
+      "step": 36776
+    },
+    {
+      "epoch": 34.31,
+      "learning_rate": 4.571128731343284e-05,
+      "loss": 0.0011,
+      "step": 36780
+    },
+    {
+      "epoch": 34.31,
+      "learning_rate": 4.571082089552239e-05,
+      "loss": 0.0,
+      "step": 36784
+    },
+    {
+      "epoch": 34.32,
+      "learning_rate": 4.5710354477611944e-05,
+      "loss": 0.0008,
+      "step": 36788
+    },
+    {
+      "epoch": 34.32,
+      "learning_rate": 4.570988805970149e-05,
+      "loss": 0.002,
+      "step": 36792
+    },
+    {
+      "epoch": 34.32,
+      "learning_rate": 4.570942164179105e-05,
+      "loss": 0.0001,
+      "step": 36796
+    },
+    {
+      "epoch": 34.33,
+      "learning_rate": 4.57089552238806e-05,
+      "loss": 0.0,
+      "step": 36800
+    },
+    {
+      "epoch": 34.33,
+      "learning_rate": 4.570848880597015e-05,
+      "loss": 0.0001,
+      "step": 36804
+    },
+    {
+      "epoch": 34.34,
+      "learning_rate": 4.57080223880597e-05,
+      "loss": 0.0005,
+      "step": 36808
+    },
+    {
+      "epoch": 34.34,
+      "learning_rate": 4.570755597014926e-05,
+      "loss": 0.0002,
+      "step": 36812
+    },
+    {
+      "epoch": 34.34,
+      "learning_rate": 4.570708955223881e-05,
+      "loss": 0.0001,
+      "step": 36816
+    },
+    {
+      "epoch": 34.35,
+      "learning_rate": 4.5706623134328356e-05,
+      "loss": 0.0001,
+      "step": 36820
+    },
+    {
+      "epoch": 34.35,
+      "learning_rate": 4.570615671641791e-05,
+      "loss": 0.0008,
+      "step": 36824
+    },
+    {
+      "epoch": 34.35,
+      "learning_rate": 4.5705690298507466e-05,
+      "loss": 0.0006,
+      "step": 36828
+    },
+    {
+      "epoch": 34.36,
+      "learning_rate": 4.5705223880597014e-05,
+      "loss": 0.0,
+      "step": 36832
+    },
+    {
+      "epoch": 34.36,
+      "learning_rate": 4.570475746268657e-05,
+      "loss": 0.0003,
+      "step": 36836
+    },
+    {
+      "epoch": 34.37,
+      "learning_rate": 4.570429104477612e-05,
+      "loss": 0.0036,
+      "step": 36840
+    },
+    {
+      "epoch": 34.37,
+      "learning_rate": 4.570382462686567e-05,
+      "loss": 0.0,
+      "step": 36844
+    },
+    {
+      "epoch": 34.37,
+      "learning_rate": 4.570335820895523e-05,
+      "loss": 0.0,
+      "step": 36848
+    },
+    {
+      "epoch": 34.38,
+      "learning_rate": 4.5702891791044775e-05,
+      "loss": 0.0,
+      "step": 36852
+    },
+    {
+      "epoch": 34.38,
+      "learning_rate": 4.570242537313433e-05,
+      "loss": 0.0,
+      "step": 36856
+    },
+    {
+      "epoch": 34.38,
+      "learning_rate": 4.5701958955223885e-05,
+      "loss": 0.001,
+      "step": 36860
+    },
+    {
+      "epoch": 34.39,
+      "learning_rate": 4.570149253731343e-05,
+      "loss": 0.0139,
+      "step": 36864
+    },
+    {
+      "epoch": 34.39,
+      "learning_rate": 4.570102611940299e-05,
+      "loss": 0.0001,
+      "step": 36868
+    },
+    {
+      "epoch": 34.4,
+      "learning_rate": 4.570055970149254e-05,
+      "loss": 0.0005,
+      "step": 36872
+    },
+    {
+      "epoch": 34.4,
+      "learning_rate": 4.570009328358209e-05,
+      "loss": 0.0002,
+      "step": 36876
+    },
+    {
+      "epoch": 34.4,
+      "learning_rate": 4.5699626865671646e-05,
+      "loss": 0.0008,
+      "step": 36880
+    },
+    {
+      "epoch": 34.41,
+      "learning_rate": 4.5699160447761194e-05,
+      "loss": 0.0003,
+      "step": 36884
+    },
+    {
+      "epoch": 34.41,
+      "learning_rate": 4.569869402985075e-05,
+      "loss": 0.0,
+      "step": 36888
+    },
+    {
+      "epoch": 34.41,
+      "learning_rate": 4.5698227611940304e-05,
+      "loss": 0.0003,
+      "step": 36892
+    },
+    {
+      "epoch": 34.42,
+      "learning_rate": 4.569776119402985e-05,
+      "loss": 0.0001,
+      "step": 36896
+    },
+    {
+      "epoch": 34.42,
+      "learning_rate": 4.56972947761194e-05,
+      "loss": 0.0002,
+      "step": 36900
+    },
+    {
+      "epoch": 34.43,
+      "learning_rate": 4.569682835820896e-05,
+      "loss": 0.0035,
+      "step": 36904
+    },
+    {
+      "epoch": 34.43,
+      "learning_rate": 4.569636194029851e-05,
+      "loss": 0.0023,
+      "step": 36908
+    },
+    {
+      "epoch": 34.43,
+      "learning_rate": 4.569589552238806e-05,
+      "loss": 0.0009,
+      "step": 36912
+    },
+    {
+      "epoch": 34.44,
+      "learning_rate": 4.569542910447761e-05,
+      "loss": 0.0073,
+      "step": 36916
+    },
+    {
+      "epoch": 34.44,
+      "learning_rate": 4.569496268656717e-05,
+      "loss": 0.0006,
+      "step": 36920
+    },
+    {
+      "epoch": 34.44,
+      "learning_rate": 4.5694496268656716e-05,
+      "loss": 0.0004,
+      "step": 36924
+    },
+    {
+      "epoch": 34.45,
+      "learning_rate": 4.569402985074627e-05,
+      "loss": 0.0003,
+      "step": 36928
+    },
+    {
+      "epoch": 34.45,
+      "learning_rate": 4.5693563432835826e-05,
+      "loss": 0.0001,
+      "step": 36932
+    },
+    {
+      "epoch": 34.46,
+      "learning_rate": 4.5693097014925374e-05,
+      "loss": 0.0001,
+      "step": 36936
+    },
+    {
+      "epoch": 34.46,
+      "learning_rate": 4.569263059701493e-05,
+      "loss": 0.0029,
+      "step": 36940
+    },
+    {
+      "epoch": 34.46,
+      "learning_rate": 4.569216417910448e-05,
+      "loss": 0.0003,
+      "step": 36944
+    },
+    {
+      "epoch": 34.47,
+      "learning_rate": 4.569169776119403e-05,
+      "loss": 0.0005,
+      "step": 36948
+    },
+    {
+      "epoch": 34.47,
+      "learning_rate": 4.569123134328359e-05,
+      "loss": 0.001,
+      "step": 36952
+    },
+    {
+      "epoch": 34.47,
+      "learning_rate": 4.5690764925373135e-05,
+      "loss": 0.0001,
+      "step": 36956
+    },
+    {
+      "epoch": 34.48,
+      "learning_rate": 4.569029850746268e-05,
+      "loss": 0.0004,
+      "step": 36960
+    },
+    {
+      "epoch": 34.48,
+      "learning_rate": 4.5689832089552245e-05,
+      "loss": 0.0003,
+      "step": 36964
+    },
+    {
+      "epoch": 34.49,
+      "learning_rate": 4.568936567164179e-05,
+      "loss": 0.0001,
+      "step": 36968
+    },
+    {
+      "epoch": 34.49,
+      "learning_rate": 4.568889925373134e-05,
+      "loss": 0.0025,
+      "step": 36972
+    },
+    {
+      "epoch": 34.49,
+      "learning_rate": 4.5688432835820896e-05,
+      "loss": 0.0008,
+      "step": 36976
+    },
+    {
+      "epoch": 34.5,
+      "learning_rate": 4.568796641791045e-05,
+      "loss": 0.0018,
+      "step": 36980
+    },
+    {
+      "epoch": 34.5,
+      "learning_rate": 4.56875e-05,
+      "loss": 0.0023,
+      "step": 36984
+    },
+    {
+      "epoch": 34.5,
+      "learning_rate": 4.5687033582089554e-05,
+      "loss": 0.0002,
+      "step": 36988
+    },
+    {
+      "epoch": 34.51,
+      "learning_rate": 4.568656716417911e-05,
+      "loss": 0.0002,
+      "step": 36992
+    },
+    {
+      "epoch": 34.51,
+      "learning_rate": 4.568610074626866e-05,
+      "loss": 0.0044,
+      "step": 36996
+    },
+    {
+      "epoch": 34.51,
+      "learning_rate": 4.568563432835821e-05,
+      "loss": 0.0001,
+      "step": 37000
+    },
+    {
+      "epoch": 34.51,
+      "eval_exact_match": 0.7504835589941973,
+      "eval_exec": 0.7794970986460348,
+      "eval_loss": 0.40531110763549805,
+      "eval_runtime": 1164.9382,
+      "eval_samples_per_second": 0.888,
+      "step": 37000
+    },
+    {
+      "epoch": 34.52,
+      "learning_rate": 4.568516791044776e-05,
+      "loss": 0.0,
+      "step": 37004
+    },
+    {
+      "epoch": 34.52,
+      "learning_rate": 4.5684701492537315e-05,
+      "loss": 0.0019,
+      "step": 37008
+    },
+    {
+      "epoch": 34.53,
+      "learning_rate": 4.568423507462687e-05,
+      "loss": 0.0003,
+      "step": 37012
+    },
+    {
+      "epoch": 34.53,
+      "learning_rate": 4.568376865671642e-05,
+      "loss": 0.0016,
+      "step": 37016
+    },
+    {
+      "epoch": 34.53,
+      "learning_rate": 4.568330223880597e-05,
+      "loss": 0.0001,
+      "step": 37020
+    },
+    {
+      "epoch": 34.54,
+      "learning_rate": 4.568283582089553e-05,
+      "loss": 0.0,
+      "step": 37024
+    },
+    {
+      "epoch": 34.54,
+      "learning_rate": 4.5682369402985076e-05,
+      "loss": 0.0001,
+      "step": 37028
+    },
+    {
+      "epoch": 34.54,
+      "learning_rate": 4.568190298507463e-05,
+      "loss": 0.0001,
+      "step": 37032
+    },
+    {
+      "epoch": 34.55,
+      "learning_rate": 4.568143656716418e-05,
+      "loss": 0.0,
+      "step": 37036
+    },
+    {
+      "epoch": 34.55,
+      "learning_rate": 4.5680970149253734e-05,
+      "loss": 0.0003,
+      "step": 37040
+    },
+    {
+      "epoch": 34.56,
+      "learning_rate": 4.568050373134329e-05,
+      "loss": 0.0035,
+      "step": 37044
+    },
+    {
+      "epoch": 34.56,
+      "learning_rate": 4.568003731343284e-05,
+      "loss": 0.0,
+      "step": 37048
+    },
+    {
+      "epoch": 34.56,
+      "learning_rate": 4.567957089552239e-05,
+      "loss": 0.0002,
+      "step": 37052
+    },
+    {
+      "epoch": 34.57,
+      "learning_rate": 4.567910447761195e-05,
+      "loss": 0.0077,
+      "step": 37056
+    },
+    {
+      "epoch": 34.57,
+      "learning_rate": 4.5678638059701495e-05,
+      "loss": 0.0022,
+      "step": 37060
+    },
+    {
+      "epoch": 34.57,
+      "learning_rate": 4.567817164179104e-05,
+      "loss": 0.0003,
+      "step": 37064
+    },
+    {
+      "epoch": 34.58,
+      "learning_rate": 4.56777052238806e-05,
+      "loss": 0.0,
+      "step": 37068
+    },
+    {
+      "epoch": 34.58,
+      "learning_rate": 4.567723880597015e-05,
+      "loss": 0.0,
+      "step": 37072
+    },
+    {
+      "epoch": 34.59,
+      "learning_rate": 4.56767723880597e-05,
+      "loss": 0.0001,
+      "step": 37076
+    },
+    {
+      "epoch": 34.59,
+      "learning_rate": 4.5676305970149256e-05,
+      "loss": 0.004,
+      "step": 37080
+    },
+    {
+      "epoch": 34.59,
+      "learning_rate": 4.567583955223881e-05,
+      "loss": 0.0001,
+      "step": 37084
+    },
+    {
+      "epoch": 34.6,
+      "learning_rate": 4.567537313432836e-05,
+      "loss": 0.0003,
+      "step": 37088
+    },
+    {
+      "epoch": 34.6,
+      "learning_rate": 4.5674906716417914e-05,
+      "loss": 0.0006,
+      "step": 37092
+    },
+    {
+      "epoch": 34.6,
+      "learning_rate": 4.567444029850746e-05,
+      "loss": 0.0001,
+      "step": 37096
+    },
+    {
+      "epoch": 34.61,
+      "learning_rate": 4.567397388059702e-05,
+      "loss": 0.0001,
+      "step": 37100
+    },
+    {
+      "epoch": 34.61,
+      "learning_rate": 4.567350746268657e-05,
+      "loss": 0.0,
+      "step": 37104
+    },
+    {
+      "epoch": 34.62,
+      "learning_rate": 4.567304104477612e-05,
+      "loss": 0.0009,
+      "step": 37108
+    },
+    {
+      "epoch": 34.62,
+      "learning_rate": 4.5672574626865675e-05,
+      "loss": 0.0002,
+      "step": 37112
+    },
+    {
+      "epoch": 34.62,
+      "learning_rate": 4.567210820895523e-05,
+      "loss": 0.0011,
+      "step": 37116
+    },
+    {
+      "epoch": 34.63,
+      "learning_rate": 4.567164179104478e-05,
+      "loss": 0.0,
+      "step": 37120
+    },
+    {
+      "epoch": 34.63,
+      "learning_rate": 4.5671175373134326e-05,
+      "loss": 0.0011,
+      "step": 37124
+    },
+    {
+      "epoch": 34.63,
+      "learning_rate": 4.567070895522388e-05,
+      "loss": 0.0004,
+      "step": 37128
+    },
+    {
+      "epoch": 34.64,
+      "learning_rate": 4.5670242537313436e-05,
+      "loss": 0.0008,
+      "step": 37132
+    },
+    {
+      "epoch": 34.64,
+      "learning_rate": 4.5669776119402984e-05,
+      "loss": 0.0001,
+      "step": 37136
+    },
+    {
+      "epoch": 34.65,
+      "learning_rate": 4.566930970149254e-05,
+      "loss": 0.0002,
+      "step": 37140
+    },
+    {
+      "epoch": 34.65,
+      "learning_rate": 4.5668843283582094e-05,
+      "loss": 0.0001,
+      "step": 37144
+    },
+    {
+      "epoch": 34.65,
+      "learning_rate": 4.566837686567164e-05,
+      "loss": 0.0009,
+      "step": 37148
+    },
+    {
+      "epoch": 34.66,
+      "learning_rate": 4.56679104477612e-05,
+      "loss": 0.0001,
+      "step": 37152
+    },
+    {
+      "epoch": 34.66,
+      "learning_rate": 4.5667444029850745e-05,
+      "loss": 0.0003,
+      "step": 37156
+    },
+    {
+      "epoch": 34.66,
+      "learning_rate": 4.56669776119403e-05,
+      "loss": 0.0029,
+      "step": 37160
+    },
+    {
+      "epoch": 34.67,
+      "learning_rate": 4.5666511194029855e-05,
+      "loss": 0.0001,
+      "step": 37164
+    },
+    {
+      "epoch": 34.67,
+      "learning_rate": 4.56660447761194e-05,
+      "loss": 0.0136,
+      "step": 37168
+    },
+    {
+      "epoch": 34.68,
+      "learning_rate": 4.566557835820896e-05,
+      "loss": 0.0019,
+      "step": 37172
+    },
+    {
+      "epoch": 34.68,
+      "learning_rate": 4.566511194029851e-05,
+      "loss": 0.0005,
+      "step": 37176
+    },
+    {
+      "epoch": 34.68,
+      "learning_rate": 4.566464552238806e-05,
+      "loss": 0.0005,
+      "step": 37180
+    },
+    {
+      "epoch": 34.69,
+      "learning_rate": 4.5664179104477616e-05,
+      "loss": 0.0008,
+      "step": 37184
+    },
+    {
+      "epoch": 34.69,
+      "learning_rate": 4.5663712686567164e-05,
+      "loss": 0.0,
+      "step": 37188
+    },
+    {
+      "epoch": 34.69,
+      "learning_rate": 4.566324626865672e-05,
+      "loss": 0.0,
+      "step": 37192
+    },
+    {
+      "epoch": 34.7,
+      "learning_rate": 4.5662779850746274e-05,
+      "loss": 0.0022,
+      "step": 37196
+    },
+    {
+      "epoch": 34.7,
+      "learning_rate": 4.566231343283582e-05,
+      "loss": 0.0001,
+      "step": 37200
+    },
+    {
+      "epoch": 34.71,
+      "learning_rate": 4.566184701492538e-05,
+      "loss": 0.0001,
+      "step": 37204
+    },
+    {
+      "epoch": 34.71,
+      "learning_rate": 4.566138059701493e-05,
+      "loss": 0.003,
+      "step": 37208
+    },
+    {
+      "epoch": 34.71,
+      "learning_rate": 4.566091417910448e-05,
+      "loss": 0.0002,
+      "step": 37212
+    },
+    {
+      "epoch": 34.72,
+      "learning_rate": 4.566044776119403e-05,
+      "loss": 0.0001,
+      "step": 37216
+    },
+    {
+      "epoch": 34.72,
+      "learning_rate": 4.565998134328359e-05,
+      "loss": 0.0001,
+      "step": 37220
+    },
+    {
+      "epoch": 34.72,
+      "learning_rate": 4.565951492537314e-05,
+      "loss": 0.005,
+      "step": 37224
+    },
+    {
+      "epoch": 34.73,
+      "learning_rate": 4.5659048507462686e-05,
+      "loss": 0.0066,
+      "step": 37228
+    },
+    {
+      "epoch": 34.73,
+      "learning_rate": 4.565858208955224e-05,
+      "loss": 0.0001,
+      "step": 37232
+    },
+    {
+      "epoch": 34.73,
+      "learning_rate": 4.5658115671641796e-05,
+      "loss": 0.0005,
+      "step": 37236
+    },
+    {
+      "epoch": 34.74,
+      "learning_rate": 4.5657649253731344e-05,
+      "loss": 0.0005,
+      "step": 37240
+    },
+    {
+      "epoch": 34.74,
+      "learning_rate": 4.56571828358209e-05,
+      "loss": 0.0,
+      "step": 37244
+    },
+    {
+      "epoch": 34.75,
+      "learning_rate": 4.565671641791045e-05,
+      "loss": 0.0042,
+      "step": 37248
+    },
+    {
+      "epoch": 34.75,
+      "learning_rate": 4.565625e-05,
+      "loss": 0.0,
+      "step": 37252
+    },
+    {
+      "epoch": 34.75,
+      "learning_rate": 4.565578358208956e-05,
+      "loss": 0.0,
+      "step": 37256
+    },
+    {
+      "epoch": 34.76,
+      "learning_rate": 4.5655317164179105e-05,
+      "loss": 0.0009,
+      "step": 37260
+    },
+    {
+      "epoch": 34.76,
+      "learning_rate": 4.565485074626866e-05,
+      "loss": 0.0001,
+      "step": 37264
+    },
+    {
+      "epoch": 34.76,
+      "learning_rate": 4.5654384328358215e-05,
+      "loss": 0.0011,
+      "step": 37268
+    },
+    {
+      "epoch": 34.77,
+      "learning_rate": 4.565391791044776e-05,
+      "loss": 0.0009,
+      "step": 37272
+    },
+    {
+      "epoch": 34.77,
+      "learning_rate": 4.565345149253731e-05,
+      "loss": 0.0,
+      "step": 37276
+    },
+    {
+      "epoch": 34.78,
+      "learning_rate": 4.565298507462687e-05,
+      "loss": 0.0012,
+      "step": 37280
+    },
+    {
+      "epoch": 34.78,
+      "learning_rate": 4.565251865671642e-05,
+      "loss": 0.0041,
+      "step": 37284
+    },
+    {
+      "epoch": 34.78,
+      "learning_rate": 4.565205223880597e-05,
+      "loss": 0.0,
+      "step": 37288
+    },
+    {
+      "epoch": 34.79,
+      "learning_rate": 4.5651585820895524e-05,
+      "loss": 0.0002,
+      "step": 37292
+    },
+    {
+      "epoch": 34.79,
+      "learning_rate": 4.565111940298508e-05,
+      "loss": 0.0001,
+      "step": 37296
+    },
+    {
+      "epoch": 34.79,
+      "learning_rate": 4.565065298507463e-05,
+      "loss": 0.0002,
+      "step": 37300
+    },
+    {
+      "epoch": 34.8,
+      "learning_rate": 4.565018656716418e-05,
+      "loss": 0.0,
+      "step": 37304
+    },
+    {
+      "epoch": 34.8,
+      "learning_rate": 4.564972014925373e-05,
+      "loss": 0.0,
+      "step": 37308
+    },
+    {
+      "epoch": 34.81,
+      "learning_rate": 4.5649253731343285e-05,
+      "loss": 0.0013,
+      "step": 37312
+    },
+    {
+      "epoch": 34.81,
+      "learning_rate": 4.564878731343284e-05,
+      "loss": 0.0001,
+      "step": 37316
+    },
+    {
+      "epoch": 34.81,
+      "learning_rate": 4.564832089552239e-05,
+      "loss": 0.0008,
+      "step": 37320
+    },
+    {
+      "epoch": 34.82,
+      "learning_rate": 4.564785447761194e-05,
+      "loss": 0.0007,
+      "step": 37324
+    },
+    {
+      "epoch": 34.82,
+      "learning_rate": 4.56473880597015e-05,
+      "loss": 0.0,
+      "step": 37328
+    },
+    {
+      "epoch": 34.82,
+      "learning_rate": 4.5646921641791046e-05,
+      "loss": 0.0019,
+      "step": 37332
+    },
+    {
+      "epoch": 34.83,
+      "learning_rate": 4.5646455223880594e-05,
+      "loss": 0.0001,
+      "step": 37336
+    },
+    {
+      "epoch": 34.83,
+      "learning_rate": 4.5645988805970156e-05,
+      "loss": 0.0046,
+      "step": 37340
+    },
+    {
+      "epoch": 34.84,
+      "learning_rate": 4.5645522388059704e-05,
+      "loss": 0.001,
+      "step": 37344
+    },
+    {
+      "epoch": 34.84,
+      "learning_rate": 4.564505597014926e-05,
+      "loss": 0.0003,
+      "step": 37348
+    },
+    {
+      "epoch": 34.84,
+      "learning_rate": 4.564458955223881e-05,
+      "loss": 0.0002,
+      "step": 37352
+    },
+    {
+      "epoch": 34.85,
+      "learning_rate": 4.564412313432836e-05,
+      "loss": 0.0004,
+      "step": 37356
+    },
+    {
+      "epoch": 34.85,
+      "learning_rate": 4.5643656716417917e-05,
+      "loss": 0.0009,
+      "step": 37360
+    },
+    {
+      "epoch": 34.85,
+      "learning_rate": 4.5643190298507465e-05,
+      "loss": 0.0,
+      "step": 37364
+    },
+    {
+      "epoch": 34.86,
+      "learning_rate": 4.564272388059701e-05,
+      "loss": 0.0001,
+      "step": 37368
+    },
+    {
+      "epoch": 34.86,
+      "learning_rate": 4.5642257462686575e-05,
+      "loss": 0.0054,
+      "step": 37372
+    },
+    {
+      "epoch": 34.87,
+      "learning_rate": 4.564179104477612e-05,
+      "loss": 0.0017,
+      "step": 37376
+    },
+    {
+      "epoch": 34.87,
+      "learning_rate": 4.564132462686567e-05,
+      "loss": 0.0002,
+      "step": 37380
+    },
+    {
+      "epoch": 34.87,
+      "learning_rate": 4.5640858208955226e-05,
+      "loss": 0.0001,
+      "step": 37384
+    },
+    {
+      "epoch": 34.88,
+      "learning_rate": 4.564039179104478e-05,
+      "loss": 0.0002,
+      "step": 37388
+    },
+    {
+      "epoch": 34.88,
+      "learning_rate": 4.563992537313433e-05,
+      "loss": 0.0001,
+      "step": 37392
+    },
+    {
+      "epoch": 34.88,
+      "learning_rate": 4.5639458955223884e-05,
+      "loss": 0.0001,
+      "step": 37396
+    },
+    {
+      "epoch": 34.89,
+      "learning_rate": 4.563899253731344e-05,
+      "loss": 0.0001,
+      "step": 37400
+    },
+    {
+      "epoch": 34.89,
+      "learning_rate": 4.563852611940299e-05,
+      "loss": 0.0001,
+      "step": 37404
+    },
+    {
+      "epoch": 34.9,
+      "learning_rate": 4.563805970149254e-05,
+      "loss": 0.0001,
+      "step": 37408
+    },
+    {
+      "epoch": 34.9,
+      "learning_rate": 4.563759328358209e-05,
+      "loss": 0.0003,
+      "step": 37412
+    },
+    {
+      "epoch": 34.9,
+      "learning_rate": 4.5637126865671645e-05,
+      "loss": 0.0,
+      "step": 37416
+    },
+    {
+      "epoch": 34.91,
+      "learning_rate": 4.56366604477612e-05,
+      "loss": 0.0,
+      "step": 37420
+    },
+    {
+      "epoch": 34.91,
+      "learning_rate": 4.563619402985075e-05,
+      "loss": 0.0013,
+      "step": 37424
+    },
+    {
+      "epoch": 34.91,
+      "learning_rate": 4.5635727611940296e-05,
+      "loss": 0.0,
+      "step": 37428
+    },
+    {
+      "epoch": 34.92,
+      "learning_rate": 4.563526119402986e-05,
+      "loss": 0.0,
+      "step": 37432
+    },
+    {
+      "epoch": 34.92,
+      "learning_rate": 4.5634794776119406e-05,
+      "loss": 0.0028,
+      "step": 37436
+    },
+    {
+      "epoch": 34.93,
+      "learning_rate": 4.5634328358208954e-05,
+      "loss": 0.0,
+      "step": 37440
+    },
+    {
+      "epoch": 34.93,
+      "learning_rate": 4.563386194029851e-05,
+      "loss": 0.0,
+      "step": 37444
+    },
+    {
+      "epoch": 34.93,
+      "learning_rate": 4.5633395522388064e-05,
+      "loss": 0.0005,
+      "step": 37448
+    },
+    {
+      "epoch": 34.94,
+      "learning_rate": 4.563292910447761e-05,
+      "loss": 0.0001,
+      "step": 37452
+    },
+    {
+      "epoch": 34.94,
+      "learning_rate": 4.5632462686567167e-05,
+      "loss": 0.0021,
+      "step": 37456
+    },
+    {
+      "epoch": 34.94,
+      "learning_rate": 4.563199626865672e-05,
+      "loss": 0.0,
+      "step": 37460
+    },
+    {
+      "epoch": 34.95,
+      "learning_rate": 4.563152985074627e-05,
+      "loss": 0.0005,
+      "step": 37464
+    },
+    {
+      "epoch": 34.95,
+      "learning_rate": 4.5631063432835825e-05,
+      "loss": 0.0005,
+      "step": 37468
+    },
+    {
+      "epoch": 34.96,
+      "learning_rate": 4.563059701492537e-05,
+      "loss": 0.0001,
+      "step": 37472
+    },
+    {
+      "epoch": 34.96,
+      "learning_rate": 4.563013059701493e-05,
+      "loss": 0.0101,
+      "step": 37476
+    },
+    {
+      "epoch": 34.96,
+      "learning_rate": 4.562966417910448e-05,
+      "loss": 0.0,
+      "step": 37480
+    },
+    {
+      "epoch": 34.97,
+      "learning_rate": 4.562919776119403e-05,
+      "loss": 0.0001,
+      "step": 37484
+    },
+    {
+      "epoch": 34.97,
+      "learning_rate": 4.562873134328358e-05,
+      "loss": 0.0044,
+      "step": 37488
+    },
+    {
+      "epoch": 34.97,
+      "learning_rate": 4.562826492537314e-05,
+      "loss": 0.0001,
+      "step": 37492
+    },
+    {
+      "epoch": 34.98,
+      "learning_rate": 4.562779850746269e-05,
+      "loss": 0.0002,
+      "step": 37496
+    },
+    {
+      "epoch": 34.98,
+      "learning_rate": 4.562733208955224e-05,
+      "loss": 0.0012,
+      "step": 37500
+    },
+    {
+      "epoch": 34.98,
+      "eval_exact_match": 0.7282398452611218,
+      "eval_exec": 0.7630560928433269,
+      "eval_loss": 0.4141908288002014,
+      "eval_runtime": 1965.9572,
+      "eval_samples_per_second": 0.526,
+      "step": 37500
+    },
+    {
+      "epoch": 34.98,
+      "learning_rate": 4.562686567164179e-05,
+      "loss": 0.0023,
+      "step": 37504
+    },
+    {
+      "epoch": 34.99,
+      "learning_rate": 4.5626399253731347e-05,
+      "loss": 0.0028,
+      "step": 37508
+    },
+    {
+      "epoch": 34.99,
+      "learning_rate": 4.56259328358209e-05,
+      "loss": 0.0001,
+      "step": 37512
+    },
+    {
+      "epoch": 35.0,
+      "learning_rate": 4.562546641791045e-05,
+      "loss": 0.0001,
+      "step": 37516
+    },
+    {
+      "epoch": 35.0,
+      "learning_rate": 4.5625e-05,
+      "loss": 0.0002,
+      "step": 37520
+    },
+    {
+      "epoch": 35.0,
+      "learning_rate": 4.562453358208956e-05,
+      "loss": 0.0001,
+      "step": 37524
+    },
+    {
+      "epoch": 35.01,
+      "learning_rate": 4.562406716417911e-05,
+      "loss": 0.0069,
+      "step": 37528
+    },
+    {
+      "epoch": 35.01,
+      "learning_rate": 4.5623600746268656e-05,
+      "loss": 0.0007,
+      "step": 37532
+    },
+    {
+      "epoch": 35.01,
+      "learning_rate": 4.562313432835821e-05,
+      "loss": 0.0001,
+      "step": 37536
+    },
+    {
+      "epoch": 35.02,
+      "learning_rate": 4.5622667910447765e-05,
+      "loss": 0.0001,
+      "step": 37540
+    },
+    {
+      "epoch": 35.02,
+      "learning_rate": 4.5622201492537314e-05,
+      "loss": 0.0001,
+      "step": 37544
+    },
+    {
+      "epoch": 35.03,
+      "learning_rate": 4.562173507462687e-05,
+      "loss": 0.0001,
+      "step": 37548
+    },
+    {
+      "epoch": 35.03,
+      "learning_rate": 4.5621268656716423e-05,
+      "loss": 0.0,
+      "step": 37552
+    },
+    {
+      "epoch": 35.03,
+      "learning_rate": 4.562080223880597e-05,
+      "loss": 0.0,
+      "step": 37556
+    },
+    {
+      "epoch": 35.04,
+      "learning_rate": 4.5620335820895526e-05,
+      "loss": 0.0,
+      "step": 37560
+    },
+    {
+      "epoch": 35.04,
+      "learning_rate": 4.5619869402985075e-05,
+      "loss": 0.0001,
+      "step": 37564
+    },
+    {
+      "epoch": 35.04,
+      "learning_rate": 4.561940298507463e-05,
+      "loss": 0.0009,
+      "step": 37568
+    },
+    {
+      "epoch": 35.05,
+      "learning_rate": 4.5618936567164184e-05,
+      "loss": 0.0,
+      "step": 37572
+    },
+    {
+      "epoch": 35.05,
+      "learning_rate": 4.561847014925373e-05,
+      "loss": 0.0001,
+      "step": 37576
+    },
+    {
+      "epoch": 35.06,
+      "learning_rate": 4.561800373134328e-05,
+      "loss": 0.0008,
+      "step": 37580
+    },
+    {
+      "epoch": 35.06,
+      "learning_rate": 4.561753731343284e-05,
+      "loss": 0.0001,
+      "step": 37584
+    },
+    {
+      "epoch": 35.06,
+      "learning_rate": 4.561707089552239e-05,
+      "loss": 0.0007,
+      "step": 37588
+    },
+    {
+      "epoch": 35.07,
+      "learning_rate": 4.561660447761194e-05,
+      "loss": 0.0001,
+      "step": 37592
+    },
+    {
+      "epoch": 35.07,
+      "learning_rate": 4.5616138059701494e-05,
+      "loss": 0.0002,
+      "step": 37596
+    },
+    {
+      "epoch": 35.07,
+      "learning_rate": 4.561567164179105e-05,
+      "loss": 0.0003,
+      "step": 37600
+    },
+    {
+      "epoch": 35.08,
+      "learning_rate": 4.5615205223880597e-05,
+      "loss": 0.0,
+      "step": 37604
+    },
+    {
+      "epoch": 35.08,
+      "learning_rate": 4.561473880597015e-05,
+      "loss": 0.0,
+      "step": 37608
+    },
+    {
+      "epoch": 35.09,
+      "learning_rate": 4.5614272388059706e-05,
+      "loss": 0.0,
+      "step": 37612
+    },
+    {
+      "epoch": 35.09,
+      "learning_rate": 4.5613805970149254e-05,
+      "loss": 0.0,
+      "step": 37616
+    },
+    {
+      "epoch": 35.09,
+      "learning_rate": 4.561333955223881e-05,
+      "loss": 0.0,
+      "step": 37620
+    },
+    {
+      "epoch": 35.1,
+      "learning_rate": 4.561287313432836e-05,
+      "loss": 0.0002,
+      "step": 37624
+    },
+    {
+      "epoch": 35.1,
+      "learning_rate": 4.561240671641791e-05,
+      "loss": 0.0,
+      "step": 37628
+    },
+    {
+      "epoch": 35.1,
+      "learning_rate": 4.561194029850747e-05,
+      "loss": 0.0,
+      "step": 37632
+    },
+    {
+      "epoch": 35.11,
+      "learning_rate": 4.5611473880597015e-05,
+      "loss": 0.0001,
+      "step": 37636
+    },
+    {
+      "epoch": 35.11,
+      "learning_rate": 4.5611007462686564e-05,
+      "loss": 0.0,
+      "step": 37640
+    },
+    {
+      "epoch": 35.12,
+      "learning_rate": 4.5610541044776125e-05,
+      "loss": 0.0,
+      "step": 37644
+    },
+    {
+      "epoch": 35.12,
+      "learning_rate": 4.5610074626865673e-05,
+      "loss": 0.0019,
+      "step": 37648
+    },
+    {
+      "epoch": 35.12,
+      "learning_rate": 4.560960820895522e-05,
+      "loss": 0.0007,
+      "step": 37652
+    },
+    {
+      "epoch": 35.13,
+      "learning_rate": 4.5609141791044776e-05,
+      "loss": 0.0001,
+      "step": 37656
+    },
+    {
+      "epoch": 35.13,
+      "learning_rate": 4.560867537313433e-05,
+      "loss": 0.0001,
+      "step": 37660
+    },
+    {
+      "epoch": 35.13,
+      "learning_rate": 4.560820895522388e-05,
+      "loss": 0.0,
+      "step": 37664
+    },
+    {
+      "epoch": 35.14,
+      "learning_rate": 4.5607742537313434e-05,
+      "loss": 0.0001,
+      "step": 37668
+    },
+    {
+      "epoch": 35.14,
+      "learning_rate": 4.560727611940299e-05,
+      "loss": 0.0004,
+      "step": 37672
+    },
+    {
+      "epoch": 35.15,
+      "learning_rate": 4.5606809701492544e-05,
+      "loss": 0.0002,
+      "step": 37676
+    },
+    {
+      "epoch": 35.15,
+      "learning_rate": 4.560634328358209e-05,
+      "loss": 0.0,
+      "step": 37680
+    },
+    {
+      "epoch": 35.15,
+      "learning_rate": 4.560587686567164e-05,
+      "loss": 0.0038,
+      "step": 37684
+    },
+    {
+      "epoch": 35.16,
+      "learning_rate": 4.5605410447761195e-05,
+      "loss": 0.0,
+      "step": 37688
+    },
+    {
+      "epoch": 35.16,
+      "learning_rate": 4.560494402985075e-05,
+      "loss": 0.0001,
+      "step": 37692
+    },
+    {
+      "epoch": 35.16,
+      "learning_rate": 4.56044776119403e-05,
+      "loss": 0.0,
+      "step": 37696
+    },
+    {
+      "epoch": 35.17,
+      "learning_rate": 4.560401119402985e-05,
+      "loss": 0.0001,
+      "step": 37700
+    },
+    {
+      "epoch": 35.17,
+      "learning_rate": 4.560354477611941e-05,
+      "loss": 0.0006,
+      "step": 37704
+    },
+    {
+      "epoch": 35.18,
+      "learning_rate": 4.5603078358208956e-05,
+      "loss": 0.0,
+      "step": 37708
+    },
+    {
+      "epoch": 35.18,
+      "learning_rate": 4.560261194029851e-05,
+      "loss": 0.0003,
+      "step": 37712
+    },
+    {
+      "epoch": 35.18,
+      "learning_rate": 4.560214552238806e-05,
+      "loss": 0.0001,
+      "step": 37716
+    },
+    {
+      "epoch": 35.19,
+      "learning_rate": 4.5601679104477614e-05,
+      "loss": 0.0001,
+      "step": 37720
+    },
+    {
+      "epoch": 35.19,
+      "learning_rate": 4.560121268656717e-05,
+      "loss": 0.0031,
+      "step": 37724
+    },
+    {
+      "epoch": 35.19,
+      "learning_rate": 4.560074626865672e-05,
+      "loss": 0.0024,
+      "step": 37728
+    },
+    {
+      "epoch": 35.2,
+      "learning_rate": 4.560027985074627e-05,
+      "loss": 0.0009,
+      "step": 37732
+    },
+    {
+      "epoch": 35.2,
+      "learning_rate": 4.559981343283583e-05,
+      "loss": 0.0,
+      "step": 37736
+    },
+    {
+      "epoch": 35.21,
+      "learning_rate": 4.5599347014925375e-05,
+      "loss": 0.0002,
+      "step": 37740
+    },
+    {
+      "epoch": 35.21,
+      "learning_rate": 4.5598880597014923e-05,
+      "loss": 0.0001,
+      "step": 37744
+    },
+    {
+      "epoch": 35.21,
+      "learning_rate": 4.559841417910448e-05,
+      "loss": 0.0006,
+      "step": 37748
+    },
+    {
+      "epoch": 35.22,
+      "learning_rate": 4.559794776119403e-05,
+      "loss": 0.0,
+      "step": 37752
+    },
+    {
+      "epoch": 35.22,
+      "learning_rate": 4.559748134328358e-05,
+      "loss": 0.0005,
+      "step": 37756
+    },
+    {
+      "epoch": 35.22,
+      "learning_rate": 4.5597014925373136e-05,
+      "loss": 0.0035,
+      "step": 37760
+    },
+    {
+      "epoch": 35.23,
+      "learning_rate": 4.559654850746269e-05,
+      "loss": 0.0001,
+      "step": 37764
+    },
+    {
+      "epoch": 35.23,
+      "learning_rate": 4.559608208955224e-05,
+      "loss": 0.0012,
+      "step": 37768
+    },
+    {
+      "epoch": 35.24,
+      "learning_rate": 4.5595615671641794e-05,
+      "loss": 0.0001,
+      "step": 37772
+    },
+    {
+      "epoch": 35.24,
+      "learning_rate": 4.559514925373134e-05,
+      "loss": 0.0022,
+      "step": 37776
+    },
+    {
+      "epoch": 35.24,
+      "learning_rate": 4.55946828358209e-05,
+      "loss": 0.0003,
+      "step": 37780
+    },
+    {
+      "epoch": 35.25,
+      "learning_rate": 4.559421641791045e-05,
+      "loss": 0.0002,
+      "step": 37784
+    },
+    {
+      "epoch": 35.25,
+      "learning_rate": 4.559375e-05,
+      "loss": 0.0003,
+      "step": 37788
+    },
+    {
+      "epoch": 35.25,
+      "learning_rate": 4.5593283582089555e-05,
+      "loss": 0.0008,
+      "step": 37792
+    },
+    {
+      "epoch": 35.26,
+      "learning_rate": 4.559281716417911e-05,
+      "loss": 0.0,
+      "step": 37796
+    },
+    {
+      "epoch": 35.26,
+      "learning_rate": 4.559235074626866e-05,
+      "loss": 0.0,
+      "step": 37800
+    },
+    {
+      "epoch": 35.26,
+      "learning_rate": 4.5591884328358206e-05,
+      "loss": 0.0,
+      "step": 37804
+    },
+    {
+      "epoch": 35.27,
+      "learning_rate": 4.559141791044776e-05,
+      "loss": 0.0002,
+      "step": 37808
+    },
+    {
+      "epoch": 35.27,
+      "learning_rate": 4.5590951492537316e-05,
+      "loss": 0.0001,
+      "step": 37812
+    },
+    {
+      "epoch": 35.28,
+      "learning_rate": 4.5590485074626864e-05,
+      "loss": 0.0009,
+      "step": 37816
+    },
+    {
+      "epoch": 35.28,
+      "learning_rate": 4.559001865671642e-05,
+      "loss": 0.0006,
+      "step": 37820
+    },
+    {
+      "epoch": 35.28,
+      "learning_rate": 4.5589552238805974e-05,
+      "loss": 0.0,
+      "step": 37824
+    },
+    {
+      "epoch": 35.29,
+      "learning_rate": 4.558908582089552e-05,
+      "loss": 0.0002,
+      "step": 37828
+    },
+    {
+      "epoch": 35.29,
+      "learning_rate": 4.558861940298508e-05,
+      "loss": 0.0002,
+      "step": 37832
+    },
+    {
+      "epoch": 35.29,
+      "learning_rate": 4.5588152985074625e-05,
+      "loss": 0.0002,
+      "step": 37836
+    },
+    {
+      "epoch": 35.3,
+      "learning_rate": 4.558768656716419e-05,
+      "loss": 0.0001,
+      "step": 37840
+    },
+    {
+      "epoch": 35.3,
+      "learning_rate": 4.5587220149253735e-05,
+      "loss": 0.0,
+      "step": 37844
+    },
+    {
+      "epoch": 35.31,
+      "learning_rate": 4.558675373134328e-05,
+      "loss": 0.0001,
+      "step": 37848
+    },
+    {
+      "epoch": 35.31,
+      "learning_rate": 4.558628731343284e-05,
+      "loss": 0.0001,
+      "step": 37852
+    },
+    {
+      "epoch": 35.31,
+      "learning_rate": 4.558582089552239e-05,
+      "loss": 0.0,
+      "step": 37856
+    },
+    {
+      "epoch": 35.32,
+      "learning_rate": 4.558535447761194e-05,
+      "loss": 0.0,
+      "step": 37860
+    },
+    {
+      "epoch": 35.32,
+      "learning_rate": 4.5584888059701496e-05,
+      "loss": 0.0001,
+      "step": 37864
+    },
+    {
+      "epoch": 35.32,
+      "learning_rate": 4.5584421641791044e-05,
+      "loss": 0.0,
+      "step": 37868
+    },
+    {
+      "epoch": 35.33,
+      "learning_rate": 4.55839552238806e-05,
+      "loss": 0.0004,
+      "step": 37872
+    },
+    {
+      "epoch": 35.33,
+      "learning_rate": 4.5583488805970154e-05,
+      "loss": 0.0001,
+      "step": 37876
+    },
+    {
+      "epoch": 35.34,
+      "learning_rate": 4.55830223880597e-05,
+      "loss": 0.0,
+      "step": 37880
+    },
+    {
+      "epoch": 35.34,
+      "learning_rate": 4.558255597014926e-05,
+      "loss": 0.0002,
+      "step": 37884
+    },
+    {
+      "epoch": 35.34,
+      "learning_rate": 4.558208955223881e-05,
+      "loss": 0.0,
+      "step": 37888
+    },
+    {
+      "epoch": 35.35,
+      "learning_rate": 4.558162313432836e-05,
+      "loss": 0.0,
+      "step": 37892
+    },
+    {
+      "epoch": 35.35,
+      "learning_rate": 4.558115671641791e-05,
+      "loss": 0.0001,
+      "step": 37896
+    },
+    {
+      "epoch": 35.35,
+      "learning_rate": 4.558069029850747e-05,
+      "loss": 0.0,
+      "step": 37900
+    },
+    {
+      "epoch": 35.36,
+      "learning_rate": 4.558022388059702e-05,
+      "loss": 0.0001,
+      "step": 37904
+    },
+    {
+      "epoch": 35.36,
+      "learning_rate": 4.5579757462686566e-05,
+      "loss": 0.0,
+      "step": 37908
+    },
+    {
+      "epoch": 35.37,
+      "learning_rate": 4.557929104477612e-05,
+      "loss": 0.0,
+      "step": 37912
+    },
+    {
+      "epoch": 35.37,
+      "learning_rate": 4.5578824626865676e-05,
+      "loss": 0.0004,
+      "step": 37916
+    },
+    {
+      "epoch": 35.37,
+      "learning_rate": 4.5578358208955224e-05,
+      "loss": 0.0003,
+      "step": 37920
+    },
+    {
+      "epoch": 35.38,
+      "learning_rate": 4.557789179104478e-05,
+      "loss": 0.0082,
+      "step": 37924
+    },
+    {
+      "epoch": 35.38,
+      "learning_rate": 4.557742537313433e-05,
+      "loss": 0.0001,
+      "step": 37928
+    },
+    {
+      "epoch": 35.38,
+      "learning_rate": 4.557695895522388e-05,
+      "loss": 0.0007,
+      "step": 37932
+    },
+    {
+      "epoch": 35.39,
+      "learning_rate": 4.557649253731344e-05,
+      "loss": 0.0001,
+      "step": 37936
+    },
+    {
+      "epoch": 35.39,
+      "learning_rate": 4.5576026119402985e-05,
+      "loss": 0.0013,
+      "step": 37940
+    },
+    {
+      "epoch": 35.4,
+      "learning_rate": 4.557555970149254e-05,
+      "loss": 0.0005,
+      "step": 37944
+    },
+    {
+      "epoch": 35.4,
+      "learning_rate": 4.5575093283582095e-05,
+      "loss": 0.0002,
+      "step": 37948
+    },
+    {
+      "epoch": 35.4,
+      "learning_rate": 4.557462686567164e-05,
+      "loss": 0.0001,
+      "step": 37952
+    },
+    {
+      "epoch": 35.41,
+      "learning_rate": 4.557416044776119e-05,
+      "loss": 0.0,
+      "step": 37956
+    },
+    {
+      "epoch": 35.41,
+      "learning_rate": 4.557369402985075e-05,
+      "loss": 0.0006,
+      "step": 37960
+    },
+    {
+      "epoch": 35.41,
+      "learning_rate": 4.55732276119403e-05,
+      "loss": 0.0,
+      "step": 37964
+    },
+    {
+      "epoch": 35.42,
+      "learning_rate": 4.557276119402985e-05,
+      "loss": 0.0001,
+      "step": 37968
+    },
+    {
+      "epoch": 35.42,
+      "learning_rate": 4.5572294776119404e-05,
+      "loss": 0.0,
+      "step": 37972
+    },
+    {
+      "epoch": 35.43,
+      "learning_rate": 4.557182835820896e-05,
+      "loss": 0.0014,
+      "step": 37976
+    },
+    {
+      "epoch": 35.43,
+      "learning_rate": 4.557136194029851e-05,
+      "loss": 0.0,
+      "step": 37980
+    },
+    {
+      "epoch": 35.43,
+      "learning_rate": 4.557089552238806e-05,
+      "loss": 0.0019,
+      "step": 37984
+    },
+    {
+      "epoch": 35.44,
+      "learning_rate": 4.557042910447761e-05,
+      "loss": 0.0021,
+      "step": 37988
+    },
+    {
+      "epoch": 35.44,
+      "learning_rate": 4.5569962686567165e-05,
+      "loss": 0.0,
+      "step": 37992
+    },
+    {
+      "epoch": 35.44,
+      "learning_rate": 4.556949626865672e-05,
+      "loss": 0.0,
+      "step": 37996
+    },
+    {
+      "epoch": 35.45,
+      "learning_rate": 4.556902985074627e-05,
+      "loss": 0.0001,
+      "step": 38000
+    },
+    {
+      "epoch": 35.45,
+      "eval_exact_match": 0.7427466150870407,
+      "eval_exec": 0.7717601547388782,
+      "eval_loss": 0.43170931935310364,
+      "eval_runtime": 1115.1121,
+      "eval_samples_per_second": 0.927,
+      "step": 38000
+    },
+    {
+      "epoch": 35.45,
+      "learning_rate": 4.556856343283582e-05,
+      "loss": 0.001,
+      "step": 38004
+    },
+    {
+      "epoch": 35.46,
+      "learning_rate": 4.556809701492538e-05,
+      "loss": 0.0019,
+      "step": 38008
+    },
+    {
+      "epoch": 35.46,
+      "learning_rate": 4.5567630597014926e-05,
+      "loss": 0.001,
+      "step": 38012
+    },
+    {
+      "epoch": 35.46,
+      "learning_rate": 4.556716417910448e-05,
+      "loss": 0.0006,
+      "step": 38016
+    },
+    {
+      "epoch": 35.47,
+      "learning_rate": 4.5566697761194036e-05,
+      "loss": 0.0,
+      "step": 38020
+    },
+    {
+      "epoch": 35.47,
+      "learning_rate": 4.5566231343283584e-05,
+      "loss": 0.0018,
+      "step": 38024
+    },
+    {
+      "epoch": 35.47,
+      "learning_rate": 4.556576492537314e-05,
+      "loss": 0.0001,
+      "step": 38028
+    },
+    {
+      "epoch": 35.48,
+      "learning_rate": 4.556529850746269e-05,
+      "loss": 0.0003,
+      "step": 38032
+    },
+    {
+      "epoch": 35.48,
+      "learning_rate": 4.556483208955224e-05,
+      "loss": 0.0003,
+      "step": 38036
+    },
+    {
+      "epoch": 35.49,
+      "learning_rate": 4.55643656716418e-05,
+      "loss": 0.0,
+      "step": 38040
+    },
+    {
+      "epoch": 35.49,
+      "learning_rate": 4.5563899253731345e-05,
+      "loss": 0.0,
+      "step": 38044
+    },
+    {
+      "epoch": 35.49,
+      "learning_rate": 4.556343283582089e-05,
+      "loss": 0.0,
+      "step": 38048
+    },
+    {
+      "epoch": 35.5,
+      "learning_rate": 4.5562966417910455e-05,
+      "loss": 0.0004,
+      "step": 38052
+    },
+    {
+      "epoch": 35.5,
+      "learning_rate": 4.55625e-05,
+      "loss": 0.0002,
+      "step": 38056
+    },
+    {
+      "epoch": 35.5,
+      "learning_rate": 4.556203358208955e-05,
+      "loss": 0.0001,
+      "step": 38060
+    },
+    {
+      "epoch": 35.51,
+      "learning_rate": 4.5561567164179106e-05,
+      "loss": 0.0,
+      "step": 38064
+    },
+    {
+      "epoch": 35.51,
+      "learning_rate": 4.556110074626866e-05,
+      "loss": 0.0002,
+      "step": 38068
+    },
+    {
+      "epoch": 35.51,
+      "learning_rate": 4.556063432835821e-05,
+      "loss": 0.0,
+      "step": 38072
+    },
+    {
+      "epoch": 35.52,
+      "learning_rate": 4.5560167910447764e-05,
+      "loss": 0.0011,
+      "step": 38076
+    },
+    {
+      "epoch": 35.52,
+      "learning_rate": 4.555970149253732e-05,
+      "loss": 0.0,
+      "step": 38080
+    },
+    {
+      "epoch": 35.53,
+      "learning_rate": 4.555923507462687e-05,
+      "loss": 0.0002,
+      "step": 38084
+    },
+    {
+      "epoch": 35.53,
+      "learning_rate": 4.555876865671642e-05,
+      "loss": 0.0001,
+      "step": 38088
+    },
+    {
+      "epoch": 35.53,
+      "learning_rate": 4.555830223880597e-05,
+      "loss": 0.0004,
+      "step": 38092
+    },
+    {
+      "epoch": 35.54,
+      "learning_rate": 4.5557835820895525e-05,
+      "loss": 0.0011,
+      "step": 38096
+    },
+    {
+      "epoch": 35.54,
+      "learning_rate": 4.555736940298508e-05,
+      "loss": 0.0,
+      "step": 38100
+    },
+    {
+      "epoch": 35.54,
+      "learning_rate": 4.555690298507463e-05,
+      "loss": 0.0001,
+      "step": 38104
+    },
+    {
+      "epoch": 35.55,
+      "learning_rate": 4.5556436567164176e-05,
+      "loss": 0.0011,
+      "step": 38108
+    },
+    {
+      "epoch": 35.55,
+      "learning_rate": 4.555597014925374e-05,
+      "loss": 0.0001,
+      "step": 38112
+    },
+    {
+      "epoch": 35.56,
+      "learning_rate": 4.5555503731343286e-05,
+      "loss": 0.0,
+      "step": 38116
+    },
+    {
+      "epoch": 35.56,
+      "learning_rate": 4.5555037313432834e-05,
+      "loss": 0.0002,
+      "step": 38120
+    },
+    {
+      "epoch": 35.56,
+      "learning_rate": 4.555457089552239e-05,
+      "loss": 0.0009,
+      "step": 38124
+    },
+    {
+      "epoch": 35.57,
+      "learning_rate": 4.5554104477611944e-05,
+      "loss": 0.0007,
+      "step": 38128
+    },
+    {
+      "epoch": 35.57,
+      "learning_rate": 4.555363805970149e-05,
+      "loss": 0.0001,
+      "step": 38132
+    },
+    {
+      "epoch": 35.57,
+      "learning_rate": 4.555317164179105e-05,
+      "loss": 0.0,
+      "step": 38136
+    },
+    {
+      "epoch": 35.58,
+      "learning_rate": 4.55527052238806e-05,
+      "loss": 0.0003,
+      "step": 38140
+    },
+    {
+      "epoch": 35.58,
+      "learning_rate": 4.555223880597015e-05,
+      "loss": 0.0007,
+      "step": 38144
+    },
+    {
+      "epoch": 35.59,
+      "learning_rate": 4.5551772388059705e-05,
+      "loss": 0.0008,
+      "step": 38148
+    },
+    {
+      "epoch": 35.59,
+      "learning_rate": 4.555130597014925e-05,
+      "loss": 0.0,
+      "step": 38152
+    },
+    {
+      "epoch": 35.59,
+      "learning_rate": 4.555083955223881e-05,
+      "loss": 0.0138,
+      "step": 38156
+    },
+    {
+      "epoch": 35.6,
+      "learning_rate": 4.555037313432836e-05,
+      "loss": 0.0019,
+      "step": 38160
+    },
+    {
+      "epoch": 35.6,
+      "learning_rate": 4.554990671641791e-05,
+      "loss": 0.0,
+      "step": 38164
+    },
+    {
+      "epoch": 35.6,
+      "learning_rate": 4.5549440298507466e-05,
+      "loss": 0.0026,
+      "step": 38168
+    },
+    {
+      "epoch": 35.61,
+      "learning_rate": 4.554897388059702e-05,
+      "loss": 0.0,
+      "step": 38172
+    },
+    {
+      "epoch": 35.61,
+      "learning_rate": 4.554850746268657e-05,
+      "loss": 0.0,
+      "step": 38176
+    },
+    {
+      "epoch": 35.62,
+      "learning_rate": 4.5548041044776124e-05,
+      "loss": 0.0003,
+      "step": 38180
+    },
+    {
+      "epoch": 35.62,
+      "learning_rate": 4.554757462686567e-05,
+      "loss": 0.0006,
+      "step": 38184
+    },
+    {
+      "epoch": 35.62,
+      "learning_rate": 4.554710820895523e-05,
+      "loss": 0.0007,
+      "step": 38188
+    },
+    {
+      "epoch": 35.63,
+      "learning_rate": 4.554664179104478e-05,
+      "loss": 0.001,
+      "step": 38192
+    },
+    {
+      "epoch": 35.63,
+      "learning_rate": 4.554617537313433e-05,
+      "loss": 0.0048,
+      "step": 38196
+    },
+    {
+      "epoch": 35.63,
+      "learning_rate": 4.554570895522388e-05,
+      "loss": 0.0004,
+      "step": 38200
+    },
+    {
+      "epoch": 35.64,
+      "learning_rate": 4.554524253731344e-05,
+      "loss": 0.0,
+      "step": 38204
+    },
+    {
+      "epoch": 35.64,
+      "learning_rate": 4.554477611940299e-05,
+      "loss": 0.0025,
+      "step": 38208
+    },
+    {
+      "epoch": 35.65,
+      "learning_rate": 4.5544309701492536e-05,
+      "loss": 0.0042,
+      "step": 38212
+    },
+    {
+      "epoch": 35.65,
+      "learning_rate": 4.554384328358209e-05,
+      "loss": 0.0001,
+      "step": 38216
+    },
+    {
+      "epoch": 35.65,
+      "learning_rate": 4.5543376865671646e-05,
+      "loss": 0.0001,
+      "step": 38220
+    },
+    {
+      "epoch": 35.66,
+      "learning_rate": 4.5542910447761194e-05,
+      "loss": 0.0016,
+      "step": 38224
+    },
+    {
+      "epoch": 35.66,
+      "learning_rate": 4.554244402985075e-05,
+      "loss": 0.0008,
+      "step": 38228
+    },
+    {
+      "epoch": 35.66,
+      "learning_rate": 4.5541977611940304e-05,
+      "loss": 0.0003,
+      "step": 38232
+    },
+    {
+      "epoch": 35.67,
+      "learning_rate": 4.554151119402985e-05,
+      "loss": 0.0009,
+      "step": 38236
+    },
+    {
+      "epoch": 35.67,
+      "learning_rate": 4.554104477611941e-05,
+      "loss": 0.0,
+      "step": 38240
+    },
+    {
+      "epoch": 35.68,
+      "learning_rate": 4.5540578358208955e-05,
+      "loss": 0.0,
+      "step": 38244
+    },
+    {
+      "epoch": 35.68,
+      "learning_rate": 4.554011194029851e-05,
+      "loss": 0.0001,
+      "step": 38248
+    },
+    {
+      "epoch": 35.68,
+      "learning_rate": 4.5539645522388065e-05,
+      "loss": 0.0003,
+      "step": 38252
+    },
+    {
+      "epoch": 35.69,
+      "learning_rate": 4.553917910447761e-05,
+      "loss": 0.0004,
+      "step": 38256
+    },
+    {
+      "epoch": 35.69,
+      "learning_rate": 4.553871268656716e-05,
+      "loss": 0.0001,
+      "step": 38260
+    },
+    {
+      "epoch": 35.69,
+      "learning_rate": 4.553824626865672e-05,
+      "loss": 0.0,
+      "step": 38264
+    },
+    {
+      "epoch": 35.7,
+      "learning_rate": 4.553777985074627e-05,
+      "loss": 0.0001,
+      "step": 38268
+    },
+    {
+      "epoch": 35.7,
+      "learning_rate": 4.553731343283582e-05,
+      "loss": 0.0001,
+      "step": 38272
+    },
+    {
+      "epoch": 35.71,
+      "learning_rate": 4.5536847014925374e-05,
+      "loss": 0.0,
+      "step": 38276
+    },
+    {
+      "epoch": 35.71,
+      "learning_rate": 4.553638059701493e-05,
+      "loss": 0.0005,
+      "step": 38280
+    },
+    {
+      "epoch": 35.71,
+      "learning_rate": 4.553591417910448e-05,
+      "loss": 0.0001,
+      "step": 38284
+    },
+    {
+      "epoch": 35.72,
+      "learning_rate": 4.553544776119403e-05,
+      "loss": 0.0013,
+      "step": 38288
+    },
+    {
+      "epoch": 35.72,
+      "learning_rate": 4.553498134328359e-05,
+      "loss": 0.0002,
+      "step": 38292
+    },
+    {
+      "epoch": 35.72,
+      "learning_rate": 4.5534514925373135e-05,
+      "loss": 0.0001,
+      "step": 38296
+    },
+    {
+      "epoch": 35.73,
+      "learning_rate": 4.553404850746269e-05,
+      "loss": 0.0001,
+      "step": 38300
+    },
+    {
+      "epoch": 35.73,
+      "learning_rate": 4.553358208955224e-05,
+      "loss": 0.0,
+      "step": 38304
+    },
+    {
+      "epoch": 35.73,
+      "learning_rate": 4.553311567164179e-05,
+      "loss": 0.0002,
+      "step": 38308
+    },
+    {
+      "epoch": 35.74,
+      "learning_rate": 4.553264925373135e-05,
+      "loss": 0.0036,
+      "step": 38312
+    },
+    {
+      "epoch": 35.74,
+      "learning_rate": 4.5532182835820896e-05,
+      "loss": 0.0001,
+      "step": 38316
+    },
+    {
+      "epoch": 35.75,
+      "learning_rate": 4.5531716417910444e-05,
+      "loss": 0.0002,
+      "step": 38320
+    },
+    {
+      "epoch": 35.75,
+      "learning_rate": 4.5531250000000006e-05,
+      "loss": 0.0003,
+      "step": 38324
+    },
+    {
+      "epoch": 35.75,
+      "learning_rate": 4.5530783582089554e-05,
+      "loss": 0.0,
+      "step": 38328
+    },
+    {
+      "epoch": 35.76,
+      "learning_rate": 4.553031716417911e-05,
+      "loss": 0.0006,
+      "step": 38332
+    },
+    {
+      "epoch": 35.76,
+      "learning_rate": 4.552985074626866e-05,
+      "loss": 0.0,
+      "step": 38336
+    },
+    {
+      "epoch": 35.76,
+      "learning_rate": 4.552938432835821e-05,
+      "loss": 0.0001,
+      "step": 38340
+    },
+    {
+      "epoch": 35.77,
+      "learning_rate": 4.5528917910447767e-05,
+      "loss": 0.0,
+      "step": 38344
+    },
+    {
+      "epoch": 35.77,
+      "learning_rate": 4.5528451492537315e-05,
+      "loss": 0.0028,
+      "step": 38348
+    },
+    {
+      "epoch": 35.78,
+      "learning_rate": 4.552798507462687e-05,
+      "loss": 0.0001,
+      "step": 38352
+    },
+    {
+      "epoch": 35.78,
+      "learning_rate": 4.5527518656716425e-05,
+      "loss": 0.0001,
+      "step": 38356
+    },
+    {
+      "epoch": 35.78,
+      "learning_rate": 4.552705223880597e-05,
+      "loss": 0.0007,
+      "step": 38360
+    },
+    {
+      "epoch": 35.79,
+      "learning_rate": 4.552658582089552e-05,
+      "loss": 0.0001,
+      "step": 38364
+    },
+    {
+      "epoch": 35.79,
+      "learning_rate": 4.5526119402985076e-05,
+      "loss": 0.0001,
+      "step": 38368
+    },
+    {
+      "epoch": 35.79,
+      "learning_rate": 4.552565298507463e-05,
+      "loss": 0.0001,
+      "step": 38372
+    },
+    {
+      "epoch": 35.8,
+      "learning_rate": 4.552518656716418e-05,
+      "loss": 0.0002,
+      "step": 38376
+    },
+    {
+      "epoch": 35.8,
+      "learning_rate": 4.5524720149253734e-05,
+      "loss": 0.0034,
+      "step": 38380
+    },
+    {
+      "epoch": 35.81,
+      "learning_rate": 4.552425373134329e-05,
+      "loss": 0.0,
+      "step": 38384
+    },
+    {
+      "epoch": 35.81,
+      "learning_rate": 4.552378731343284e-05,
+      "loss": 0.0005,
+      "step": 38388
+    },
+    {
+      "epoch": 35.81,
+      "learning_rate": 4.552332089552239e-05,
+      "loss": 0.0,
+      "step": 38392
+    },
+    {
+      "epoch": 35.82,
+      "learning_rate": 4.552285447761194e-05,
+      "loss": 0.0001,
+      "step": 38396
+    },
+    {
+      "epoch": 35.82,
+      "learning_rate": 4.5522388059701495e-05,
+      "loss": 0.0,
+      "step": 38400
+    },
+    {
+      "epoch": 35.82,
+      "learning_rate": 4.552192164179105e-05,
+      "loss": 0.0,
+      "step": 38404
+    },
+    {
+      "epoch": 35.83,
+      "learning_rate": 4.55214552238806e-05,
+      "loss": 0.0,
+      "step": 38408
+    },
+    {
+      "epoch": 35.83,
+      "learning_rate": 4.552098880597015e-05,
+      "loss": 0.0026,
+      "step": 38412
+    },
+    {
+      "epoch": 35.84,
+      "learning_rate": 4.552052238805971e-05,
+      "loss": 0.0008,
+      "step": 38416
+    },
+    {
+      "epoch": 35.84,
+      "learning_rate": 4.5520055970149256e-05,
+      "loss": 0.0,
+      "step": 38420
+    },
+    {
+      "epoch": 35.84,
+      "learning_rate": 4.5519589552238804e-05,
+      "loss": 0.0,
+      "step": 38424
+    },
+    {
+      "epoch": 35.85,
+      "learning_rate": 4.551912313432836e-05,
+      "loss": 0.0,
+      "step": 38428
+    },
+    {
+      "epoch": 35.85,
+      "learning_rate": 4.5518656716417914e-05,
+      "loss": 0.0002,
+      "step": 38432
+    },
+    {
+      "epoch": 35.85,
+      "learning_rate": 4.551819029850746e-05,
+      "loss": 0.0,
+      "step": 38436
+    },
+    {
+      "epoch": 35.86,
+      "learning_rate": 4.5517723880597017e-05,
+      "loss": 0.0011,
+      "step": 38440
+    },
+    {
+      "epoch": 35.86,
+      "learning_rate": 4.551725746268657e-05,
+      "loss": 0.0001,
+      "step": 38444
+    },
+    {
+      "epoch": 35.87,
+      "learning_rate": 4.551679104477612e-05,
+      "loss": 0.0,
+      "step": 38448
+    },
+    {
+      "epoch": 35.87,
+      "learning_rate": 4.5516324626865675e-05,
+      "loss": 0.0001,
+      "step": 38452
+    },
+    {
+      "epoch": 35.87,
+      "learning_rate": 4.551585820895522e-05,
+      "loss": 0.0003,
+      "step": 38456
+    },
+    {
+      "epoch": 35.88,
+      "learning_rate": 4.551539179104478e-05,
+      "loss": 0.0002,
+      "step": 38460
+    },
+    {
+      "epoch": 35.88,
+      "learning_rate": 4.551492537313433e-05,
+      "loss": 0.0004,
+      "step": 38464
+    },
+    {
+      "epoch": 35.88,
+      "learning_rate": 4.551445895522388e-05,
+      "loss": 0.0,
+      "step": 38468
+    },
+    {
+      "epoch": 35.89,
+      "learning_rate": 4.5513992537313436e-05,
+      "loss": 0.0001,
+      "step": 38472
+    },
+    {
+      "epoch": 35.89,
+      "learning_rate": 4.551352611940299e-05,
+      "loss": 0.0,
+      "step": 38476
+    },
+    {
+      "epoch": 35.9,
+      "learning_rate": 4.551305970149254e-05,
+      "loss": 0.0,
+      "step": 38480
+    },
+    {
+      "epoch": 35.9,
+      "learning_rate": 4.551259328358209e-05,
+      "loss": 0.0001,
+      "step": 38484
+    },
+    {
+      "epoch": 35.9,
+      "learning_rate": 4.551212686567164e-05,
+      "loss": 0.0001,
+      "step": 38488
+    },
+    {
+      "epoch": 35.91,
+      "learning_rate": 4.5511660447761197e-05,
+      "loss": 0.0011,
+      "step": 38492
+    },
+    {
+      "epoch": 35.91,
+      "learning_rate": 4.551119402985075e-05,
+      "loss": 0.0003,
+      "step": 38496
+    },
+    {
+      "epoch": 35.91,
+      "learning_rate": 4.55107276119403e-05,
+      "loss": 0.0001,
+      "step": 38500
+    },
+    {
+      "epoch": 35.91,
+      "eval_exact_match": 0.7543520309477756,
+      "eval_exec": 0.7794970986460348,
+      "eval_loss": 0.45771145820617676,
+      "eval_runtime": 1150.8415,
+      "eval_samples_per_second": 0.898,
+      "step": 38500
+    },
+    {
+      "epoch": 35.92,
+      "learning_rate": 4.5510261194029854e-05,
+      "loss": 0.0,
+      "step": 38504
+    },
+    {
+      "epoch": 35.92,
+      "learning_rate": 4.550979477611941e-05,
+      "loss": 0.0,
+      "step": 38508
+    },
+    {
+      "epoch": 35.93,
+      "learning_rate": 4.550932835820896e-05,
+      "loss": 0.0005,
+      "step": 38512
+    },
+    {
+      "epoch": 35.93,
+      "learning_rate": 4.5508861940298506e-05,
+      "loss": 0.0,
+      "step": 38516
+    },
+    {
+      "epoch": 35.93,
+      "learning_rate": 4.550839552238807e-05,
+      "loss": 0.0001,
+      "step": 38520
+    },
+    {
+      "epoch": 35.94,
+      "learning_rate": 4.5507929104477615e-05,
+      "loss": 0.0026,
+      "step": 38524
+    },
+    {
+      "epoch": 35.94,
+      "learning_rate": 4.5507462686567164e-05,
+      "loss": 0.0004,
+      "step": 38528
+    },
+    {
+      "epoch": 35.94,
+      "learning_rate": 4.550699626865672e-05,
+      "loss": 0.0002,
+      "step": 38532
+    },
+    {
+      "epoch": 35.95,
+      "learning_rate": 4.550652985074627e-05,
+      "loss": 0.0,
+      "step": 38536
+    },
+    {
+      "epoch": 35.95,
+      "learning_rate": 4.550606343283582e-05,
+      "loss": 0.0,
+      "step": 38540
+    },
+    {
+      "epoch": 35.96,
+      "learning_rate": 4.5505597014925376e-05,
+      "loss": 0.0015,
+      "step": 38544
+    },
+    {
+      "epoch": 35.96,
+      "learning_rate": 4.5505130597014925e-05,
+      "loss": 0.0002,
+      "step": 38548
+    },
+    {
+      "epoch": 35.96,
+      "learning_rate": 4.550466417910448e-05,
+      "loss": 0.0001,
+      "step": 38552
+    },
+    {
+      "epoch": 35.97,
+      "learning_rate": 4.5504197761194034e-05,
+      "loss": 0.0001,
+      "step": 38556
+    },
+    {
+      "epoch": 35.97,
+      "learning_rate": 4.550373134328358e-05,
+      "loss": 0.0,
+      "step": 38560
+    },
+    {
+      "epoch": 35.97,
+      "learning_rate": 4.550326492537314e-05,
+      "loss": 0.0,
+      "step": 38564
+    },
+    {
+      "epoch": 35.98,
+      "learning_rate": 4.550279850746269e-05,
+      "loss": 0.0002,
+      "step": 38568
+    },
+    {
+      "epoch": 35.98,
+      "learning_rate": 4.550233208955224e-05,
+      "loss": 0.0,
+      "step": 38572
+    },
+    {
+      "epoch": 35.98,
+      "learning_rate": 4.550186567164179e-05,
+      "loss": 0.0,
+      "step": 38576
+    },
+    {
+      "epoch": 35.99,
+      "learning_rate": 4.550139925373135e-05,
+      "loss": 0.0055,
+      "step": 38580
+    },
+    {
+      "epoch": 35.99,
+      "learning_rate": 4.55009328358209e-05,
+      "loss": 0.0001,
+      "step": 38584
+    },
+    {
+      "epoch": 36.0,
+      "learning_rate": 4.5500466417910447e-05,
+      "loss": 0.0002,
+      "step": 38588
+    },
+    {
+      "epoch": 36.0,
+      "learning_rate": 4.55e-05,
+      "loss": 0.0001,
+      "step": 38592
+    },
+    {
+      "epoch": 36.0,
+      "learning_rate": 4.5499533582089556e-05,
+      "loss": 0.0011,
+      "step": 38596
+    },
+    {
+      "epoch": 36.01,
+      "learning_rate": 4.5499067164179104e-05,
+      "loss": 0.0008,
+      "step": 38600
+    },
+    {
+      "epoch": 36.01,
+      "learning_rate": 4.549860074626866e-05,
+      "loss": 0.0,
+      "step": 38604
+    },
+    {
+      "epoch": 36.01,
+      "learning_rate": 4.549813432835821e-05,
+      "loss": 0.0,
+      "step": 38608
+    },
+    {
+      "epoch": 36.02,
+      "learning_rate": 4.549766791044776e-05,
+      "loss": 0.0,
+      "step": 38612
+    },
+    {
+      "epoch": 36.02,
+      "learning_rate": 4.549720149253732e-05,
+      "loss": 0.0001,
+      "step": 38616
+    },
+    {
+      "epoch": 36.03,
+      "learning_rate": 4.5496735074626865e-05,
+      "loss": 0.0002,
+      "step": 38620
+    },
+    {
+      "epoch": 36.03,
+      "learning_rate": 4.549626865671642e-05,
+      "loss": 0.0,
+      "step": 38624
+    },
+    {
+      "epoch": 36.03,
+      "learning_rate": 4.5495802238805975e-05,
+      "loss": 0.0008,
+      "step": 38628
+    },
+    {
+      "epoch": 36.04,
+      "learning_rate": 4.5495335820895523e-05,
+      "loss": 0.0006,
+      "step": 38632
+    },
+    {
+      "epoch": 36.04,
+      "learning_rate": 4.549486940298507e-05,
+      "loss": 0.0002,
+      "step": 38636
+    },
+    {
+      "epoch": 36.04,
+      "learning_rate": 4.549440298507463e-05,
+      "loss": 0.0,
+      "step": 38640
+    },
+    {
+      "epoch": 36.05,
+      "learning_rate": 4.549393656716418e-05,
+      "loss": 0.0,
+      "step": 38644
+    },
+    {
+      "epoch": 36.05,
+      "learning_rate": 4.5493470149253736e-05,
+      "loss": 0.0001,
+      "step": 38648
+    },
+    {
+      "epoch": 36.06,
+      "learning_rate": 4.5493003731343284e-05,
+      "loss": 0.0,
+      "step": 38652
+    },
+    {
+      "epoch": 36.06,
+      "learning_rate": 4.549253731343284e-05,
+      "loss": 0.0,
+      "step": 38656
+    },
+    {
+      "epoch": 36.06,
+      "learning_rate": 4.5492070895522394e-05,
+      "loss": 0.0,
+      "step": 38660
+    },
+    {
+      "epoch": 36.07,
+      "learning_rate": 4.549160447761194e-05,
+      "loss": 0.0006,
+      "step": 38664
+    },
+    {
+      "epoch": 36.07,
+      "learning_rate": 4.549113805970149e-05,
+      "loss": 0.0,
+      "step": 38668
+    },
+    {
+      "epoch": 36.07,
+      "learning_rate": 4.549067164179105e-05,
+      "loss": 0.0003,
+      "step": 38672
+    },
+    {
+      "epoch": 36.08,
+      "learning_rate": 4.54902052238806e-05,
+      "loss": 0.0006,
+      "step": 38676
+    },
+    {
+      "epoch": 36.08,
+      "learning_rate": 4.548973880597015e-05,
+      "loss": 0.0,
+      "step": 38680
+    },
+    {
+      "epoch": 36.09,
+      "learning_rate": 4.54892723880597e-05,
+      "loss": 0.0,
+      "step": 38684
+    },
+    {
+      "epoch": 36.09,
+      "learning_rate": 4.548880597014926e-05,
+      "loss": 0.001,
+      "step": 38688
+    },
+    {
+      "epoch": 36.09,
+      "learning_rate": 4.5488339552238806e-05,
+      "loss": 0.0001,
+      "step": 38692
+    },
+    {
+      "epoch": 36.1,
+      "learning_rate": 4.548787313432836e-05,
+      "loss": 0.0,
+      "step": 38696
+    },
+    {
+      "epoch": 36.1,
+      "learning_rate": 4.5487406716417916e-05,
+      "loss": 0.0014,
+      "step": 38700
+    },
+    {
+      "epoch": 36.1,
+      "learning_rate": 4.5486940298507464e-05,
+      "loss": 0.0001,
+      "step": 38704
+    },
+    {
+      "epoch": 36.11,
+      "learning_rate": 4.548647388059702e-05,
+      "loss": 0.0002,
+      "step": 38708
+    },
+    {
+      "epoch": 36.11,
+      "learning_rate": 4.548600746268657e-05,
+      "loss": 0.0002,
+      "step": 38712
+    },
+    {
+      "epoch": 36.12,
+      "learning_rate": 4.548554104477612e-05,
+      "loss": 0.0001,
+      "step": 38716
+    },
+    {
+      "epoch": 36.12,
+      "learning_rate": 4.548507462686568e-05,
+      "loss": 0.0017,
+      "step": 38720
+    },
+    {
+      "epoch": 36.12,
+      "learning_rate": 4.5484608208955225e-05,
+      "loss": 0.0003,
+      "step": 38724
+    },
+    {
+      "epoch": 36.13,
+      "learning_rate": 4.5484141791044773e-05,
+      "loss": 0.0176,
+      "step": 38728
+    },
+    {
+      "epoch": 36.13,
+      "learning_rate": 4.5483675373134335e-05,
+      "loss": 0.0012,
+      "step": 38732
+    },
+    {
+      "epoch": 36.13,
+      "learning_rate": 4.548320895522388e-05,
+      "loss": 0.0001,
+      "step": 38736
+    },
+    {
+      "epoch": 36.14,
+      "learning_rate": 4.548274253731343e-05,
+      "loss": 0.0043,
+      "step": 38740
+    },
+    {
+      "epoch": 36.14,
+      "learning_rate": 4.5482276119402986e-05,
+      "loss": 0.0004,
+      "step": 38744
+    },
+    {
+      "epoch": 36.15,
+      "learning_rate": 4.548180970149254e-05,
+      "loss": 0.0002,
+      "step": 38748
+    },
+    {
+      "epoch": 36.15,
+      "learning_rate": 4.548134328358209e-05,
+      "loss": 0.0001,
+      "step": 38752
+    },
+    {
+      "epoch": 36.15,
+      "learning_rate": 4.5480876865671644e-05,
+      "loss": 0.0,
+      "step": 38756
+    },
+    {
+      "epoch": 36.16,
+      "learning_rate": 4.54804104477612e-05,
+      "loss": 0.0,
+      "step": 38760
+    },
+    {
+      "epoch": 36.16,
+      "learning_rate": 4.547994402985075e-05,
+      "loss": 0.0,
+      "step": 38764
+    },
+    {
+      "epoch": 36.16,
+      "learning_rate": 4.54794776119403e-05,
+      "loss": 0.0251,
+      "step": 38768
+    },
+    {
+      "epoch": 36.17,
+      "learning_rate": 4.547901119402985e-05,
+      "loss": 0.001,
+      "step": 38772
+    },
+    {
+      "epoch": 36.17,
+      "learning_rate": 4.5478544776119405e-05,
+      "loss": 0.0011,
+      "step": 38776
+    },
+    {
+      "epoch": 36.18,
+      "learning_rate": 4.547807835820896e-05,
+      "loss": 0.0,
+      "step": 38780
+    },
+    {
+      "epoch": 36.18,
+      "learning_rate": 4.547761194029851e-05,
+      "loss": 0.0,
+      "step": 38784
+    },
+    {
+      "epoch": 36.18,
+      "learning_rate": 4.5477145522388056e-05,
+      "loss": 0.0001,
+      "step": 38788
+    },
+    {
+      "epoch": 36.19,
+      "learning_rate": 4.547667910447762e-05,
+      "loss": 0.0,
+      "step": 38792
+    },
+    {
+      "epoch": 36.19,
+      "learning_rate": 4.5476212686567166e-05,
+      "loss": 0.0001,
+      "step": 38796
+    },
+    {
+      "epoch": 36.19,
+      "learning_rate": 4.5475746268656714e-05,
+      "loss": 0.0001,
+      "step": 38800
+    },
+    {
+      "epoch": 36.2,
+      "learning_rate": 4.547527985074627e-05,
+      "loss": 0.0001,
+      "step": 38804
+    },
+    {
+      "epoch": 36.2,
+      "learning_rate": 4.5474813432835824e-05,
+      "loss": 0.0001,
+      "step": 38808
+    },
+    {
+      "epoch": 36.21,
+      "learning_rate": 4.547434701492538e-05,
+      "loss": 0.0001,
+      "step": 38812
+    },
+    {
+      "epoch": 36.21,
+      "learning_rate": 4.547388059701493e-05,
+      "loss": 0.0001,
+      "step": 38816
+    },
+    {
+      "epoch": 36.21,
+      "learning_rate": 4.547341417910448e-05,
+      "loss": 0.0004,
+      "step": 38820
+    },
+    {
+      "epoch": 36.22,
+      "learning_rate": 4.547294776119404e-05,
+      "loss": 0.001,
+      "step": 38824
+    },
+    {
+      "epoch": 36.22,
+      "learning_rate": 4.5472481343283585e-05,
+      "loss": 0.0011,
+      "step": 38828
+    },
+    {
+      "epoch": 36.22,
+      "learning_rate": 4.547201492537313e-05,
+      "loss": 0.001,
+      "step": 38832
+    },
+    {
+      "epoch": 36.23,
+      "learning_rate": 4.547154850746269e-05,
+      "loss": 0.0,
+      "step": 38836
+    },
+    {
+      "epoch": 36.23,
+      "learning_rate": 4.547108208955224e-05,
+      "loss": 0.0002,
+      "step": 38840
+    },
+    {
+      "epoch": 36.24,
+      "learning_rate": 4.547061567164179e-05,
+      "loss": 0.0001,
+      "step": 38844
+    },
+    {
+      "epoch": 36.24,
+      "learning_rate": 4.5470149253731346e-05,
+      "loss": 0.0001,
+      "step": 38848
+    },
+    {
+      "epoch": 36.24,
+      "learning_rate": 4.54696828358209e-05,
+      "loss": 0.0035,
+      "step": 38852
+    },
+    {
+      "epoch": 36.25,
+      "learning_rate": 4.546921641791045e-05,
+      "loss": 0.0001,
+      "step": 38856
+    },
+    {
+      "epoch": 36.25,
+      "learning_rate": 4.5468750000000004e-05,
+      "loss": 0.0008,
+      "step": 38860
+    },
+    {
+      "epoch": 36.25,
+      "learning_rate": 4.546828358208955e-05,
+      "loss": 0.0001,
+      "step": 38864
+    },
+    {
+      "epoch": 36.26,
+      "learning_rate": 4.546781716417911e-05,
+      "loss": 0.0001,
+      "step": 38868
+    },
+    {
+      "epoch": 36.26,
+      "learning_rate": 4.546735074626866e-05,
+      "loss": 0.0006,
+      "step": 38872
+    },
+    {
+      "epoch": 36.26,
+      "learning_rate": 4.546688432835821e-05,
+      "loss": 0.0001,
+      "step": 38876
+    },
+    {
+      "epoch": 36.27,
+      "learning_rate": 4.5466417910447765e-05,
+      "loss": 0.0,
+      "step": 38880
+    },
+    {
+      "epoch": 36.27,
+      "learning_rate": 4.546595149253732e-05,
+      "loss": 0.0008,
+      "step": 38884
+    },
+    {
+      "epoch": 36.28,
+      "learning_rate": 4.546548507462687e-05,
+      "loss": 0.0001,
+      "step": 38888
+    },
+    {
+      "epoch": 36.28,
+      "learning_rate": 4.5465018656716416e-05,
+      "loss": 0.0017,
+      "step": 38892
+    },
+    {
+      "epoch": 36.28,
+      "learning_rate": 4.546455223880597e-05,
+      "loss": 0.0,
+      "step": 38896
+    },
+    {
+      "epoch": 36.29,
+      "learning_rate": 4.5464085820895526e-05,
+      "loss": 0.0043,
+      "step": 38900
+    },
+    {
+      "epoch": 36.29,
+      "learning_rate": 4.5463619402985074e-05,
+      "loss": 0.0006,
+      "step": 38904
+    },
+    {
+      "epoch": 36.29,
+      "learning_rate": 4.546315298507463e-05,
+      "loss": 0.0,
+      "step": 38908
+    },
+    {
+      "epoch": 36.3,
+      "learning_rate": 4.5462686567164184e-05,
+      "loss": 0.0001,
+      "step": 38912
+    },
+    {
+      "epoch": 36.3,
+      "learning_rate": 4.546222014925373e-05,
+      "loss": 0.0014,
+      "step": 38916
+    },
+    {
+      "epoch": 36.31,
+      "learning_rate": 4.546175373134329e-05,
+      "loss": 0.0001,
+      "step": 38920
+    },
+    {
+      "epoch": 36.31,
+      "learning_rate": 4.5461287313432835e-05,
+      "loss": 0.0006,
+      "step": 38924
+    },
+    {
+      "epoch": 36.31,
+      "learning_rate": 4.546082089552239e-05,
+      "loss": 0.0004,
+      "step": 38928
+    },
+    {
+      "epoch": 36.32,
+      "learning_rate": 4.5460354477611945e-05,
+      "loss": 0.0022,
+      "step": 38932
+    },
+    {
+      "epoch": 36.32,
+      "learning_rate": 4.545988805970149e-05,
+      "loss": 0.0,
+      "step": 38936
+    },
+    {
+      "epoch": 36.32,
+      "learning_rate": 4.545942164179104e-05,
+      "loss": 0.0,
+      "step": 38940
+    },
+    {
+      "epoch": 36.33,
+      "learning_rate": 4.54589552238806e-05,
+      "loss": 0.0007,
+      "step": 38944
+    },
+    {
+      "epoch": 36.33,
+      "learning_rate": 4.545848880597015e-05,
+      "loss": 0.0001,
+      "step": 38948
+    },
+    {
+      "epoch": 36.34,
+      "learning_rate": 4.54580223880597e-05,
+      "loss": 0.0027,
+      "step": 38952
+    },
+    {
+      "epoch": 36.34,
+      "learning_rate": 4.5457555970149254e-05,
+      "loss": 0.0004,
+      "step": 38956
+    },
+    {
+      "epoch": 36.34,
+      "learning_rate": 4.545708955223881e-05,
+      "loss": 0.0001,
+      "step": 38960
+    },
+    {
+      "epoch": 36.35,
+      "learning_rate": 4.545662313432836e-05,
+      "loss": 0.0001,
+      "step": 38964
+    },
+    {
+      "epoch": 36.35,
+      "learning_rate": 4.545615671641791e-05,
+      "loss": 0.0035,
+      "step": 38968
+    },
+    {
+      "epoch": 36.35,
+      "learning_rate": 4.545569029850747e-05,
+      "loss": 0.0002,
+      "step": 38972
+    },
+    {
+      "epoch": 36.36,
+      "learning_rate": 4.545522388059702e-05,
+      "loss": 0.0072,
+      "step": 38976
+    },
+    {
+      "epoch": 36.36,
+      "learning_rate": 4.545475746268657e-05,
+      "loss": 0.0,
+      "step": 38980
+    },
+    {
+      "epoch": 36.37,
+      "learning_rate": 4.545429104477612e-05,
+      "loss": 0.0,
+      "step": 38984
+    },
+    {
+      "epoch": 36.37,
+      "learning_rate": 4.545382462686568e-05,
+      "loss": 0.0001,
+      "step": 38988
+    },
+    {
+      "epoch": 36.37,
+      "learning_rate": 4.545335820895523e-05,
+      "loss": 0.0007,
+      "step": 38992
+    },
+    {
+      "epoch": 36.38,
+      "learning_rate": 4.5452891791044776e-05,
+      "loss": 0.0015,
+      "step": 38996
+    },
+    {
+      "epoch": 36.38,
+      "learning_rate": 4.545242537313433e-05,
+      "loss": 0.0001,
+      "step": 39000
+    },
+    {
+      "epoch": 36.38,
+      "eval_exact_match": 0.7456479690522244,
+      "eval_exec": 0.769825918762089,
+      "eval_loss": 0.4195860028266907,
+      "eval_runtime": 1174.4998,
+      "eval_samples_per_second": 0.88,
+      "step": 39000
+    },
+    {
+      "epoch": 36.38,
+      "learning_rate": 4.5451958955223886e-05,
+      "loss": 0.0005,
+      "step": 39004
+    },
+    {
+      "epoch": 36.39,
+      "learning_rate": 4.5451492537313434e-05,
+      "loss": 0.0,
+      "step": 39008
+    },
+    {
+      "epoch": 36.39,
+      "learning_rate": 4.545102611940299e-05,
+      "loss": 0.0001,
+      "step": 39012
+    },
+    {
+      "epoch": 36.4,
+      "learning_rate": 4.545055970149254e-05,
+      "loss": 0.0,
+      "step": 39016
+    },
+    {
+      "epoch": 36.4,
+      "learning_rate": 4.545009328358209e-05,
+      "loss": 0.0003,
+      "step": 39020
+    },
+    {
+      "epoch": 36.4,
+      "learning_rate": 4.544962686567165e-05,
+      "loss": 0.0,
+      "step": 39024
+    },
+    {
+      "epoch": 36.41,
+      "learning_rate": 4.5449160447761195e-05,
+      "loss": 0.0001,
+      "step": 39028
+    },
+    {
+      "epoch": 36.41,
+      "learning_rate": 4.544869402985075e-05,
+      "loss": 0.0001,
+      "step": 39032
+    },
+    {
+      "epoch": 36.41,
+      "learning_rate": 4.5448227611940305e-05,
+      "loss": 0.0033,
+      "step": 39036
+    },
+    {
+      "epoch": 36.42,
+      "learning_rate": 4.544776119402985e-05,
+      "loss": 0.0002,
+      "step": 39040
+    },
+    {
+      "epoch": 36.42,
+      "learning_rate": 4.54472947761194e-05,
+      "loss": 0.0001,
+      "step": 39044
+    },
+    {
+      "epoch": 36.43,
+      "learning_rate": 4.5446828358208956e-05,
+      "loss": 0.0,
+      "step": 39048
+    },
+    {
+      "epoch": 36.43,
+      "learning_rate": 4.544636194029851e-05,
+      "loss": 0.0002,
+      "step": 39052
+    },
+    {
+      "epoch": 36.43,
+      "learning_rate": 4.544589552238806e-05,
+      "loss": 0.0,
+      "step": 39056
+    },
+    {
+      "epoch": 36.44,
+      "learning_rate": 4.5445429104477614e-05,
+      "loss": 0.0002,
+      "step": 39060
+    },
+    {
+      "epoch": 36.44,
+      "learning_rate": 4.544496268656717e-05,
+      "loss": 0.0009,
+      "step": 39064
+    },
+    {
+      "epoch": 36.44,
+      "learning_rate": 4.544449626865672e-05,
+      "loss": 0.0001,
+      "step": 39068
+    },
+    {
+      "epoch": 36.45,
+      "learning_rate": 4.544402985074627e-05,
+      "loss": 0.0,
+      "step": 39072
+    },
+    {
+      "epoch": 36.45,
+      "learning_rate": 4.544356343283582e-05,
+      "loss": 0.0001,
+      "step": 39076
+    },
+    {
+      "epoch": 36.46,
+      "learning_rate": 4.5443097014925375e-05,
+      "loss": 0.0,
+      "step": 39080
+    },
+    {
+      "epoch": 36.46,
+      "learning_rate": 4.544263059701493e-05,
+      "loss": 0.0,
+      "step": 39084
+    },
+    {
+      "epoch": 36.46,
+      "learning_rate": 4.544216417910448e-05,
+      "loss": 0.0002,
+      "step": 39088
+    },
+    {
+      "epoch": 36.47,
+      "learning_rate": 4.544169776119403e-05,
+      "loss": 0.003,
+      "step": 39092
+    },
+    {
+      "epoch": 36.47,
+      "learning_rate": 4.544123134328359e-05,
+      "loss": 0.0001,
+      "step": 39096
+    },
+    {
+      "epoch": 36.47,
+      "learning_rate": 4.5440764925373136e-05,
+      "loss": 0.0002,
+      "step": 39100
+    },
+    {
+      "epoch": 36.48,
+      "learning_rate": 4.5440298507462684e-05,
+      "loss": 0.0001,
+      "step": 39104
+    },
+    {
+      "epoch": 36.48,
+      "learning_rate": 4.543983208955224e-05,
+      "loss": 0.0004,
+      "step": 39108
+    },
+    {
+      "epoch": 36.49,
+      "learning_rate": 4.5439365671641794e-05,
+      "loss": 0.0192,
+      "step": 39112
+    },
+    {
+      "epoch": 36.49,
+      "learning_rate": 4.543889925373134e-05,
+      "loss": 0.0006,
+      "step": 39116
+    },
+    {
+      "epoch": 36.49,
+      "learning_rate": 4.54384328358209e-05,
+      "loss": 0.0017,
+      "step": 39120
+    },
+    {
+      "epoch": 36.5,
+      "learning_rate": 4.543796641791045e-05,
+      "loss": 0.0049,
+      "step": 39124
+    },
+    {
+      "epoch": 36.5,
+      "learning_rate": 4.54375e-05,
+      "loss": 0.0001,
+      "step": 39128
+    },
+    {
+      "epoch": 36.5,
+      "learning_rate": 4.5437033582089555e-05,
+      "loss": 0.0003,
+      "step": 39132
+    },
+    {
+      "epoch": 36.51,
+      "learning_rate": 4.54365671641791e-05,
+      "loss": 0.0,
+      "step": 39136
+    },
+    {
+      "epoch": 36.51,
+      "learning_rate": 4.5436100746268665e-05,
+      "loss": 0.0001,
+      "step": 39140
+    },
+    {
+      "epoch": 36.51,
+      "learning_rate": 4.543563432835821e-05,
+      "loss": 0.0001,
+      "step": 39144
+    },
+    {
+      "epoch": 36.52,
+      "learning_rate": 4.543516791044776e-05,
+      "loss": 0.0001,
+      "step": 39148
+    },
+    {
+      "epoch": 36.52,
+      "learning_rate": 4.5434701492537316e-05,
+      "loss": 0.0001,
+      "step": 39152
+    },
+    {
+      "epoch": 36.53,
+      "learning_rate": 4.543423507462687e-05,
+      "loss": 0.0001,
+      "step": 39156
+    },
+    {
+      "epoch": 36.53,
+      "learning_rate": 4.543376865671642e-05,
+      "loss": 0.0001,
+      "step": 39160
+    },
+    {
+      "epoch": 36.53,
+      "learning_rate": 4.5433302238805974e-05,
+      "loss": 0.0,
+      "step": 39164
+    },
+    {
+      "epoch": 36.54,
+      "learning_rate": 4.543283582089552e-05,
+      "loss": 0.0002,
+      "step": 39168
+    },
+    {
+      "epoch": 36.54,
+      "learning_rate": 4.543236940298508e-05,
+      "loss": 0.0,
+      "step": 39172
+    },
+    {
+      "epoch": 36.54,
+      "learning_rate": 4.543190298507463e-05,
+      "loss": 0.0012,
+      "step": 39176
+    },
+    {
+      "epoch": 36.55,
+      "learning_rate": 4.543143656716418e-05,
+      "loss": 0.0001,
+      "step": 39180
+    },
+    {
+      "epoch": 36.55,
+      "learning_rate": 4.5430970149253735e-05,
+      "loss": 0.0009,
+      "step": 39184
+    },
+    {
+      "epoch": 36.56,
+      "learning_rate": 4.543050373134329e-05,
+      "loss": 0.0001,
+      "step": 39188
+    },
+    {
+      "epoch": 36.56,
+      "learning_rate": 4.543003731343284e-05,
+      "loss": 0.0006,
+      "step": 39192
+    },
+    {
+      "epoch": 36.56,
+      "learning_rate": 4.5429570895522386e-05,
+      "loss": 0.0001,
+      "step": 39196
+    },
+    {
+      "epoch": 36.57,
+      "learning_rate": 4.542910447761195e-05,
+      "loss": 0.0,
+      "step": 39200
+    },
+    {
+      "epoch": 36.57,
+      "learning_rate": 4.5428638059701496e-05,
+      "loss": 0.0002,
+      "step": 39204
+    },
+    {
+      "epoch": 36.57,
+      "learning_rate": 4.5428171641791044e-05,
+      "loss": 0.0,
+      "step": 39208
+    },
+    {
+      "epoch": 36.58,
+      "learning_rate": 4.54277052238806e-05,
+      "loss": 0.0001,
+      "step": 39212
+    },
+    {
+      "epoch": 36.58,
+      "learning_rate": 4.5427238805970154e-05,
+      "loss": 0.0,
+      "step": 39216
+    },
+    {
+      "epoch": 36.59,
+      "learning_rate": 4.54267723880597e-05,
+      "loss": 0.0001,
+      "step": 39220
+    },
+    {
+      "epoch": 36.59,
+      "learning_rate": 4.542630597014926e-05,
+      "loss": 0.0001,
+      "step": 39224
+    },
+    {
+      "epoch": 36.59,
+      "learning_rate": 4.5425839552238805e-05,
+      "loss": 0.0,
+      "step": 39228
+    },
+    {
+      "epoch": 36.6,
+      "learning_rate": 4.542537313432836e-05,
+      "loss": 0.0,
+      "step": 39232
+    },
+    {
+      "epoch": 36.6,
+      "learning_rate": 4.5424906716417915e-05,
+      "loss": 0.0001,
+      "step": 39236
+    },
+    {
+      "epoch": 36.6,
+      "learning_rate": 4.542444029850746e-05,
+      "loss": 0.0025,
+      "step": 39240
+    },
+    {
+      "epoch": 36.61,
+      "learning_rate": 4.542397388059702e-05,
+      "loss": 0.0007,
+      "step": 39244
+    },
+    {
+      "epoch": 36.61,
+      "learning_rate": 4.542350746268657e-05,
+      "loss": 0.0016,
+      "step": 39248
+    },
+    {
+      "epoch": 36.62,
+      "learning_rate": 4.542304104477612e-05,
+      "loss": 0.0003,
+      "step": 39252
+    },
+    {
+      "epoch": 36.62,
+      "learning_rate": 4.542257462686567e-05,
+      "loss": 0.0,
+      "step": 39256
+    },
+    {
+      "epoch": 36.62,
+      "learning_rate": 4.542210820895523e-05,
+      "loss": 0.0,
+      "step": 39260
+    },
+    {
+      "epoch": 36.63,
+      "learning_rate": 4.542164179104478e-05,
+      "loss": 0.0004,
+      "step": 39264
+    },
+    {
+      "epoch": 36.63,
+      "learning_rate": 4.542117537313433e-05,
+      "loss": 0.0013,
+      "step": 39268
+    },
+    {
+      "epoch": 36.63,
+      "learning_rate": 4.542070895522388e-05,
+      "loss": 0.0,
+      "step": 39272
+    },
+    {
+      "epoch": 36.64,
+      "learning_rate": 4.542024253731344e-05,
+      "loss": 0.0,
+      "step": 39276
+    },
+    {
+      "epoch": 36.64,
+      "learning_rate": 4.5419776119402985e-05,
+      "loss": 0.0001,
+      "step": 39280
+    },
+    {
+      "epoch": 36.65,
+      "learning_rate": 4.541930970149254e-05,
+      "loss": 0.0001,
+      "step": 39284
+    },
+    {
+      "epoch": 36.65,
+      "learning_rate": 4.541884328358209e-05,
+      "loss": 0.0,
+      "step": 39288
+    },
+    {
+      "epoch": 36.65,
+      "learning_rate": 4.541837686567164e-05,
+      "loss": 0.0,
+      "step": 39292
+    },
+    {
+      "epoch": 36.66,
+      "learning_rate": 4.54179104477612e-05,
+      "loss": 0.0,
+      "step": 39296
+    },
+    {
+      "epoch": 36.66,
+      "learning_rate": 4.5417444029850746e-05,
+      "loss": 0.0001,
+      "step": 39300
+    },
+    {
+      "epoch": 36.66,
+      "learning_rate": 4.54169776119403e-05,
+      "loss": 0.0014,
+      "step": 39304
+    },
+    {
+      "epoch": 36.67,
+      "learning_rate": 4.5416511194029856e-05,
+      "loss": 0.0,
+      "step": 39308
+    },
+    {
+      "epoch": 36.67,
+      "learning_rate": 4.5416044776119404e-05,
+      "loss": 0.0003,
+      "step": 39312
+    },
+    {
+      "epoch": 36.68,
+      "learning_rate": 4.541557835820896e-05,
+      "loss": 0.0,
+      "step": 39316
+    },
+    {
+      "epoch": 36.68,
+      "learning_rate": 4.5415111940298514e-05,
+      "loss": 0.0046,
+      "step": 39320
+    },
+    {
+      "epoch": 36.68,
+      "learning_rate": 4.541464552238806e-05,
+      "loss": 0.0013,
+      "step": 39324
+    },
+    {
+      "epoch": 36.69,
+      "learning_rate": 4.5414179104477617e-05,
+      "loss": 0.0006,
+      "step": 39328
+    },
+    {
+      "epoch": 36.69,
+      "learning_rate": 4.5413712686567165e-05,
+      "loss": 0.0003,
+      "step": 39332
+    },
+    {
+      "epoch": 36.69,
+      "learning_rate": 4.541324626865672e-05,
+      "loss": 0.0001,
+      "step": 39336
+    },
+    {
+      "epoch": 36.7,
+      "learning_rate": 4.5412779850746274e-05,
+      "loss": 0.0006,
+      "step": 39340
+    },
+    {
+      "epoch": 36.7,
+      "learning_rate": 4.541231343283582e-05,
+      "loss": 0.0,
+      "step": 39344
+    },
+    {
+      "epoch": 36.71,
+      "learning_rate": 4.541184701492537e-05,
+      "loss": 0.0007,
+      "step": 39348
+    },
+    {
+      "epoch": 36.71,
+      "learning_rate": 4.541138059701493e-05,
+      "loss": 0.0008,
+      "step": 39352
+    },
+    {
+      "epoch": 36.71,
+      "learning_rate": 4.541091417910448e-05,
+      "loss": 0.0,
+      "step": 39356
+    },
+    {
+      "epoch": 36.72,
+      "learning_rate": 4.541044776119403e-05,
+      "loss": 0.0,
+      "step": 39360
+    },
+    {
+      "epoch": 36.72,
+      "learning_rate": 4.5409981343283584e-05,
+      "loss": 0.0,
+      "step": 39364
+    },
+    {
+      "epoch": 36.72,
+      "learning_rate": 4.540951492537314e-05,
+      "loss": 0.0002,
+      "step": 39368
+    },
+    {
+      "epoch": 36.73,
+      "learning_rate": 4.540904850746269e-05,
+      "loss": 0.0005,
+      "step": 39372
+    },
+    {
+      "epoch": 36.73,
+      "learning_rate": 4.540858208955224e-05,
+      "loss": 0.0001,
+      "step": 39376
+    },
+    {
+      "epoch": 36.73,
+      "learning_rate": 4.5408115671641796e-05,
+      "loss": 0.0001,
+      "step": 39380
+    },
+    {
+      "epoch": 36.74,
+      "learning_rate": 4.5407649253731345e-05,
+      "loss": 0.0,
+      "step": 39384
+    },
+    {
+      "epoch": 36.74,
+      "learning_rate": 4.54071828358209e-05,
+      "loss": 0.0,
+      "step": 39388
+    },
+    {
+      "epoch": 36.75,
+      "learning_rate": 4.540671641791045e-05,
+      "loss": 0.0,
+      "step": 39392
+    },
+    {
+      "epoch": 36.75,
+      "learning_rate": 4.540625e-05,
+      "loss": 0.0005,
+      "step": 39396
+    },
+    {
+      "epoch": 36.75,
+      "learning_rate": 4.540578358208956e-05,
+      "loss": 0.0004,
+      "step": 39400
+    },
+    {
+      "epoch": 36.76,
+      "learning_rate": 4.5405317164179106e-05,
+      "loss": 0.0001,
+      "step": 39404
+    },
+    {
+      "epoch": 36.76,
+      "learning_rate": 4.5404850746268654e-05,
+      "loss": 0.0,
+      "step": 39408
+    },
+    {
+      "epoch": 36.76,
+      "learning_rate": 4.5404384328358215e-05,
+      "loss": 0.0,
+      "step": 39412
+    },
+    {
+      "epoch": 36.77,
+      "learning_rate": 4.5403917910447764e-05,
+      "loss": 0.0001,
+      "step": 39416
+    },
+    {
+      "epoch": 36.77,
+      "learning_rate": 4.540345149253731e-05,
+      "loss": 0.0,
+      "step": 39420
+    },
+    {
+      "epoch": 36.78,
+      "learning_rate": 4.5402985074626867e-05,
+      "loss": 0.0001,
+      "step": 39424
+    },
+    {
+      "epoch": 36.78,
+      "learning_rate": 4.540251865671642e-05,
+      "loss": 0.0,
+      "step": 39428
+    },
+    {
+      "epoch": 36.78,
+      "learning_rate": 4.540205223880597e-05,
+      "loss": 0.0009,
+      "step": 39432
+    },
+    {
+      "epoch": 36.79,
+      "learning_rate": 4.5401585820895525e-05,
+      "loss": 0.0,
+      "step": 39436
+    },
+    {
+      "epoch": 36.79,
+      "learning_rate": 4.540111940298508e-05,
+      "loss": 0.0,
+      "step": 39440
+    },
+    {
+      "epoch": 36.79,
+      "learning_rate": 4.540065298507463e-05,
+      "loss": 0.0,
+      "step": 39444
+    },
+    {
+      "epoch": 36.8,
+      "learning_rate": 4.540018656716418e-05,
+      "loss": 0.0001,
+      "step": 39448
+    },
+    {
+      "epoch": 36.8,
+      "learning_rate": 4.539972014925373e-05,
+      "loss": 0.0027,
+      "step": 39452
+    },
+    {
+      "epoch": 36.81,
+      "learning_rate": 4.5399253731343286e-05,
+      "loss": 0.0,
+      "step": 39456
+    },
+    {
+      "epoch": 36.81,
+      "learning_rate": 4.539878731343284e-05,
+      "loss": 0.0002,
+      "step": 39460
+    },
+    {
+      "epoch": 36.81,
+      "learning_rate": 4.539832089552239e-05,
+      "loss": 0.0001,
+      "step": 39464
+    },
+    {
+      "epoch": 36.82,
+      "learning_rate": 4.5397854477611943e-05,
+      "loss": 0.0,
+      "step": 39468
+    },
+    {
+      "epoch": 36.82,
+      "learning_rate": 4.53973880597015e-05,
+      "loss": 0.0009,
+      "step": 39472
+    },
+    {
+      "epoch": 36.82,
+      "learning_rate": 4.5396921641791046e-05,
+      "loss": 0.0038,
+      "step": 39476
+    },
+    {
+      "epoch": 36.83,
+      "learning_rate": 4.53964552238806e-05,
+      "loss": 0.0022,
+      "step": 39480
+    },
+    {
+      "epoch": 36.83,
+      "learning_rate": 4.539598880597015e-05,
+      "loss": 0.0005,
+      "step": 39484
+    },
+    {
+      "epoch": 36.84,
+      "learning_rate": 4.5395522388059704e-05,
+      "loss": 0.0,
+      "step": 39488
+    },
+    {
+      "epoch": 36.84,
+      "learning_rate": 4.539505597014926e-05,
+      "loss": 0.0001,
+      "step": 39492
+    },
+    {
+      "epoch": 36.84,
+      "learning_rate": 4.539458955223881e-05,
+      "loss": 0.0,
+      "step": 39496
+    },
+    {
+      "epoch": 36.85,
+      "learning_rate": 4.539412313432836e-05,
+      "loss": 0.0001,
+      "step": 39500
+    },
+    {
+      "epoch": 36.85,
+      "eval_exact_match": 0.7495164410058027,
+      "eval_exec": 0.7852998065764023,
+      "eval_loss": 0.4335377514362335,
+      "eval_runtime": 1568.2028,
+      "eval_samples_per_second": 0.659,
+      "step": 39500
+    },
+    {
+      "epoch": 36.85,
+      "learning_rate": 4.539365671641792e-05,
+      "loss": 0.0002,
+      "step": 39504
+    },
+    {
+      "epoch": 36.85,
+      "learning_rate": 4.5393190298507465e-05,
+      "loss": 0.0,
+      "step": 39508
+    },
+    {
+      "epoch": 36.86,
+      "learning_rate": 4.5392723880597014e-05,
+      "loss": 0.0001,
+      "step": 39512
+    },
+    {
+      "epoch": 36.86,
+      "learning_rate": 4.539225746268657e-05,
+      "loss": 0.0041,
+      "step": 39516
+    },
+    {
+      "epoch": 36.87,
+      "learning_rate": 4.539179104477612e-05,
+      "loss": 0.0001,
+      "step": 39520
+    },
+    {
+      "epoch": 36.87,
+      "learning_rate": 4.539132462686567e-05,
+      "loss": 0.0001,
+      "step": 39524
+    },
+    {
+      "epoch": 36.87,
+      "learning_rate": 4.5390858208955226e-05,
+      "loss": 0.0,
+      "step": 39528
+    },
+    {
+      "epoch": 36.88,
+      "learning_rate": 4.539039179104478e-05,
+      "loss": 0.0055,
+      "step": 39532
+    },
+    {
+      "epoch": 36.88,
+      "learning_rate": 4.538992537313433e-05,
+      "loss": 0.0002,
+      "step": 39536
+    },
+    {
+      "epoch": 36.88,
+      "learning_rate": 4.5389458955223884e-05,
+      "loss": 0.0001,
+      "step": 39540
+    },
+    {
+      "epoch": 36.89,
+      "learning_rate": 4.538899253731343e-05,
+      "loss": 0.0002,
+      "step": 39544
+    },
+    {
+      "epoch": 36.89,
+      "learning_rate": 4.538852611940299e-05,
+      "loss": 0.0,
+      "step": 39548
+    },
+    {
+      "epoch": 36.9,
+      "learning_rate": 4.538805970149254e-05,
+      "loss": 0.0,
+      "step": 39552
+    },
+    {
+      "epoch": 36.9,
+      "learning_rate": 4.538759328358209e-05,
+      "loss": 0.0,
+      "step": 39556
+    },
+    {
+      "epoch": 36.9,
+      "learning_rate": 4.5387126865671645e-05,
+      "loss": 0.0015,
+      "step": 39560
+    },
+    {
+      "epoch": 36.91,
+      "learning_rate": 4.53866604477612e-05,
+      "loss": 0.0001,
+      "step": 39564
+    },
+    {
+      "epoch": 36.91,
+      "learning_rate": 4.538619402985075e-05,
+      "loss": 0.0001,
+      "step": 39568
+    },
+    {
+      "epoch": 36.91,
+      "learning_rate": 4.5385727611940297e-05,
+      "loss": 0.0041,
+      "step": 39572
+    },
+    {
+      "epoch": 36.92,
+      "learning_rate": 4.538526119402985e-05,
+      "loss": 0.0001,
+      "step": 39576
+    },
+    {
+      "epoch": 36.92,
+      "learning_rate": 4.5384794776119406e-05,
+      "loss": 0.0001,
+      "step": 39580
+    },
+    {
+      "epoch": 36.93,
+      "learning_rate": 4.5384328358208954e-05,
+      "loss": 0.0004,
+      "step": 39584
+    },
+    {
+      "epoch": 36.93,
+      "learning_rate": 4.538386194029851e-05,
+      "loss": 0.0002,
+      "step": 39588
+    },
+    {
+      "epoch": 36.93,
+      "learning_rate": 4.5383395522388064e-05,
+      "loss": 0.0001,
+      "step": 39592
+    },
+    {
+      "epoch": 36.94,
+      "learning_rate": 4.538292910447761e-05,
+      "loss": 0.0,
+      "step": 39596
+    },
+    {
+      "epoch": 36.94,
+      "learning_rate": 4.538246268656717e-05,
+      "loss": 0.0011,
+      "step": 39600
+    },
+    {
+      "epoch": 36.94,
+      "learning_rate": 4.5381996268656715e-05,
+      "loss": 0.0004,
+      "step": 39604
+    },
+    {
+      "epoch": 36.95,
+      "learning_rate": 4.538152985074627e-05,
+      "loss": 0.0002,
+      "step": 39608
+    },
+    {
+      "epoch": 36.95,
+      "learning_rate": 4.5381063432835825e-05,
+      "loss": 0.0101,
+      "step": 39612
+    },
+    {
+      "epoch": 36.96,
+      "learning_rate": 4.538059701492537e-05,
+      "loss": 0.0006,
+      "step": 39616
+    },
+    {
+      "epoch": 36.96,
+      "learning_rate": 4.538013059701492e-05,
+      "loss": 0.0001,
+      "step": 39620
+    },
+    {
+      "epoch": 36.96,
+      "learning_rate": 4.537966417910448e-05,
+      "loss": 0.0042,
+      "step": 39624
+    },
+    {
+      "epoch": 36.97,
+      "learning_rate": 4.537919776119403e-05,
+      "loss": 0.0004,
+      "step": 39628
+    },
+    {
+      "epoch": 36.97,
+      "learning_rate": 4.5378731343283586e-05,
+      "loss": 0.0002,
+      "step": 39632
+    },
+    {
+      "epoch": 36.97,
+      "learning_rate": 4.5378264925373134e-05,
+      "loss": 0.001,
+      "step": 39636
+    },
+    {
+      "epoch": 36.98,
+      "learning_rate": 4.537779850746269e-05,
+      "loss": 0.0011,
+      "step": 39640
+    },
+    {
+      "epoch": 36.98,
+      "learning_rate": 4.5377332089552244e-05,
+      "loss": 0.0018,
+      "step": 39644
+    },
+    {
+      "epoch": 36.98,
+      "learning_rate": 4.537686567164179e-05,
+      "loss": 0.0001,
+      "step": 39648
+    },
+    {
+      "epoch": 36.99,
+      "learning_rate": 4.537639925373135e-05,
+      "loss": 0.0,
+      "step": 39652
+    },
+    {
+      "epoch": 36.99,
+      "learning_rate": 4.53759328358209e-05,
+      "loss": 0.0001,
+      "step": 39656
+    },
+    {
+      "epoch": 37.0,
+      "learning_rate": 4.537546641791045e-05,
+      "loss": 0.0,
+      "step": 39660
+    },
+    {
+      "epoch": 37.0,
+      "learning_rate": 4.5375e-05,
+      "loss": 0.0,
+      "step": 39664
+    },
+    {
+      "epoch": 37.0,
+      "learning_rate": 4.537453358208956e-05,
+      "loss": 0.0002,
+      "step": 39668
+    },
+    {
+      "epoch": 37.01,
+      "learning_rate": 4.537406716417911e-05,
+      "loss": 0.0,
+      "step": 39672
+    },
+    {
+      "epoch": 37.01,
+      "learning_rate": 4.5373600746268656e-05,
+      "loss": 0.0,
+      "step": 39676
+    },
+    {
+      "epoch": 37.01,
+      "learning_rate": 4.537313432835821e-05,
+      "loss": 0.0,
+      "step": 39680
+    },
+    {
+      "epoch": 37.02,
+      "learning_rate": 4.5372667910447766e-05,
+      "loss": 0.0,
+      "step": 39684
+    },
+    {
+      "epoch": 37.02,
+      "learning_rate": 4.5372201492537314e-05,
+      "loss": 0.0001,
+      "step": 39688
+    },
+    {
+      "epoch": 37.03,
+      "learning_rate": 4.537173507462687e-05,
+      "loss": 0.0005,
+      "step": 39692
+    },
+    {
+      "epoch": 37.03,
+      "learning_rate": 4.537126865671642e-05,
+      "loss": 0.0,
+      "step": 39696
+    },
+    {
+      "epoch": 37.03,
+      "learning_rate": 4.537080223880597e-05,
+      "loss": 0.0,
+      "step": 39700
+    },
+    {
+      "epoch": 37.04,
+      "learning_rate": 4.537033582089553e-05,
+      "loss": 0.0011,
+      "step": 39704
+    },
+    {
+      "epoch": 37.04,
+      "learning_rate": 4.5369869402985075e-05,
+      "loss": 0.0004,
+      "step": 39708
+    },
+    {
+      "epoch": 37.04,
+      "learning_rate": 4.536940298507463e-05,
+      "loss": 0.0002,
+      "step": 39712
+    },
+    {
+      "epoch": 37.05,
+      "learning_rate": 4.5368936567164185e-05,
+      "loss": 0.0001,
+      "step": 39716
+    },
+    {
+      "epoch": 37.05,
+      "learning_rate": 4.536847014925373e-05,
+      "loss": 0.0,
+      "step": 39720
+    },
+    {
+      "epoch": 37.06,
+      "learning_rate": 4.536800373134328e-05,
+      "loss": 0.0,
+      "step": 39724
+    },
+    {
+      "epoch": 37.06,
+      "learning_rate": 4.5367537313432836e-05,
+      "loss": 0.0006,
+      "step": 39728
+    },
+    {
+      "epoch": 37.06,
+      "learning_rate": 4.536707089552239e-05,
+      "loss": 0.0001,
+      "step": 39732
+    },
+    {
+      "epoch": 37.07,
+      "learning_rate": 4.536660447761194e-05,
+      "loss": 0.0005,
+      "step": 39736
+    },
+    {
+      "epoch": 37.07,
+      "learning_rate": 4.5366138059701494e-05,
+      "loss": 0.0002,
+      "step": 39740
+    },
+    {
+      "epoch": 37.07,
+      "learning_rate": 4.536567164179105e-05,
+      "loss": 0.0,
+      "step": 39744
+    },
+    {
+      "epoch": 37.08,
+      "learning_rate": 4.53652052238806e-05,
+      "loss": 0.0088,
+      "step": 39748
+    },
+    {
+      "epoch": 37.08,
+      "learning_rate": 4.536473880597015e-05,
+      "loss": 0.0005,
+      "step": 39752
+    },
+    {
+      "epoch": 37.09,
+      "learning_rate": 4.53642723880597e-05,
+      "loss": 0.0001,
+      "step": 39756
+    },
+    {
+      "epoch": 37.09,
+      "learning_rate": 4.5363805970149255e-05,
+      "loss": 0.0,
+      "step": 39760
+    },
+    {
+      "epoch": 37.09,
+      "learning_rate": 4.536333955223881e-05,
+      "loss": 0.0002,
+      "step": 39764
+    },
+    {
+      "epoch": 37.1,
+      "learning_rate": 4.536287313432836e-05,
+      "loss": 0.0011,
+      "step": 39768
+    },
+    {
+      "epoch": 37.1,
+      "learning_rate": 4.536240671641791e-05,
+      "loss": 0.0,
+      "step": 39772
+    },
+    {
+      "epoch": 37.1,
+      "learning_rate": 4.536194029850747e-05,
+      "loss": 0.0,
+      "step": 39776
+    },
+    {
+      "epoch": 37.11,
+      "learning_rate": 4.5361473880597016e-05,
+      "loss": 0.0001,
+      "step": 39780
+    },
+    {
+      "epoch": 37.11,
+      "learning_rate": 4.5361007462686564e-05,
+      "loss": 0.0004,
+      "step": 39784
+    },
+    {
+      "epoch": 37.12,
+      "learning_rate": 4.536054104477612e-05,
+      "loss": 0.0015,
+      "step": 39788
+    },
+    {
+      "epoch": 37.12,
+      "learning_rate": 4.5360074626865674e-05,
+      "loss": 0.0009,
+      "step": 39792
+    },
+    {
+      "epoch": 37.12,
+      "learning_rate": 4.535960820895523e-05,
+      "loss": 0.0,
+      "step": 39796
+    },
+    {
+      "epoch": 37.13,
+      "learning_rate": 4.535914179104478e-05,
+      "loss": 0.0,
+      "step": 39800
+    },
+    {
+      "epoch": 37.13,
+      "learning_rate": 4.535867537313433e-05,
+      "loss": 0.0018,
+      "step": 39804
+    },
+    {
+      "epoch": 37.13,
+      "learning_rate": 4.535820895522389e-05,
+      "loss": 0.0001,
+      "step": 39808
+    },
+    {
+      "epoch": 37.14,
+      "learning_rate": 4.5357742537313435e-05,
+      "loss": 0.0003,
+      "step": 39812
+    },
+    {
+      "epoch": 37.14,
+      "learning_rate": 4.535727611940298e-05,
+      "loss": 0.002,
+      "step": 39816
+    },
+    {
+      "epoch": 37.15,
+      "learning_rate": 4.5356809701492545e-05,
+      "loss": 0.0006,
+      "step": 39820
+    },
+    {
+      "epoch": 37.15,
+      "learning_rate": 4.535634328358209e-05,
+      "loss": 0.001,
+      "step": 39824
+    },
+    {
+      "epoch": 37.15,
+      "learning_rate": 4.535587686567164e-05,
+      "loss": 0.0003,
+      "step": 39828
+    },
+    {
+      "epoch": 37.16,
+      "learning_rate": 4.5355410447761196e-05,
+      "loss": 0.0011,
+      "step": 39832
+    },
+    {
+      "epoch": 37.16,
+      "learning_rate": 4.535494402985075e-05,
+      "loss": 0.0001,
+      "step": 39836
+    },
+    {
+      "epoch": 37.16,
+      "learning_rate": 4.53544776119403e-05,
+      "loss": 0.0002,
+      "step": 39840
+    },
+    {
+      "epoch": 37.17,
+      "learning_rate": 4.5354011194029854e-05,
+      "loss": 0.0001,
+      "step": 39844
+    },
+    {
+      "epoch": 37.17,
+      "learning_rate": 4.53535447761194e-05,
+      "loss": 0.0,
+      "step": 39848
+    },
+    {
+      "epoch": 37.18,
+      "learning_rate": 4.535307835820896e-05,
+      "loss": 0.0,
+      "step": 39852
+    },
+    {
+      "epoch": 37.18,
+      "learning_rate": 4.535261194029851e-05,
+      "loss": 0.0001,
+      "step": 39856
+    },
+    {
+      "epoch": 37.18,
+      "learning_rate": 4.535214552238806e-05,
+      "loss": 0.0,
+      "step": 39860
+    },
+    {
+      "epoch": 37.19,
+      "learning_rate": 4.5351679104477615e-05,
+      "loss": 0.0002,
+      "step": 39864
+    },
+    {
+      "epoch": 37.19,
+      "learning_rate": 4.535121268656717e-05,
+      "loss": 0.001,
+      "step": 39868
+    },
+    {
+      "epoch": 37.19,
+      "learning_rate": 4.535074626865672e-05,
+      "loss": 0.0001,
+      "step": 39872
+    },
+    {
+      "epoch": 37.2,
+      "learning_rate": 4.5350279850746266e-05,
+      "loss": 0.0003,
+      "step": 39876
+    },
+    {
+      "epoch": 37.2,
+      "learning_rate": 4.534981343283583e-05,
+      "loss": 0.0,
+      "step": 39880
+    },
+    {
+      "epoch": 37.21,
+      "learning_rate": 4.5349347014925376e-05,
+      "loss": 0.0006,
+      "step": 39884
+    },
+    {
+      "epoch": 37.21,
+      "learning_rate": 4.5348880597014924e-05,
+      "loss": 0.0006,
+      "step": 39888
+    },
+    {
+      "epoch": 37.21,
+      "learning_rate": 4.534841417910448e-05,
+      "loss": 0.0009,
+      "step": 39892
+    },
+    {
+      "epoch": 37.22,
+      "learning_rate": 4.5347947761194034e-05,
+      "loss": 0.0,
+      "step": 39896
+    },
+    {
+      "epoch": 37.22,
+      "learning_rate": 4.534748134328358e-05,
+      "loss": 0.0002,
+      "step": 39900
+    },
+    {
+      "epoch": 37.22,
+      "learning_rate": 4.534701492537314e-05,
+      "loss": 0.0,
+      "step": 39904
+    },
+    {
+      "epoch": 37.23,
+      "learning_rate": 4.5346548507462685e-05,
+      "loss": 0.0004,
+      "step": 39908
+    },
+    {
+      "epoch": 37.23,
+      "learning_rate": 4.534608208955224e-05,
+      "loss": 0.0,
+      "step": 39912
+    },
+    {
+      "epoch": 37.24,
+      "learning_rate": 4.5345615671641795e-05,
+      "loss": 0.0,
+      "step": 39916
+    },
+    {
+      "epoch": 37.24,
+      "learning_rate": 4.534514925373134e-05,
+      "loss": 0.0002,
+      "step": 39920
+    },
+    {
+      "epoch": 37.24,
+      "learning_rate": 4.53446828358209e-05,
+      "loss": 0.0001,
+      "step": 39924
+    },
+    {
+      "epoch": 37.25,
+      "learning_rate": 4.534421641791045e-05,
+      "loss": 0.0015,
+      "step": 39928
+    },
+    {
+      "epoch": 37.25,
+      "learning_rate": 4.534375e-05,
+      "loss": 0.0,
+      "step": 39932
+    },
+    {
+      "epoch": 37.25,
+      "learning_rate": 4.534328358208955e-05,
+      "loss": 0.0,
+      "step": 39936
+    },
+    {
+      "epoch": 37.26,
+      "learning_rate": 4.534281716417911e-05,
+      "loss": 0.0,
+      "step": 39940
+    },
+    {
+      "epoch": 37.26,
+      "learning_rate": 4.534235074626866e-05,
+      "loss": 0.0,
+      "step": 39944
+    },
+    {
+      "epoch": 37.26,
+      "learning_rate": 4.534188432835821e-05,
+      "loss": 0.0003,
+      "step": 39948
+    },
+    {
+      "epoch": 37.27,
+      "learning_rate": 4.534141791044776e-05,
+      "loss": 0.0,
+      "step": 39952
+    },
+    {
+      "epoch": 37.27,
+      "learning_rate": 4.534095149253732e-05,
+      "loss": 0.0006,
+      "step": 39956
+    },
+    {
+      "epoch": 37.28,
+      "learning_rate": 4.534048507462687e-05,
+      "loss": 0.0017,
+      "step": 39960
+    },
+    {
+      "epoch": 37.28,
+      "learning_rate": 4.534001865671642e-05,
+      "loss": 0.0028,
+      "step": 39964
+    },
+    {
+      "epoch": 37.28,
+      "learning_rate": 4.533955223880597e-05,
+      "loss": 0.0019,
+      "step": 39968
+    },
+    {
+      "epoch": 37.29,
+      "learning_rate": 4.533908582089553e-05,
+      "loss": 0.0001,
+      "step": 39972
+    },
+    {
+      "epoch": 37.29,
+      "learning_rate": 4.533861940298508e-05,
+      "loss": 0.0009,
+      "step": 39976
+    },
+    {
+      "epoch": 37.29,
+      "learning_rate": 4.5338152985074626e-05,
+      "loss": 0.0059,
+      "step": 39980
+    },
+    {
+      "epoch": 37.3,
+      "learning_rate": 4.533768656716418e-05,
+      "loss": 0.0,
+      "step": 39984
+    },
+    {
+      "epoch": 37.3,
+      "learning_rate": 4.5337220149253736e-05,
+      "loss": 0.0034,
+      "step": 39988
+    },
+    {
+      "epoch": 37.31,
+      "learning_rate": 4.5336753731343284e-05,
+      "loss": 0.0,
+      "step": 39992
+    },
+    {
+      "epoch": 37.31,
+      "learning_rate": 4.533628731343284e-05,
+      "loss": 0.0012,
+      "step": 39996
+    },
+    {
+      "epoch": 37.31,
+      "learning_rate": 4.5335820895522394e-05,
+      "loss": 0.0001,
+      "step": 40000
+    },
+    {
+      "epoch": 37.31,
+      "eval_exact_match": 0.7282398452611218,
+      "eval_exec": 0.7514506769825918,
+      "eval_loss": 0.42848145961761475,
+      "eval_runtime": 1107.5429,
+      "eval_samples_per_second": 0.934,
+      "step": 40000
+    },
+    {
+      "epoch": 37.32,
+      "learning_rate": 4.533535447761194e-05,
+      "loss": 0.0,
+      "step": 40004
+    },
+    {
+      "epoch": 37.32,
+      "learning_rate": 4.53348880597015e-05,
+      "loss": 0.0039,
+      "step": 40008
+    },
+    {
+      "epoch": 37.32,
+      "learning_rate": 4.5334421641791045e-05,
+      "loss": 0.0006,
+      "step": 40012
+    },
+    {
+      "epoch": 37.33,
+      "learning_rate": 4.53339552238806e-05,
+      "loss": 0.0014,
+      "step": 40016
+    },
+    {
+      "epoch": 37.33,
+      "learning_rate": 4.5333488805970155e-05,
+      "loss": 0.0025,
+      "step": 40020
+    },
+    {
+      "epoch": 37.34,
+      "learning_rate": 4.53330223880597e-05,
+      "loss": 0.0002,
+      "step": 40024
+    },
+    {
+      "epoch": 37.34,
+      "learning_rate": 4.533255597014925e-05,
+      "loss": 0.0001,
+      "step": 40028
+    },
+    {
+      "epoch": 37.34,
+      "learning_rate": 4.533208955223881e-05,
+      "loss": 0.0017,
+      "step": 40032
+    },
+    {
+      "epoch": 37.35,
+      "learning_rate": 4.533162313432836e-05,
+      "loss": 0.0,
+      "step": 40036
+    },
+    {
+      "epoch": 37.35,
+      "learning_rate": 4.533115671641791e-05,
+      "loss": 0.0001,
+      "step": 40040
+    },
+    {
+      "epoch": 37.35,
+      "learning_rate": 4.5330690298507464e-05,
+      "loss": 0.0001,
+      "step": 40044
+    },
+    {
+      "epoch": 37.36,
+      "learning_rate": 4.533022388059702e-05,
+      "loss": 0.0,
+      "step": 40048
+    },
+    {
+      "epoch": 37.36,
+      "learning_rate": 4.532975746268657e-05,
+      "loss": 0.0002,
+      "step": 40052
+    },
+    {
+      "epoch": 37.37,
+      "learning_rate": 4.532929104477612e-05,
+      "loss": 0.0016,
+      "step": 40056
+    },
+    {
+      "epoch": 37.37,
+      "learning_rate": 4.532882462686568e-05,
+      "loss": 0.0,
+      "step": 40060
+    },
+    {
+      "epoch": 37.37,
+      "learning_rate": 4.5328358208955225e-05,
+      "loss": 0.0003,
+      "step": 40064
+    },
+    {
+      "epoch": 37.38,
+      "learning_rate": 4.532789179104478e-05,
+      "loss": 0.0003,
+      "step": 40068
+    },
+    {
+      "epoch": 37.38,
+      "learning_rate": 4.532742537313433e-05,
+      "loss": 0.0012,
+      "step": 40072
+    },
+    {
+      "epoch": 37.38,
+      "learning_rate": 4.532695895522388e-05,
+      "loss": 0.0,
+      "step": 40076
+    },
+    {
+      "epoch": 37.39,
+      "learning_rate": 4.532649253731344e-05,
+      "loss": 0.0,
+      "step": 40080
+    },
+    {
+      "epoch": 37.39,
+      "learning_rate": 4.5326026119402986e-05,
+      "loss": 0.0002,
+      "step": 40084
+    },
+    {
+      "epoch": 37.4,
+      "learning_rate": 4.5325559701492534e-05,
+      "loss": 0.0001,
+      "step": 40088
+    },
+    {
+      "epoch": 37.4,
+      "learning_rate": 4.5325093283582096e-05,
+      "loss": 0.0004,
+      "step": 40092
+    },
+    {
+      "epoch": 37.4,
+      "learning_rate": 4.5324626865671644e-05,
+      "loss": 0.0038,
+      "step": 40096
+    },
+    {
+      "epoch": 37.41,
+      "learning_rate": 4.532416044776119e-05,
+      "loss": 0.0,
+      "step": 40100
+    },
+    {
+      "epoch": 37.41,
+      "learning_rate": 4.532369402985075e-05,
+      "loss": 0.001,
+      "step": 40104
+    },
+    {
+      "epoch": 37.41,
+      "learning_rate": 4.53232276119403e-05,
+      "loss": 0.0088,
+      "step": 40108
+    },
+    {
+      "epoch": 37.42,
+      "learning_rate": 4.532276119402985e-05,
+      "loss": 0.0,
+      "step": 40112
+    },
+    {
+      "epoch": 37.42,
+      "learning_rate": 4.5322294776119405e-05,
+      "loss": 0.001,
+      "step": 40116
+    },
+    {
+      "epoch": 37.43,
+      "learning_rate": 4.532182835820896e-05,
+      "loss": 0.0009,
+      "step": 40120
+    },
+    {
+      "epoch": 37.43,
+      "learning_rate": 4.5321361940298515e-05,
+      "loss": 0.0002,
+      "step": 40124
+    },
+    {
+      "epoch": 37.43,
+      "learning_rate": 4.532089552238806e-05,
+      "loss": 0.0002,
+      "step": 40128
+    },
+    {
+      "epoch": 37.44,
+      "learning_rate": 4.532042910447761e-05,
+      "loss": 0.0,
+      "step": 40132
+    },
+    {
+      "epoch": 37.44,
+      "learning_rate": 4.5319962686567166e-05,
+      "loss": 0.0004,
+      "step": 40136
+    },
+    {
+      "epoch": 37.44,
+      "learning_rate": 4.531949626865672e-05,
+      "loss": 0.0,
+      "step": 40140
+    },
+    {
+      "epoch": 37.45,
+      "learning_rate": 4.531902985074627e-05,
+      "loss": 0.0,
+      "step": 40144
+    },
+    {
+      "epoch": 37.45,
+      "learning_rate": 4.5318563432835824e-05,
+      "loss": 0.0,
+      "step": 40148
+    },
+    {
+      "epoch": 37.46,
+      "learning_rate": 4.531809701492538e-05,
+      "loss": 0.0001,
+      "step": 40152
+    },
+    {
+      "epoch": 37.46,
+      "learning_rate": 4.531763059701493e-05,
+      "loss": 0.0,
+      "step": 40156
+    },
+    {
+      "epoch": 37.46,
+      "learning_rate": 4.531716417910448e-05,
+      "loss": 0.0,
+      "step": 40160
+    },
+    {
+      "epoch": 37.47,
+      "learning_rate": 4.531669776119403e-05,
+      "loss": 0.0004,
+      "step": 40164
+    },
+    {
+      "epoch": 37.47,
+      "learning_rate": 4.5316231343283585e-05,
+      "loss": 0.0003,
+      "step": 40168
+    },
+    {
+      "epoch": 37.47,
+      "learning_rate": 4.531576492537314e-05,
+      "loss": 0.002,
+      "step": 40172
+    },
+    {
+      "epoch": 37.48,
+      "learning_rate": 4.531529850746269e-05,
+      "loss": 0.0,
+      "step": 40176
+    },
+    {
+      "epoch": 37.48,
+      "learning_rate": 4.531483208955224e-05,
+      "loss": 0.0005,
+      "step": 40180
+    },
+    {
+      "epoch": 37.49,
+      "learning_rate": 4.53143656716418e-05,
+      "loss": 0.0011,
+      "step": 40184
+    },
+    {
+      "epoch": 37.49,
+      "learning_rate": 4.5313899253731346e-05,
+      "loss": 0.0002,
+      "step": 40188
+    },
+    {
+      "epoch": 37.49,
+      "learning_rate": 4.5313432835820894e-05,
+      "loss": 0.0003,
+      "step": 40192
+    },
+    {
+      "epoch": 37.5,
+      "learning_rate": 4.531296641791045e-05,
+      "loss": 0.0002,
+      "step": 40196
+    },
+    {
+      "epoch": 37.5,
+      "learning_rate": 4.5312500000000004e-05,
+      "loss": 0.0,
+      "step": 40200
+    },
+    {
+      "epoch": 37.5,
+      "learning_rate": 4.531203358208955e-05,
+      "loss": 0.0,
+      "step": 40204
+    },
+    {
+      "epoch": 37.51,
+      "learning_rate": 4.531156716417911e-05,
+      "loss": 0.0011,
+      "step": 40208
+    },
+    {
+      "epoch": 37.51,
+      "learning_rate": 4.531110074626866e-05,
+      "loss": 0.0003,
+      "step": 40212
+    },
+    {
+      "epoch": 37.51,
+      "learning_rate": 4.531063432835821e-05,
+      "loss": 0.0001,
+      "step": 40216
+    },
+    {
+      "epoch": 37.52,
+      "learning_rate": 4.5310167910447765e-05,
+      "loss": 0.0002,
+      "step": 40220
+    },
+    {
+      "epoch": 37.52,
+      "learning_rate": 4.530970149253731e-05,
+      "loss": 0.0004,
+      "step": 40224
+    },
+    {
+      "epoch": 37.53,
+      "learning_rate": 4.530923507462687e-05,
+      "loss": 0.0001,
+      "step": 40228
+    },
+    {
+      "epoch": 37.53,
+      "learning_rate": 4.530876865671642e-05,
+      "loss": 0.0,
+      "step": 40232
+    },
+    {
+      "epoch": 37.53,
+      "learning_rate": 4.530830223880597e-05,
+      "loss": 0.0,
+      "step": 40236
+    },
+    {
+      "epoch": 37.54,
+      "learning_rate": 4.5307835820895526e-05,
+      "loss": 0.0,
+      "step": 40240
+    },
+    {
+      "epoch": 37.54,
+      "learning_rate": 4.530736940298508e-05,
+      "loss": 0.0001,
+      "step": 40244
+    },
+    {
+      "epoch": 37.54,
+      "learning_rate": 4.530690298507463e-05,
+      "loss": 0.0001,
+      "step": 40248
+    },
+    {
+      "epoch": 37.55,
+      "learning_rate": 4.530643656716418e-05,
+      "loss": 0.0005,
+      "step": 40252
+    },
+    {
+      "epoch": 37.55,
+      "learning_rate": 4.530597014925373e-05,
+      "loss": 0.0003,
+      "step": 40256
+    },
+    {
+      "epoch": 37.56,
+      "learning_rate": 4.5305503731343287e-05,
+      "loss": 0.0009,
+      "step": 40260
+    },
+    {
+      "epoch": 37.56,
+      "learning_rate": 4.5305037313432835e-05,
+      "loss": 0.0,
+      "step": 40264
+    },
+    {
+      "epoch": 37.56,
+      "learning_rate": 4.530457089552239e-05,
+      "loss": 0.0017,
+      "step": 40268
+    },
+    {
+      "epoch": 37.57,
+      "learning_rate": 4.5304104477611945e-05,
+      "loss": 0.0001,
+      "step": 40272
+    },
+    {
+      "epoch": 37.57,
+      "learning_rate": 4.530363805970149e-05,
+      "loss": 0.0,
+      "step": 40276
+    },
+    {
+      "epoch": 37.57,
+      "learning_rate": 4.530317164179105e-05,
+      "loss": 0.0001,
+      "step": 40280
+    },
+    {
+      "epoch": 37.58,
+      "learning_rate": 4.5302705223880596e-05,
+      "loss": 0.0002,
+      "step": 40284
+    },
+    {
+      "epoch": 37.58,
+      "learning_rate": 4.530223880597016e-05,
+      "loss": 0.0019,
+      "step": 40288
+    },
+    {
+      "epoch": 37.59,
+      "learning_rate": 4.5301772388059706e-05,
+      "loss": 0.0,
+      "step": 40292
+    },
+    {
+      "epoch": 37.59,
+      "learning_rate": 4.5301305970149254e-05,
+      "loss": 0.0,
+      "step": 40296
+    },
+    {
+      "epoch": 37.59,
+      "learning_rate": 4.530083955223881e-05,
+      "loss": 0.0057,
+      "step": 40300
+    },
+    {
+      "epoch": 37.6,
+      "learning_rate": 4.5300373134328363e-05,
+      "loss": 0.0,
+      "step": 40304
+    },
+    {
+      "epoch": 37.6,
+      "learning_rate": 4.529990671641791e-05,
+      "loss": 0.0,
+      "step": 40308
+    },
+    {
+      "epoch": 37.6,
+      "learning_rate": 4.5299440298507467e-05,
+      "loss": 0.0036,
+      "step": 40312
+    },
+    {
+      "epoch": 37.61,
+      "learning_rate": 4.5298973880597015e-05,
+      "loss": 0.0002,
+      "step": 40316
+    },
+    {
+      "epoch": 37.61,
+      "learning_rate": 4.529850746268657e-05,
+      "loss": 0.0001,
+      "step": 40320
+    },
+    {
+      "epoch": 37.62,
+      "learning_rate": 4.5298041044776124e-05,
+      "loss": 0.0006,
+      "step": 40324
+    },
+    {
+      "epoch": 37.62,
+      "learning_rate": 4.529757462686567e-05,
+      "loss": 0.0001,
+      "step": 40328
+    },
+    {
+      "epoch": 37.62,
+      "learning_rate": 4.529710820895523e-05,
+      "loss": 0.0,
+      "step": 40332
+    },
+    {
+      "epoch": 37.63,
+      "learning_rate": 4.529664179104478e-05,
+      "loss": 0.0013,
+      "step": 40336
+    },
+    {
+      "epoch": 37.63,
+      "learning_rate": 4.529617537313433e-05,
+      "loss": 0.0,
+      "step": 40340
+    },
+    {
+      "epoch": 37.63,
+      "learning_rate": 4.529570895522388e-05,
+      "loss": 0.0001,
+      "step": 40344
+    },
+    {
+      "epoch": 37.64,
+      "learning_rate": 4.529524253731344e-05,
+      "loss": 0.0,
+      "step": 40348
+    },
+    {
+      "epoch": 37.64,
+      "learning_rate": 4.529477611940299e-05,
+      "loss": 0.0002,
+      "step": 40352
+    },
+    {
+      "epoch": 37.65,
+      "learning_rate": 4.529430970149254e-05,
+      "loss": 0.0017,
+      "step": 40356
+    },
+    {
+      "epoch": 37.65,
+      "learning_rate": 4.529384328358209e-05,
+      "loss": 0.0001,
+      "step": 40360
+    },
+    {
+      "epoch": 37.65,
+      "learning_rate": 4.5293376865671646e-05,
+      "loss": 0.0,
+      "step": 40364
+    },
+    {
+      "epoch": 37.66,
+      "learning_rate": 4.5292910447761195e-05,
+      "loss": 0.0007,
+      "step": 40368
+    },
+    {
+      "epoch": 37.66,
+      "learning_rate": 4.529244402985075e-05,
+      "loss": 0.0,
+      "step": 40372
+    },
+    {
+      "epoch": 37.66,
+      "learning_rate": 4.52919776119403e-05,
+      "loss": 0.0,
+      "step": 40376
+    },
+    {
+      "epoch": 37.67,
+      "learning_rate": 4.529151119402985e-05,
+      "loss": 0.0,
+      "step": 40380
+    },
+    {
+      "epoch": 37.67,
+      "learning_rate": 4.529104477611941e-05,
+      "loss": 0.0001,
+      "step": 40384
+    },
+    {
+      "epoch": 37.68,
+      "learning_rate": 4.5290578358208956e-05,
+      "loss": 0.0013,
+      "step": 40388
+    },
+    {
+      "epoch": 37.68,
+      "learning_rate": 4.529011194029851e-05,
+      "loss": 0.02,
+      "step": 40392
+    },
+    {
+      "epoch": 37.68,
+      "learning_rate": 4.5289645522388065e-05,
+      "loss": 0.0,
+      "step": 40396
+    },
+    {
+      "epoch": 37.69,
+      "learning_rate": 4.5289179104477614e-05,
+      "loss": 0.0,
+      "step": 40400
+    },
+    {
+      "epoch": 37.69,
+      "learning_rate": 4.528871268656716e-05,
+      "loss": 0.0001,
+      "step": 40404
+    },
+    {
+      "epoch": 37.69,
+      "learning_rate": 4.5288246268656717e-05,
+      "loss": 0.0001,
+      "step": 40408
+    },
+    {
+      "epoch": 37.7,
+      "learning_rate": 4.528777985074627e-05,
+      "loss": 0.0001,
+      "step": 40412
+    },
+    {
+      "epoch": 37.7,
+      "learning_rate": 4.528731343283582e-05,
+      "loss": 0.0001,
+      "step": 40416
+    },
+    {
+      "epoch": 37.71,
+      "learning_rate": 4.5286847014925374e-05,
+      "loss": 0.0,
+      "step": 40420
+    },
+    {
+      "epoch": 37.71,
+      "learning_rate": 4.528638059701493e-05,
+      "loss": 0.0002,
+      "step": 40424
+    },
+    {
+      "epoch": 37.71,
+      "learning_rate": 4.528591417910448e-05,
+      "loss": 0.0071,
+      "step": 40428
+    },
+    {
+      "epoch": 37.72,
+      "learning_rate": 4.528544776119403e-05,
+      "loss": 0.0001,
+      "step": 40432
+    },
+    {
+      "epoch": 37.72,
+      "learning_rate": 4.528498134328358e-05,
+      "loss": 0.0001,
+      "step": 40436
+    },
+    {
+      "epoch": 37.72,
+      "learning_rate": 4.5284514925373135e-05,
+      "loss": 0.0004,
+      "step": 40440
+    },
+    {
+      "epoch": 37.73,
+      "learning_rate": 4.528404850746269e-05,
+      "loss": 0.0006,
+      "step": 40444
+    },
+    {
+      "epoch": 37.73,
+      "learning_rate": 4.528358208955224e-05,
+      "loss": 0.0,
+      "step": 40448
+    },
+    {
+      "epoch": 37.73,
+      "learning_rate": 4.5283115671641793e-05,
+      "loss": 0.0,
+      "step": 40452
+    },
+    {
+      "epoch": 37.74,
+      "learning_rate": 4.528264925373135e-05,
+      "loss": 0.0001,
+      "step": 40456
+    },
+    {
+      "epoch": 37.74,
+      "learning_rate": 4.5282182835820896e-05,
+      "loss": 0.0001,
+      "step": 40460
+    },
+    {
+      "epoch": 37.75,
+      "learning_rate": 4.528171641791045e-05,
+      "loss": 0.0003,
+      "step": 40464
+    },
+    {
+      "epoch": 37.75,
+      "learning_rate": 4.528125e-05,
+      "loss": 0.0015,
+      "step": 40468
+    },
+    {
+      "epoch": 37.75,
+      "learning_rate": 4.5280783582089554e-05,
+      "loss": 0.0029,
+      "step": 40472
+    },
+    {
+      "epoch": 37.76,
+      "learning_rate": 4.528031716417911e-05,
+      "loss": 0.0001,
+      "step": 40476
+    },
+    {
+      "epoch": 37.76,
+      "learning_rate": 4.527985074626866e-05,
+      "loss": 0.0,
+      "step": 40480
+    },
+    {
+      "epoch": 37.76,
+      "learning_rate": 4.527938432835821e-05,
+      "loss": 0.0002,
+      "step": 40484
+    },
+    {
+      "epoch": 37.77,
+      "learning_rate": 4.527891791044777e-05,
+      "loss": 0.0014,
+      "step": 40488
+    },
+    {
+      "epoch": 37.77,
+      "learning_rate": 4.5278451492537315e-05,
+      "loss": 0.0001,
+      "step": 40492
+    },
+    {
+      "epoch": 37.78,
+      "learning_rate": 4.5277985074626864e-05,
+      "loss": 0.0,
+      "step": 40496
+    },
+    {
+      "epoch": 37.78,
+      "learning_rate": 4.5277518656716425e-05,
+      "loss": 0.012,
+      "step": 40500
+    },
+    {
+      "epoch": 37.78,
+      "eval_exact_match": 0.7379110251450677,
+      "eval_exec": 0.769825918762089,
+      "eval_loss": 0.41693997383117676,
+      "eval_runtime": 1159.3081,
+      "eval_samples_per_second": 0.892,
+      "step": 40500
+    },
+    {
+      "epoch": 37.78,
+      "learning_rate": 4.527705223880597e-05,
+      "loss": 0.0001,
+      "step": 40504
+    },
+    {
+      "epoch": 37.79,
+      "learning_rate": 4.527658582089552e-05,
+      "loss": 0.0,
+      "step": 40508
+    },
+    {
+      "epoch": 37.79,
+      "learning_rate": 4.5276119402985076e-05,
+      "loss": 0.0003,
+      "step": 40512
+    },
+    {
+      "epoch": 37.79,
+      "learning_rate": 4.527565298507463e-05,
+      "loss": 0.0001,
+      "step": 40516
+    },
+    {
+      "epoch": 37.8,
+      "learning_rate": 4.527518656716418e-05,
+      "loss": 0.0007,
+      "step": 40520
+    },
+    {
+      "epoch": 37.8,
+      "learning_rate": 4.5274720149253734e-05,
+      "loss": 0.0003,
+      "step": 40524
+    },
+    {
+      "epoch": 37.81,
+      "learning_rate": 4.527425373134328e-05,
+      "loss": 0.0,
+      "step": 40528
+    },
+    {
+      "epoch": 37.81,
+      "learning_rate": 4.527378731343284e-05,
+      "loss": 0.0003,
+      "step": 40532
+    },
+    {
+      "epoch": 37.81,
+      "learning_rate": 4.527332089552239e-05,
+      "loss": 0.0007,
+      "step": 40536
+    },
+    {
+      "epoch": 37.82,
+      "learning_rate": 4.527285447761194e-05,
+      "loss": 0.0001,
+      "step": 40540
+    },
+    {
+      "epoch": 37.82,
+      "learning_rate": 4.5272388059701495e-05,
+      "loss": 0.0029,
+      "step": 40544
+    },
+    {
+      "epoch": 37.82,
+      "learning_rate": 4.527192164179105e-05,
+      "loss": 0.0,
+      "step": 40548
+    },
+    {
+      "epoch": 37.83,
+      "learning_rate": 4.52714552238806e-05,
+      "loss": 0.0005,
+      "step": 40552
+    },
+    {
+      "epoch": 37.83,
+      "learning_rate": 4.5270988805970146e-05,
+      "loss": 0.0001,
+      "step": 40556
+    },
+    {
+      "epoch": 37.84,
+      "learning_rate": 4.527052238805971e-05,
+      "loss": 0.0002,
+      "step": 40560
+    },
+    {
+      "epoch": 37.84,
+      "learning_rate": 4.5270055970149256e-05,
+      "loss": 0.0,
+      "step": 40564
+    },
+    {
+      "epoch": 37.84,
+      "learning_rate": 4.5269589552238804e-05,
+      "loss": 0.0,
+      "step": 40568
+    },
+    {
+      "epoch": 37.85,
+      "learning_rate": 4.526912313432836e-05,
+      "loss": 0.0019,
+      "step": 40572
+    },
+    {
+      "epoch": 37.85,
+      "learning_rate": 4.5268656716417914e-05,
+      "loss": 0.0003,
+      "step": 40576
+    },
+    {
+      "epoch": 37.85,
+      "learning_rate": 4.526819029850746e-05,
+      "loss": 0.0002,
+      "step": 40580
+    },
+    {
+      "epoch": 37.86,
+      "learning_rate": 4.526772388059702e-05,
+      "loss": 0.0005,
+      "step": 40584
+    },
+    {
+      "epoch": 37.86,
+      "learning_rate": 4.5267257462686565e-05,
+      "loss": 0.0,
+      "step": 40588
+    },
+    {
+      "epoch": 37.87,
+      "learning_rate": 4.526679104477612e-05,
+      "loss": 0.0014,
+      "step": 40592
+    },
+    {
+      "epoch": 37.87,
+      "learning_rate": 4.5266324626865675e-05,
+      "loss": 0.0,
+      "step": 40596
+    },
+    {
+      "epoch": 37.87,
+      "learning_rate": 4.526585820895522e-05,
+      "loss": 0.0076,
+      "step": 40600
+    },
+    {
+      "epoch": 37.88,
+      "learning_rate": 4.526539179104478e-05,
+      "loss": 0.0003,
+      "step": 40604
+    },
+    {
+      "epoch": 37.88,
+      "learning_rate": 4.526492537313433e-05,
+      "loss": 0.0031,
+      "step": 40608
+    },
+    {
+      "epoch": 37.88,
+      "learning_rate": 4.526445895522388e-05,
+      "loss": 0.0001,
+      "step": 40612
+    },
+    {
+      "epoch": 37.89,
+      "learning_rate": 4.5263992537313436e-05,
+      "loss": 0.0002,
+      "step": 40616
+    },
+    {
+      "epoch": 37.89,
+      "learning_rate": 4.526352611940299e-05,
+      "loss": 0.0002,
+      "step": 40620
+    },
+    {
+      "epoch": 37.9,
+      "learning_rate": 4.526305970149254e-05,
+      "loss": 0.0004,
+      "step": 40624
+    },
+    {
+      "epoch": 37.9,
+      "learning_rate": 4.5262593283582094e-05,
+      "loss": 0.0,
+      "step": 40628
+    },
+    {
+      "epoch": 37.9,
+      "learning_rate": 4.526212686567164e-05,
+      "loss": 0.0001,
+      "step": 40632
+    },
+    {
+      "epoch": 37.91,
+      "learning_rate": 4.52616604477612e-05,
+      "loss": 0.0016,
+      "step": 40636
+    },
+    {
+      "epoch": 37.91,
+      "learning_rate": 4.526119402985075e-05,
+      "loss": 0.0,
+      "step": 40640
+    },
+    {
+      "epoch": 37.91,
+      "learning_rate": 4.52607276119403e-05,
+      "loss": 0.0002,
+      "step": 40644
+    },
+    {
+      "epoch": 37.92,
+      "learning_rate": 4.526026119402985e-05,
+      "loss": 0.0,
+      "step": 40648
+    },
+    {
+      "epoch": 37.92,
+      "learning_rate": 4.525979477611941e-05,
+      "loss": 0.0011,
+      "step": 40652
+    },
+    {
+      "epoch": 37.93,
+      "learning_rate": 4.525932835820896e-05,
+      "loss": 0.0,
+      "step": 40656
+    },
+    {
+      "epoch": 37.93,
+      "learning_rate": 4.5258861940298506e-05,
+      "loss": 0.0002,
+      "step": 40660
+    },
+    {
+      "epoch": 37.93,
+      "learning_rate": 4.525839552238806e-05,
+      "loss": 0.0008,
+      "step": 40664
+    },
+    {
+      "epoch": 37.94,
+      "learning_rate": 4.5257929104477616e-05,
+      "loss": 0.0001,
+      "step": 40668
+    },
+    {
+      "epoch": 37.94,
+      "learning_rate": 4.5257462686567164e-05,
+      "loss": 0.0021,
+      "step": 40672
+    },
+    {
+      "epoch": 37.94,
+      "learning_rate": 4.525699626865672e-05,
+      "loss": 0.0029,
+      "step": 40676
+    },
+    {
+      "epoch": 37.95,
+      "learning_rate": 4.5256529850746274e-05,
+      "loss": 0.0002,
+      "step": 40680
+    },
+    {
+      "epoch": 37.95,
+      "learning_rate": 4.525606343283582e-05,
+      "loss": 0.0024,
+      "step": 40684
+    },
+    {
+      "epoch": 37.96,
+      "learning_rate": 4.525559701492538e-05,
+      "loss": 0.0222,
+      "step": 40688
+    },
+    {
+      "epoch": 37.96,
+      "learning_rate": 4.5255130597014925e-05,
+      "loss": 0.0001,
+      "step": 40692
+    },
+    {
+      "epoch": 37.96,
+      "learning_rate": 4.525466417910448e-05,
+      "loss": 0.0001,
+      "step": 40696
+    },
+    {
+      "epoch": 37.97,
+      "learning_rate": 4.5254197761194035e-05,
+      "loss": 0.0037,
+      "step": 40700
+    },
+    {
+      "epoch": 37.97,
+      "learning_rate": 4.525373134328358e-05,
+      "loss": 0.0008,
+      "step": 40704
+    },
+    {
+      "epoch": 37.97,
+      "learning_rate": 4.525326492537313e-05,
+      "loss": 0.0002,
+      "step": 40708
+    },
+    {
+      "epoch": 37.98,
+      "learning_rate": 4.525279850746269e-05,
+      "loss": 0.0007,
+      "step": 40712
+    },
+    {
+      "epoch": 37.98,
+      "learning_rate": 4.525233208955224e-05,
+      "loss": 0.0,
+      "step": 40716
+    },
+    {
+      "epoch": 37.98,
+      "learning_rate": 4.525186567164179e-05,
+      "loss": 0.0,
+      "step": 40720
+    },
+    {
+      "epoch": 37.99,
+      "learning_rate": 4.5251399253731344e-05,
+      "loss": 0.0,
+      "step": 40724
+    },
+    {
+      "epoch": 37.99,
+      "learning_rate": 4.52509328358209e-05,
+      "loss": 0.0001,
+      "step": 40728
+    },
+    {
+      "epoch": 38.0,
+      "learning_rate": 4.525046641791045e-05,
+      "loss": 0.001,
+      "step": 40732
+    },
+    {
+      "epoch": 38.0,
+      "learning_rate": 4.525e-05,
+      "loss": 0.0001,
+      "step": 40736
+    },
+    {
+      "epoch": 38.0,
+      "learning_rate": 4.524953358208956e-05,
+      "loss": 0.0,
+      "step": 40740
+    },
+    {
+      "epoch": 38.01,
+      "learning_rate": 4.5249067164179105e-05,
+      "loss": 0.0,
+      "step": 40744
+    },
+    {
+      "epoch": 38.01,
+      "learning_rate": 4.524860074626866e-05,
+      "loss": 0.0,
+      "step": 40748
+    },
+    {
+      "epoch": 38.01,
+      "learning_rate": 4.524813432835821e-05,
+      "loss": 0.0,
+      "step": 40752
+    },
+    {
+      "epoch": 38.02,
+      "learning_rate": 4.524766791044776e-05,
+      "loss": 0.0001,
+      "step": 40756
+    },
+    {
+      "epoch": 38.02,
+      "learning_rate": 4.524720149253732e-05,
+      "loss": 0.0,
+      "step": 40760
+    },
+    {
+      "epoch": 38.03,
+      "learning_rate": 4.5246735074626866e-05,
+      "loss": 0.0001,
+      "step": 40764
+    },
+    {
+      "epoch": 38.03,
+      "learning_rate": 4.5246268656716414e-05,
+      "loss": 0.0001,
+      "step": 40768
+    },
+    {
+      "epoch": 38.03,
+      "learning_rate": 4.5245802238805976e-05,
+      "loss": 0.0001,
+      "step": 40772
+    },
+    {
+      "epoch": 38.04,
+      "learning_rate": 4.5245335820895524e-05,
+      "loss": 0.0012,
+      "step": 40776
+    },
+    {
+      "epoch": 38.04,
+      "learning_rate": 4.524486940298508e-05,
+      "loss": 0.0003,
+      "step": 40780
+    },
+    {
+      "epoch": 38.04,
+      "learning_rate": 4.524440298507463e-05,
+      "loss": 0.0003,
+      "step": 40784
+    },
+    {
+      "epoch": 38.05,
+      "learning_rate": 4.524393656716418e-05,
+      "loss": 0.0001,
+      "step": 40788
+    },
+    {
+      "epoch": 38.05,
+      "learning_rate": 4.524347014925374e-05,
+      "loss": 0.0014,
+      "step": 40792
+    },
+    {
+      "epoch": 38.06,
+      "learning_rate": 4.5243003731343285e-05,
+      "loss": 0.0007,
+      "step": 40796
+    },
+    {
+      "epoch": 38.06,
+      "learning_rate": 4.524253731343284e-05,
+      "loss": 0.0045,
+      "step": 40800
+    },
+    {
+      "epoch": 38.06,
+      "learning_rate": 4.5242070895522395e-05,
+      "loss": 0.0002,
+      "step": 40804
+    },
+    {
+      "epoch": 38.07,
+      "learning_rate": 4.524160447761194e-05,
+      "loss": 0.0011,
+      "step": 40808
+    },
+    {
+      "epoch": 38.07,
+      "learning_rate": 4.524113805970149e-05,
+      "loss": 0.0001,
+      "step": 40812
+    },
+    {
+      "epoch": 38.07,
+      "learning_rate": 4.5240671641791046e-05,
+      "loss": 0.0,
+      "step": 40816
+    },
+    {
+      "epoch": 38.08,
+      "learning_rate": 4.52402052238806e-05,
+      "loss": 0.0006,
+      "step": 40820
+    },
+    {
+      "epoch": 38.08,
+      "learning_rate": 4.523973880597015e-05,
+      "loss": 0.001,
+      "step": 40824
+    },
+    {
+      "epoch": 38.09,
+      "learning_rate": 4.5239272388059704e-05,
+      "loss": 0.0,
+      "step": 40828
+    },
+    {
+      "epoch": 38.09,
+      "learning_rate": 4.523880597014926e-05,
+      "loss": 0.0,
+      "step": 40832
+    },
+    {
+      "epoch": 38.09,
+      "learning_rate": 4.523833955223881e-05,
+      "loss": 0.0007,
+      "step": 40836
+    },
+    {
+      "epoch": 38.1,
+      "learning_rate": 4.523787313432836e-05,
+      "loss": 0.0,
+      "step": 40840
+    },
+    {
+      "epoch": 38.1,
+      "learning_rate": 4.523740671641791e-05,
+      "loss": 0.0003,
+      "step": 40844
+    },
+    {
+      "epoch": 38.1,
+      "learning_rate": 4.5236940298507465e-05,
+      "loss": 0.0009,
+      "step": 40848
+    },
+    {
+      "epoch": 38.11,
+      "learning_rate": 4.523647388059702e-05,
+      "loss": 0.0041,
+      "step": 40852
+    },
+    {
+      "epoch": 38.11,
+      "learning_rate": 4.523600746268657e-05,
+      "loss": 0.0011,
+      "step": 40856
+    },
+    {
+      "epoch": 38.12,
+      "learning_rate": 4.523554104477612e-05,
+      "loss": 0.0,
+      "step": 40860
+    },
+    {
+      "epoch": 38.12,
+      "learning_rate": 4.523507462686568e-05,
+      "loss": 0.0002,
+      "step": 40864
+    },
+    {
+      "epoch": 38.12,
+      "learning_rate": 4.5234608208955226e-05,
+      "loss": 0.0001,
+      "step": 40868
+    },
+    {
+      "epoch": 38.13,
+      "learning_rate": 4.5234141791044774e-05,
+      "loss": 0.0,
+      "step": 40872
+    },
+    {
+      "epoch": 38.13,
+      "learning_rate": 4.523367537313433e-05,
+      "loss": 0.0001,
+      "step": 40876
+    },
+    {
+      "epoch": 38.13,
+      "learning_rate": 4.5233208955223884e-05,
+      "loss": 0.0,
+      "step": 40880
+    },
+    {
+      "epoch": 38.14,
+      "learning_rate": 4.523274253731343e-05,
+      "loss": 0.0003,
+      "step": 40884
+    },
+    {
+      "epoch": 38.14,
+      "learning_rate": 4.523227611940299e-05,
+      "loss": 0.0001,
+      "step": 40888
+    },
+    {
+      "epoch": 38.15,
+      "learning_rate": 4.523180970149254e-05,
+      "loss": 0.0015,
+      "step": 40892
+    },
+    {
+      "epoch": 38.15,
+      "learning_rate": 4.523134328358209e-05,
+      "loss": 0.0003,
+      "step": 40896
+    },
+    {
+      "epoch": 38.15,
+      "learning_rate": 4.5230876865671645e-05,
+      "loss": 0.001,
+      "step": 40900
+    },
+    {
+      "epoch": 38.16,
+      "learning_rate": 4.523041044776119e-05,
+      "loss": 0.0016,
+      "step": 40904
+    },
+    {
+      "epoch": 38.16,
+      "learning_rate": 4.522994402985075e-05,
+      "loss": 0.0023,
+      "step": 40908
+    },
+    {
+      "epoch": 38.16,
+      "learning_rate": 4.52294776119403e-05,
+      "loss": 0.0,
+      "step": 40912
+    },
+    {
+      "epoch": 38.17,
+      "learning_rate": 4.522901119402985e-05,
+      "loss": 0.0004,
+      "step": 40916
+    },
+    {
+      "epoch": 38.17,
+      "learning_rate": 4.5228544776119406e-05,
+      "loss": 0.0001,
+      "step": 40920
+    },
+    {
+      "epoch": 38.18,
+      "learning_rate": 4.522807835820896e-05,
+      "loss": 0.0,
+      "step": 40924
+    },
+    {
+      "epoch": 38.18,
+      "learning_rate": 4.522761194029851e-05,
+      "loss": 0.0017,
+      "step": 40928
+    },
+    {
+      "epoch": 38.18,
+      "learning_rate": 4.522714552238806e-05,
+      "loss": 0.0002,
+      "step": 40932
+    },
+    {
+      "epoch": 38.19,
+      "learning_rate": 4.522667910447761e-05,
+      "loss": 0.0008,
+      "step": 40936
+    },
+    {
+      "epoch": 38.19,
+      "learning_rate": 4.522621268656717e-05,
+      "loss": 0.0008,
+      "step": 40940
+    },
+    {
+      "epoch": 38.19,
+      "learning_rate": 4.522574626865672e-05,
+      "loss": 0.0,
+      "step": 40944
+    },
+    {
+      "epoch": 38.2,
+      "learning_rate": 4.522527985074627e-05,
+      "loss": 0.0019,
+      "step": 40948
+    },
+    {
+      "epoch": 38.2,
+      "learning_rate": 4.5224813432835825e-05,
+      "loss": 0.0011,
+      "step": 40952
+    },
+    {
+      "epoch": 38.21,
+      "learning_rate": 4.522434701492538e-05,
+      "loss": 0.0005,
+      "step": 40956
+    },
+    {
+      "epoch": 38.21,
+      "learning_rate": 4.522388059701493e-05,
+      "loss": 0.0,
+      "step": 40960
+    },
+    {
+      "epoch": 38.21,
+      "learning_rate": 4.5223414179104476e-05,
+      "loss": 0.0003,
+      "step": 40964
+    },
+    {
+      "epoch": 38.22,
+      "learning_rate": 4.522294776119404e-05,
+      "loss": 0.0,
+      "step": 40968
+    },
+    {
+      "epoch": 38.22,
+      "learning_rate": 4.5222481343283586e-05,
+      "loss": 0.0027,
+      "step": 40972
+    },
+    {
+      "epoch": 38.22,
+      "learning_rate": 4.5222014925373134e-05,
+      "loss": 0.0,
+      "step": 40976
+    },
+    {
+      "epoch": 38.23,
+      "learning_rate": 4.522154850746269e-05,
+      "loss": 0.0003,
+      "step": 40980
+    },
+    {
+      "epoch": 38.23,
+      "learning_rate": 4.5221082089552244e-05,
+      "loss": 0.0002,
+      "step": 40984
+    },
+    {
+      "epoch": 38.24,
+      "learning_rate": 4.522061567164179e-05,
+      "loss": 0.0001,
+      "step": 40988
+    },
+    {
+      "epoch": 38.24,
+      "learning_rate": 4.522014925373135e-05,
+      "loss": 0.0,
+      "step": 40992
+    },
+    {
+      "epoch": 38.24,
+      "learning_rate": 4.5219682835820895e-05,
+      "loss": 0.0001,
+      "step": 40996
+    },
+    {
+      "epoch": 38.25,
+      "learning_rate": 4.521921641791045e-05,
+      "loss": 0.0,
+      "step": 41000
+    },
+    {
+      "epoch": 38.25,
+      "eval_exact_match": 0.7359767891682786,
+      "eval_exec": 0.7649903288201161,
+      "eval_loss": 0.43792104721069336,
+      "eval_runtime": 1152.442,
+      "eval_samples_per_second": 0.897,
+      "step": 41000
+    },
+    {
+      "epoch": 38.25,
+      "learning_rate": 4.5218750000000005e-05,
+      "loss": 0.0001,
+      "step": 41004
+    },
+    {
+      "epoch": 38.25,
+      "learning_rate": 4.521828358208955e-05,
+      "loss": 0.0003,
+      "step": 41008
+    },
+    {
+      "epoch": 38.26,
+      "learning_rate": 4.521781716417911e-05,
+      "loss": 0.0,
+      "step": 41012
+    },
+    {
+      "epoch": 38.26,
+      "learning_rate": 4.521735074626866e-05,
+      "loss": 0.0003,
+      "step": 41016
+    },
+    {
+      "epoch": 38.26,
+      "learning_rate": 4.521688432835821e-05,
+      "loss": 0.0002,
+      "step": 41020
+    },
+    {
+      "epoch": 38.27,
+      "learning_rate": 4.521641791044776e-05,
+      "loss": 0.0002,
+      "step": 41024
+    },
+    {
+      "epoch": 38.27,
+      "learning_rate": 4.521595149253732e-05,
+      "loss": 0.0,
+      "step": 41028
+    },
+    {
+      "epoch": 38.28,
+      "learning_rate": 4.521548507462687e-05,
+      "loss": 0.0004,
+      "step": 41032
+    },
+    {
+      "epoch": 38.28,
+      "learning_rate": 4.521501865671642e-05,
+      "loss": 0.0,
+      "step": 41036
+    },
+    {
+      "epoch": 38.28,
+      "learning_rate": 4.521455223880597e-05,
+      "loss": 0.0008,
+      "step": 41040
+    },
+    {
+      "epoch": 38.29,
+      "learning_rate": 4.521408582089553e-05,
+      "loss": 0.0001,
+      "step": 41044
+    },
+    {
+      "epoch": 38.29,
+      "learning_rate": 4.5213619402985075e-05,
+      "loss": 0.0,
+      "step": 41048
+    },
+    {
+      "epoch": 38.29,
+      "learning_rate": 4.521315298507463e-05,
+      "loss": 0.0,
+      "step": 41052
+    },
+    {
+      "epoch": 38.3,
+      "learning_rate": 4.521268656716418e-05,
+      "loss": 0.0,
+      "step": 41056
+    },
+    {
+      "epoch": 38.3,
+      "learning_rate": 4.521222014925373e-05,
+      "loss": 0.0003,
+      "step": 41060
+    },
+    {
+      "epoch": 38.31,
+      "learning_rate": 4.521175373134329e-05,
+      "loss": 0.0001,
+      "step": 41064
+    },
+    {
+      "epoch": 38.31,
+      "learning_rate": 4.5211287313432836e-05,
+      "loss": 0.0003,
+      "step": 41068
+    },
+    {
+      "epoch": 38.31,
+      "learning_rate": 4.521082089552239e-05,
+      "loss": 0.0001,
+      "step": 41072
+    },
+    {
+      "epoch": 38.32,
+      "learning_rate": 4.5210354477611946e-05,
+      "loss": 0.0004,
+      "step": 41076
+    },
+    {
+      "epoch": 38.32,
+      "learning_rate": 4.5209888059701494e-05,
+      "loss": 0.0073,
+      "step": 41080
+    },
+    {
+      "epoch": 38.32,
+      "learning_rate": 4.520942164179104e-05,
+      "loss": 0.0,
+      "step": 41084
+    },
+    {
+      "epoch": 38.33,
+      "learning_rate": 4.5208955223880604e-05,
+      "loss": 0.0018,
+      "step": 41088
+    },
+    {
+      "epoch": 38.33,
+      "learning_rate": 4.520848880597015e-05,
+      "loss": 0.0053,
+      "step": 41092
+    },
+    {
+      "epoch": 38.34,
+      "learning_rate": 4.52080223880597e-05,
+      "loss": 0.0001,
+      "step": 41096
+    },
+    {
+      "epoch": 38.34,
+      "learning_rate": 4.5207555970149255e-05,
+      "loss": 0.0004,
+      "step": 41100
+    },
+    {
+      "epoch": 38.34,
+      "learning_rate": 4.520708955223881e-05,
+      "loss": 0.0002,
+      "step": 41104
+    },
+    {
+      "epoch": 38.35,
+      "learning_rate": 4.5206623134328365e-05,
+      "loss": 0.0,
+      "step": 41108
+    },
+    {
+      "epoch": 38.35,
+      "learning_rate": 4.520615671641791e-05,
+      "loss": 0.0,
+      "step": 41112
+    },
+    {
+      "epoch": 38.35,
+      "learning_rate": 4.520569029850746e-05,
+      "loss": 0.0016,
+      "step": 41116
+    },
+    {
+      "epoch": 38.36,
+      "learning_rate": 4.520522388059702e-05,
+      "loss": 0.0,
+      "step": 41120
+    },
+    {
+      "epoch": 38.36,
+      "learning_rate": 4.520475746268657e-05,
+      "loss": 0.0,
+      "step": 41124
+    },
+    {
+      "epoch": 38.37,
+      "learning_rate": 4.520429104477612e-05,
+      "loss": 0.0001,
+      "step": 41128
+    },
+    {
+      "epoch": 38.37,
+      "learning_rate": 4.5203824626865674e-05,
+      "loss": 0.0,
+      "step": 41132
+    },
+    {
+      "epoch": 38.37,
+      "learning_rate": 4.520335820895523e-05,
+      "loss": 0.0,
+      "step": 41136
+    },
+    {
+      "epoch": 38.38,
+      "learning_rate": 4.520289179104478e-05,
+      "loss": 0.0,
+      "step": 41140
+    },
+    {
+      "epoch": 38.38,
+      "learning_rate": 4.520242537313433e-05,
+      "loss": 0.0,
+      "step": 41144
+    },
+    {
+      "epoch": 38.38,
+      "learning_rate": 4.520195895522388e-05,
+      "loss": 0.0002,
+      "step": 41148
+    },
+    {
+      "epoch": 38.39,
+      "learning_rate": 4.5201492537313435e-05,
+      "loss": 0.0001,
+      "step": 41152
+    },
+    {
+      "epoch": 38.39,
+      "learning_rate": 4.520102611940299e-05,
+      "loss": 0.0018,
+      "step": 41156
+    },
+    {
+      "epoch": 38.4,
+      "learning_rate": 4.520055970149254e-05,
+      "loss": 0.0,
+      "step": 41160
+    },
+    {
+      "epoch": 38.4,
+      "learning_rate": 4.520009328358209e-05,
+      "loss": 0.0,
+      "step": 41164
+    },
+    {
+      "epoch": 38.4,
+      "learning_rate": 4.519962686567165e-05,
+      "loss": 0.0032,
+      "step": 41168
+    },
+    {
+      "epoch": 38.41,
+      "learning_rate": 4.5199160447761196e-05,
+      "loss": 0.0019,
+      "step": 41172
+    },
+    {
+      "epoch": 38.41,
+      "learning_rate": 4.5198694029850744e-05,
+      "loss": 0.0002,
+      "step": 41176
+    },
+    {
+      "epoch": 38.41,
+      "learning_rate": 4.5198227611940306e-05,
+      "loss": 0.0,
+      "step": 41180
+    },
+    {
+      "epoch": 38.42,
+      "learning_rate": 4.5197761194029854e-05,
+      "loss": 0.0001,
+      "step": 41184
+    },
+    {
+      "epoch": 38.42,
+      "learning_rate": 4.51972947761194e-05,
+      "loss": 0.0002,
+      "step": 41188
+    },
+    {
+      "epoch": 38.43,
+      "learning_rate": 4.519682835820896e-05,
+      "loss": 0.0034,
+      "step": 41192
+    },
+    {
+      "epoch": 38.43,
+      "learning_rate": 4.519636194029851e-05,
+      "loss": 0.0011,
+      "step": 41196
+    },
+    {
+      "epoch": 38.43,
+      "learning_rate": 4.519589552238806e-05,
+      "loss": 0.0003,
+      "step": 41200
+    },
+    {
+      "epoch": 38.44,
+      "learning_rate": 4.5195429104477615e-05,
+      "loss": 0.0002,
+      "step": 41204
+    },
+    {
+      "epoch": 38.44,
+      "learning_rate": 4.519496268656716e-05,
+      "loss": 0.0008,
+      "step": 41208
+    },
+    {
+      "epoch": 38.44,
+      "learning_rate": 4.519449626865672e-05,
+      "loss": 0.0,
+      "step": 41212
+    },
+    {
+      "epoch": 38.45,
+      "learning_rate": 4.519402985074627e-05,
+      "loss": 0.0001,
+      "step": 41216
+    },
+    {
+      "epoch": 38.45,
+      "learning_rate": 4.519356343283582e-05,
+      "loss": 0.0002,
+      "step": 41220
+    },
+    {
+      "epoch": 38.46,
+      "learning_rate": 4.5193097014925376e-05,
+      "loss": 0.0,
+      "step": 41224
+    },
+    {
+      "epoch": 38.46,
+      "learning_rate": 4.519263059701493e-05,
+      "loss": 0.0001,
+      "step": 41228
+    },
+    {
+      "epoch": 38.46,
+      "learning_rate": 4.519216417910448e-05,
+      "loss": 0.0005,
+      "step": 41232
+    },
+    {
+      "epoch": 38.47,
+      "learning_rate": 4.519169776119403e-05,
+      "loss": 0.0,
+      "step": 41236
+    },
+    {
+      "epoch": 38.47,
+      "learning_rate": 4.519123134328359e-05,
+      "loss": 0.0,
+      "step": 41240
+    },
+    {
+      "epoch": 38.47,
+      "learning_rate": 4.5190764925373137e-05,
+      "loss": 0.0,
+      "step": 41244
+    },
+    {
+      "epoch": 38.48,
+      "learning_rate": 4.5190298507462685e-05,
+      "loss": 0.0001,
+      "step": 41248
+    },
+    {
+      "epoch": 38.48,
+      "learning_rate": 4.518983208955224e-05,
+      "loss": 0.0003,
+      "step": 41252
+    },
+    {
+      "epoch": 38.49,
+      "learning_rate": 4.5189365671641795e-05,
+      "loss": 0.0,
+      "step": 41256
+    },
+    {
+      "epoch": 38.49,
+      "learning_rate": 4.518889925373134e-05,
+      "loss": 0.0,
+      "step": 41260
+    },
+    {
+      "epoch": 38.49,
+      "learning_rate": 4.51884328358209e-05,
+      "loss": 0.0,
+      "step": 41264
+    },
+    {
+      "epoch": 38.5,
+      "learning_rate": 4.5187966417910446e-05,
+      "loss": 0.0,
+      "step": 41268
+    },
+    {
+      "epoch": 38.5,
+      "learning_rate": 4.518750000000001e-05,
+      "loss": 0.0001,
+      "step": 41272
+    },
+    {
+      "epoch": 38.5,
+      "learning_rate": 4.5187033582089556e-05,
+      "loss": 0.0044,
+      "step": 41276
+    },
+    {
+      "epoch": 38.51,
+      "learning_rate": 4.5186567164179104e-05,
+      "loss": 0.0001,
+      "step": 41280
+    },
+    {
+      "epoch": 38.51,
+      "learning_rate": 4.518610074626866e-05,
+      "loss": 0.0,
+      "step": 41284
+    },
+    {
+      "epoch": 38.51,
+      "learning_rate": 4.5185634328358213e-05,
+      "loss": 0.0,
+      "step": 41288
+    },
+    {
+      "epoch": 38.52,
+      "learning_rate": 4.518516791044776e-05,
+      "loss": 0.0022,
+      "step": 41292
+    },
+    {
+      "epoch": 38.52,
+      "learning_rate": 4.5184701492537317e-05,
+      "loss": 0.0013,
+      "step": 41296
+    },
+    {
+      "epoch": 38.53,
+      "learning_rate": 4.518423507462687e-05,
+      "loss": 0.0003,
+      "step": 41300
+    },
+    {
+      "epoch": 38.53,
+      "learning_rate": 4.518376865671642e-05,
+      "loss": 0.0,
+      "step": 41304
+    },
+    {
+      "epoch": 38.53,
+      "learning_rate": 4.5183302238805974e-05,
+      "loss": 0.0,
+      "step": 41308
+    },
+    {
+      "epoch": 38.54,
+      "learning_rate": 4.518283582089552e-05,
+      "loss": 0.0001,
+      "step": 41312
+    },
+    {
+      "epoch": 38.54,
+      "learning_rate": 4.518236940298508e-05,
+      "loss": 0.0003,
+      "step": 41316
+    },
+    {
+      "epoch": 38.54,
+      "learning_rate": 4.518190298507463e-05,
+      "loss": 0.0,
+      "step": 41320
+    },
+    {
+      "epoch": 38.55,
+      "learning_rate": 4.518143656716418e-05,
+      "loss": 0.0,
+      "step": 41324
+    },
+    {
+      "epoch": 38.55,
+      "learning_rate": 4.518097014925373e-05,
+      "loss": 0.0002,
+      "step": 41328
+    },
+    {
+      "epoch": 38.56,
+      "learning_rate": 4.518050373134329e-05,
+      "loss": 0.0066,
+      "step": 41332
+    },
+    {
+      "epoch": 38.56,
+      "learning_rate": 4.518003731343284e-05,
+      "loss": 0.0001,
+      "step": 41336
+    },
+    {
+      "epoch": 38.56,
+      "learning_rate": 4.5179570895522387e-05,
+      "loss": 0.0005,
+      "step": 41340
+    },
+    {
+      "epoch": 38.57,
+      "learning_rate": 4.517910447761194e-05,
+      "loss": 0.0003,
+      "step": 41344
+    },
+    {
+      "epoch": 38.57,
+      "learning_rate": 4.5178638059701496e-05,
+      "loss": 0.0009,
+      "step": 41348
+    },
+    {
+      "epoch": 38.57,
+      "learning_rate": 4.5178171641791045e-05,
+      "loss": 0.0,
+      "step": 41352
+    },
+    {
+      "epoch": 38.58,
+      "learning_rate": 4.51777052238806e-05,
+      "loss": 0.0,
+      "step": 41356
+    },
+    {
+      "epoch": 38.58,
+      "learning_rate": 4.5177238805970154e-05,
+      "loss": 0.0002,
+      "step": 41360
+    },
+    {
+      "epoch": 38.59,
+      "learning_rate": 4.51767723880597e-05,
+      "loss": 0.0,
+      "step": 41364
+    },
+    {
+      "epoch": 38.59,
+      "learning_rate": 4.517630597014926e-05,
+      "loss": 0.0001,
+      "step": 41368
+    },
+    {
+      "epoch": 38.59,
+      "learning_rate": 4.5175839552238806e-05,
+      "loss": 0.0,
+      "step": 41372
+    },
+    {
+      "epoch": 38.6,
+      "learning_rate": 4.517537313432836e-05,
+      "loss": 0.0,
+      "step": 41376
+    },
+    {
+      "epoch": 38.6,
+      "learning_rate": 4.5174906716417915e-05,
+      "loss": 0.0,
+      "step": 41380
+    },
+    {
+      "epoch": 38.6,
+      "learning_rate": 4.5174440298507463e-05,
+      "loss": 0.0012,
+      "step": 41384
+    },
+    {
+      "epoch": 38.61,
+      "learning_rate": 4.517397388059701e-05,
+      "loss": 0.0001,
+      "step": 41388
+    },
+    {
+      "epoch": 38.61,
+      "learning_rate": 4.517350746268657e-05,
+      "loss": 0.0,
+      "step": 41392
+    },
+    {
+      "epoch": 38.62,
+      "learning_rate": 4.517304104477612e-05,
+      "loss": 0.0001,
+      "step": 41396
+    },
+    {
+      "epoch": 38.62,
+      "learning_rate": 4.517257462686567e-05,
+      "loss": 0.0001,
+      "step": 41400
+    },
+    {
+      "epoch": 38.62,
+      "learning_rate": 4.5172108208955224e-05,
+      "loss": 0.0016,
+      "step": 41404
+    },
+    {
+      "epoch": 38.63,
+      "learning_rate": 4.517164179104478e-05,
+      "loss": 0.0024,
+      "step": 41408
+    },
+    {
+      "epoch": 38.63,
+      "learning_rate": 4.517117537313433e-05,
+      "loss": 0.0001,
+      "step": 41412
+    },
+    {
+      "epoch": 38.63,
+      "learning_rate": 4.517070895522388e-05,
+      "loss": 0.0,
+      "step": 41416
+    },
+    {
+      "epoch": 38.64,
+      "learning_rate": 4.517024253731344e-05,
+      "loss": 0.0019,
+      "step": 41420
+    },
+    {
+      "epoch": 38.64,
+      "learning_rate": 4.5169776119402985e-05,
+      "loss": 0.0,
+      "step": 41424
+    },
+    {
+      "epoch": 38.65,
+      "learning_rate": 4.516930970149254e-05,
+      "loss": 0.0032,
+      "step": 41428
+    },
+    {
+      "epoch": 38.65,
+      "learning_rate": 4.516884328358209e-05,
+      "loss": 0.0009,
+      "step": 41432
+    },
+    {
+      "epoch": 38.65,
+      "learning_rate": 4.5168376865671643e-05,
+      "loss": 0.0049,
+      "step": 41436
+    },
+    {
+      "epoch": 38.66,
+      "learning_rate": 4.51679104477612e-05,
+      "loss": 0.0,
+      "step": 41440
+    },
+    {
+      "epoch": 38.66,
+      "learning_rate": 4.5167444029850746e-05,
+      "loss": 0.0004,
+      "step": 41444
+    },
+    {
+      "epoch": 38.66,
+      "learning_rate": 4.51669776119403e-05,
+      "loss": 0.0015,
+      "step": 41448
+    },
+    {
+      "epoch": 38.67,
+      "learning_rate": 4.5166511194029856e-05,
+      "loss": 0.0,
+      "step": 41452
+    },
+    {
+      "epoch": 38.67,
+      "learning_rate": 4.5166044776119404e-05,
+      "loss": 0.0008,
+      "step": 41456
+    },
+    {
+      "epoch": 38.68,
+      "learning_rate": 4.516557835820896e-05,
+      "loss": 0.0,
+      "step": 41460
+    },
+    {
+      "epoch": 38.68,
+      "learning_rate": 4.516511194029851e-05,
+      "loss": 0.0,
+      "step": 41464
+    },
+    {
+      "epoch": 38.68,
+      "learning_rate": 4.516464552238806e-05,
+      "loss": 0.0,
+      "step": 41468
+    },
+    {
+      "epoch": 38.69,
+      "learning_rate": 4.516417910447762e-05,
+      "loss": 0.0008,
+      "step": 41472
+    },
+    {
+      "epoch": 38.69,
+      "learning_rate": 4.5163712686567165e-05,
+      "loss": 0.0001,
+      "step": 41476
+    },
+    {
+      "epoch": 38.69,
+      "learning_rate": 4.516324626865672e-05,
+      "loss": 0.0002,
+      "step": 41480
+    },
+    {
+      "epoch": 38.7,
+      "learning_rate": 4.5162779850746275e-05,
+      "loss": 0.0001,
+      "step": 41484
+    },
+    {
+      "epoch": 38.7,
+      "learning_rate": 4.516231343283582e-05,
+      "loss": 0.0001,
+      "step": 41488
+    },
+    {
+      "epoch": 38.71,
+      "learning_rate": 4.516184701492537e-05,
+      "loss": 0.0001,
+      "step": 41492
+    },
+    {
+      "epoch": 38.71,
+      "learning_rate": 4.5161380597014926e-05,
+      "loss": 0.0,
+      "step": 41496
+    },
+    {
+      "epoch": 38.71,
+      "learning_rate": 4.516091417910448e-05,
+      "loss": 0.0035,
+      "step": 41500
+    },
+    {
+      "epoch": 38.71,
+      "eval_exact_match": 0.7263056092843327,
+      "eval_exec": 0.7562862669245648,
+      "eval_loss": 0.440681517124176,
+      "eval_runtime": 1097.6911,
+      "eval_samples_per_second": 0.942,
+      "step": 41500
+    },
+    {
+      "epoch": 38.72,
+      "learning_rate": 4.516044776119403e-05,
+      "loss": 0.0,
+      "step": 41504
+    },
+    {
+      "epoch": 38.72,
+      "learning_rate": 4.5159981343283584e-05,
+      "loss": 0.0001,
+      "step": 41508
+    },
+    {
+      "epoch": 38.72,
+      "learning_rate": 4.515951492537314e-05,
+      "loss": 0.0014,
+      "step": 41512
+    },
+    {
+      "epoch": 38.73,
+      "learning_rate": 4.515904850746269e-05,
+      "loss": 0.0014,
+      "step": 41516
+    },
+    {
+      "epoch": 38.73,
+      "learning_rate": 4.515858208955224e-05,
+      "loss": 0.0,
+      "step": 41520
+    },
+    {
+      "epoch": 38.73,
+      "learning_rate": 4.515811567164179e-05,
+      "loss": 0.0002,
+      "step": 41524
+    },
+    {
+      "epoch": 38.74,
+      "learning_rate": 4.5157649253731345e-05,
+      "loss": 0.002,
+      "step": 41528
+    },
+    {
+      "epoch": 38.74,
+      "learning_rate": 4.51571828358209e-05,
+      "loss": 0.0015,
+      "step": 41532
+    },
+    {
+      "epoch": 38.75,
+      "learning_rate": 4.515671641791045e-05,
+      "loss": 0.0,
+      "step": 41536
+    },
+    {
+      "epoch": 38.75,
+      "learning_rate": 4.515625e-05,
+      "loss": 0.0,
+      "step": 41540
+    },
+    {
+      "epoch": 38.75,
+      "learning_rate": 4.515578358208956e-05,
+      "loss": 0.0013,
+      "step": 41544
+    },
+    {
+      "epoch": 38.76,
+      "learning_rate": 4.5155317164179106e-05,
+      "loss": 0.0,
+      "step": 41548
+    },
+    {
+      "epoch": 38.76,
+      "learning_rate": 4.5154850746268654e-05,
+      "loss": 0.0001,
+      "step": 41552
+    },
+    {
+      "epoch": 38.76,
+      "learning_rate": 4.515438432835821e-05,
+      "loss": 0.0022,
+      "step": 41556
+    },
+    {
+      "epoch": 38.77,
+      "learning_rate": 4.5153917910447764e-05,
+      "loss": 0.0011,
+      "step": 41560
+    },
+    {
+      "epoch": 38.77,
+      "learning_rate": 4.515345149253731e-05,
+      "loss": 0.0001,
+      "step": 41564
+    },
+    {
+      "epoch": 38.78,
+      "learning_rate": 4.515298507462687e-05,
+      "loss": 0.0001,
+      "step": 41568
+    },
+    {
+      "epoch": 38.78,
+      "learning_rate": 4.515251865671642e-05,
+      "loss": 0.0004,
+      "step": 41572
+    },
+    {
+      "epoch": 38.78,
+      "learning_rate": 4.515205223880597e-05,
+      "loss": 0.0,
+      "step": 41576
+    },
+    {
+      "epoch": 38.79,
+      "learning_rate": 4.5151585820895525e-05,
+      "loss": 0.0,
+      "step": 41580
+    },
+    {
+      "epoch": 38.79,
+      "learning_rate": 4.515111940298507e-05,
+      "loss": 0.0006,
+      "step": 41584
+    },
+    {
+      "epoch": 38.79,
+      "learning_rate": 4.515065298507463e-05,
+      "loss": 0.0001,
+      "step": 41588
+    },
+    {
+      "epoch": 38.8,
+      "learning_rate": 4.515018656716418e-05,
+      "loss": 0.0016,
+      "step": 41592
+    },
+    {
+      "epoch": 38.8,
+      "learning_rate": 4.514972014925373e-05,
+      "loss": 0.0007,
+      "step": 41596
+    },
+    {
+      "epoch": 38.81,
+      "learning_rate": 4.5149253731343286e-05,
+      "loss": 0.0,
+      "step": 41600
+    },
+    {
+      "epoch": 38.81,
+      "learning_rate": 4.514878731343284e-05,
+      "loss": 0.0,
+      "step": 41604
+    },
+    {
+      "epoch": 38.81,
+      "learning_rate": 4.514832089552239e-05,
+      "loss": 0.0014,
+      "step": 41608
+    },
+    {
+      "epoch": 38.82,
+      "learning_rate": 4.5147854477611944e-05,
+      "loss": 0.0007,
+      "step": 41612
+    },
+    {
+      "epoch": 38.82,
+      "learning_rate": 4.514738805970149e-05,
+      "loss": 0.0,
+      "step": 41616
+    },
+    {
+      "epoch": 38.82,
+      "learning_rate": 4.514692164179105e-05,
+      "loss": 0.0002,
+      "step": 41620
+    },
+    {
+      "epoch": 38.83,
+      "learning_rate": 4.51464552238806e-05,
+      "loss": 0.0001,
+      "step": 41624
+    },
+    {
+      "epoch": 38.83,
+      "learning_rate": 4.514598880597015e-05,
+      "loss": 0.0015,
+      "step": 41628
+    },
+    {
+      "epoch": 38.84,
+      "learning_rate": 4.5145522388059705e-05,
+      "loss": 0.0,
+      "step": 41632
+    },
+    {
+      "epoch": 38.84,
+      "learning_rate": 4.514505597014926e-05,
+      "loss": 0.0023,
+      "step": 41636
+    },
+    {
+      "epoch": 38.84,
+      "learning_rate": 4.514458955223881e-05,
+      "loss": 0.0008,
+      "step": 41640
+    },
+    {
+      "epoch": 38.85,
+      "learning_rate": 4.5144123134328356e-05,
+      "loss": 0.0001,
+      "step": 41644
+    },
+    {
+      "epoch": 38.85,
+      "learning_rate": 4.514365671641792e-05,
+      "loss": 0.0004,
+      "step": 41648
+    },
+    {
+      "epoch": 38.85,
+      "learning_rate": 4.5143190298507466e-05,
+      "loss": 0.0014,
+      "step": 41652
+    },
+    {
+      "epoch": 38.86,
+      "learning_rate": 4.5142723880597014e-05,
+      "loss": 0.0,
+      "step": 41656
+    },
+    {
+      "epoch": 38.86,
+      "learning_rate": 4.514225746268657e-05,
+      "loss": 0.0041,
+      "step": 41660
+    },
+    {
+      "epoch": 38.87,
+      "learning_rate": 4.5141791044776124e-05,
+      "loss": 0.0014,
+      "step": 41664
+    },
+    {
+      "epoch": 38.87,
+      "learning_rate": 4.514132462686567e-05,
+      "loss": 0.0019,
+      "step": 41668
+    },
+    {
+      "epoch": 38.87,
+      "learning_rate": 4.514085820895523e-05,
+      "loss": 0.0,
+      "step": 41672
+    },
+    {
+      "epoch": 38.88,
+      "learning_rate": 4.5140391791044775e-05,
+      "loss": 0.0001,
+      "step": 41676
+    },
+    {
+      "epoch": 38.88,
+      "learning_rate": 4.513992537313433e-05,
+      "loss": 0.0018,
+      "step": 41680
+    },
+    {
+      "epoch": 38.88,
+      "learning_rate": 4.5139458955223885e-05,
+      "loss": 0.0001,
+      "step": 41684
+    },
+    {
+      "epoch": 38.89,
+      "learning_rate": 4.513899253731343e-05,
+      "loss": 0.0001,
+      "step": 41688
+    },
+    {
+      "epoch": 38.89,
+      "learning_rate": 4.513852611940299e-05,
+      "loss": 0.0001,
+      "step": 41692
+    },
+    {
+      "epoch": 38.9,
+      "learning_rate": 4.513805970149254e-05,
+      "loss": 0.0003,
+      "step": 41696
+    },
+    {
+      "epoch": 38.9,
+      "learning_rate": 4.513759328358209e-05,
+      "loss": 0.0003,
+      "step": 41700
+    },
+    {
+      "epoch": 38.9,
+      "learning_rate": 4.513712686567164e-05,
+      "loss": 0.0004,
+      "step": 41704
+    },
+    {
+      "epoch": 38.91,
+      "learning_rate": 4.51366604477612e-05,
+      "loss": 0.0001,
+      "step": 41708
+    },
+    {
+      "epoch": 38.91,
+      "learning_rate": 4.513619402985075e-05,
+      "loss": 0.0031,
+      "step": 41712
+    },
+    {
+      "epoch": 38.91,
+      "learning_rate": 4.51357276119403e-05,
+      "loss": 0.0006,
+      "step": 41716
+    },
+    {
+      "epoch": 38.92,
+      "learning_rate": 4.513526119402985e-05,
+      "loss": 0.0009,
+      "step": 41720
+    },
+    {
+      "epoch": 38.92,
+      "learning_rate": 4.513479477611941e-05,
+      "loss": 0.0001,
+      "step": 41724
+    },
+    {
+      "epoch": 38.93,
+      "learning_rate": 4.5134328358208955e-05,
+      "loss": 0.0,
+      "step": 41728
+    },
+    {
+      "epoch": 38.93,
+      "learning_rate": 4.513386194029851e-05,
+      "loss": 0.0048,
+      "step": 41732
+    },
+    {
+      "epoch": 38.93,
+      "learning_rate": 4.513339552238806e-05,
+      "loss": 0.0001,
+      "step": 41736
+    },
+    {
+      "epoch": 38.94,
+      "learning_rate": 4.513292910447761e-05,
+      "loss": 0.0022,
+      "step": 41740
+    },
+    {
+      "epoch": 38.94,
+      "learning_rate": 4.513246268656717e-05,
+      "loss": 0.0,
+      "step": 41744
+    },
+    {
+      "epoch": 38.94,
+      "learning_rate": 4.5131996268656716e-05,
+      "loss": 0.0003,
+      "step": 41748
+    },
+    {
+      "epoch": 38.95,
+      "learning_rate": 4.513152985074627e-05,
+      "loss": 0.0146,
+      "step": 41752
+    },
+    {
+      "epoch": 38.95,
+      "learning_rate": 4.5131063432835826e-05,
+      "loss": 0.0001,
+      "step": 41756
+    },
+    {
+      "epoch": 38.96,
+      "learning_rate": 4.5130597014925374e-05,
+      "loss": 0.0,
+      "step": 41760
+    },
+    {
+      "epoch": 38.96,
+      "learning_rate": 4.513013059701493e-05,
+      "loss": 0.0001,
+      "step": 41764
+    },
+    {
+      "epoch": 38.96,
+      "learning_rate": 4.5129664179104484e-05,
+      "loss": 0.0,
+      "step": 41768
+    },
+    {
+      "epoch": 38.97,
+      "learning_rate": 4.512919776119403e-05,
+      "loss": 0.0001,
+      "step": 41772
+    },
+    {
+      "epoch": 38.97,
+      "learning_rate": 4.512873134328359e-05,
+      "loss": 0.0052,
+      "step": 41776
+    },
+    {
+      "epoch": 38.97,
+      "learning_rate": 4.5128264925373135e-05,
+      "loss": 0.0008,
+      "step": 41780
+    },
+    {
+      "epoch": 38.98,
+      "learning_rate": 4.512779850746269e-05,
+      "loss": 0.0001,
+      "step": 41784
+    },
+    {
+      "epoch": 38.98,
+      "learning_rate": 4.5127332089552245e-05,
+      "loss": 0.0001,
+      "step": 41788
+    },
+    {
+      "epoch": 38.98,
+      "learning_rate": 4.512686567164179e-05,
+      "loss": 0.0,
+      "step": 41792
+    },
+    {
+      "epoch": 38.99,
+      "learning_rate": 4.512639925373134e-05,
+      "loss": 0.0007,
+      "step": 41796
+    },
+    {
+      "epoch": 38.99,
+      "learning_rate": 4.51259328358209e-05,
+      "loss": 0.0001,
+      "step": 41800
+    },
+    {
+      "epoch": 39.0,
+      "learning_rate": 4.512546641791045e-05,
+      "loss": 0.0001,
+      "step": 41804
+    },
+    {
+      "epoch": 39.0,
+      "learning_rate": 4.5125e-05,
+      "loss": 0.0,
+      "step": 41808
+    },
+    {
+      "epoch": 39.0,
+      "learning_rate": 4.5124533582089554e-05,
+      "loss": 0.0,
+      "step": 41812
+    },
+    {
+      "epoch": 39.01,
+      "learning_rate": 4.512406716417911e-05,
+      "loss": 0.0001,
+      "step": 41816
+    },
+    {
+      "epoch": 39.01,
+      "learning_rate": 4.512360074626866e-05,
+      "loss": 0.0003,
+      "step": 41820
+    },
+    {
+      "epoch": 39.01,
+      "learning_rate": 4.512313432835821e-05,
+      "loss": 0.0001,
+      "step": 41824
+    },
+    {
+      "epoch": 39.02,
+      "learning_rate": 4.512266791044776e-05,
+      "loss": 0.0002,
+      "step": 41828
+    },
+    {
+      "epoch": 39.02,
+      "learning_rate": 4.5122201492537315e-05,
+      "loss": 0.0001,
+      "step": 41832
+    },
+    {
+      "epoch": 39.03,
+      "learning_rate": 4.512173507462687e-05,
+      "loss": 0.0,
+      "step": 41836
+    },
+    {
+      "epoch": 39.03,
+      "learning_rate": 4.512126865671642e-05,
+      "loss": 0.0009,
+      "step": 41840
+    },
+    {
+      "epoch": 39.03,
+      "learning_rate": 4.512080223880597e-05,
+      "loss": 0.0002,
+      "step": 41844
+    },
+    {
+      "epoch": 39.04,
+      "learning_rate": 4.512033582089553e-05,
+      "loss": 0.0073,
+      "step": 41848
+    },
+    {
+      "epoch": 39.04,
+      "learning_rate": 4.5119869402985076e-05,
+      "loss": 0.0,
+      "step": 41852
+    },
+    {
+      "epoch": 39.04,
+      "learning_rate": 4.5119402985074624e-05,
+      "loss": 0.0,
+      "step": 41856
+    },
+    {
+      "epoch": 39.05,
+      "learning_rate": 4.5118936567164186e-05,
+      "loss": 0.0001,
+      "step": 41860
+    },
+    {
+      "epoch": 39.05,
+      "learning_rate": 4.5118470149253734e-05,
+      "loss": 0.0,
+      "step": 41864
+    },
+    {
+      "epoch": 39.06,
+      "learning_rate": 4.511800373134328e-05,
+      "loss": 0.0001,
+      "step": 41868
+    },
+    {
+      "epoch": 39.06,
+      "learning_rate": 4.511753731343284e-05,
+      "loss": 0.0,
+      "step": 41872
+    },
+    {
+      "epoch": 39.06,
+      "learning_rate": 4.511707089552239e-05,
+      "loss": 0.0,
+      "step": 41876
+    },
+    {
+      "epoch": 39.07,
+      "learning_rate": 4.511660447761194e-05,
+      "loss": 0.0087,
+      "step": 41880
+    },
+    {
+      "epoch": 39.07,
+      "learning_rate": 4.5116138059701495e-05,
+      "loss": 0.0001,
+      "step": 41884
+    },
+    {
+      "epoch": 39.07,
+      "learning_rate": 4.511567164179104e-05,
+      "loss": 0.0001,
+      "step": 41888
+    },
+    {
+      "epoch": 39.08,
+      "learning_rate": 4.51152052238806e-05,
+      "loss": 0.0,
+      "step": 41892
+    },
+    {
+      "epoch": 39.08,
+      "learning_rate": 4.511473880597015e-05,
+      "loss": 0.0004,
+      "step": 41896
+    },
+    {
+      "epoch": 39.09,
+      "learning_rate": 4.51142723880597e-05,
+      "loss": 0.0007,
+      "step": 41900
+    },
+    {
+      "epoch": 39.09,
+      "learning_rate": 4.5113805970149256e-05,
+      "loss": 0.0012,
+      "step": 41904
+    },
+    {
+      "epoch": 39.09,
+      "learning_rate": 4.511333955223881e-05,
+      "loss": 0.0003,
+      "step": 41908
+    },
+    {
+      "epoch": 39.1,
+      "learning_rate": 4.511287313432836e-05,
+      "loss": 0.0005,
+      "step": 41912
+    },
+    {
+      "epoch": 39.1,
+      "learning_rate": 4.5112406716417914e-05,
+      "loss": 0.0007,
+      "step": 41916
+    },
+    {
+      "epoch": 39.1,
+      "learning_rate": 4.511194029850747e-05,
+      "loss": 0.0,
+      "step": 41920
+    },
+    {
+      "epoch": 39.11,
+      "learning_rate": 4.511147388059702e-05,
+      "loss": 0.0,
+      "step": 41924
+    },
+    {
+      "epoch": 39.11,
+      "learning_rate": 4.511100746268657e-05,
+      "loss": 0.0003,
+      "step": 41928
+    },
+    {
+      "epoch": 39.12,
+      "learning_rate": 4.511054104477612e-05,
+      "loss": 0.0015,
+      "step": 41932
+    },
+    {
+      "epoch": 39.12,
+      "learning_rate": 4.5110074626865675e-05,
+      "loss": 0.0,
+      "step": 41936
+    },
+    {
+      "epoch": 39.12,
+      "learning_rate": 4.510960820895523e-05,
+      "loss": 0.0,
+      "step": 41940
+    },
+    {
+      "epoch": 39.13,
+      "learning_rate": 4.510914179104478e-05,
+      "loss": 0.0004,
+      "step": 41944
+    },
+    {
+      "epoch": 39.13,
+      "learning_rate": 4.5108675373134326e-05,
+      "loss": 0.0,
+      "step": 41948
+    },
+    {
+      "epoch": 39.13,
+      "learning_rate": 4.510820895522389e-05,
+      "loss": 0.0001,
+      "step": 41952
+    },
+    {
+      "epoch": 39.14,
+      "learning_rate": 4.5107742537313436e-05,
+      "loss": 0.0001,
+      "step": 41956
+    },
+    {
+      "epoch": 39.14,
+      "learning_rate": 4.5107276119402984e-05,
+      "loss": 0.0001,
+      "step": 41960
+    },
+    {
+      "epoch": 39.15,
+      "learning_rate": 4.510680970149254e-05,
+      "loss": 0.0001,
+      "step": 41964
+    },
+    {
+      "epoch": 39.15,
+      "learning_rate": 4.5106343283582094e-05,
+      "loss": 0.005,
+      "step": 41968
+    },
+    {
+      "epoch": 39.15,
+      "learning_rate": 4.510587686567164e-05,
+      "loss": 0.0011,
+      "step": 41972
+    },
+    {
+      "epoch": 39.16,
+      "learning_rate": 4.51054104477612e-05,
+      "loss": 0.0019,
+      "step": 41976
+    },
+    {
+      "epoch": 39.16,
+      "learning_rate": 4.510494402985075e-05,
+      "loss": 0.0013,
+      "step": 41980
+    },
+    {
+      "epoch": 39.16,
+      "learning_rate": 4.51044776119403e-05,
+      "loss": 0.0001,
+      "step": 41984
+    },
+    {
+      "epoch": 39.17,
+      "learning_rate": 4.5104011194029855e-05,
+      "loss": 0.0,
+      "step": 41988
+    },
+    {
+      "epoch": 39.17,
+      "learning_rate": 4.51035447761194e-05,
+      "loss": 0.0001,
+      "step": 41992
+    },
+    {
+      "epoch": 39.18,
+      "learning_rate": 4.510307835820896e-05,
+      "loss": 0.0001,
+      "step": 41996
+    },
+    {
+      "epoch": 39.18,
+      "learning_rate": 4.510261194029851e-05,
+      "loss": 0.0001,
+      "step": 42000
+    },
+    {
+      "epoch": 39.18,
+      "eval_exact_match": 0.7350096711798839,
+      "eval_exec": 0.7630560928433269,
+      "eval_loss": 0.40719056129455566,
+      "eval_runtime": 1122.0735,
+      "eval_samples_per_second": 0.922,
+      "step": 42000
+    },
+    {
+      "epoch": 39.18,
+      "learning_rate": 4.510214552238806e-05,
+      "loss": 0.0,
+      "step": 42004
+    },
+    {
+      "epoch": 39.19,
+      "learning_rate": 4.510167910447761e-05,
+      "loss": 0.0002,
+      "step": 42008
+    },
+    {
+      "epoch": 39.19,
+      "learning_rate": 4.510121268656717e-05,
+      "loss": 0.0005,
+      "step": 42012
+    },
+    {
+      "epoch": 39.19,
+      "learning_rate": 4.510074626865672e-05,
+      "loss": 0.0002,
+      "step": 42016
+    },
+    {
+      "epoch": 39.2,
+      "learning_rate": 4.510027985074627e-05,
+      "loss": 0.0,
+      "step": 42020
+    },
+    {
+      "epoch": 39.2,
+      "learning_rate": 4.509981343283582e-05,
+      "loss": 0.0,
+      "step": 42024
+    },
+    {
+      "epoch": 39.21,
+      "learning_rate": 4.509934701492538e-05,
+      "loss": 0.0,
+      "step": 42028
+    },
+    {
+      "epoch": 39.21,
+      "learning_rate": 4.5098880597014925e-05,
+      "loss": 0.0,
+      "step": 42032
+    },
+    {
+      "epoch": 39.21,
+      "learning_rate": 4.509841417910448e-05,
+      "loss": 0.0064,
+      "step": 42036
+    },
+    {
+      "epoch": 39.22,
+      "learning_rate": 4.5097947761194035e-05,
+      "loss": 0.0003,
+      "step": 42040
+    },
+    {
+      "epoch": 39.22,
+      "learning_rate": 4.509748134328358e-05,
+      "loss": 0.0094,
+      "step": 42044
+    },
+    {
+      "epoch": 39.22,
+      "learning_rate": 4.509701492537314e-05,
+      "loss": 0.0001,
+      "step": 42048
+    },
+    {
+      "epoch": 39.23,
+      "learning_rate": 4.5096548507462686e-05,
+      "loss": 0.0002,
+      "step": 42052
+    },
+    {
+      "epoch": 39.23,
+      "learning_rate": 4.509608208955224e-05,
+      "loss": 0.0006,
+      "step": 42056
+    },
+    {
+      "epoch": 39.24,
+      "learning_rate": 4.5095615671641796e-05,
+      "loss": 0.0011,
+      "step": 42060
+    },
+    {
+      "epoch": 39.24,
+      "learning_rate": 4.5095149253731344e-05,
+      "loss": 0.0003,
+      "step": 42064
+    },
+    {
+      "epoch": 39.24,
+      "learning_rate": 4.509468283582089e-05,
+      "loss": 0.0008,
+      "step": 42068
+    },
+    {
+      "epoch": 39.25,
+      "learning_rate": 4.5094216417910454e-05,
+      "loss": 0.0,
+      "step": 42072
+    },
+    {
+      "epoch": 39.25,
+      "learning_rate": 4.509375e-05,
+      "loss": 0.0001,
+      "step": 42076
+    },
+    {
+      "epoch": 39.25,
+      "learning_rate": 4.509328358208956e-05,
+      "loss": 0.0001,
+      "step": 42080
+    },
+    {
+      "epoch": 39.26,
+      "learning_rate": 4.5092817164179105e-05,
+      "loss": 0.0,
+      "step": 42084
+    },
+    {
+      "epoch": 39.26,
+      "learning_rate": 4.509235074626866e-05,
+      "loss": 0.0,
+      "step": 42088
+    },
+    {
+      "epoch": 39.26,
+      "learning_rate": 4.5091884328358215e-05,
+      "loss": 0.0019,
+      "step": 42092
+    },
+    {
+      "epoch": 39.27,
+      "learning_rate": 4.509141791044776e-05,
+      "loss": 0.0,
+      "step": 42096
+    },
+    {
+      "epoch": 39.27,
+      "learning_rate": 4.509095149253732e-05,
+      "loss": 0.0014,
+      "step": 42100
+    },
+    {
+      "epoch": 39.28,
+      "learning_rate": 4.509048507462687e-05,
+      "loss": 0.0001,
+      "step": 42104
+    },
+    {
+      "epoch": 39.28,
+      "learning_rate": 4.509001865671642e-05,
+      "loss": 0.0002,
+      "step": 42108
+    },
+    {
+      "epoch": 39.28,
+      "learning_rate": 4.508955223880597e-05,
+      "loss": 0.0001,
+      "step": 42112
+    },
+    {
+      "epoch": 39.29,
+      "learning_rate": 4.5089085820895524e-05,
+      "loss": 0.0046,
+      "step": 42116
+    },
+    {
+      "epoch": 39.29,
+      "learning_rate": 4.508861940298508e-05,
+      "loss": 0.0,
+      "step": 42120
+    },
+    {
+      "epoch": 39.29,
+      "learning_rate": 4.508815298507463e-05,
+      "loss": 0.0,
+      "step": 42124
+    },
+    {
+      "epoch": 39.3,
+      "learning_rate": 4.508768656716418e-05,
+      "loss": 0.0001,
+      "step": 42128
+    },
+    {
+      "epoch": 39.3,
+      "learning_rate": 4.5087220149253737e-05,
+      "loss": 0.0006,
+      "step": 42132
+    },
+    {
+      "epoch": 39.31,
+      "learning_rate": 4.5086753731343285e-05,
+      "loss": 0.0001,
+      "step": 42136
+    },
+    {
+      "epoch": 39.31,
+      "learning_rate": 4.508628731343284e-05,
+      "loss": 0.0,
+      "step": 42140
+    },
+    {
+      "epoch": 39.31,
+      "learning_rate": 4.508582089552239e-05,
+      "loss": 0.0008,
+      "step": 42144
+    },
+    {
+      "epoch": 39.32,
+      "learning_rate": 4.508535447761194e-05,
+      "loss": 0.0,
+      "step": 42148
+    },
+    {
+      "epoch": 39.32,
+      "learning_rate": 4.50848880597015e-05,
+      "loss": 0.0,
+      "step": 42152
+    },
+    {
+      "epoch": 39.32,
+      "learning_rate": 4.5084421641791046e-05,
+      "loss": 0.002,
+      "step": 42156
+    },
+    {
+      "epoch": 39.33,
+      "learning_rate": 4.50839552238806e-05,
+      "loss": 0.0001,
+      "step": 42160
+    },
+    {
+      "epoch": 39.33,
+      "learning_rate": 4.5083488805970155e-05,
+      "loss": 0.0001,
+      "step": 42164
+    },
+    {
+      "epoch": 39.34,
+      "learning_rate": 4.5083022388059704e-05,
+      "loss": 0.0012,
+      "step": 42168
+    },
+    {
+      "epoch": 39.34,
+      "learning_rate": 4.508255597014925e-05,
+      "loss": 0.0014,
+      "step": 42172
+    },
+    {
+      "epoch": 39.34,
+      "learning_rate": 4.508208955223881e-05,
+      "loss": 0.0,
+      "step": 42176
+    },
+    {
+      "epoch": 39.35,
+      "learning_rate": 4.508162313432836e-05,
+      "loss": 0.0024,
+      "step": 42180
+    },
+    {
+      "epoch": 39.35,
+      "learning_rate": 4.508115671641791e-05,
+      "loss": 0.0,
+      "step": 42184
+    },
+    {
+      "epoch": 39.35,
+      "learning_rate": 4.5080690298507465e-05,
+      "loss": 0.0,
+      "step": 42188
+    },
+    {
+      "epoch": 39.36,
+      "learning_rate": 4.508022388059702e-05,
+      "loss": 0.0,
+      "step": 42192
+    },
+    {
+      "epoch": 39.36,
+      "learning_rate": 4.507975746268657e-05,
+      "loss": 0.0001,
+      "step": 42196
+    },
+    {
+      "epoch": 39.37,
+      "learning_rate": 4.507929104477612e-05,
+      "loss": 0.0,
+      "step": 42200
+    },
+    {
+      "epoch": 39.37,
+      "learning_rate": 4.507882462686567e-05,
+      "loss": 0.0002,
+      "step": 42204
+    },
+    {
+      "epoch": 39.37,
+      "learning_rate": 4.5078358208955226e-05,
+      "loss": 0.0,
+      "step": 42208
+    },
+    {
+      "epoch": 39.38,
+      "learning_rate": 4.507789179104478e-05,
+      "loss": 0.0001,
+      "step": 42212
+    },
+    {
+      "epoch": 39.38,
+      "learning_rate": 4.507742537313433e-05,
+      "loss": 0.0,
+      "step": 42216
+    },
+    {
+      "epoch": 39.38,
+      "learning_rate": 4.5076958955223884e-05,
+      "loss": 0.0,
+      "step": 42220
+    },
+    {
+      "epoch": 39.39,
+      "learning_rate": 4.507649253731344e-05,
+      "loss": 0.0002,
+      "step": 42224
+    },
+    {
+      "epoch": 39.39,
+      "learning_rate": 4.5076026119402987e-05,
+      "loss": 0.0001,
+      "step": 42228
+    },
+    {
+      "epoch": 39.4,
+      "learning_rate": 4.5075559701492535e-05,
+      "loss": 0.0001,
+      "step": 42232
+    },
+    {
+      "epoch": 39.4,
+      "learning_rate": 4.507509328358209e-05,
+      "loss": 0.0,
+      "step": 42236
+    },
+    {
+      "epoch": 39.4,
+      "learning_rate": 4.5074626865671645e-05,
+      "loss": 0.0,
+      "step": 42240
+    },
+    {
+      "epoch": 39.41,
+      "learning_rate": 4.50741604477612e-05,
+      "loss": 0.0002,
+      "step": 42244
+    },
+    {
+      "epoch": 39.41,
+      "learning_rate": 4.507369402985075e-05,
+      "loss": 0.0,
+      "step": 42248
+    },
+    {
+      "epoch": 39.41,
+      "learning_rate": 4.50732276119403e-05,
+      "loss": 0.0,
+      "step": 42252
+    },
+    {
+      "epoch": 39.42,
+      "learning_rate": 4.507276119402986e-05,
+      "loss": 0.0021,
+      "step": 42256
+    },
+    {
+      "epoch": 39.42,
+      "learning_rate": 4.5072294776119406e-05,
+      "loss": 0.0,
+      "step": 42260
+    },
+    {
+      "epoch": 39.43,
+      "learning_rate": 4.5071828358208954e-05,
+      "loss": 0.0011,
+      "step": 42264
+    },
+    {
+      "epoch": 39.43,
+      "learning_rate": 4.5071361940298515e-05,
+      "loss": 0.0001,
+      "step": 42268
+    },
+    {
+      "epoch": 39.43,
+      "learning_rate": 4.5070895522388063e-05,
+      "loss": 0.0,
+      "step": 42272
+    },
+    {
+      "epoch": 39.44,
+      "learning_rate": 4.507042910447761e-05,
+      "loss": 0.0,
+      "step": 42276
+    },
+    {
+      "epoch": 39.44,
+      "learning_rate": 4.5069962686567166e-05,
+      "loss": 0.0004,
+      "step": 42280
+    },
+    {
+      "epoch": 39.44,
+      "learning_rate": 4.506949626865672e-05,
+      "loss": 0.003,
+      "step": 42284
+    },
+    {
+      "epoch": 39.45,
+      "learning_rate": 4.506902985074627e-05,
+      "loss": 0.0007,
+      "step": 42288
+    },
+    {
+      "epoch": 39.45,
+      "learning_rate": 4.5068563432835824e-05,
+      "loss": 0.0001,
+      "step": 42292
+    },
+    {
+      "epoch": 39.46,
+      "learning_rate": 4.506809701492537e-05,
+      "loss": 0.0,
+      "step": 42296
+    },
+    {
+      "epoch": 39.46,
+      "learning_rate": 4.506763059701493e-05,
+      "loss": 0.0012,
+      "step": 42300
+    },
+    {
+      "epoch": 39.46,
+      "learning_rate": 4.506716417910448e-05,
+      "loss": 0.0109,
+      "step": 42304
+    },
+    {
+      "epoch": 39.47,
+      "learning_rate": 4.506669776119403e-05,
+      "loss": 0.0,
+      "step": 42308
+    },
+    {
+      "epoch": 39.47,
+      "learning_rate": 4.5066231343283585e-05,
+      "loss": 0.0053,
+      "step": 42312
+    },
+    {
+      "epoch": 39.47,
+      "learning_rate": 4.506576492537314e-05,
+      "loss": 0.0001,
+      "step": 42316
+    },
+    {
+      "epoch": 39.48,
+      "learning_rate": 4.506529850746269e-05,
+      "loss": 0.0001,
+      "step": 42320
+    },
+    {
+      "epoch": 39.48,
+      "learning_rate": 4.5064832089552237e-05,
+      "loss": 0.0001,
+      "step": 42324
+    },
+    {
+      "epoch": 39.49,
+      "learning_rate": 4.50643656716418e-05,
+      "loss": 0.0002,
+      "step": 42328
+    },
+    {
+      "epoch": 39.49,
+      "learning_rate": 4.5063899253731346e-05,
+      "loss": 0.0001,
+      "step": 42332
+    },
+    {
+      "epoch": 39.49,
+      "learning_rate": 4.5063432835820895e-05,
+      "loss": 0.0,
+      "step": 42336
+    },
+    {
+      "epoch": 39.5,
+      "learning_rate": 4.506296641791045e-05,
+      "loss": 0.0001,
+      "step": 42340
+    },
+    {
+      "epoch": 39.5,
+      "learning_rate": 4.5062500000000004e-05,
+      "loss": 0.0037,
+      "step": 42344
+    },
+    {
+      "epoch": 39.5,
+      "learning_rate": 4.506203358208955e-05,
+      "loss": 0.0001,
+      "step": 42348
+    },
+    {
+      "epoch": 39.51,
+      "learning_rate": 4.506156716417911e-05,
+      "loss": 0.0001,
+      "step": 42352
+    },
+    {
+      "epoch": 39.51,
+      "learning_rate": 4.5061100746268656e-05,
+      "loss": 0.0,
+      "step": 42356
+    },
+    {
+      "epoch": 39.51,
+      "learning_rate": 4.506063432835821e-05,
+      "loss": 0.0,
+      "step": 42360
+    },
+    {
+      "epoch": 39.52,
+      "learning_rate": 4.5060167910447765e-05,
+      "loss": 0.0,
+      "step": 42364
+    },
+    {
+      "epoch": 39.52,
+      "learning_rate": 4.5059701492537313e-05,
+      "loss": 0.0001,
+      "step": 42368
+    },
+    {
+      "epoch": 39.53,
+      "learning_rate": 4.505923507462687e-05,
+      "loss": 0.0,
+      "step": 42372
+    },
+    {
+      "epoch": 39.53,
+      "learning_rate": 4.505876865671642e-05,
+      "loss": 0.0001,
+      "step": 42376
+    },
+    {
+      "epoch": 39.53,
+      "learning_rate": 4.505830223880597e-05,
+      "loss": 0.0,
+      "step": 42380
+    },
+    {
+      "epoch": 39.54,
+      "learning_rate": 4.505783582089552e-05,
+      "loss": 0.0001,
+      "step": 42384
+    },
+    {
+      "epoch": 39.54,
+      "learning_rate": 4.505736940298508e-05,
+      "loss": 0.0001,
+      "step": 42388
+    },
+    {
+      "epoch": 39.54,
+      "learning_rate": 4.505690298507463e-05,
+      "loss": 0.0004,
+      "step": 42392
+    },
+    {
+      "epoch": 39.55,
+      "learning_rate": 4.505643656716418e-05,
+      "loss": 0.0005,
+      "step": 42396
+    },
+    {
+      "epoch": 39.55,
+      "learning_rate": 4.505597014925373e-05,
+      "loss": 0.0,
+      "step": 42400
+    },
+    {
+      "epoch": 39.56,
+      "learning_rate": 4.505550373134329e-05,
+      "loss": 0.0008,
+      "step": 42404
+    },
+    {
+      "epoch": 39.56,
+      "learning_rate": 4.505503731343284e-05,
+      "loss": 0.0,
+      "step": 42408
+    },
+    {
+      "epoch": 39.56,
+      "learning_rate": 4.505457089552239e-05,
+      "loss": 0.0002,
+      "step": 42412
+    },
+    {
+      "epoch": 39.57,
+      "learning_rate": 4.505410447761194e-05,
+      "loss": 0.0011,
+      "step": 42416
+    },
+    {
+      "epoch": 39.57,
+      "learning_rate": 4.50536380597015e-05,
+      "loss": 0.0007,
+      "step": 42420
+    },
+    {
+      "epoch": 39.57,
+      "learning_rate": 4.505317164179105e-05,
+      "loss": 0.0,
+      "step": 42424
+    },
+    {
+      "epoch": 39.58,
+      "learning_rate": 4.5052705223880596e-05,
+      "loss": 0.0,
+      "step": 42428
+    },
+    {
+      "epoch": 39.58,
+      "learning_rate": 4.505223880597015e-05,
+      "loss": 0.0001,
+      "step": 42432
+    },
+    {
+      "epoch": 39.59,
+      "learning_rate": 4.5051772388059706e-05,
+      "loss": 0.0001,
+      "step": 42436
+    },
+    {
+      "epoch": 39.59,
+      "learning_rate": 4.5051305970149254e-05,
+      "loss": 0.0003,
+      "step": 42440
+    },
+    {
+      "epoch": 39.59,
+      "learning_rate": 4.505083955223881e-05,
+      "loss": 0.0003,
+      "step": 42444
+    },
+    {
+      "epoch": 39.6,
+      "learning_rate": 4.5050373134328364e-05,
+      "loss": 0.0001,
+      "step": 42448
+    },
+    {
+      "epoch": 39.6,
+      "learning_rate": 4.504990671641791e-05,
+      "loss": 0.0006,
+      "step": 42452
+    },
+    {
+      "epoch": 39.6,
+      "learning_rate": 4.504944029850747e-05,
+      "loss": 0.0,
+      "step": 42456
+    },
+    {
+      "epoch": 39.61,
+      "learning_rate": 4.5048973880597015e-05,
+      "loss": 0.0004,
+      "step": 42460
+    },
+    {
+      "epoch": 39.61,
+      "learning_rate": 4.504850746268657e-05,
+      "loss": 0.0001,
+      "step": 42464
+    },
+    {
+      "epoch": 39.62,
+      "learning_rate": 4.5048041044776125e-05,
+      "loss": 0.0,
+      "step": 42468
+    },
+    {
+      "epoch": 39.62,
+      "learning_rate": 4.504757462686567e-05,
+      "loss": 0.0,
+      "step": 42472
+    },
+    {
+      "epoch": 39.62,
+      "learning_rate": 4.504710820895522e-05,
+      "loss": 0.0008,
+      "step": 42476
+    },
+    {
+      "epoch": 39.63,
+      "learning_rate": 4.504664179104478e-05,
+      "loss": 0.0,
+      "step": 42480
+    },
+    {
+      "epoch": 39.63,
+      "learning_rate": 4.504617537313433e-05,
+      "loss": 0.0015,
+      "step": 42484
+    },
+    {
+      "epoch": 39.63,
+      "learning_rate": 4.504570895522388e-05,
+      "loss": 0.0004,
+      "step": 42488
+    },
+    {
+      "epoch": 39.64,
+      "learning_rate": 4.5045242537313434e-05,
+      "loss": 0.0001,
+      "step": 42492
+    },
+    {
+      "epoch": 39.64,
+      "learning_rate": 4.504477611940299e-05,
+      "loss": 0.0002,
+      "step": 42496
+    },
+    {
+      "epoch": 39.65,
+      "learning_rate": 4.504430970149254e-05,
+      "loss": 0.0002,
+      "step": 42500
+    },
+    {
+      "epoch": 39.65,
+      "eval_exact_match": 0.7253384912959381,
+      "eval_exec": 0.758220502901354,
+      "eval_loss": 0.42694738507270813,
+      "eval_runtime": 1134.6223,
+      "eval_samples_per_second": 0.911,
+      "step": 42500
+    },
+    {
+      "epoch": 39.65,
+      "learning_rate": 4.504384328358209e-05,
+      "loss": 0.0001,
+      "step": 42504
+    },
+    {
+      "epoch": 39.65,
+      "learning_rate": 4.504337686567164e-05,
+      "loss": 0.0,
+      "step": 42508
+    },
+    {
+      "epoch": 39.66,
+      "learning_rate": 4.5042910447761195e-05,
+      "loss": 0.0,
+      "step": 42512
+    },
+    {
+      "epoch": 39.66,
+      "learning_rate": 4.504244402985075e-05,
+      "loss": 0.0001,
+      "step": 42516
+    },
+    {
+      "epoch": 39.66,
+      "learning_rate": 4.50419776119403e-05,
+      "loss": 0.0,
+      "step": 42520
+    },
+    {
+      "epoch": 39.67,
+      "learning_rate": 4.504151119402985e-05,
+      "loss": 0.0001,
+      "step": 42524
+    },
+    {
+      "epoch": 39.67,
+      "learning_rate": 4.504104477611941e-05,
+      "loss": 0.0,
+      "step": 42528
+    },
+    {
+      "epoch": 39.68,
+      "learning_rate": 4.5040578358208956e-05,
+      "loss": 0.0,
+      "step": 42532
+    },
+    {
+      "epoch": 39.68,
+      "learning_rate": 4.5040111940298504e-05,
+      "loss": 0.0016,
+      "step": 42536
+    },
+    {
+      "epoch": 39.68,
+      "learning_rate": 4.5039645522388066e-05,
+      "loss": 0.0002,
+      "step": 42540
+    },
+    {
+      "epoch": 39.69,
+      "learning_rate": 4.5039179104477614e-05,
+      "loss": 0.0003,
+      "step": 42544
+    },
+    {
+      "epoch": 39.69,
+      "learning_rate": 4.503871268656716e-05,
+      "loss": 0.0,
+      "step": 42548
+    },
+    {
+      "epoch": 39.69,
+      "learning_rate": 4.503824626865672e-05,
+      "loss": 0.0002,
+      "step": 42552
+    },
+    {
+      "epoch": 39.7,
+      "learning_rate": 4.503777985074627e-05,
+      "loss": 0.0,
+      "step": 42556
+    },
+    {
+      "epoch": 39.7,
+      "learning_rate": 4.503731343283582e-05,
+      "loss": 0.0,
+      "step": 42560
+    },
+    {
+      "epoch": 39.71,
+      "learning_rate": 4.5036847014925375e-05,
+      "loss": 0.0002,
+      "step": 42564
+    },
+    {
+      "epoch": 39.71,
+      "learning_rate": 4.503638059701492e-05,
+      "loss": 0.0001,
+      "step": 42568
+    },
+    {
+      "epoch": 39.71,
+      "learning_rate": 4.5035914179104485e-05,
+      "loss": 0.0001,
+      "step": 42572
+    },
+    {
+      "epoch": 39.72,
+      "learning_rate": 4.503544776119403e-05,
+      "loss": 0.0008,
+      "step": 42576
+    },
+    {
+      "epoch": 39.72,
+      "learning_rate": 4.503498134328358e-05,
+      "loss": 0.0001,
+      "step": 42580
+    },
+    {
+      "epoch": 39.72,
+      "learning_rate": 4.5034514925373136e-05,
+      "loss": 0.0006,
+      "step": 42584
+    },
+    {
+      "epoch": 39.73,
+      "learning_rate": 4.503404850746269e-05,
+      "loss": 0.0005,
+      "step": 42588
+    },
+    {
+      "epoch": 39.73,
+      "learning_rate": 4.503358208955224e-05,
+      "loss": 0.0,
+      "step": 42592
+    },
+    {
+      "epoch": 39.73,
+      "learning_rate": 4.5033115671641794e-05,
+      "loss": 0.0,
+      "step": 42596
+    },
+    {
+      "epoch": 39.74,
+      "learning_rate": 4.503264925373135e-05,
+      "loss": 0.0001,
+      "step": 42600
+    },
+    {
+      "epoch": 39.74,
+      "learning_rate": 4.50321828358209e-05,
+      "loss": 0.0,
+      "step": 42604
+    },
+    {
+      "epoch": 39.75,
+      "learning_rate": 4.503171641791045e-05,
+      "loss": 0.0004,
+      "step": 42608
+    },
+    {
+      "epoch": 39.75,
+      "learning_rate": 4.503125e-05,
+      "loss": 0.0001,
+      "step": 42612
+    },
+    {
+      "epoch": 39.75,
+      "learning_rate": 4.5030783582089555e-05,
+      "loss": 0.0002,
+      "step": 42616
+    },
+    {
+      "epoch": 39.76,
+      "learning_rate": 4.503031716417911e-05,
+      "loss": 0.0,
+      "step": 42620
+    },
+    {
+      "epoch": 39.76,
+      "learning_rate": 4.502985074626866e-05,
+      "loss": 0.0,
+      "step": 42624
+    },
+    {
+      "epoch": 39.76,
+      "learning_rate": 4.5029384328358206e-05,
+      "loss": 0.0,
+      "step": 42628
+    },
+    {
+      "epoch": 39.77,
+      "learning_rate": 4.502891791044777e-05,
+      "loss": 0.0,
+      "step": 42632
+    },
+    {
+      "epoch": 39.77,
+      "learning_rate": 4.5028451492537316e-05,
+      "loss": 0.0,
+      "step": 42636
+    },
+    {
+      "epoch": 39.78,
+      "learning_rate": 4.5027985074626864e-05,
+      "loss": 0.0001,
+      "step": 42640
+    },
+    {
+      "epoch": 39.78,
+      "learning_rate": 4.502751865671642e-05,
+      "loss": 0.0004,
+      "step": 42644
+    },
+    {
+      "epoch": 39.78,
+      "learning_rate": 4.5027052238805974e-05,
+      "loss": 0.0,
+      "step": 42648
+    },
+    {
+      "epoch": 39.79,
+      "learning_rate": 4.502658582089552e-05,
+      "loss": 0.0,
+      "step": 42652
+    },
+    {
+      "epoch": 39.79,
+      "learning_rate": 4.502611940298508e-05,
+      "loss": 0.0,
+      "step": 42656
+    },
+    {
+      "epoch": 39.79,
+      "learning_rate": 4.502565298507463e-05,
+      "loss": 0.0002,
+      "step": 42660
+    },
+    {
+      "epoch": 39.8,
+      "learning_rate": 4.502518656716418e-05,
+      "loss": 0.0,
+      "step": 42664
+    },
+    {
+      "epoch": 39.8,
+      "learning_rate": 4.5024720149253735e-05,
+      "loss": 0.0007,
+      "step": 42668
+    },
+    {
+      "epoch": 39.81,
+      "learning_rate": 4.502425373134328e-05,
+      "loss": 0.0002,
+      "step": 42672
+    },
+    {
+      "epoch": 39.81,
+      "learning_rate": 4.502378731343284e-05,
+      "loss": 0.0,
+      "step": 42676
+    },
+    {
+      "epoch": 39.81,
+      "learning_rate": 4.502332089552239e-05,
+      "loss": 0.0,
+      "step": 42680
+    },
+    {
+      "epoch": 39.82,
+      "learning_rate": 4.502285447761194e-05,
+      "loss": 0.0008,
+      "step": 42684
+    },
+    {
+      "epoch": 39.82,
+      "learning_rate": 4.502238805970149e-05,
+      "loss": 0.0002,
+      "step": 42688
+    },
+    {
+      "epoch": 39.82,
+      "learning_rate": 4.502192164179105e-05,
+      "loss": 0.0004,
+      "step": 42692
+    },
+    {
+      "epoch": 39.83,
+      "learning_rate": 4.50214552238806e-05,
+      "loss": 0.0031,
+      "step": 42696
+    },
+    {
+      "epoch": 39.83,
+      "learning_rate": 4.502098880597015e-05,
+      "loss": 0.0,
+      "step": 42700
+    },
+    {
+      "epoch": 39.84,
+      "learning_rate": 4.50205223880597e-05,
+      "loss": 0.0003,
+      "step": 42704
+    },
+    {
+      "epoch": 39.84,
+      "learning_rate": 4.502005597014926e-05,
+      "loss": 0.0,
+      "step": 42708
+    },
+    {
+      "epoch": 39.84,
+      "learning_rate": 4.5019589552238805e-05,
+      "loss": 0.001,
+      "step": 42712
+    },
+    {
+      "epoch": 39.85,
+      "learning_rate": 4.501912313432836e-05,
+      "loss": 0.0,
+      "step": 42716
+    },
+    {
+      "epoch": 39.85,
+      "learning_rate": 4.5018656716417915e-05,
+      "loss": 0.0,
+      "step": 42720
+    },
+    {
+      "epoch": 39.85,
+      "learning_rate": 4.501819029850746e-05,
+      "loss": 0.0,
+      "step": 42724
+    },
+    {
+      "epoch": 39.86,
+      "learning_rate": 4.501772388059702e-05,
+      "loss": 0.0001,
+      "step": 42728
+    },
+    {
+      "epoch": 39.86,
+      "learning_rate": 4.5017257462686566e-05,
+      "loss": 0.0026,
+      "step": 42732
+    },
+    {
+      "epoch": 39.87,
+      "learning_rate": 4.501679104477612e-05,
+      "loss": 0.0008,
+      "step": 42736
+    },
+    {
+      "epoch": 39.87,
+      "learning_rate": 4.5016324626865676e-05,
+      "loss": 0.0,
+      "step": 42740
+    },
+    {
+      "epoch": 39.87,
+      "learning_rate": 4.5015858208955224e-05,
+      "loss": 0.0,
+      "step": 42744
+    },
+    {
+      "epoch": 39.88,
+      "learning_rate": 4.501539179104478e-05,
+      "loss": 0.0001,
+      "step": 42748
+    },
+    {
+      "epoch": 39.88,
+      "learning_rate": 4.5014925373134334e-05,
+      "loss": 0.0,
+      "step": 42752
+    },
+    {
+      "epoch": 39.88,
+      "learning_rate": 4.501445895522388e-05,
+      "loss": 0.0,
+      "step": 42756
+    },
+    {
+      "epoch": 39.89,
+      "learning_rate": 4.501399253731344e-05,
+      "loss": 0.0001,
+      "step": 42760
+    },
+    {
+      "epoch": 39.89,
+      "learning_rate": 4.5013526119402985e-05,
+      "loss": 0.0055,
+      "step": 42764
+    },
+    {
+      "epoch": 39.9,
+      "learning_rate": 4.501305970149254e-05,
+      "loss": 0.0,
+      "step": 42768
+    },
+    {
+      "epoch": 39.9,
+      "learning_rate": 4.5012593283582095e-05,
+      "loss": 0.0006,
+      "step": 42772
+    },
+    {
+      "epoch": 39.9,
+      "learning_rate": 4.501212686567164e-05,
+      "loss": 0.0,
+      "step": 42776
+    },
+    {
+      "epoch": 39.91,
+      "learning_rate": 4.50116604477612e-05,
+      "loss": 0.0004,
+      "step": 42780
+    },
+    {
+      "epoch": 39.91,
+      "learning_rate": 4.501119402985075e-05,
+      "loss": 0.0001,
+      "step": 42784
+    },
+    {
+      "epoch": 39.91,
+      "learning_rate": 4.50107276119403e-05,
+      "loss": 0.0001,
+      "step": 42788
+    },
+    {
+      "epoch": 39.92,
+      "learning_rate": 4.501026119402985e-05,
+      "loss": 0.0006,
+      "step": 42792
+    },
+    {
+      "epoch": 39.92,
+      "learning_rate": 4.5009794776119404e-05,
+      "loss": 0.0028,
+      "step": 42796
+    },
+    {
+      "epoch": 39.93,
+      "learning_rate": 4.500932835820896e-05,
+      "loss": 0.0006,
+      "step": 42800
+    },
+    {
+      "epoch": 39.93,
+      "learning_rate": 4.500886194029851e-05,
+      "loss": 0.0002,
+      "step": 42804
+    },
+    {
+      "epoch": 39.93,
+      "learning_rate": 4.500839552238806e-05,
+      "loss": 0.0001,
+      "step": 42808
+    },
+    {
+      "epoch": 39.94,
+      "learning_rate": 4.500792910447762e-05,
+      "loss": 0.0001,
+      "step": 42812
+    },
+    {
+      "epoch": 39.94,
+      "learning_rate": 4.5007462686567165e-05,
+      "loss": 0.0001,
+      "step": 42816
+    },
+    {
+      "epoch": 39.94,
+      "learning_rate": 4.500699626865672e-05,
+      "loss": 0.0,
+      "step": 42820
+    },
+    {
+      "epoch": 39.95,
+      "learning_rate": 4.500652985074627e-05,
+      "loss": 0.0002,
+      "step": 42824
+    },
+    {
+      "epoch": 39.95,
+      "learning_rate": 4.500606343283582e-05,
+      "loss": 0.0001,
+      "step": 42828
+    },
+    {
+      "epoch": 39.96,
+      "learning_rate": 4.500559701492538e-05,
+      "loss": 0.0004,
+      "step": 42832
+    },
+    {
+      "epoch": 39.96,
+      "learning_rate": 4.5005130597014926e-05,
+      "loss": 0.0,
+      "step": 42836
+    },
+    {
+      "epoch": 39.96,
+      "learning_rate": 4.500466417910448e-05,
+      "loss": 0.0006,
+      "step": 42840
+    },
+    {
+      "epoch": 39.97,
+      "learning_rate": 4.5004197761194036e-05,
+      "loss": 0.0023,
+      "step": 42844
+    },
+    {
+      "epoch": 39.97,
+      "learning_rate": 4.5003731343283584e-05,
+      "loss": 0.0,
+      "step": 42848
+    },
+    {
+      "epoch": 39.97,
+      "learning_rate": 4.500326492537313e-05,
+      "loss": 0.0045,
+      "step": 42852
+    },
+    {
+      "epoch": 39.98,
+      "learning_rate": 4.500279850746269e-05,
+      "loss": 0.0001,
+      "step": 42856
+    },
+    {
+      "epoch": 39.98,
+      "learning_rate": 4.500233208955224e-05,
+      "loss": 0.0001,
+      "step": 42860
+    },
+    {
+      "epoch": 39.98,
+      "learning_rate": 4.500186567164179e-05,
+      "loss": 0.0188,
+      "step": 42864
+    },
+    {
+      "epoch": 39.99,
+      "learning_rate": 4.5001399253731345e-05,
+      "loss": 0.0,
+      "step": 42868
+    },
+    {
+      "epoch": 39.99,
+      "learning_rate": 4.50009328358209e-05,
+      "loss": 0.0,
+      "step": 42872
+    },
+    {
+      "epoch": 40.0,
+      "learning_rate": 4.500046641791045e-05,
+      "loss": 0.0005,
+      "step": 42876
+    },
+    {
+      "epoch": 40.0,
+      "learning_rate": 4.5e-05,
+      "loss": 0.0057,
+      "step": 42880
+    },
+    {
+      "epoch": 40.0,
+      "learning_rate": 4.499953358208955e-05,
+      "loss": 0.0001,
+      "step": 42884
+    },
+    {
+      "epoch": 40.01,
+      "learning_rate": 4.4999067164179106e-05,
+      "loss": 0.0001,
+      "step": 42888
+    },
+    {
+      "epoch": 40.01,
+      "learning_rate": 4.499860074626866e-05,
+      "loss": 0.0,
+      "step": 42892
+    },
+    {
+      "epoch": 40.01,
+      "learning_rate": 4.499813432835821e-05,
+      "loss": 0.0,
+      "step": 42896
+    },
+    {
+      "epoch": 40.02,
+      "learning_rate": 4.4997667910447764e-05,
+      "loss": 0.0,
+      "step": 42900
+    },
+    {
+      "epoch": 40.02,
+      "learning_rate": 4.499720149253732e-05,
+      "loss": 0.0,
+      "step": 42904
+    },
+    {
+      "epoch": 40.03,
+      "learning_rate": 4.499673507462687e-05,
+      "loss": 0.0001,
+      "step": 42908
+    },
+    {
+      "epoch": 40.03,
+      "learning_rate": 4.499626865671642e-05,
+      "loss": 0.0015,
+      "step": 42912
+    },
+    {
+      "epoch": 40.03,
+      "learning_rate": 4.499580223880597e-05,
+      "loss": 0.001,
+      "step": 42916
+    },
+    {
+      "epoch": 40.04,
+      "learning_rate": 4.4995335820895525e-05,
+      "loss": 0.0001,
+      "step": 42920
+    },
+    {
+      "epoch": 40.04,
+      "learning_rate": 4.499486940298508e-05,
+      "loss": 0.0004,
+      "step": 42924
+    },
+    {
+      "epoch": 40.04,
+      "learning_rate": 4.499440298507463e-05,
+      "loss": 0.0,
+      "step": 42928
+    },
+    {
+      "epoch": 40.05,
+      "learning_rate": 4.499393656716418e-05,
+      "loss": 0.0004,
+      "step": 42932
+    },
+    {
+      "epoch": 40.05,
+      "learning_rate": 4.499347014925374e-05,
+      "loss": 0.0017,
+      "step": 42936
+    },
+    {
+      "epoch": 40.06,
+      "learning_rate": 4.4993003731343286e-05,
+      "loss": 0.0006,
+      "step": 42940
+    },
+    {
+      "epoch": 40.06,
+      "learning_rate": 4.4992537313432834e-05,
+      "loss": 0.0,
+      "step": 42944
+    },
+    {
+      "epoch": 40.06,
+      "learning_rate": 4.4992070895522396e-05,
+      "loss": 0.0004,
+      "step": 42948
+    },
+    {
+      "epoch": 40.07,
+      "learning_rate": 4.4991604477611944e-05,
+      "loss": 0.0,
+      "step": 42952
+    },
+    {
+      "epoch": 40.07,
+      "learning_rate": 4.499113805970149e-05,
+      "loss": 0.0001,
+      "step": 42956
+    },
+    {
+      "epoch": 40.07,
+      "learning_rate": 4.499067164179105e-05,
+      "loss": 0.0,
+      "step": 42960
+    },
+    {
+      "epoch": 40.08,
+      "learning_rate": 4.49902052238806e-05,
+      "loss": 0.0,
+      "step": 42964
+    },
+    {
+      "epoch": 40.08,
+      "learning_rate": 4.498973880597015e-05,
+      "loss": 0.0002,
+      "step": 42968
+    },
+    {
+      "epoch": 40.09,
+      "learning_rate": 4.4989272388059705e-05,
+      "loss": 0.0001,
+      "step": 42972
+    },
+    {
+      "epoch": 40.09,
+      "learning_rate": 4.498880597014925e-05,
+      "loss": 0.0001,
+      "step": 42976
+    },
+    {
+      "epoch": 40.09,
+      "learning_rate": 4.498833955223881e-05,
+      "loss": 0.0,
+      "step": 42980
+    },
+    {
+      "epoch": 40.1,
+      "learning_rate": 4.498787313432836e-05,
+      "loss": 0.0001,
+      "step": 42984
+    },
+    {
+      "epoch": 40.1,
+      "learning_rate": 4.498740671641791e-05,
+      "loss": 0.0,
+      "step": 42988
+    },
+    {
+      "epoch": 40.1,
+      "learning_rate": 4.4986940298507466e-05,
+      "loss": 0.0,
+      "step": 42992
+    },
+    {
+      "epoch": 40.11,
+      "learning_rate": 4.498647388059702e-05,
+      "loss": 0.0001,
+      "step": 42996
+    },
+    {
+      "epoch": 40.11,
+      "learning_rate": 4.498600746268657e-05,
+      "loss": 0.0,
+      "step": 43000
+    },
+    {
+      "epoch": 40.11,
+      "eval_exact_match": 0.7475822050290135,
+      "eval_exec": 0.7736943907156673,
+      "eval_loss": 0.4348256289958954,
+      "eval_runtime": 1267.9199,
+      "eval_samples_per_second": 0.816,
+      "step": 43000
+    },
+    {
+      "epoch": 40.12,
+      "learning_rate": 4.498554104477612e-05,
+      "loss": 0.0,
+      "step": 43004
+    },
+    {
+      "epoch": 40.12,
+      "learning_rate": 4.498507462686568e-05,
+      "loss": 0.0,
+      "step": 43008
+    },
+    {
+      "epoch": 40.12,
+      "learning_rate": 4.498460820895523e-05,
+      "loss": 0.0,
+      "step": 43012
+    },
+    {
+      "epoch": 40.13,
+      "learning_rate": 4.4984141791044775e-05,
+      "loss": 0.0013,
+      "step": 43016
+    },
+    {
+      "epoch": 40.13,
+      "learning_rate": 4.498367537313433e-05,
+      "loss": 0.0,
+      "step": 43020
+    },
+    {
+      "epoch": 40.13,
+      "learning_rate": 4.4983208955223885e-05,
+      "loss": 0.0,
+      "step": 43024
+    },
+    {
+      "epoch": 40.14,
+      "learning_rate": 4.498274253731343e-05,
+      "loss": 0.0024,
+      "step": 43028
+    },
+    {
+      "epoch": 40.14,
+      "learning_rate": 4.498227611940299e-05,
+      "loss": 0.0,
+      "step": 43032
+    },
+    {
+      "epoch": 40.15,
+      "learning_rate": 4.4981809701492536e-05,
+      "loss": 0.0001,
+      "step": 43036
+    },
+    {
+      "epoch": 40.15,
+      "learning_rate": 4.498134328358209e-05,
+      "loss": 0.0,
+      "step": 43040
+    },
+    {
+      "epoch": 40.15,
+      "learning_rate": 4.4980876865671646e-05,
+      "loss": 0.0001,
+      "step": 43044
+    },
+    {
+      "epoch": 40.16,
+      "learning_rate": 4.4980410447761194e-05,
+      "loss": 0.0001,
+      "step": 43048
+    },
+    {
+      "epoch": 40.16,
+      "learning_rate": 4.497994402985075e-05,
+      "loss": 0.0,
+      "step": 43052
+    },
+    {
+      "epoch": 40.16,
+      "learning_rate": 4.4979477611940304e-05,
+      "loss": 0.0,
+      "step": 43056
+    },
+    {
+      "epoch": 40.17,
+      "learning_rate": 4.497901119402985e-05,
+      "loss": 0.0,
+      "step": 43060
+    },
+    {
+      "epoch": 40.17,
+      "learning_rate": 4.497854477611941e-05,
+      "loss": 0.0,
+      "step": 43064
+    },
+    {
+      "epoch": 40.18,
+      "learning_rate": 4.497807835820896e-05,
+      "loss": 0.0012,
+      "step": 43068
+    },
+    {
+      "epoch": 40.18,
+      "learning_rate": 4.497761194029851e-05,
+      "loss": 0.0,
+      "step": 43072
+    },
+    {
+      "epoch": 40.18,
+      "learning_rate": 4.4977145522388065e-05,
+      "loss": 0.0006,
+      "step": 43076
+    },
+    {
+      "epoch": 40.19,
+      "learning_rate": 4.497667910447761e-05,
+      "loss": 0.0001,
+      "step": 43080
+    },
+    {
+      "epoch": 40.19,
+      "learning_rate": 4.497621268656717e-05,
+      "loss": 0.0,
+      "step": 43084
+    },
+    {
+      "epoch": 40.19,
+      "learning_rate": 4.497574626865672e-05,
+      "loss": 0.0,
+      "step": 43088
+    },
+    {
+      "epoch": 40.2,
+      "learning_rate": 4.497527985074627e-05,
+      "loss": 0.0001,
+      "step": 43092
+    },
+    {
+      "epoch": 40.2,
+      "learning_rate": 4.497481343283582e-05,
+      "loss": 0.0,
+      "step": 43096
+    },
+    {
+      "epoch": 40.21,
+      "learning_rate": 4.497434701492538e-05,
+      "loss": 0.0,
+      "step": 43100
+    },
+    {
+      "epoch": 40.21,
+      "learning_rate": 4.497388059701493e-05,
+      "loss": 0.0004,
+      "step": 43104
+    },
+    {
+      "epoch": 40.21,
+      "learning_rate": 4.497341417910448e-05,
+      "loss": 0.0,
+      "step": 43108
+    },
+    {
+      "epoch": 40.22,
+      "learning_rate": 4.497294776119403e-05,
+      "loss": 0.0,
+      "step": 43112
+    },
+    {
+      "epoch": 40.22,
+      "learning_rate": 4.4972481343283587e-05,
+      "loss": 0.0,
+      "step": 43116
+    },
+    {
+      "epoch": 40.22,
+      "learning_rate": 4.4972014925373135e-05,
+      "loss": 0.0003,
+      "step": 43120
+    },
+    {
+      "epoch": 40.23,
+      "learning_rate": 4.497154850746269e-05,
+      "loss": 0.0002,
+      "step": 43124
+    },
+    {
+      "epoch": 40.23,
+      "learning_rate": 4.4971082089552244e-05,
+      "loss": 0.0001,
+      "step": 43128
+    },
+    {
+      "epoch": 40.24,
+      "learning_rate": 4.497061567164179e-05,
+      "loss": 0.0003,
+      "step": 43132
+    },
+    {
+      "epoch": 40.24,
+      "learning_rate": 4.497014925373135e-05,
+      "loss": 0.0,
+      "step": 43136
+    },
+    {
+      "epoch": 40.24,
+      "learning_rate": 4.4969682835820896e-05,
+      "loss": 0.0018,
+      "step": 43140
+    },
+    {
+      "epoch": 40.25,
+      "learning_rate": 4.496921641791045e-05,
+      "loss": 0.0003,
+      "step": 43144
+    },
+    {
+      "epoch": 40.25,
+      "learning_rate": 4.4968750000000005e-05,
+      "loss": 0.0001,
+      "step": 43148
+    },
+    {
+      "epoch": 40.25,
+      "learning_rate": 4.4968283582089554e-05,
+      "loss": 0.0001,
+      "step": 43152
+    },
+    {
+      "epoch": 40.26,
+      "learning_rate": 4.49678171641791e-05,
+      "loss": 0.0,
+      "step": 43156
+    },
+    {
+      "epoch": 40.26,
+      "learning_rate": 4.4967350746268663e-05,
+      "loss": 0.0,
+      "step": 43160
+    },
+    {
+      "epoch": 40.26,
+      "learning_rate": 4.496688432835821e-05,
+      "loss": 0.0,
+      "step": 43164
+    },
+    {
+      "epoch": 40.27,
+      "learning_rate": 4.496641791044776e-05,
+      "loss": 0.0013,
+      "step": 43168
+    },
+    {
+      "epoch": 40.27,
+      "learning_rate": 4.4965951492537315e-05,
+      "loss": 0.0022,
+      "step": 43172
+    },
+    {
+      "epoch": 40.28,
+      "learning_rate": 4.496548507462687e-05,
+      "loss": 0.0003,
+      "step": 43176
+    },
+    {
+      "epoch": 40.28,
+      "learning_rate": 4.496501865671642e-05,
+      "loss": 0.0008,
+      "step": 43180
+    },
+    {
+      "epoch": 40.28,
+      "learning_rate": 4.496455223880597e-05,
+      "loss": 0.0,
+      "step": 43184
+    },
+    {
+      "epoch": 40.29,
+      "learning_rate": 4.496408582089552e-05,
+      "loss": 0.0001,
+      "step": 43188
+    },
+    {
+      "epoch": 40.29,
+      "learning_rate": 4.4963619402985076e-05,
+      "loss": 0.0,
+      "step": 43192
+    },
+    {
+      "epoch": 40.29,
+      "learning_rate": 4.496315298507463e-05,
+      "loss": 0.0001,
+      "step": 43196
+    },
+    {
+      "epoch": 40.3,
+      "learning_rate": 4.496268656716418e-05,
+      "loss": 0.0,
+      "step": 43200
+    },
+    {
+      "epoch": 40.3,
+      "learning_rate": 4.4962220149253734e-05,
+      "loss": 0.0001,
+      "step": 43204
+    },
+    {
+      "epoch": 40.31,
+      "learning_rate": 4.496175373134329e-05,
+      "loss": 0.0001,
+      "step": 43208
+    },
+    {
+      "epoch": 40.31,
+      "learning_rate": 4.4961287313432837e-05,
+      "loss": 0.0,
+      "step": 43212
+    },
+    {
+      "epoch": 40.31,
+      "learning_rate": 4.4960820895522385e-05,
+      "loss": 0.0008,
+      "step": 43216
+    },
+    {
+      "epoch": 40.32,
+      "learning_rate": 4.4960354477611946e-05,
+      "loss": 0.0002,
+      "step": 43220
+    },
+    {
+      "epoch": 40.32,
+      "learning_rate": 4.4959888059701494e-05,
+      "loss": 0.0,
+      "step": 43224
+    },
+    {
+      "epoch": 40.32,
+      "learning_rate": 4.495942164179105e-05,
+      "loss": 0.0025,
+      "step": 43228
+    },
+    {
+      "epoch": 40.33,
+      "learning_rate": 4.49589552238806e-05,
+      "loss": 0.0001,
+      "step": 43232
+    },
+    {
+      "epoch": 40.33,
+      "learning_rate": 4.495848880597015e-05,
+      "loss": 0.0,
+      "step": 43236
+    },
+    {
+      "epoch": 40.34,
+      "learning_rate": 4.495802238805971e-05,
+      "loss": 0.0032,
+      "step": 43240
+    },
+    {
+      "epoch": 40.34,
+      "learning_rate": 4.4957555970149255e-05,
+      "loss": 0.0,
+      "step": 43244
+    },
+    {
+      "epoch": 40.34,
+      "learning_rate": 4.4957089552238804e-05,
+      "loss": 0.0,
+      "step": 43248
+    },
+    {
+      "epoch": 40.35,
+      "learning_rate": 4.4956623134328365e-05,
+      "loss": 0.0097,
+      "step": 43252
+    },
+    {
+      "epoch": 40.35,
+      "learning_rate": 4.4956156716417913e-05,
+      "loss": 0.0077,
+      "step": 43256
+    },
+    {
+      "epoch": 40.35,
+      "learning_rate": 4.495569029850746e-05,
+      "loss": 0.0002,
+      "step": 43260
+    },
+    {
+      "epoch": 40.36,
+      "learning_rate": 4.4955223880597016e-05,
+      "loss": 0.0002,
+      "step": 43264
+    },
+    {
+      "epoch": 40.36,
+      "learning_rate": 4.495475746268657e-05,
+      "loss": 0.0,
+      "step": 43268
+    },
+    {
+      "epoch": 40.37,
+      "learning_rate": 4.495429104477612e-05,
+      "loss": 0.0004,
+      "step": 43272
+    },
+    {
+      "epoch": 40.37,
+      "learning_rate": 4.4953824626865674e-05,
+      "loss": 0.0,
+      "step": 43276
+    },
+    {
+      "epoch": 40.37,
+      "learning_rate": 4.495335820895523e-05,
+      "loss": 0.0011,
+      "step": 43280
+    },
+    {
+      "epoch": 40.38,
+      "learning_rate": 4.495289179104478e-05,
+      "loss": 0.0002,
+      "step": 43284
+    },
+    {
+      "epoch": 40.38,
+      "learning_rate": 4.495242537313433e-05,
+      "loss": 0.0,
+      "step": 43288
+    },
+    {
+      "epoch": 40.38,
+      "learning_rate": 4.495195895522388e-05,
+      "loss": 0.0,
+      "step": 43292
+    },
+    {
+      "epoch": 40.39,
+      "learning_rate": 4.4951492537313435e-05,
+      "loss": 0.0039,
+      "step": 43296
+    },
+    {
+      "epoch": 40.39,
+      "learning_rate": 4.495102611940299e-05,
+      "loss": 0.0004,
+      "step": 43300
+    },
+    {
+      "epoch": 40.4,
+      "learning_rate": 4.495055970149254e-05,
+      "loss": 0.0003,
+      "step": 43304
+    },
+    {
+      "epoch": 40.4,
+      "learning_rate": 4.4950093283582087e-05,
+      "loss": 0.0001,
+      "step": 43308
+    },
+    {
+      "epoch": 40.4,
+      "learning_rate": 4.494962686567165e-05,
+      "loss": 0.0001,
+      "step": 43312
+    },
+    {
+      "epoch": 40.41,
+      "learning_rate": 4.4949160447761196e-05,
+      "loss": 0.0006,
+      "step": 43316
+    },
+    {
+      "epoch": 40.41,
+      "learning_rate": 4.4948694029850745e-05,
+      "loss": 0.0,
+      "step": 43320
+    },
+    {
+      "epoch": 40.41,
+      "learning_rate": 4.49482276119403e-05,
+      "loss": 0.0,
+      "step": 43324
+    },
+    {
+      "epoch": 40.42,
+      "learning_rate": 4.4947761194029854e-05,
+      "loss": 0.0001,
+      "step": 43328
+    },
+    {
+      "epoch": 40.42,
+      "learning_rate": 4.49472947761194e-05,
+      "loss": 0.0,
+      "step": 43332
+    },
+    {
+      "epoch": 40.43,
+      "learning_rate": 4.494682835820896e-05,
+      "loss": 0.0058,
+      "step": 43336
+    },
+    {
+      "epoch": 40.43,
+      "learning_rate": 4.494636194029851e-05,
+      "loss": 0.0001,
+      "step": 43340
+    },
+    {
+      "epoch": 40.43,
+      "learning_rate": 4.494589552238806e-05,
+      "loss": 0.0001,
+      "step": 43344
+    },
+    {
+      "epoch": 40.44,
+      "learning_rate": 4.4945429104477615e-05,
+      "loss": 0.0006,
+      "step": 43348
+    },
+    {
+      "epoch": 40.44,
+      "learning_rate": 4.4944962686567163e-05,
+      "loss": 0.0006,
+      "step": 43352
+    },
+    {
+      "epoch": 40.44,
+      "learning_rate": 4.494449626865672e-05,
+      "loss": 0.0001,
+      "step": 43356
+    },
+    {
+      "epoch": 40.45,
+      "learning_rate": 4.494402985074627e-05,
+      "loss": 0.0,
+      "step": 43360
+    },
+    {
+      "epoch": 40.45,
+      "learning_rate": 4.494356343283582e-05,
+      "loss": 0.0013,
+      "step": 43364
+    },
+    {
+      "epoch": 40.46,
+      "learning_rate": 4.494309701492537e-05,
+      "loss": 0.0009,
+      "step": 43368
+    },
+    {
+      "epoch": 40.46,
+      "learning_rate": 4.494263059701493e-05,
+      "loss": 0.0001,
+      "step": 43372
+    },
+    {
+      "epoch": 40.46,
+      "learning_rate": 4.494216417910448e-05,
+      "loss": 0.0,
+      "step": 43376
+    },
+    {
+      "epoch": 40.47,
+      "learning_rate": 4.494169776119403e-05,
+      "loss": 0.0,
+      "step": 43380
+    },
+    {
+      "epoch": 40.47,
+      "learning_rate": 4.494123134328358e-05,
+      "loss": 0.0001,
+      "step": 43384
+    },
+    {
+      "epoch": 40.47,
+      "learning_rate": 4.494076492537314e-05,
+      "loss": 0.0,
+      "step": 43388
+    },
+    {
+      "epoch": 40.48,
+      "learning_rate": 4.494029850746269e-05,
+      "loss": 0.0,
+      "step": 43392
+    },
+    {
+      "epoch": 40.48,
+      "learning_rate": 4.493983208955224e-05,
+      "loss": 0.001,
+      "step": 43396
+    },
+    {
+      "epoch": 40.49,
+      "learning_rate": 4.4939365671641795e-05,
+      "loss": 0.0,
+      "step": 43400
+    },
+    {
+      "epoch": 40.49,
+      "learning_rate": 4.493889925373135e-05,
+      "loss": 0.0,
+      "step": 43404
+    },
+    {
+      "epoch": 40.49,
+      "learning_rate": 4.49384328358209e-05,
+      "loss": 0.0,
+      "step": 43408
+    },
+    {
+      "epoch": 40.5,
+      "learning_rate": 4.4937966417910446e-05,
+      "loss": 0.0002,
+      "step": 43412
+    },
+    {
+      "epoch": 40.5,
+      "learning_rate": 4.49375e-05,
+      "loss": 0.0018,
+      "step": 43416
+    },
+    {
+      "epoch": 40.5,
+      "learning_rate": 4.4937033582089556e-05,
+      "loss": 0.0,
+      "step": 43420
+    },
+    {
+      "epoch": 40.51,
+      "learning_rate": 4.4936567164179104e-05,
+      "loss": 0.003,
+      "step": 43424
+    },
+    {
+      "epoch": 40.51,
+      "learning_rate": 4.493610074626866e-05,
+      "loss": 0.0008,
+      "step": 43428
+    },
+    {
+      "epoch": 40.51,
+      "learning_rate": 4.4935634328358214e-05,
+      "loss": 0.0,
+      "step": 43432
+    },
+    {
+      "epoch": 40.52,
+      "learning_rate": 4.493516791044776e-05,
+      "loss": 0.009,
+      "step": 43436
+    },
+    {
+      "epoch": 40.52,
+      "learning_rate": 4.493470149253732e-05,
+      "loss": 0.0,
+      "step": 43440
+    },
+    {
+      "epoch": 40.53,
+      "learning_rate": 4.4934235074626865e-05,
+      "loss": 0.0,
+      "step": 43444
+    },
+    {
+      "epoch": 40.53,
+      "learning_rate": 4.493376865671642e-05,
+      "loss": 0.0,
+      "step": 43448
+    },
+    {
+      "epoch": 40.53,
+      "learning_rate": 4.4933302238805975e-05,
+      "loss": 0.0,
+      "step": 43452
+    },
+    {
+      "epoch": 40.54,
+      "learning_rate": 4.493283582089552e-05,
+      "loss": 0.0004,
+      "step": 43456
+    },
+    {
+      "epoch": 40.54,
+      "learning_rate": 4.493236940298508e-05,
+      "loss": 0.0004,
+      "step": 43460
+    },
+    {
+      "epoch": 40.54,
+      "learning_rate": 4.493190298507463e-05,
+      "loss": 0.0,
+      "step": 43464
+    },
+    {
+      "epoch": 40.55,
+      "learning_rate": 4.493143656716418e-05,
+      "loss": 0.0,
+      "step": 43468
+    },
+    {
+      "epoch": 40.55,
+      "learning_rate": 4.493097014925373e-05,
+      "loss": 0.0,
+      "step": 43472
+    },
+    {
+      "epoch": 40.56,
+      "learning_rate": 4.4930503731343284e-05,
+      "loss": 0.0001,
+      "step": 43476
+    },
+    {
+      "epoch": 40.56,
+      "learning_rate": 4.493003731343284e-05,
+      "loss": 0.0008,
+      "step": 43480
+    },
+    {
+      "epoch": 40.56,
+      "learning_rate": 4.492957089552239e-05,
+      "loss": 0.0001,
+      "step": 43484
+    },
+    {
+      "epoch": 40.57,
+      "learning_rate": 4.492910447761194e-05,
+      "loss": 0.0,
+      "step": 43488
+    },
+    {
+      "epoch": 40.57,
+      "learning_rate": 4.49286380597015e-05,
+      "loss": 0.0012,
+      "step": 43492
+    },
+    {
+      "epoch": 40.57,
+      "learning_rate": 4.4928171641791045e-05,
+      "loss": 0.0,
+      "step": 43496
+    },
+    {
+      "epoch": 40.58,
+      "learning_rate": 4.49277052238806e-05,
+      "loss": 0.0013,
+      "step": 43500
+    },
+    {
+      "epoch": 40.58,
+      "eval_exact_match": 0.7292069632495164,
+      "eval_exec": 0.758220502901354,
+      "eval_loss": 0.4297082722187042,
+      "eval_runtime": 1140.8475,
+      "eval_samples_per_second": 0.906,
+      "step": 43500
+    },
+    {
+      "epoch": 40.58,
+      "learning_rate": 4.492723880597015e-05,
+      "loss": 0.0,
+      "step": 43504
+    },
+    {
+      "epoch": 40.59,
+      "learning_rate": 4.49267723880597e-05,
+      "loss": 0.0001,
+      "step": 43508
+    },
+    {
+      "epoch": 40.59,
+      "learning_rate": 4.492630597014926e-05,
+      "loss": 0.0003,
+      "step": 43512
+    },
+    {
+      "epoch": 40.59,
+      "learning_rate": 4.4925839552238806e-05,
+      "loss": 0.0001,
+      "step": 43516
+    },
+    {
+      "epoch": 40.6,
+      "learning_rate": 4.492537313432836e-05,
+      "loss": 0.0108,
+      "step": 43520
+    },
+    {
+      "epoch": 40.6,
+      "learning_rate": 4.4924906716417916e-05,
+      "loss": 0.0,
+      "step": 43524
+    },
+    {
+      "epoch": 40.6,
+      "learning_rate": 4.4924440298507464e-05,
+      "loss": 0.0001,
+      "step": 43528
+    },
+    {
+      "epoch": 40.61,
+      "learning_rate": 4.492397388059701e-05,
+      "loss": 0.0,
+      "step": 43532
+    },
+    {
+      "epoch": 40.61,
+      "learning_rate": 4.492350746268657e-05,
+      "loss": 0.0001,
+      "step": 43536
+    },
+    {
+      "epoch": 40.62,
+      "learning_rate": 4.492304104477612e-05,
+      "loss": 0.0,
+      "step": 43540
+    },
+    {
+      "epoch": 40.62,
+      "learning_rate": 4.492257462686567e-05,
+      "loss": 0.005,
+      "step": 43544
+    },
+    {
+      "epoch": 40.62,
+      "learning_rate": 4.4922108208955225e-05,
+      "loss": 0.0,
+      "step": 43548
+    },
+    {
+      "epoch": 40.63,
+      "learning_rate": 4.492164179104478e-05,
+      "loss": 0.0001,
+      "step": 43552
+    },
+    {
+      "epoch": 40.63,
+      "learning_rate": 4.4921175373134335e-05,
+      "loss": 0.0003,
+      "step": 43556
+    },
+    {
+      "epoch": 40.63,
+      "learning_rate": 4.492070895522388e-05,
+      "loss": 0.0024,
+      "step": 43560
+    },
+    {
+      "epoch": 40.64,
+      "learning_rate": 4.492024253731343e-05,
+      "loss": 0.0004,
+      "step": 43564
+    },
+    {
+      "epoch": 40.64,
+      "learning_rate": 4.491977611940299e-05,
+      "loss": 0.0016,
+      "step": 43568
+    },
+    {
+      "epoch": 40.65,
+      "learning_rate": 4.491930970149254e-05,
+      "loss": 0.0014,
+      "step": 43572
+    },
+    {
+      "epoch": 40.65,
+      "learning_rate": 4.491884328358209e-05,
+      "loss": 0.0,
+      "step": 43576
+    },
+    {
+      "epoch": 40.65,
+      "learning_rate": 4.4918376865671644e-05,
+      "loss": 0.0018,
+      "step": 43580
+    },
+    {
+      "epoch": 40.66,
+      "learning_rate": 4.49179104477612e-05,
+      "loss": 0.0004,
+      "step": 43584
+    },
+    {
+      "epoch": 40.66,
+      "learning_rate": 4.491744402985075e-05,
+      "loss": 0.0006,
+      "step": 43588
+    },
+    {
+      "epoch": 40.66,
+      "learning_rate": 4.49169776119403e-05,
+      "loss": 0.0,
+      "step": 43592
+    },
+    {
+      "epoch": 40.67,
+      "learning_rate": 4.491651119402985e-05,
+      "loss": 0.0,
+      "step": 43596
+    },
+    {
+      "epoch": 40.67,
+      "learning_rate": 4.4916044776119405e-05,
+      "loss": 0.0014,
+      "step": 43600
+    },
+    {
+      "epoch": 40.68,
+      "learning_rate": 4.491557835820896e-05,
+      "loss": 0.0001,
+      "step": 43604
+    },
+    {
+      "epoch": 40.68,
+      "learning_rate": 4.491511194029851e-05,
+      "loss": 0.0,
+      "step": 43608
+    },
+    {
+      "epoch": 40.68,
+      "learning_rate": 4.491464552238806e-05,
+      "loss": 0.0,
+      "step": 43612
+    },
+    {
+      "epoch": 40.69,
+      "learning_rate": 4.491417910447762e-05,
+      "loss": 0.0,
+      "step": 43616
+    },
+    {
+      "epoch": 40.69,
+      "learning_rate": 4.4913712686567166e-05,
+      "loss": 0.0028,
+      "step": 43620
+    },
+    {
+      "epoch": 40.69,
+      "learning_rate": 4.4913246268656714e-05,
+      "loss": 0.0,
+      "step": 43624
+    },
+    {
+      "epoch": 40.7,
+      "learning_rate": 4.4912779850746276e-05,
+      "loss": 0.0,
+      "step": 43628
+    },
+    {
+      "epoch": 40.7,
+      "learning_rate": 4.4912313432835824e-05,
+      "loss": 0.0001,
+      "step": 43632
+    },
+    {
+      "epoch": 40.71,
+      "learning_rate": 4.491184701492537e-05,
+      "loss": 0.0001,
+      "step": 43636
+    },
+    {
+      "epoch": 40.71,
+      "learning_rate": 4.491138059701493e-05,
+      "loss": 0.0,
+      "step": 43640
+    },
+    {
+      "epoch": 40.71,
+      "learning_rate": 4.491091417910448e-05,
+      "loss": 0.0,
+      "step": 43644
+    },
+    {
+      "epoch": 40.72,
+      "learning_rate": 4.491044776119403e-05,
+      "loss": 0.0,
+      "step": 43648
+    },
+    {
+      "epoch": 40.72,
+      "learning_rate": 4.4909981343283585e-05,
+      "loss": 0.0002,
+      "step": 43652
+    },
+    {
+      "epoch": 40.72,
+      "learning_rate": 4.490951492537313e-05,
+      "loss": 0.0,
+      "step": 43656
+    },
+    {
+      "epoch": 40.73,
+      "learning_rate": 4.490904850746269e-05,
+      "loss": 0.0001,
+      "step": 43660
+    },
+    {
+      "epoch": 40.73,
+      "learning_rate": 4.490858208955224e-05,
+      "loss": 0.0001,
+      "step": 43664
+    },
+    {
+      "epoch": 40.73,
+      "learning_rate": 4.490811567164179e-05,
+      "loss": 0.0,
+      "step": 43668
+    },
+    {
+      "epoch": 40.74,
+      "learning_rate": 4.4907649253731346e-05,
+      "loss": 0.0,
+      "step": 43672
+    },
+    {
+      "epoch": 40.74,
+      "learning_rate": 4.49071828358209e-05,
+      "loss": 0.0042,
+      "step": 43676
+    },
+    {
+      "epoch": 40.75,
+      "learning_rate": 4.490671641791045e-05,
+      "loss": 0.0,
+      "step": 43680
+    },
+    {
+      "epoch": 40.75,
+      "learning_rate": 4.490625e-05,
+      "loss": 0.0001,
+      "step": 43684
+    },
+    {
+      "epoch": 40.75,
+      "learning_rate": 4.490578358208956e-05,
+      "loss": 0.0001,
+      "step": 43688
+    },
+    {
+      "epoch": 40.76,
+      "learning_rate": 4.490531716417911e-05,
+      "loss": 0.0002,
+      "step": 43692
+    },
+    {
+      "epoch": 40.76,
+      "learning_rate": 4.4904850746268655e-05,
+      "loss": 0.0001,
+      "step": 43696
+    },
+    {
+      "epoch": 40.76,
+      "learning_rate": 4.490438432835821e-05,
+      "loss": 0.0003,
+      "step": 43700
+    },
+    {
+      "epoch": 40.77,
+      "learning_rate": 4.4903917910447765e-05,
+      "loss": 0.0,
+      "step": 43704
+    },
+    {
+      "epoch": 40.77,
+      "learning_rate": 4.490345149253731e-05,
+      "loss": 0.0001,
+      "step": 43708
+    },
+    {
+      "epoch": 40.78,
+      "learning_rate": 4.490298507462687e-05,
+      "loss": 0.0,
+      "step": 43712
+    },
+    {
+      "epoch": 40.78,
+      "learning_rate": 4.4902518656716416e-05,
+      "loss": 0.0007,
+      "step": 43716
+    },
+    {
+      "epoch": 40.78,
+      "learning_rate": 4.490205223880598e-05,
+      "loss": 0.0001,
+      "step": 43720
+    },
+    {
+      "epoch": 40.79,
+      "learning_rate": 4.4901585820895526e-05,
+      "loss": 0.0003,
+      "step": 43724
+    },
+    {
+      "epoch": 40.79,
+      "learning_rate": 4.4901119402985074e-05,
+      "loss": 0.0,
+      "step": 43728
+    },
+    {
+      "epoch": 40.79,
+      "learning_rate": 4.490065298507463e-05,
+      "loss": 0.0001,
+      "step": 43732
+    },
+    {
+      "epoch": 40.8,
+      "learning_rate": 4.4900186567164184e-05,
+      "loss": 0.0014,
+      "step": 43736
+    },
+    {
+      "epoch": 40.8,
+      "learning_rate": 4.489972014925373e-05,
+      "loss": 0.0,
+      "step": 43740
+    },
+    {
+      "epoch": 40.81,
+      "learning_rate": 4.489925373134329e-05,
+      "loss": 0.0016,
+      "step": 43744
+    },
+    {
+      "epoch": 40.81,
+      "learning_rate": 4.489878731343284e-05,
+      "loss": 0.0,
+      "step": 43748
+    },
+    {
+      "epoch": 40.81,
+      "learning_rate": 4.489832089552239e-05,
+      "loss": 0.0,
+      "step": 43752
+    },
+    {
+      "epoch": 40.82,
+      "learning_rate": 4.4897854477611945e-05,
+      "loss": 0.0,
+      "step": 43756
+    },
+    {
+      "epoch": 40.82,
+      "learning_rate": 4.489738805970149e-05,
+      "loss": 0.0,
+      "step": 43760
+    },
+    {
+      "epoch": 40.82,
+      "learning_rate": 4.489692164179105e-05,
+      "loss": 0.0001,
+      "step": 43764
+    },
+    {
+      "epoch": 40.83,
+      "learning_rate": 4.48964552238806e-05,
+      "loss": 0.0001,
+      "step": 43768
+    },
+    {
+      "epoch": 40.83,
+      "learning_rate": 4.489598880597015e-05,
+      "loss": 0.0,
+      "step": 43772
+    },
+    {
+      "epoch": 40.84,
+      "learning_rate": 4.48955223880597e-05,
+      "loss": 0.0,
+      "step": 43776
+    },
+    {
+      "epoch": 40.84,
+      "learning_rate": 4.489505597014926e-05,
+      "loss": 0.0,
+      "step": 43780
+    },
+    {
+      "epoch": 40.84,
+      "learning_rate": 4.489458955223881e-05,
+      "loss": 0.0,
+      "step": 43784
+    },
+    {
+      "epoch": 40.85,
+      "learning_rate": 4.489412313432836e-05,
+      "loss": 0.0,
+      "step": 43788
+    },
+    {
+      "epoch": 40.85,
+      "learning_rate": 4.489365671641791e-05,
+      "loss": 0.0009,
+      "step": 43792
+    },
+    {
+      "epoch": 40.85,
+      "learning_rate": 4.489319029850747e-05,
+      "loss": 0.0001,
+      "step": 43796
+    },
+    {
+      "epoch": 40.86,
+      "learning_rate": 4.4892723880597015e-05,
+      "loss": 0.0002,
+      "step": 43800
+    },
+    {
+      "epoch": 40.86,
+      "learning_rate": 4.489225746268657e-05,
+      "loss": 0.0,
+      "step": 43804
+    },
+    {
+      "epoch": 40.87,
+      "learning_rate": 4.4891791044776125e-05,
+      "loss": 0.0022,
+      "step": 43808
+    },
+    {
+      "epoch": 40.87,
+      "learning_rate": 4.489132462686567e-05,
+      "loss": 0.0,
+      "step": 43812
+    },
+    {
+      "epoch": 40.87,
+      "learning_rate": 4.489085820895523e-05,
+      "loss": 0.0,
+      "step": 43816
+    },
+    {
+      "epoch": 40.88,
+      "learning_rate": 4.4890391791044776e-05,
+      "loss": 0.0017,
+      "step": 43820
+    },
+    {
+      "epoch": 40.88,
+      "learning_rate": 4.488992537313433e-05,
+      "loss": 0.0,
+      "step": 43824
+    },
+    {
+      "epoch": 40.88,
+      "learning_rate": 4.4889458955223886e-05,
+      "loss": 0.0002,
+      "step": 43828
+    },
+    {
+      "epoch": 40.89,
+      "learning_rate": 4.4888992537313434e-05,
+      "loss": 0.0004,
+      "step": 43832
+    },
+    {
+      "epoch": 40.89,
+      "learning_rate": 4.488852611940298e-05,
+      "loss": 0.0003,
+      "step": 43836
+    },
+    {
+      "epoch": 40.9,
+      "learning_rate": 4.4888059701492544e-05,
+      "loss": 0.0,
+      "step": 43840
+    },
+    {
+      "epoch": 40.9,
+      "learning_rate": 4.488759328358209e-05,
+      "loss": 0.0001,
+      "step": 43844
+    },
+    {
+      "epoch": 40.9,
+      "learning_rate": 4.488712686567164e-05,
+      "loss": 0.0009,
+      "step": 43848
+    },
+    {
+      "epoch": 40.91,
+      "learning_rate": 4.4886660447761195e-05,
+      "loss": 0.0002,
+      "step": 43852
+    },
+    {
+      "epoch": 40.91,
+      "learning_rate": 4.488619402985075e-05,
+      "loss": 0.0005,
+      "step": 43856
+    },
+    {
+      "epoch": 40.91,
+      "learning_rate": 4.48857276119403e-05,
+      "loss": 0.0003,
+      "step": 43860
+    },
+    {
+      "epoch": 40.92,
+      "learning_rate": 4.488526119402985e-05,
+      "loss": 0.0,
+      "step": 43864
+    },
+    {
+      "epoch": 40.92,
+      "learning_rate": 4.48847947761194e-05,
+      "loss": 0.003,
+      "step": 43868
+    },
+    {
+      "epoch": 40.93,
+      "learning_rate": 4.4884328358208956e-05,
+      "loss": 0.0,
+      "step": 43872
+    },
+    {
+      "epoch": 40.93,
+      "learning_rate": 4.488386194029851e-05,
+      "loss": 0.0001,
+      "step": 43876
+    },
+    {
+      "epoch": 40.93,
+      "learning_rate": 4.488339552238806e-05,
+      "loss": 0.0018,
+      "step": 43880
+    },
+    {
+      "epoch": 40.94,
+      "learning_rate": 4.4882929104477614e-05,
+      "loss": 0.0003,
+      "step": 43884
+    },
+    {
+      "epoch": 40.94,
+      "learning_rate": 4.488246268656717e-05,
+      "loss": 0.0004,
+      "step": 43888
+    },
+    {
+      "epoch": 40.94,
+      "learning_rate": 4.488199626865672e-05,
+      "loss": 0.0,
+      "step": 43892
+    },
+    {
+      "epoch": 40.95,
+      "learning_rate": 4.488152985074627e-05,
+      "loss": 0.0,
+      "step": 43896
+    },
+    {
+      "epoch": 40.95,
+      "learning_rate": 4.488106343283583e-05,
+      "loss": 0.0001,
+      "step": 43900
+    },
+    {
+      "epoch": 40.96,
+      "learning_rate": 4.4880597014925375e-05,
+      "loss": 0.0001,
+      "step": 43904
+    },
+    {
+      "epoch": 40.96,
+      "learning_rate": 4.488013059701493e-05,
+      "loss": 0.0,
+      "step": 43908
+    },
+    {
+      "epoch": 40.96,
+      "learning_rate": 4.487966417910448e-05,
+      "loss": 0.0,
+      "step": 43912
+    },
+    {
+      "epoch": 40.97,
+      "learning_rate": 4.487919776119403e-05,
+      "loss": 0.0001,
+      "step": 43916
+    },
+    {
+      "epoch": 40.97,
+      "learning_rate": 4.487873134328359e-05,
+      "loss": 0.0001,
+      "step": 43920
+    },
+    {
+      "epoch": 40.97,
+      "learning_rate": 4.4878264925373136e-05,
+      "loss": 0.0001,
+      "step": 43924
+    },
+    {
+      "epoch": 40.98,
+      "learning_rate": 4.4877798507462684e-05,
+      "loss": 0.0003,
+      "step": 43928
+    },
+    {
+      "epoch": 40.98,
+      "learning_rate": 4.4877332089552246e-05,
+      "loss": 0.0009,
+      "step": 43932
+    },
+    {
+      "epoch": 40.98,
+      "learning_rate": 4.4876865671641794e-05,
+      "loss": 0.0008,
+      "step": 43936
+    },
+    {
+      "epoch": 40.99,
+      "learning_rate": 4.487639925373134e-05,
+      "loss": 0.0021,
+      "step": 43940
+    },
+    {
+      "epoch": 40.99,
+      "learning_rate": 4.48759328358209e-05,
+      "loss": 0.0,
+      "step": 43944
+    },
+    {
+      "epoch": 41.0,
+      "learning_rate": 4.487546641791045e-05,
+      "loss": 0.0003,
+      "step": 43948
+    },
+    {
+      "epoch": 41.0,
+      "learning_rate": 4.4875e-05,
+      "loss": 0.0001,
+      "step": 43952
+    },
+    {
+      "epoch": 41.0,
+      "learning_rate": 4.4874533582089555e-05,
+      "loss": 0.0001,
+      "step": 43956
+    },
+    {
+      "epoch": 41.01,
+      "learning_rate": 4.487406716417911e-05,
+      "loss": 0.0003,
+      "step": 43960
+    },
+    {
+      "epoch": 41.01,
+      "learning_rate": 4.487360074626866e-05,
+      "loss": 0.0,
+      "step": 43964
+    },
+    {
+      "epoch": 41.01,
+      "learning_rate": 4.487313432835821e-05,
+      "loss": 0.0003,
+      "step": 43968
+    },
+    {
+      "epoch": 41.02,
+      "learning_rate": 4.487266791044776e-05,
+      "loss": 0.0003,
+      "step": 43972
+    },
+    {
+      "epoch": 41.02,
+      "learning_rate": 4.4872201492537316e-05,
+      "loss": 0.0,
+      "step": 43976
+    },
+    {
+      "epoch": 41.03,
+      "learning_rate": 4.487173507462687e-05,
+      "loss": 0.0017,
+      "step": 43980
+    },
+    {
+      "epoch": 41.03,
+      "learning_rate": 4.487126865671642e-05,
+      "loss": 0.0002,
+      "step": 43984
+    },
+    {
+      "epoch": 41.03,
+      "learning_rate": 4.487080223880597e-05,
+      "loss": 0.0002,
+      "step": 43988
+    },
+    {
+      "epoch": 41.04,
+      "learning_rate": 4.487033582089553e-05,
+      "loss": 0.0,
+      "step": 43992
+    },
+    {
+      "epoch": 41.04,
+      "learning_rate": 4.486986940298508e-05,
+      "loss": 0.0002,
+      "step": 43996
+    },
+    {
+      "epoch": 41.04,
+      "learning_rate": 4.4869402985074625e-05,
+      "loss": 0.0001,
+      "step": 44000
+    },
+    {
+      "epoch": 41.04,
+      "eval_exact_match": 0.7379110251450677,
+      "eval_exec": 0.7543520309477756,
+      "eval_loss": 0.43217331171035767,
+      "eval_runtime": 1129.1199,
+      "eval_samples_per_second": 0.916,
+      "step": 44000
+    },
+    {
+      "epoch": 41.05,
+      "learning_rate": 4.486893656716418e-05,
+      "loss": 0.0008,
+      "step": 44004
+    },
+    {
+      "epoch": 41.05,
+      "learning_rate": 4.4868470149253735e-05,
+      "loss": 0.0,
+      "step": 44008
+    },
+    {
+      "epoch": 41.06,
+      "learning_rate": 4.486800373134328e-05,
+      "loss": 0.0001,
+      "step": 44012
+    },
+    {
+      "epoch": 41.06,
+      "learning_rate": 4.486753731343284e-05,
+      "loss": 0.0,
+      "step": 44016
+    },
+    {
+      "epoch": 41.06,
+      "learning_rate": 4.486707089552239e-05,
+      "loss": 0.0,
+      "step": 44020
+    },
+    {
+      "epoch": 41.07,
+      "learning_rate": 4.486660447761194e-05,
+      "loss": 0.0,
+      "step": 44024
+    },
+    {
+      "epoch": 41.07,
+      "learning_rate": 4.4866138059701496e-05,
+      "loss": 0.0001,
+      "step": 44028
+    },
+    {
+      "epoch": 41.07,
+      "learning_rate": 4.4865671641791044e-05,
+      "loss": 0.0,
+      "step": 44032
+    },
+    {
+      "epoch": 41.08,
+      "learning_rate": 4.48652052238806e-05,
+      "loss": 0.0,
+      "step": 44036
+    },
+    {
+      "epoch": 41.08,
+      "learning_rate": 4.4864738805970154e-05,
+      "loss": 0.0,
+      "step": 44040
+    },
+    {
+      "epoch": 41.09,
+      "learning_rate": 4.48642723880597e-05,
+      "loss": 0.0001,
+      "step": 44044
+    },
+    {
+      "epoch": 41.09,
+      "learning_rate": 4.4863805970149257e-05,
+      "loss": 0.0,
+      "step": 44048
+    },
+    {
+      "epoch": 41.09,
+      "learning_rate": 4.486333955223881e-05,
+      "loss": 0.0011,
+      "step": 44052
+    },
+    {
+      "epoch": 41.1,
+      "learning_rate": 4.486287313432836e-05,
+      "loss": 0.0001,
+      "step": 44056
+    },
+    {
+      "epoch": 41.1,
+      "learning_rate": 4.4862406716417915e-05,
+      "loss": 0.0003,
+      "step": 44060
+    },
+    {
+      "epoch": 41.1,
+      "learning_rate": 4.486194029850746e-05,
+      "loss": 0.0003,
+      "step": 44064
+    },
+    {
+      "epoch": 41.11,
+      "learning_rate": 4.486147388059702e-05,
+      "loss": 0.0001,
+      "step": 44068
+    },
+    {
+      "epoch": 41.11,
+      "learning_rate": 4.486100746268657e-05,
+      "loss": 0.0,
+      "step": 44072
+    },
+    {
+      "epoch": 41.12,
+      "learning_rate": 4.486054104477612e-05,
+      "loss": 0.0,
+      "step": 44076
+    },
+    {
+      "epoch": 41.12,
+      "learning_rate": 4.4860074626865676e-05,
+      "loss": 0.0,
+      "step": 44080
+    },
+    {
+      "epoch": 41.12,
+      "learning_rate": 4.485960820895523e-05,
+      "loss": 0.0,
+      "step": 44084
+    },
+    {
+      "epoch": 41.13,
+      "learning_rate": 4.485914179104478e-05,
+      "loss": 0.0001,
+      "step": 44088
+    },
+    {
+      "epoch": 41.13,
+      "learning_rate": 4.485867537313433e-05,
+      "loss": 0.0,
+      "step": 44092
+    },
+    {
+      "epoch": 41.13,
+      "learning_rate": 4.485820895522388e-05,
+      "loss": 0.0001,
+      "step": 44096
+    },
+    {
+      "epoch": 41.14,
+      "learning_rate": 4.4857742537313437e-05,
+      "loss": 0.0,
+      "step": 44100
+    },
+    {
+      "epoch": 41.14,
+      "learning_rate": 4.4857276119402985e-05,
+      "loss": 0.0001,
+      "step": 44104
+    },
+    {
+      "epoch": 41.15,
+      "learning_rate": 4.485680970149254e-05,
+      "loss": 0.0,
+      "step": 44108
+    },
+    {
+      "epoch": 41.15,
+      "learning_rate": 4.4856343283582094e-05,
+      "loss": 0.0001,
+      "step": 44112
+    },
+    {
+      "epoch": 41.15,
+      "learning_rate": 4.485587686567164e-05,
+      "loss": 0.0,
+      "step": 44116
+    },
+    {
+      "epoch": 41.16,
+      "learning_rate": 4.48554104477612e-05,
+      "loss": 0.0,
+      "step": 44120
+    },
+    {
+      "epoch": 41.16,
+      "learning_rate": 4.4854944029850746e-05,
+      "loss": 0.0,
+      "step": 44124
+    },
+    {
+      "epoch": 41.16,
+      "learning_rate": 4.48544776119403e-05,
+      "loss": 0.0002,
+      "step": 44128
+    },
+    {
+      "epoch": 41.17,
+      "learning_rate": 4.4854011194029855e-05,
+      "loss": 0.0001,
+      "step": 44132
+    },
+    {
+      "epoch": 41.17,
+      "learning_rate": 4.4853544776119404e-05,
+      "loss": 0.0,
+      "step": 44136
+    },
+    {
+      "epoch": 41.18,
+      "learning_rate": 4.485307835820896e-05,
+      "loss": 0.0009,
+      "step": 44140
+    },
+    {
+      "epoch": 41.18,
+      "learning_rate": 4.4852611940298513e-05,
+      "loss": 0.0006,
+      "step": 44144
+    },
+    {
+      "epoch": 41.18,
+      "learning_rate": 4.485214552238806e-05,
+      "loss": 0.0,
+      "step": 44148
+    },
+    {
+      "epoch": 41.19,
+      "learning_rate": 4.485167910447761e-05,
+      "loss": 0.0001,
+      "step": 44152
+    },
+    {
+      "epoch": 41.19,
+      "learning_rate": 4.4851212686567165e-05,
+      "loss": 0.0001,
+      "step": 44156
+    },
+    {
+      "epoch": 41.19,
+      "learning_rate": 4.485074626865672e-05,
+      "loss": 0.0,
+      "step": 44160
+    },
+    {
+      "epoch": 41.2,
+      "learning_rate": 4.485027985074627e-05,
+      "loss": 0.0,
+      "step": 44164
+    },
+    {
+      "epoch": 41.2,
+      "learning_rate": 4.484981343283582e-05,
+      "loss": 0.0,
+      "step": 44168
+    },
+    {
+      "epoch": 41.21,
+      "learning_rate": 4.484934701492538e-05,
+      "loss": 0.0,
+      "step": 44172
+    },
+    {
+      "epoch": 41.21,
+      "learning_rate": 4.4848880597014926e-05,
+      "loss": 0.0,
+      "step": 44176
+    },
+    {
+      "epoch": 41.21,
+      "learning_rate": 4.484841417910448e-05,
+      "loss": 0.0,
+      "step": 44180
+    },
+    {
+      "epoch": 41.22,
+      "learning_rate": 4.484794776119403e-05,
+      "loss": 0.0023,
+      "step": 44184
+    },
+    {
+      "epoch": 41.22,
+      "learning_rate": 4.4847481343283583e-05,
+      "loss": 0.0003,
+      "step": 44188
+    },
+    {
+      "epoch": 41.22,
+      "learning_rate": 4.484701492537314e-05,
+      "loss": 0.0,
+      "step": 44192
+    },
+    {
+      "epoch": 41.23,
+      "learning_rate": 4.4846548507462687e-05,
+      "loss": 0.0001,
+      "step": 44196
+    },
+    {
+      "epoch": 41.23,
+      "learning_rate": 4.484608208955224e-05,
+      "loss": 0.0001,
+      "step": 44200
+    },
+    {
+      "epoch": 41.24,
+      "learning_rate": 4.4845615671641796e-05,
+      "loss": 0.0002,
+      "step": 44204
+    },
+    {
+      "epoch": 41.24,
+      "learning_rate": 4.4845149253731344e-05,
+      "loss": 0.0,
+      "step": 44208
+    },
+    {
+      "epoch": 41.24,
+      "learning_rate": 4.48446828358209e-05,
+      "loss": 0.0,
+      "step": 44212
+    },
+    {
+      "epoch": 41.25,
+      "learning_rate": 4.484421641791045e-05,
+      "loss": 0.0,
+      "step": 44216
+    },
+    {
+      "epoch": 41.25,
+      "learning_rate": 4.484375e-05,
+      "loss": 0.0,
+      "step": 44220
+    },
+    {
+      "epoch": 41.25,
+      "learning_rate": 4.484328358208956e-05,
+      "loss": 0.0004,
+      "step": 44224
+    },
+    {
+      "epoch": 41.26,
+      "learning_rate": 4.4842817164179105e-05,
+      "loss": 0.0,
+      "step": 44228
+    },
+    {
+      "epoch": 41.26,
+      "learning_rate": 4.484235074626866e-05,
+      "loss": 0.0015,
+      "step": 44232
+    },
+    {
+      "epoch": 41.26,
+      "learning_rate": 4.4841884328358215e-05,
+      "loss": 0.0,
+      "step": 44236
+    },
+    {
+      "epoch": 41.27,
+      "learning_rate": 4.4841417910447763e-05,
+      "loss": 0.0005,
+      "step": 44240
+    },
+    {
+      "epoch": 41.27,
+      "learning_rate": 4.484095149253731e-05,
+      "loss": 0.0,
+      "step": 44244
+    },
+    {
+      "epoch": 41.28,
+      "learning_rate": 4.484048507462687e-05,
+      "loss": 0.0,
+      "step": 44248
+    },
+    {
+      "epoch": 41.28,
+      "learning_rate": 4.484001865671642e-05,
+      "loss": 0.0001,
+      "step": 44252
+    },
+    {
+      "epoch": 41.28,
+      "learning_rate": 4.483955223880597e-05,
+      "loss": 0.0,
+      "step": 44256
+    },
+    {
+      "epoch": 41.29,
+      "learning_rate": 4.4839085820895524e-05,
+      "loss": 0.0001,
+      "step": 44260
+    },
+    {
+      "epoch": 41.29,
+      "learning_rate": 4.483861940298508e-05,
+      "loss": 0.0003,
+      "step": 44264
+    },
+    {
+      "epoch": 41.29,
+      "learning_rate": 4.483815298507463e-05,
+      "loss": 0.0,
+      "step": 44268
+    },
+    {
+      "epoch": 41.3,
+      "learning_rate": 4.483768656716418e-05,
+      "loss": 0.0,
+      "step": 44272
+    },
+    {
+      "epoch": 41.3,
+      "learning_rate": 4.483722014925373e-05,
+      "loss": 0.0001,
+      "step": 44276
+    },
+    {
+      "epoch": 41.31,
+      "learning_rate": 4.4836753731343285e-05,
+      "loss": 0.0006,
+      "step": 44280
+    },
+    {
+      "epoch": 41.31,
+      "learning_rate": 4.483628731343284e-05,
+      "loss": 0.0,
+      "step": 44284
+    },
+    {
+      "epoch": 41.31,
+      "learning_rate": 4.483582089552239e-05,
+      "loss": 0.0,
+      "step": 44288
+    },
+    {
+      "epoch": 41.32,
+      "learning_rate": 4.483535447761194e-05,
+      "loss": 0.0001,
+      "step": 44292
+    },
+    {
+      "epoch": 41.32,
+      "learning_rate": 4.48348880597015e-05,
+      "loss": 0.0001,
+      "step": 44296
+    },
+    {
+      "epoch": 41.32,
+      "learning_rate": 4.4834421641791046e-05,
+      "loss": 0.0,
+      "step": 44300
+    },
+    {
+      "epoch": 41.33,
+      "learning_rate": 4.4833955223880594e-05,
+      "loss": 0.002,
+      "step": 44304
+    },
+    {
+      "epoch": 41.33,
+      "learning_rate": 4.4833488805970156e-05,
+      "loss": 0.0003,
+      "step": 44308
+    },
+    {
+      "epoch": 41.34,
+      "learning_rate": 4.4833022388059704e-05,
+      "loss": 0.0,
+      "step": 44312
+    },
+    {
+      "epoch": 41.34,
+      "learning_rate": 4.483255597014925e-05,
+      "loss": 0.0,
+      "step": 44316
+    },
+    {
+      "epoch": 41.34,
+      "learning_rate": 4.483208955223881e-05,
+      "loss": 0.0039,
+      "step": 44320
+    },
+    {
+      "epoch": 41.35,
+      "learning_rate": 4.483162313432836e-05,
+      "loss": 0.0029,
+      "step": 44324
+    },
+    {
+      "epoch": 41.35,
+      "learning_rate": 4.483115671641791e-05,
+      "loss": 0.0016,
+      "step": 44328
+    },
+    {
+      "epoch": 41.35,
+      "learning_rate": 4.4830690298507465e-05,
+      "loss": 0.0003,
+      "step": 44332
+    },
+    {
+      "epoch": 41.36,
+      "learning_rate": 4.4830223880597013e-05,
+      "loss": 0.0001,
+      "step": 44336
+    },
+    {
+      "epoch": 41.36,
+      "learning_rate": 4.482975746268657e-05,
+      "loss": 0.0,
+      "step": 44340
+    },
+    {
+      "epoch": 41.37,
+      "learning_rate": 4.482929104477612e-05,
+      "loss": 0.0001,
+      "step": 44344
+    },
+    {
+      "epoch": 41.37,
+      "learning_rate": 4.482882462686567e-05,
+      "loss": 0.0034,
+      "step": 44348
+    },
+    {
+      "epoch": 41.37,
+      "learning_rate": 4.4828358208955226e-05,
+      "loss": 0.0032,
+      "step": 44352
+    },
+    {
+      "epoch": 41.38,
+      "learning_rate": 4.482789179104478e-05,
+      "loss": 0.0004,
+      "step": 44356
+    },
+    {
+      "epoch": 41.38,
+      "learning_rate": 4.482742537313433e-05,
+      "loss": 0.0,
+      "step": 44360
+    },
+    {
+      "epoch": 41.38,
+      "learning_rate": 4.482695895522388e-05,
+      "loss": 0.0013,
+      "step": 44364
+    },
+    {
+      "epoch": 41.39,
+      "learning_rate": 4.482649253731344e-05,
+      "loss": 0.0055,
+      "step": 44368
+    },
+    {
+      "epoch": 41.39,
+      "learning_rate": 4.482602611940299e-05,
+      "loss": 0.0,
+      "step": 44372
+    },
+    {
+      "epoch": 41.4,
+      "learning_rate": 4.482555970149254e-05,
+      "loss": 0.0,
+      "step": 44376
+    },
+    {
+      "epoch": 41.4,
+      "learning_rate": 4.482509328358209e-05,
+      "loss": 0.0028,
+      "step": 44380
+    },
+    {
+      "epoch": 41.4,
+      "learning_rate": 4.4824626865671645e-05,
+      "loss": 0.0,
+      "step": 44384
+    },
+    {
+      "epoch": 41.41,
+      "learning_rate": 4.48241604477612e-05,
+      "loss": 0.0,
+      "step": 44388
+    },
+    {
+      "epoch": 41.41,
+      "learning_rate": 4.482369402985075e-05,
+      "loss": 0.0001,
+      "step": 44392
+    },
+    {
+      "epoch": 41.41,
+      "learning_rate": 4.4823227611940296e-05,
+      "loss": 0.0,
+      "step": 44396
+    },
+    {
+      "epoch": 41.42,
+      "learning_rate": 4.482276119402986e-05,
+      "loss": 0.0,
+      "step": 44400
+    },
+    {
+      "epoch": 41.42,
+      "learning_rate": 4.4822294776119406e-05,
+      "loss": 0.0,
+      "step": 44404
+    },
+    {
+      "epoch": 41.43,
+      "learning_rate": 4.4821828358208954e-05,
+      "loss": 0.0,
+      "step": 44408
+    },
+    {
+      "epoch": 41.43,
+      "learning_rate": 4.482136194029851e-05,
+      "loss": 0.0004,
+      "step": 44412
+    },
+    {
+      "epoch": 41.43,
+      "learning_rate": 4.4820895522388064e-05,
+      "loss": 0.0,
+      "step": 44416
+    },
+    {
+      "epoch": 41.44,
+      "learning_rate": 4.482042910447761e-05,
+      "loss": 0.0004,
+      "step": 44420
+    },
+    {
+      "epoch": 41.44,
+      "learning_rate": 4.481996268656717e-05,
+      "loss": 0.0,
+      "step": 44424
+    },
+    {
+      "epoch": 41.44,
+      "learning_rate": 4.481949626865672e-05,
+      "loss": 0.0,
+      "step": 44428
+    },
+    {
+      "epoch": 41.45,
+      "learning_rate": 4.481902985074627e-05,
+      "loss": 0.0029,
+      "step": 44432
+    },
+    {
+      "epoch": 41.45,
+      "learning_rate": 4.4818563432835825e-05,
+      "loss": 0.0,
+      "step": 44436
+    },
+    {
+      "epoch": 41.46,
+      "learning_rate": 4.481809701492537e-05,
+      "loss": 0.0001,
+      "step": 44440
+    },
+    {
+      "epoch": 41.46,
+      "learning_rate": 4.481763059701493e-05,
+      "loss": 0.0004,
+      "step": 44444
+    },
+    {
+      "epoch": 41.46,
+      "learning_rate": 4.481716417910448e-05,
+      "loss": 0.0,
+      "step": 44448
+    },
+    {
+      "epoch": 41.47,
+      "learning_rate": 4.481669776119403e-05,
+      "loss": 0.0,
+      "step": 44452
+    },
+    {
+      "epoch": 41.47,
+      "learning_rate": 4.481623134328358e-05,
+      "loss": 0.0001,
+      "step": 44456
+    },
+    {
+      "epoch": 41.47,
+      "learning_rate": 4.481576492537314e-05,
+      "loss": 0.0001,
+      "step": 44460
+    },
+    {
+      "epoch": 41.48,
+      "learning_rate": 4.481529850746269e-05,
+      "loss": 0.0,
+      "step": 44464
+    },
+    {
+      "epoch": 41.48,
+      "learning_rate": 4.481483208955224e-05,
+      "loss": 0.0,
+      "step": 44468
+    },
+    {
+      "epoch": 41.49,
+      "learning_rate": 4.481436567164179e-05,
+      "loss": 0.0009,
+      "step": 44472
+    },
+    {
+      "epoch": 41.49,
+      "learning_rate": 4.481389925373135e-05,
+      "loss": 0.0001,
+      "step": 44476
+    },
+    {
+      "epoch": 41.49,
+      "learning_rate": 4.4813432835820895e-05,
+      "loss": 0.0,
+      "step": 44480
+    },
+    {
+      "epoch": 41.5,
+      "learning_rate": 4.481296641791045e-05,
+      "loss": 0.0,
+      "step": 44484
+    },
+    {
+      "epoch": 41.5,
+      "learning_rate": 4.4812500000000005e-05,
+      "loss": 0.0,
+      "step": 44488
+    },
+    {
+      "epoch": 41.5,
+      "learning_rate": 4.481203358208955e-05,
+      "loss": 0.0001,
+      "step": 44492
+    },
+    {
+      "epoch": 41.51,
+      "learning_rate": 4.481156716417911e-05,
+      "loss": 0.0027,
+      "step": 44496
+    },
+    {
+      "epoch": 41.51,
+      "learning_rate": 4.4811100746268656e-05,
+      "loss": 0.0001,
+      "step": 44500
+    },
+    {
+      "epoch": 41.51,
+      "eval_exact_match": 0.7350096711798839,
+      "eval_exec": 0.7543520309477756,
+      "eval_loss": 0.42939507961273193,
+      "eval_runtime": 1157.4066,
+      "eval_samples_per_second": 0.893,
+      "step": 44500
+    },
+    {
+      "epoch": 41.51,
+      "learning_rate": 4.481063432835821e-05,
+      "loss": 0.0,
+      "step": 44504
+    },
+    {
+      "epoch": 41.52,
+      "learning_rate": 4.4810167910447766e-05,
+      "loss": 0.0,
+      "step": 44508
+    },
+    {
+      "epoch": 41.52,
+      "learning_rate": 4.4809701492537314e-05,
+      "loss": 0.0,
+      "step": 44512
+    },
+    {
+      "epoch": 41.53,
+      "learning_rate": 4.480923507462686e-05,
+      "loss": 0.0003,
+      "step": 44516
+    },
+    {
+      "epoch": 41.53,
+      "learning_rate": 4.4808768656716424e-05,
+      "loss": 0.0,
+      "step": 44520
+    },
+    {
+      "epoch": 41.53,
+      "learning_rate": 4.480830223880597e-05,
+      "loss": 0.0011,
+      "step": 44524
+    },
+    {
+      "epoch": 41.54,
+      "learning_rate": 4.480783582089552e-05,
+      "loss": 0.0003,
+      "step": 44528
+    },
+    {
+      "epoch": 41.54,
+      "learning_rate": 4.4807369402985075e-05,
+      "loss": 0.0,
+      "step": 44532
+    },
+    {
+      "epoch": 41.54,
+      "learning_rate": 4.480690298507463e-05,
+      "loss": 0.0002,
+      "step": 44536
+    },
+    {
+      "epoch": 41.55,
+      "learning_rate": 4.4806436567164185e-05,
+      "loss": 0.0012,
+      "step": 44540
+    },
+    {
+      "epoch": 41.55,
+      "learning_rate": 4.480597014925373e-05,
+      "loss": 0.0001,
+      "step": 44544
+    },
+    {
+      "epoch": 41.56,
+      "learning_rate": 4.480550373134328e-05,
+      "loss": 0.0,
+      "step": 44548
+    },
+    {
+      "epoch": 41.56,
+      "learning_rate": 4.480503731343284e-05,
+      "loss": 0.0,
+      "step": 44552
+    },
+    {
+      "epoch": 41.56,
+      "learning_rate": 4.480457089552239e-05,
+      "loss": 0.0,
+      "step": 44556
+    },
+    {
+      "epoch": 41.57,
+      "learning_rate": 4.480410447761194e-05,
+      "loss": 0.0,
+      "step": 44560
+    },
+    {
+      "epoch": 41.57,
+      "learning_rate": 4.4803638059701494e-05,
+      "loss": 0.0001,
+      "step": 44564
+    },
+    {
+      "epoch": 41.57,
+      "learning_rate": 4.480317164179105e-05,
+      "loss": 0.0,
+      "step": 44568
+    },
+    {
+      "epoch": 41.58,
+      "learning_rate": 4.48027052238806e-05,
+      "loss": 0.0,
+      "step": 44572
+    },
+    {
+      "epoch": 41.58,
+      "learning_rate": 4.480223880597015e-05,
+      "loss": 0.0001,
+      "step": 44576
+    },
+    {
+      "epoch": 41.59,
+      "learning_rate": 4.480177238805971e-05,
+      "loss": 0.0003,
+      "step": 44580
+    },
+    {
+      "epoch": 41.59,
+      "learning_rate": 4.4801305970149255e-05,
+      "loss": 0.0,
+      "step": 44584
+    },
+    {
+      "epoch": 41.59,
+      "learning_rate": 4.480083955223881e-05,
+      "loss": 0.0013,
+      "step": 44588
+    },
+    {
+      "epoch": 41.6,
+      "learning_rate": 4.480037313432836e-05,
+      "loss": 0.0021,
+      "step": 44592
+    },
+    {
+      "epoch": 41.6,
+      "learning_rate": 4.479990671641791e-05,
+      "loss": 0.0,
+      "step": 44596
+    },
+    {
+      "epoch": 41.6,
+      "learning_rate": 4.479944029850747e-05,
+      "loss": 0.0001,
+      "step": 44600
+    },
+    {
+      "epoch": 41.61,
+      "learning_rate": 4.4798973880597016e-05,
+      "loss": 0.0043,
+      "step": 44604
+    },
+    {
+      "epoch": 41.61,
+      "learning_rate": 4.4798507462686564e-05,
+      "loss": 0.0014,
+      "step": 44608
+    },
+    {
+      "epoch": 41.62,
+      "learning_rate": 4.4798041044776126e-05,
+      "loss": 0.0001,
+      "step": 44612
+    },
+    {
+      "epoch": 41.62,
+      "learning_rate": 4.4797574626865674e-05,
+      "loss": 0.0002,
+      "step": 44616
+    },
+    {
+      "epoch": 41.62,
+      "learning_rate": 4.479710820895522e-05,
+      "loss": 0.0001,
+      "step": 44620
+    },
+    {
+      "epoch": 41.63,
+      "learning_rate": 4.479664179104478e-05,
+      "loss": 0.0001,
+      "step": 44624
+    },
+    {
+      "epoch": 41.63,
+      "learning_rate": 4.479617537313433e-05,
+      "loss": 0.0007,
+      "step": 44628
+    },
+    {
+      "epoch": 41.63,
+      "learning_rate": 4.479570895522388e-05,
+      "loss": 0.0006,
+      "step": 44632
+    },
+    {
+      "epoch": 41.64,
+      "learning_rate": 4.4795242537313435e-05,
+      "loss": 0.0002,
+      "step": 44636
+    },
+    {
+      "epoch": 41.64,
+      "learning_rate": 4.479477611940299e-05,
+      "loss": 0.0001,
+      "step": 44640
+    },
+    {
+      "epoch": 41.65,
+      "learning_rate": 4.479430970149254e-05,
+      "loss": 0.0,
+      "step": 44644
+    },
+    {
+      "epoch": 41.65,
+      "learning_rate": 4.479384328358209e-05,
+      "loss": 0.0001,
+      "step": 44648
+    },
+    {
+      "epoch": 41.65,
+      "learning_rate": 4.479337686567164e-05,
+      "loss": 0.0,
+      "step": 44652
+    },
+    {
+      "epoch": 41.66,
+      "learning_rate": 4.4792910447761196e-05,
+      "loss": 0.0,
+      "step": 44656
+    },
+    {
+      "epoch": 41.66,
+      "learning_rate": 4.479244402985075e-05,
+      "loss": 0.0016,
+      "step": 44660
+    },
+    {
+      "epoch": 41.66,
+      "learning_rate": 4.47919776119403e-05,
+      "loss": 0.0019,
+      "step": 44664
+    },
+    {
+      "epoch": 41.67,
+      "learning_rate": 4.479151119402985e-05,
+      "loss": 0.0009,
+      "step": 44668
+    },
+    {
+      "epoch": 41.67,
+      "learning_rate": 4.479104477611941e-05,
+      "loss": 0.0006,
+      "step": 44672
+    },
+    {
+      "epoch": 41.68,
+      "learning_rate": 4.479057835820896e-05,
+      "loss": 0.0,
+      "step": 44676
+    },
+    {
+      "epoch": 41.68,
+      "learning_rate": 4.4790111940298505e-05,
+      "loss": 0.0,
+      "step": 44680
+    },
+    {
+      "epoch": 41.68,
+      "learning_rate": 4.478964552238806e-05,
+      "loss": 0.0001,
+      "step": 44684
+    },
+    {
+      "epoch": 41.69,
+      "learning_rate": 4.4789179104477615e-05,
+      "loss": 0.0008,
+      "step": 44688
+    },
+    {
+      "epoch": 41.69,
+      "learning_rate": 4.478871268656716e-05,
+      "loss": 0.0019,
+      "step": 44692
+    },
+    {
+      "epoch": 41.69,
+      "learning_rate": 4.478824626865672e-05,
+      "loss": 0.0001,
+      "step": 44696
+    },
+    {
+      "epoch": 41.7,
+      "learning_rate": 4.478777985074627e-05,
+      "loss": 0.0,
+      "step": 44700
+    },
+    {
+      "epoch": 41.7,
+      "learning_rate": 4.478731343283583e-05,
+      "loss": 0.0,
+      "step": 44704
+    },
+    {
+      "epoch": 41.71,
+      "learning_rate": 4.4786847014925376e-05,
+      "loss": 0.0001,
+      "step": 44708
+    },
+    {
+      "epoch": 41.71,
+      "learning_rate": 4.4786380597014924e-05,
+      "loss": 0.0004,
+      "step": 44712
+    },
+    {
+      "epoch": 41.71,
+      "learning_rate": 4.478591417910448e-05,
+      "loss": 0.0005,
+      "step": 44716
+    },
+    {
+      "epoch": 41.72,
+      "learning_rate": 4.4785447761194034e-05,
+      "loss": 0.0002,
+      "step": 44720
+    },
+    {
+      "epoch": 41.72,
+      "learning_rate": 4.478498134328358e-05,
+      "loss": 0.0002,
+      "step": 44724
+    },
+    {
+      "epoch": 41.72,
+      "learning_rate": 4.478451492537314e-05,
+      "loss": 0.0001,
+      "step": 44728
+    },
+    {
+      "epoch": 41.73,
+      "learning_rate": 4.478404850746269e-05,
+      "loss": 0.0,
+      "step": 44732
+    },
+    {
+      "epoch": 41.73,
+      "learning_rate": 4.478358208955224e-05,
+      "loss": 0.0019,
+      "step": 44736
+    },
+    {
+      "epoch": 41.73,
+      "learning_rate": 4.4783115671641795e-05,
+      "loss": 0.0001,
+      "step": 44740
+    },
+    {
+      "epoch": 41.74,
+      "learning_rate": 4.478264925373134e-05,
+      "loss": 0.0006,
+      "step": 44744
+    },
+    {
+      "epoch": 41.74,
+      "learning_rate": 4.47821828358209e-05,
+      "loss": 0.0004,
+      "step": 44748
+    },
+    {
+      "epoch": 41.75,
+      "learning_rate": 4.478171641791045e-05,
+      "loss": 0.0001,
+      "step": 44752
+    },
+    {
+      "epoch": 41.75,
+      "learning_rate": 4.478125e-05,
+      "loss": 0.0,
+      "step": 44756
+    },
+    {
+      "epoch": 41.75,
+      "learning_rate": 4.4780783582089556e-05,
+      "loss": 0.0001,
+      "step": 44760
+    },
+    {
+      "epoch": 41.76,
+      "learning_rate": 4.478031716417911e-05,
+      "loss": 0.0,
+      "step": 44764
+    },
+    {
+      "epoch": 41.76,
+      "learning_rate": 4.477985074626866e-05,
+      "loss": 0.0,
+      "step": 44768
+    },
+    {
+      "epoch": 41.76,
+      "learning_rate": 4.477938432835821e-05,
+      "loss": 0.0,
+      "step": 44772
+    },
+    {
+      "epoch": 41.77,
+      "learning_rate": 4.477891791044776e-05,
+      "loss": 0.002,
+      "step": 44776
+    },
+    {
+      "epoch": 41.77,
+      "learning_rate": 4.477845149253732e-05,
+      "loss": 0.0001,
+      "step": 44780
+    },
+    {
+      "epoch": 41.78,
+      "learning_rate": 4.4777985074626865e-05,
+      "loss": 0.0,
+      "step": 44784
+    },
+    {
+      "epoch": 41.78,
+      "learning_rate": 4.477751865671642e-05,
+      "loss": 0.0,
+      "step": 44788
+    },
+    {
+      "epoch": 41.78,
+      "learning_rate": 4.4777052238805975e-05,
+      "loss": 0.0,
+      "step": 44792
+    },
+    {
+      "epoch": 41.79,
+      "learning_rate": 4.477658582089552e-05,
+      "loss": 0.0029,
+      "step": 44796
+    },
+    {
+      "epoch": 41.79,
+      "learning_rate": 4.477611940298508e-05,
+      "loss": 0.0,
+      "step": 44800
+    },
+    {
+      "epoch": 41.79,
+      "learning_rate": 4.4775652985074626e-05,
+      "loss": 0.0002,
+      "step": 44804
+    },
+    {
+      "epoch": 41.8,
+      "learning_rate": 4.477518656716418e-05,
+      "loss": 0.0,
+      "step": 44808
+    },
+    {
+      "epoch": 41.8,
+      "learning_rate": 4.4774720149253736e-05,
+      "loss": 0.0001,
+      "step": 44812
+    },
+    {
+      "epoch": 41.81,
+      "learning_rate": 4.4774253731343284e-05,
+      "loss": 0.0,
+      "step": 44816
+    },
+    {
+      "epoch": 41.81,
+      "learning_rate": 4.477378731343284e-05,
+      "loss": 0.0,
+      "step": 44820
+    },
+    {
+      "epoch": 41.81,
+      "learning_rate": 4.4773320895522394e-05,
+      "loss": 0.0006,
+      "step": 44824
+    },
+    {
+      "epoch": 41.82,
+      "learning_rate": 4.477285447761194e-05,
+      "loss": 0.0006,
+      "step": 44828
+    },
+    {
+      "epoch": 41.82,
+      "learning_rate": 4.477238805970149e-05,
+      "loss": 0.0013,
+      "step": 44832
+    },
+    {
+      "epoch": 41.82,
+      "learning_rate": 4.4771921641791045e-05,
+      "loss": 0.0014,
+      "step": 44836
+    },
+    {
+      "epoch": 41.83,
+      "learning_rate": 4.47714552238806e-05,
+      "loss": 0.0001,
+      "step": 44840
+    },
+    {
+      "epoch": 41.83,
+      "learning_rate": 4.477098880597015e-05,
+      "loss": 0.0001,
+      "step": 44844
+    },
+    {
+      "epoch": 41.84,
+      "learning_rate": 4.47705223880597e-05,
+      "loss": 0.0,
+      "step": 44848
+    },
+    {
+      "epoch": 41.84,
+      "learning_rate": 4.477005597014926e-05,
+      "loss": 0.0001,
+      "step": 44852
+    },
+    {
+      "epoch": 41.84,
+      "learning_rate": 4.4769589552238806e-05,
+      "loss": 0.003,
+      "step": 44856
+    },
+    {
+      "epoch": 41.85,
+      "learning_rate": 4.476912313432836e-05,
+      "loss": 0.0001,
+      "step": 44860
+    },
+    {
+      "epoch": 41.85,
+      "learning_rate": 4.476865671641791e-05,
+      "loss": 0.0,
+      "step": 44864
+    },
+    {
+      "epoch": 41.85,
+      "learning_rate": 4.476819029850747e-05,
+      "loss": 0.0001,
+      "step": 44868
+    },
+    {
+      "epoch": 41.86,
+      "learning_rate": 4.476772388059702e-05,
+      "loss": 0.0054,
+      "step": 44872
+    },
+    {
+      "epoch": 41.86,
+      "learning_rate": 4.476725746268657e-05,
+      "loss": 0.0003,
+      "step": 44876
+    },
+    {
+      "epoch": 41.87,
+      "learning_rate": 4.476679104477612e-05,
+      "loss": 0.0001,
+      "step": 44880
+    },
+    {
+      "epoch": 41.87,
+      "learning_rate": 4.476632462686568e-05,
+      "loss": 0.0001,
+      "step": 44884
+    },
+    {
+      "epoch": 41.87,
+      "learning_rate": 4.4765858208955225e-05,
+      "loss": 0.0,
+      "step": 44888
+    },
+    {
+      "epoch": 41.88,
+      "learning_rate": 4.476539179104478e-05,
+      "loss": 0.0001,
+      "step": 44892
+    },
+    {
+      "epoch": 41.88,
+      "learning_rate": 4.476492537313433e-05,
+      "loss": 0.0,
+      "step": 44896
+    },
+    {
+      "epoch": 41.88,
+      "learning_rate": 4.476445895522388e-05,
+      "loss": 0.0,
+      "step": 44900
+    },
+    {
+      "epoch": 41.89,
+      "learning_rate": 4.476399253731344e-05,
+      "loss": 0.0001,
+      "step": 44904
+    },
+    {
+      "epoch": 41.89,
+      "learning_rate": 4.4763526119402986e-05,
+      "loss": 0.0,
+      "step": 44908
+    },
+    {
+      "epoch": 41.9,
+      "learning_rate": 4.476305970149254e-05,
+      "loss": 0.0026,
+      "step": 44912
+    },
+    {
+      "epoch": 41.9,
+      "learning_rate": 4.4762593283582096e-05,
+      "loss": 0.0,
+      "step": 44916
+    },
+    {
+      "epoch": 41.9,
+      "learning_rate": 4.4762126865671644e-05,
+      "loss": 0.0,
+      "step": 44920
+    },
+    {
+      "epoch": 41.91,
+      "learning_rate": 4.476166044776119e-05,
+      "loss": 0.0001,
+      "step": 44924
+    },
+    {
+      "epoch": 41.91,
+      "learning_rate": 4.4761194029850754e-05,
+      "loss": 0.0001,
+      "step": 44928
+    },
+    {
+      "epoch": 41.91,
+      "learning_rate": 4.47607276119403e-05,
+      "loss": 0.0,
+      "step": 44932
+    },
+    {
+      "epoch": 41.92,
+      "learning_rate": 4.476026119402985e-05,
+      "loss": 0.0,
+      "step": 44936
+    },
+    {
+      "epoch": 41.92,
+      "learning_rate": 4.4759794776119405e-05,
+      "loss": 0.0032,
+      "step": 44940
+    },
+    {
+      "epoch": 41.93,
+      "learning_rate": 4.475932835820896e-05,
+      "loss": 0.001,
+      "step": 44944
+    },
+    {
+      "epoch": 41.93,
+      "learning_rate": 4.475886194029851e-05,
+      "loss": 0.0,
+      "step": 44948
+    },
+    {
+      "epoch": 41.93,
+      "learning_rate": 4.475839552238806e-05,
+      "loss": 0.0,
+      "step": 44952
+    },
+    {
+      "epoch": 41.94,
+      "learning_rate": 4.475792910447761e-05,
+      "loss": 0.0,
+      "step": 44956
+    },
+    {
+      "epoch": 41.94,
+      "learning_rate": 4.4757462686567166e-05,
+      "loss": 0.0,
+      "step": 44960
+    },
+    {
+      "epoch": 41.94,
+      "learning_rate": 4.475699626865672e-05,
+      "loss": 0.0045,
+      "step": 44964
+    },
+    {
+      "epoch": 41.95,
+      "learning_rate": 4.475652985074627e-05,
+      "loss": 0.0001,
+      "step": 44968
+    },
+    {
+      "epoch": 41.95,
+      "learning_rate": 4.4756063432835824e-05,
+      "loss": 0.0,
+      "step": 44972
+    },
+    {
+      "epoch": 41.96,
+      "learning_rate": 4.475559701492538e-05,
+      "loss": 0.0,
+      "step": 44976
+    },
+    {
+      "epoch": 41.96,
+      "learning_rate": 4.475513059701493e-05,
+      "loss": 0.0002,
+      "step": 44980
+    },
+    {
+      "epoch": 41.96,
+      "learning_rate": 4.4754664179104475e-05,
+      "loss": 0.0,
+      "step": 44984
+    },
+    {
+      "epoch": 41.97,
+      "learning_rate": 4.4754197761194036e-05,
+      "loss": 0.0015,
+      "step": 44988
+    },
+    {
+      "epoch": 41.97,
+      "learning_rate": 4.4753731343283585e-05,
+      "loss": 0.0,
+      "step": 44992
+    },
+    {
+      "epoch": 41.97,
+      "learning_rate": 4.475326492537313e-05,
+      "loss": 0.0,
+      "step": 44996
+    },
+    {
+      "epoch": 41.98,
+      "learning_rate": 4.475279850746269e-05,
+      "loss": 0.0,
+      "step": 45000
+    },
+    {
+      "epoch": 41.98,
+      "eval_exact_match": 0.7340425531914894,
+      "eval_exec": 0.758220502901354,
+      "eval_loss": 0.4176974892616272,
+      "eval_runtime": 1170.0858,
+      "eval_samples_per_second": 0.884,
+      "step": 45000
+    },
+    {
+      "epoch": 41.98,
+      "learning_rate": 4.475233208955224e-05,
+      "loss": 0.0081,
+      "step": 45004
+    },
+    {
+      "epoch": 41.98,
+      "learning_rate": 4.475186567164179e-05,
+      "loss": 0.0001,
+      "step": 45008
+    },
+    {
+      "epoch": 41.99,
+      "learning_rate": 4.4751399253731346e-05,
+      "loss": 0.0001,
+      "step": 45012
+    },
+    {
+      "epoch": 41.99,
+      "learning_rate": 4.4750932835820894e-05,
+      "loss": 0.0002,
+      "step": 45016
+    },
+    {
+      "epoch": 42.0,
+      "learning_rate": 4.475046641791045e-05,
+      "loss": 0.0002,
+      "step": 45020
+    },
+    {
+      "epoch": 42.0,
+      "learning_rate": 4.4750000000000004e-05,
+      "loss": 0.0002,
+      "step": 45024
+    },
+    {
+      "epoch": 42.0,
+      "learning_rate": 4.474953358208955e-05,
+      "loss": 0.0011,
+      "step": 45028
+    },
+    {
+      "epoch": 42.01,
+      "learning_rate": 4.4749067164179107e-05,
+      "loss": 0.0001,
+      "step": 45032
+    },
+    {
+      "epoch": 42.01,
+      "learning_rate": 4.474860074626866e-05,
+      "loss": 0.0,
+      "step": 45036
+    },
+    {
+      "epoch": 42.01,
+      "learning_rate": 4.474813432835821e-05,
+      "loss": 0.0003,
+      "step": 45040
+    },
+    {
+      "epoch": 42.02,
+      "learning_rate": 4.4747667910447765e-05,
+      "loss": 0.0004,
+      "step": 45044
+    },
+    {
+      "epoch": 42.02,
+      "learning_rate": 4.474720149253732e-05,
+      "loss": 0.0,
+      "step": 45048
+    },
+    {
+      "epoch": 42.03,
+      "learning_rate": 4.474673507462687e-05,
+      "loss": 0.0,
+      "step": 45052
+    },
+    {
+      "epoch": 42.03,
+      "learning_rate": 4.474626865671642e-05,
+      "loss": 0.0,
+      "step": 45056
+    },
+    {
+      "epoch": 42.03,
+      "learning_rate": 4.474580223880597e-05,
+      "loss": 0.0,
+      "step": 45060
+    },
+    {
+      "epoch": 42.04,
+      "learning_rate": 4.4745335820895526e-05,
+      "loss": 0.0,
+      "step": 45064
+    },
+    {
+      "epoch": 42.04,
+      "learning_rate": 4.474486940298508e-05,
+      "loss": 0.0001,
+      "step": 45068
+    },
+    {
+      "epoch": 42.04,
+      "learning_rate": 4.474440298507463e-05,
+      "loss": 0.002,
+      "step": 45072
+    },
+    {
+      "epoch": 42.05,
+      "learning_rate": 4.474393656716418e-05,
+      "loss": 0.0003,
+      "step": 45076
+    },
+    {
+      "epoch": 42.05,
+      "learning_rate": 4.474347014925374e-05,
+      "loss": 0.0001,
+      "step": 45080
+    },
+    {
+      "epoch": 42.06,
+      "learning_rate": 4.4743003731343286e-05,
+      "loss": 0.0003,
+      "step": 45084
+    },
+    {
+      "epoch": 42.06,
+      "learning_rate": 4.4742537313432835e-05,
+      "loss": 0.0027,
+      "step": 45088
+    },
+    {
+      "epoch": 42.06,
+      "learning_rate": 4.474207089552239e-05,
+      "loss": 0.0005,
+      "step": 45092
+    },
+    {
+      "epoch": 42.07,
+      "learning_rate": 4.4741604477611944e-05,
+      "loss": 0.0,
+      "step": 45096
+    },
+    {
+      "epoch": 42.07,
+      "learning_rate": 4.474113805970149e-05,
+      "loss": 0.0014,
+      "step": 45100
+    },
+    {
+      "epoch": 42.07,
+      "learning_rate": 4.474067164179105e-05,
+      "loss": 0.0001,
+      "step": 45104
+    },
+    {
+      "epoch": 42.08,
+      "learning_rate": 4.47402052238806e-05,
+      "loss": 0.0013,
+      "step": 45108
+    },
+    {
+      "epoch": 42.08,
+      "learning_rate": 4.473973880597015e-05,
+      "loss": 0.0,
+      "step": 45112
+    },
+    {
+      "epoch": 42.09,
+      "learning_rate": 4.4739272388059705e-05,
+      "loss": 0.0,
+      "step": 45116
+    },
+    {
+      "epoch": 42.09,
+      "learning_rate": 4.4738805970149254e-05,
+      "loss": 0.0001,
+      "step": 45120
+    },
+    {
+      "epoch": 42.09,
+      "learning_rate": 4.473833955223881e-05,
+      "loss": 0.0002,
+      "step": 45124
+    },
+    {
+      "epoch": 42.1,
+      "learning_rate": 4.473787313432836e-05,
+      "loss": 0.0008,
+      "step": 45128
+    },
+    {
+      "epoch": 42.1,
+      "learning_rate": 4.473740671641791e-05,
+      "loss": 0.0001,
+      "step": 45132
+    },
+    {
+      "epoch": 42.1,
+      "learning_rate": 4.473694029850746e-05,
+      "loss": 0.0,
+      "step": 45136
+    },
+    {
+      "epoch": 42.11,
+      "learning_rate": 4.473647388059702e-05,
+      "loss": 0.0,
+      "step": 45140
+    },
+    {
+      "epoch": 42.11,
+      "learning_rate": 4.473600746268657e-05,
+      "loss": 0.0014,
+      "step": 45144
+    },
+    {
+      "epoch": 42.12,
+      "learning_rate": 4.473554104477612e-05,
+      "loss": 0.0001,
+      "step": 45148
+    },
+    {
+      "epoch": 42.12,
+      "learning_rate": 4.473507462686567e-05,
+      "loss": 0.0015,
+      "step": 45152
+    },
+    {
+      "epoch": 42.12,
+      "learning_rate": 4.473460820895523e-05,
+      "loss": 0.0,
+      "step": 45156
+    },
+    {
+      "epoch": 42.13,
+      "learning_rate": 4.4734141791044776e-05,
+      "loss": 0.0055,
+      "step": 45160
+    },
+    {
+      "epoch": 42.13,
+      "learning_rate": 4.473367537313433e-05,
+      "loss": 0.0001,
+      "step": 45164
+    },
+    {
+      "epoch": 42.13,
+      "learning_rate": 4.4733208955223885e-05,
+      "loss": 0.0002,
+      "step": 45168
+    },
+    {
+      "epoch": 42.14,
+      "learning_rate": 4.4732742537313433e-05,
+      "loss": 0.0001,
+      "step": 45172
+    },
+    {
+      "epoch": 42.14,
+      "learning_rate": 4.473227611940299e-05,
+      "loss": 0.0001,
+      "step": 45176
+    },
+    {
+      "epoch": 42.15,
+      "learning_rate": 4.4731809701492537e-05,
+      "loss": 0.0,
+      "step": 45180
+    },
+    {
+      "epoch": 42.15,
+      "learning_rate": 4.473134328358209e-05,
+      "loss": 0.0,
+      "step": 45184
+    },
+    {
+      "epoch": 42.15,
+      "learning_rate": 4.4730876865671646e-05,
+      "loss": 0.0,
+      "step": 45188
+    },
+    {
+      "epoch": 42.16,
+      "learning_rate": 4.4730410447761194e-05,
+      "loss": 0.0003,
+      "step": 45192
+    },
+    {
+      "epoch": 42.16,
+      "learning_rate": 4.472994402985075e-05,
+      "loss": 0.0033,
+      "step": 45196
+    },
+    {
+      "epoch": 42.16,
+      "learning_rate": 4.4729477611940304e-05,
+      "loss": 0.0011,
+      "step": 45200
+    },
+    {
+      "epoch": 42.17,
+      "learning_rate": 4.472901119402985e-05,
+      "loss": 0.0,
+      "step": 45204
+    },
+    {
+      "epoch": 42.17,
+      "learning_rate": 4.472854477611941e-05,
+      "loss": 0.0017,
+      "step": 45208
+    },
+    {
+      "epoch": 42.18,
+      "learning_rate": 4.4728078358208955e-05,
+      "loss": 0.0001,
+      "step": 45212
+    },
+    {
+      "epoch": 42.18,
+      "learning_rate": 4.472761194029851e-05,
+      "loss": 0.0001,
+      "step": 45216
+    },
+    {
+      "epoch": 42.18,
+      "learning_rate": 4.4727145522388065e-05,
+      "loss": 0.0,
+      "step": 45220
+    },
+    {
+      "epoch": 42.19,
+      "learning_rate": 4.4726679104477613e-05,
+      "loss": 0.0,
+      "step": 45224
+    },
+    {
+      "epoch": 42.19,
+      "learning_rate": 4.472621268656716e-05,
+      "loss": 0.0,
+      "step": 45228
+    },
+    {
+      "epoch": 42.19,
+      "learning_rate": 4.472574626865672e-05,
+      "loss": 0.0005,
+      "step": 45232
+    },
+    {
+      "epoch": 42.2,
+      "learning_rate": 4.472527985074627e-05,
+      "loss": 0.0001,
+      "step": 45236
+    },
+    {
+      "epoch": 42.2,
+      "learning_rate": 4.472481343283582e-05,
+      "loss": 0.0001,
+      "step": 45240
+    },
+    {
+      "epoch": 42.21,
+      "learning_rate": 4.4724347014925374e-05,
+      "loss": 0.001,
+      "step": 45244
+    },
+    {
+      "epoch": 42.21,
+      "learning_rate": 4.472388059701493e-05,
+      "loss": 0.0001,
+      "step": 45248
+    },
+    {
+      "epoch": 42.21,
+      "learning_rate": 4.472341417910448e-05,
+      "loss": 0.0,
+      "step": 45252
+    },
+    {
+      "epoch": 42.22,
+      "learning_rate": 4.472294776119403e-05,
+      "loss": 0.0018,
+      "step": 45256
+    },
+    {
+      "epoch": 42.22,
+      "learning_rate": 4.472248134328359e-05,
+      "loss": 0.0006,
+      "step": 45260
+    },
+    {
+      "epoch": 42.22,
+      "learning_rate": 4.4722014925373135e-05,
+      "loss": 0.0,
+      "step": 45264
+    },
+    {
+      "epoch": 42.23,
+      "learning_rate": 4.472154850746269e-05,
+      "loss": 0.0001,
+      "step": 45268
+    },
+    {
+      "epoch": 42.23,
+      "learning_rate": 4.472108208955224e-05,
+      "loss": 0.0004,
+      "step": 45272
+    },
+    {
+      "epoch": 42.24,
+      "learning_rate": 4.472061567164179e-05,
+      "loss": 0.0028,
+      "step": 45276
+    },
+    {
+      "epoch": 42.24,
+      "learning_rate": 4.472014925373135e-05,
+      "loss": 0.0,
+      "step": 45280
+    },
+    {
+      "epoch": 42.24,
+      "learning_rate": 4.4719682835820896e-05,
+      "loss": 0.0001,
+      "step": 45284
+    },
+    {
+      "epoch": 42.25,
+      "learning_rate": 4.4719216417910444e-05,
+      "loss": 0.0,
+      "step": 45288
+    },
+    {
+      "epoch": 42.25,
+      "learning_rate": 4.4718750000000006e-05,
+      "loss": 0.0003,
+      "step": 45292
+    },
+    {
+      "epoch": 42.25,
+      "learning_rate": 4.4718283582089554e-05,
+      "loss": 0.0,
+      "step": 45296
+    },
+    {
+      "epoch": 42.26,
+      "learning_rate": 4.47178171641791e-05,
+      "loss": 0.004,
+      "step": 45300
+    },
+    {
+      "epoch": 42.26,
+      "learning_rate": 4.471735074626866e-05,
+      "loss": 0.0,
+      "step": 45304
+    },
+    {
+      "epoch": 42.26,
+      "learning_rate": 4.471688432835821e-05,
+      "loss": 0.0,
+      "step": 45308
+    },
+    {
+      "epoch": 42.27,
+      "learning_rate": 4.471641791044776e-05,
+      "loss": 0.0014,
+      "step": 45312
+    },
+    {
+      "epoch": 42.27,
+      "learning_rate": 4.4715951492537315e-05,
+      "loss": 0.0,
+      "step": 45316
+    },
+    {
+      "epoch": 42.28,
+      "learning_rate": 4.471548507462687e-05,
+      "loss": 0.0,
+      "step": 45320
+    },
+    {
+      "epoch": 42.28,
+      "learning_rate": 4.471501865671642e-05,
+      "loss": 0.0001,
+      "step": 45324
+    },
+    {
+      "epoch": 42.28,
+      "learning_rate": 4.471455223880597e-05,
+      "loss": 0.0005,
+      "step": 45328
+    },
+    {
+      "epoch": 42.29,
+      "learning_rate": 4.471408582089552e-05,
+      "loss": 0.0,
+      "step": 45332
+    },
+    {
+      "epoch": 42.29,
+      "learning_rate": 4.4713619402985076e-05,
+      "loss": 0.0002,
+      "step": 45336
+    },
+    {
+      "epoch": 42.29,
+      "learning_rate": 4.471315298507463e-05,
+      "loss": 0.0,
+      "step": 45340
+    },
+    {
+      "epoch": 42.3,
+      "learning_rate": 4.471268656716418e-05,
+      "loss": 0.0031,
+      "step": 45344
+    },
+    {
+      "epoch": 42.3,
+      "learning_rate": 4.4712220149253734e-05,
+      "loss": 0.0,
+      "step": 45348
+    },
+    {
+      "epoch": 42.31,
+      "learning_rate": 4.471175373134329e-05,
+      "loss": 0.0,
+      "step": 45352
+    },
+    {
+      "epoch": 42.31,
+      "learning_rate": 4.471128731343284e-05,
+      "loss": 0.0001,
+      "step": 45356
+    },
+    {
+      "epoch": 42.31,
+      "learning_rate": 4.471082089552239e-05,
+      "loss": 0.0,
+      "step": 45360
+    },
+    {
+      "epoch": 42.32,
+      "learning_rate": 4.471035447761194e-05,
+      "loss": 0.0,
+      "step": 45364
+    },
+    {
+      "epoch": 42.32,
+      "learning_rate": 4.4709888059701495e-05,
+      "loss": 0.0001,
+      "step": 45368
+    },
+    {
+      "epoch": 42.32,
+      "learning_rate": 4.470942164179105e-05,
+      "loss": 0.0006,
+      "step": 45372
+    },
+    {
+      "epoch": 42.33,
+      "learning_rate": 4.47089552238806e-05,
+      "loss": 0.0,
+      "step": 45376
+    },
+    {
+      "epoch": 42.33,
+      "learning_rate": 4.470848880597015e-05,
+      "loss": 0.0045,
+      "step": 45380
+    },
+    {
+      "epoch": 42.34,
+      "learning_rate": 4.470802238805971e-05,
+      "loss": 0.0004,
+      "step": 45384
+    },
+    {
+      "epoch": 42.34,
+      "learning_rate": 4.4707555970149256e-05,
+      "loss": 0.0,
+      "step": 45388
+    },
+    {
+      "epoch": 42.34,
+      "learning_rate": 4.4707089552238804e-05,
+      "loss": 0.0006,
+      "step": 45392
+    },
+    {
+      "epoch": 42.35,
+      "learning_rate": 4.470662313432836e-05,
+      "loss": 0.0,
+      "step": 45396
+    },
+    {
+      "epoch": 42.35,
+      "learning_rate": 4.4706156716417914e-05,
+      "loss": 0.0001,
+      "step": 45400
+    },
+    {
+      "epoch": 42.35,
+      "learning_rate": 4.470569029850746e-05,
+      "loss": 0.0,
+      "step": 45404
+    },
+    {
+      "epoch": 42.36,
+      "learning_rate": 4.470522388059702e-05,
+      "loss": 0.0005,
+      "step": 45408
+    },
+    {
+      "epoch": 42.36,
+      "learning_rate": 4.470475746268657e-05,
+      "loss": 0.0001,
+      "step": 45412
+    },
+    {
+      "epoch": 42.37,
+      "learning_rate": 4.470429104477612e-05,
+      "loss": 0.0004,
+      "step": 45416
+    },
+    {
+      "epoch": 42.37,
+      "learning_rate": 4.4703824626865675e-05,
+      "loss": 0.0001,
+      "step": 45420
+    },
+    {
+      "epoch": 42.37,
+      "learning_rate": 4.470335820895522e-05,
+      "loss": 0.0008,
+      "step": 45424
+    },
+    {
+      "epoch": 42.38,
+      "learning_rate": 4.470289179104478e-05,
+      "loss": 0.0,
+      "step": 45428
+    },
+    {
+      "epoch": 42.38,
+      "learning_rate": 4.470242537313433e-05,
+      "loss": 0.0001,
+      "step": 45432
+    },
+    {
+      "epoch": 42.38,
+      "learning_rate": 4.470195895522388e-05,
+      "loss": 0.0001,
+      "step": 45436
+    },
+    {
+      "epoch": 42.39,
+      "learning_rate": 4.4701492537313436e-05,
+      "loss": 0.0002,
+      "step": 45440
+    },
+    {
+      "epoch": 42.39,
+      "learning_rate": 4.470102611940299e-05,
+      "loss": 0.0,
+      "step": 45444
+    },
+    {
+      "epoch": 42.4,
+      "learning_rate": 4.470055970149254e-05,
+      "loss": 0.0,
+      "step": 45448
+    },
+    {
+      "epoch": 42.4,
+      "learning_rate": 4.470009328358209e-05,
+      "loss": 0.0,
+      "step": 45452
+    },
+    {
+      "epoch": 42.4,
+      "learning_rate": 4.469962686567164e-05,
+      "loss": 0.0,
+      "step": 45456
+    },
+    {
+      "epoch": 42.41,
+      "learning_rate": 4.46991604477612e-05,
+      "loss": 0.0,
+      "step": 45460
+    },
+    {
+      "epoch": 42.41,
+      "learning_rate": 4.4698694029850745e-05,
+      "loss": 0.0,
+      "step": 45464
+    },
+    {
+      "epoch": 42.41,
+      "learning_rate": 4.46982276119403e-05,
+      "loss": 0.0,
+      "step": 45468
+    },
+    {
+      "epoch": 42.42,
+      "learning_rate": 4.4697761194029855e-05,
+      "loss": 0.0001,
+      "step": 45472
+    },
+    {
+      "epoch": 42.42,
+      "learning_rate": 4.46972947761194e-05,
+      "loss": 0.0001,
+      "step": 45476
+    },
+    {
+      "epoch": 42.43,
+      "learning_rate": 4.469682835820896e-05,
+      "loss": 0.0,
+      "step": 45480
+    },
+    {
+      "epoch": 42.43,
+      "learning_rate": 4.4696361940298506e-05,
+      "loss": 0.001,
+      "step": 45484
+    },
+    {
+      "epoch": 42.43,
+      "learning_rate": 4.469589552238806e-05,
+      "loss": 0.0001,
+      "step": 45488
+    },
+    {
+      "epoch": 42.44,
+      "learning_rate": 4.4695429104477616e-05,
+      "loss": 0.0,
+      "step": 45492
+    },
+    {
+      "epoch": 42.44,
+      "learning_rate": 4.4694962686567164e-05,
+      "loss": 0.0,
+      "step": 45496
+    },
+    {
+      "epoch": 42.44,
+      "learning_rate": 4.469449626865672e-05,
+      "loss": 0.0,
+      "step": 45500
+    },
+    {
+      "epoch": 42.44,
+      "eval_exact_match": 0.7485493230174082,
+      "eval_exec": 0.7669245647969052,
+      "eval_loss": 0.42073753476142883,
+      "eval_runtime": 1143.3015,
+      "eval_samples_per_second": 0.904,
+      "step": 45500
+    },
+    {
+      "epoch": 42.45,
+      "learning_rate": 4.4694029850746274e-05,
+      "loss": 0.0001,
+      "step": 45504
+    },
+    {
+      "epoch": 42.45,
+      "learning_rate": 4.469356343283582e-05,
+      "loss": 0.0002,
+      "step": 45508
+    },
+    {
+      "epoch": 42.46,
+      "learning_rate": 4.469309701492538e-05,
+      "loss": 0.0001,
+      "step": 45512
+    },
+    {
+      "epoch": 42.46,
+      "learning_rate": 4.4692630597014925e-05,
+      "loss": 0.0,
+      "step": 45516
+    },
+    {
+      "epoch": 42.46,
+      "learning_rate": 4.469216417910448e-05,
+      "loss": 0.0001,
+      "step": 45520
+    },
+    {
+      "epoch": 42.47,
+      "learning_rate": 4.4691697761194035e-05,
+      "loss": 0.0001,
+      "step": 45524
+    },
+    {
+      "epoch": 42.47,
+      "learning_rate": 4.469123134328358e-05,
+      "loss": 0.0003,
+      "step": 45528
+    },
+    {
+      "epoch": 42.47,
+      "learning_rate": 4.469076492537314e-05,
+      "loss": 0.0,
+      "step": 45532
+    },
+    {
+      "epoch": 42.48,
+      "learning_rate": 4.469029850746269e-05,
+      "loss": 0.001,
+      "step": 45536
+    },
+    {
+      "epoch": 42.48,
+      "learning_rate": 4.468983208955224e-05,
+      "loss": 0.0009,
+      "step": 45540
+    },
+    {
+      "epoch": 42.49,
+      "learning_rate": 4.468936567164179e-05,
+      "loss": 0.0001,
+      "step": 45544
+    },
+    {
+      "epoch": 42.49,
+      "learning_rate": 4.468889925373135e-05,
+      "loss": 0.0,
+      "step": 45548
+    },
+    {
+      "epoch": 42.49,
+      "learning_rate": 4.46884328358209e-05,
+      "loss": 0.0004,
+      "step": 45552
+    },
+    {
+      "epoch": 42.5,
+      "learning_rate": 4.468796641791045e-05,
+      "loss": 0.0,
+      "step": 45556
+    },
+    {
+      "epoch": 42.5,
+      "learning_rate": 4.46875e-05,
+      "loss": 0.0,
+      "step": 45560
+    },
+    {
+      "epoch": 42.5,
+      "learning_rate": 4.468703358208956e-05,
+      "loss": 0.0068,
+      "step": 45564
+    },
+    {
+      "epoch": 42.51,
+      "learning_rate": 4.4686567164179105e-05,
+      "loss": 0.0,
+      "step": 45568
+    },
+    {
+      "epoch": 42.51,
+      "learning_rate": 4.468610074626866e-05,
+      "loss": 0.0,
+      "step": 45572
+    },
+    {
+      "epoch": 42.51,
+      "learning_rate": 4.468563432835821e-05,
+      "loss": 0.0,
+      "step": 45576
+    },
+    {
+      "epoch": 42.52,
+      "learning_rate": 4.468516791044776e-05,
+      "loss": 0.0,
+      "step": 45580
+    },
+    {
+      "epoch": 42.52,
+      "learning_rate": 4.468470149253732e-05,
+      "loss": 0.0,
+      "step": 45584
+    },
+    {
+      "epoch": 42.53,
+      "learning_rate": 4.4684235074626866e-05,
+      "loss": 0.0003,
+      "step": 45588
+    },
+    {
+      "epoch": 42.53,
+      "learning_rate": 4.468376865671642e-05,
+      "loss": 0.0,
+      "step": 45592
+    },
+    {
+      "epoch": 42.53,
+      "learning_rate": 4.4683302238805976e-05,
+      "loss": 0.0002,
+      "step": 45596
+    },
+    {
+      "epoch": 42.54,
+      "learning_rate": 4.4682835820895524e-05,
+      "loss": 0.0001,
+      "step": 45600
+    },
+    {
+      "epoch": 42.54,
+      "learning_rate": 4.468236940298507e-05,
+      "loss": 0.0002,
+      "step": 45604
+    },
+    {
+      "epoch": 42.54,
+      "learning_rate": 4.4681902985074634e-05,
+      "loss": 0.0,
+      "step": 45608
+    },
+    {
+      "epoch": 42.55,
+      "learning_rate": 4.468143656716418e-05,
+      "loss": 0.002,
+      "step": 45612
+    },
+    {
+      "epoch": 42.55,
+      "learning_rate": 4.468097014925373e-05,
+      "loss": 0.0003,
+      "step": 45616
+    },
+    {
+      "epoch": 42.56,
+      "learning_rate": 4.4680503731343285e-05,
+      "loss": 0.0,
+      "step": 45620
+    },
+    {
+      "epoch": 42.56,
+      "learning_rate": 4.468003731343284e-05,
+      "loss": 0.0,
+      "step": 45624
+    },
+    {
+      "epoch": 42.56,
+      "learning_rate": 4.467957089552239e-05,
+      "loss": 0.0001,
+      "step": 45628
+    },
+    {
+      "epoch": 42.57,
+      "learning_rate": 4.467910447761194e-05,
+      "loss": 0.0001,
+      "step": 45632
+    },
+    {
+      "epoch": 42.57,
+      "learning_rate": 4.467863805970149e-05,
+      "loss": 0.0,
+      "step": 45636
+    },
+    {
+      "epoch": 42.57,
+      "learning_rate": 4.4678171641791046e-05,
+      "loss": 0.0,
+      "step": 45640
+    },
+    {
+      "epoch": 42.58,
+      "learning_rate": 4.46777052238806e-05,
+      "loss": 0.0,
+      "step": 45644
+    },
+    {
+      "epoch": 42.58,
+      "learning_rate": 4.467723880597015e-05,
+      "loss": 0.0,
+      "step": 45648
+    },
+    {
+      "epoch": 42.59,
+      "learning_rate": 4.4676772388059704e-05,
+      "loss": 0.0,
+      "step": 45652
+    },
+    {
+      "epoch": 42.59,
+      "learning_rate": 4.467630597014926e-05,
+      "loss": 0.0031,
+      "step": 45656
+    },
+    {
+      "epoch": 42.59,
+      "learning_rate": 4.467583955223881e-05,
+      "loss": 0.0,
+      "step": 45660
+    },
+    {
+      "epoch": 42.6,
+      "learning_rate": 4.4675373134328355e-05,
+      "loss": 0.0002,
+      "step": 45664
+    },
+    {
+      "epoch": 42.6,
+      "learning_rate": 4.467490671641792e-05,
+      "loss": 0.0,
+      "step": 45668
+    },
+    {
+      "epoch": 42.6,
+      "learning_rate": 4.4674440298507465e-05,
+      "loss": 0.0006,
+      "step": 45672
+    },
+    {
+      "epoch": 42.61,
+      "learning_rate": 4.467397388059702e-05,
+      "loss": 0.0003,
+      "step": 45676
+    },
+    {
+      "epoch": 42.61,
+      "learning_rate": 4.467350746268657e-05,
+      "loss": 0.0002,
+      "step": 45680
+    },
+    {
+      "epoch": 42.62,
+      "learning_rate": 4.467304104477612e-05,
+      "loss": 0.0,
+      "step": 45684
+    },
+    {
+      "epoch": 42.62,
+      "learning_rate": 4.467257462686568e-05,
+      "loss": 0.0,
+      "step": 45688
+    },
+    {
+      "epoch": 42.62,
+      "learning_rate": 4.4672108208955226e-05,
+      "loss": 0.0,
+      "step": 45692
+    },
+    {
+      "epoch": 42.63,
+      "learning_rate": 4.4671641791044774e-05,
+      "loss": 0.0,
+      "step": 45696
+    },
+    {
+      "epoch": 42.63,
+      "learning_rate": 4.4671175373134336e-05,
+      "loss": 0.0002,
+      "step": 45700
+    },
+    {
+      "epoch": 42.63,
+      "learning_rate": 4.4670708955223884e-05,
+      "loss": 0.0,
+      "step": 45704
+    },
+    {
+      "epoch": 42.64,
+      "learning_rate": 4.467024253731343e-05,
+      "loss": 0.0001,
+      "step": 45708
+    },
+    {
+      "epoch": 42.64,
+      "learning_rate": 4.466977611940299e-05,
+      "loss": 0.0,
+      "step": 45712
+    },
+    {
+      "epoch": 42.65,
+      "learning_rate": 4.466930970149254e-05,
+      "loss": 0.0,
+      "step": 45716
+    },
+    {
+      "epoch": 42.65,
+      "learning_rate": 4.466884328358209e-05,
+      "loss": 0.0002,
+      "step": 45720
+    },
+    {
+      "epoch": 42.65,
+      "learning_rate": 4.4668376865671645e-05,
+      "loss": 0.0002,
+      "step": 45724
+    },
+    {
+      "epoch": 42.66,
+      "learning_rate": 4.46679104477612e-05,
+      "loss": 0.0031,
+      "step": 45728
+    },
+    {
+      "epoch": 42.66,
+      "learning_rate": 4.466744402985075e-05,
+      "loss": 0.0003,
+      "step": 45732
+    },
+    {
+      "epoch": 42.66,
+      "learning_rate": 4.46669776119403e-05,
+      "loss": 0.0001,
+      "step": 45736
+    },
+    {
+      "epoch": 42.67,
+      "learning_rate": 4.466651119402985e-05,
+      "loss": 0.0,
+      "step": 45740
+    },
+    {
+      "epoch": 42.67,
+      "learning_rate": 4.4666044776119406e-05,
+      "loss": 0.0,
+      "step": 45744
+    },
+    {
+      "epoch": 42.68,
+      "learning_rate": 4.466557835820896e-05,
+      "loss": 0.0001,
+      "step": 45748
+    },
+    {
+      "epoch": 42.68,
+      "learning_rate": 4.466511194029851e-05,
+      "loss": 0.0001,
+      "step": 45752
+    },
+    {
+      "epoch": 42.68,
+      "learning_rate": 4.466464552238806e-05,
+      "loss": 0.0,
+      "step": 45756
+    },
+    {
+      "epoch": 42.69,
+      "learning_rate": 4.466417910447762e-05,
+      "loss": 0.0,
+      "step": 45760
+    },
+    {
+      "epoch": 42.69,
+      "learning_rate": 4.466371268656717e-05,
+      "loss": 0.0004,
+      "step": 45764
+    },
+    {
+      "epoch": 42.69,
+      "learning_rate": 4.4663246268656715e-05,
+      "loss": 0.0001,
+      "step": 45768
+    },
+    {
+      "epoch": 42.7,
+      "learning_rate": 4.466277985074627e-05,
+      "loss": 0.0004,
+      "step": 45772
+    },
+    {
+      "epoch": 42.7,
+      "learning_rate": 4.4662313432835825e-05,
+      "loss": 0.0003,
+      "step": 45776
+    },
+    {
+      "epoch": 42.71,
+      "learning_rate": 4.466184701492537e-05,
+      "loss": 0.0,
+      "step": 45780
+    },
+    {
+      "epoch": 42.71,
+      "learning_rate": 4.466138059701493e-05,
+      "loss": 0.0,
+      "step": 45784
+    },
+    {
+      "epoch": 42.71,
+      "learning_rate": 4.466091417910448e-05,
+      "loss": 0.0001,
+      "step": 45788
+    },
+    {
+      "epoch": 42.72,
+      "learning_rate": 4.466044776119403e-05,
+      "loss": 0.0,
+      "step": 45792
+    },
+    {
+      "epoch": 42.72,
+      "learning_rate": 4.4659981343283586e-05,
+      "loss": 0.0001,
+      "step": 45796
+    },
+    {
+      "epoch": 42.72,
+      "learning_rate": 4.4659514925373134e-05,
+      "loss": 0.0002,
+      "step": 45800
+    },
+    {
+      "epoch": 42.73,
+      "learning_rate": 4.465904850746269e-05,
+      "loss": 0.0003,
+      "step": 45804
+    },
+    {
+      "epoch": 42.73,
+      "learning_rate": 4.4658582089552244e-05,
+      "loss": 0.0,
+      "step": 45808
+    },
+    {
+      "epoch": 42.73,
+      "learning_rate": 4.465811567164179e-05,
+      "loss": 0.0002,
+      "step": 45812
+    },
+    {
+      "epoch": 42.74,
+      "learning_rate": 4.465764925373134e-05,
+      "loss": 0.0,
+      "step": 45816
+    },
+    {
+      "epoch": 42.74,
+      "learning_rate": 4.46571828358209e-05,
+      "loss": 0.0016,
+      "step": 45820
+    },
+    {
+      "epoch": 42.75,
+      "learning_rate": 4.465671641791045e-05,
+      "loss": 0.0005,
+      "step": 45824
+    },
+    {
+      "epoch": 42.75,
+      "learning_rate": 4.465625e-05,
+      "loss": 0.0003,
+      "step": 45828
+    },
+    {
+      "epoch": 42.75,
+      "learning_rate": 4.465578358208955e-05,
+      "loss": 0.0,
+      "step": 45832
+    },
+    {
+      "epoch": 42.76,
+      "learning_rate": 4.465531716417911e-05,
+      "loss": 0.0033,
+      "step": 45836
+    },
+    {
+      "epoch": 42.76,
+      "learning_rate": 4.465485074626866e-05,
+      "loss": 0.0,
+      "step": 45840
+    },
+    {
+      "epoch": 42.76,
+      "learning_rate": 4.465438432835821e-05,
+      "loss": 0.0001,
+      "step": 45844
+    },
+    {
+      "epoch": 42.77,
+      "learning_rate": 4.4653917910447766e-05,
+      "loss": 0.0,
+      "step": 45848
+    },
+    {
+      "epoch": 42.77,
+      "learning_rate": 4.465345149253732e-05,
+      "loss": 0.0007,
+      "step": 45852
+    },
+    {
+      "epoch": 42.78,
+      "learning_rate": 4.465298507462687e-05,
+      "loss": 0.0,
+      "step": 45856
+    },
+    {
+      "epoch": 42.78,
+      "learning_rate": 4.465251865671642e-05,
+      "loss": 0.0,
+      "step": 45860
+    },
+    {
+      "epoch": 42.78,
+      "learning_rate": 4.465205223880597e-05,
+      "loss": 0.0,
+      "step": 45864
+    },
+    {
+      "epoch": 42.79,
+      "learning_rate": 4.465158582089553e-05,
+      "loss": 0.0,
+      "step": 45868
+    },
+    {
+      "epoch": 42.79,
+      "learning_rate": 4.4651119402985075e-05,
+      "loss": 0.0001,
+      "step": 45872
+    },
+    {
+      "epoch": 42.79,
+      "learning_rate": 4.465065298507463e-05,
+      "loss": 0.0013,
+      "step": 45876
+    },
+    {
+      "epoch": 42.8,
+      "learning_rate": 4.4650186567164185e-05,
+      "loss": 0.0,
+      "step": 45880
+    },
+    {
+      "epoch": 42.8,
+      "learning_rate": 4.464972014925373e-05,
+      "loss": 0.0,
+      "step": 45884
+    },
+    {
+      "epoch": 42.81,
+      "learning_rate": 4.464925373134329e-05,
+      "loss": 0.0,
+      "step": 45888
+    },
+    {
+      "epoch": 42.81,
+      "learning_rate": 4.4648787313432836e-05,
+      "loss": 0.0005,
+      "step": 45892
+    },
+    {
+      "epoch": 42.81,
+      "learning_rate": 4.464832089552239e-05,
+      "loss": 0.0,
+      "step": 45896
+    },
+    {
+      "epoch": 42.82,
+      "learning_rate": 4.4647854477611946e-05,
+      "loss": 0.0001,
+      "step": 45900
+    },
+    {
+      "epoch": 42.82,
+      "learning_rate": 4.4647388059701494e-05,
+      "loss": 0.0,
+      "step": 45904
+    },
+    {
+      "epoch": 42.82,
+      "learning_rate": 4.464692164179104e-05,
+      "loss": 0.0105,
+      "step": 45908
+    },
+    {
+      "epoch": 42.83,
+      "learning_rate": 4.4646455223880603e-05,
+      "loss": 0.0007,
+      "step": 45912
+    },
+    {
+      "epoch": 42.83,
+      "learning_rate": 4.464598880597015e-05,
+      "loss": 0.0002,
+      "step": 45916
+    },
+    {
+      "epoch": 42.84,
+      "learning_rate": 4.46455223880597e-05,
+      "loss": 0.0,
+      "step": 45920
+    },
+    {
+      "epoch": 42.84,
+      "learning_rate": 4.4645055970149255e-05,
+      "loss": 0.0043,
+      "step": 45924
+    },
+    {
+      "epoch": 42.84,
+      "learning_rate": 4.464458955223881e-05,
+      "loss": 0.0,
+      "step": 45928
+    },
+    {
+      "epoch": 42.85,
+      "learning_rate": 4.464412313432836e-05,
+      "loss": 0.0002,
+      "step": 45932
+    },
+    {
+      "epoch": 42.85,
+      "learning_rate": 4.464365671641791e-05,
+      "loss": 0.0033,
+      "step": 45936
+    },
+    {
+      "epoch": 42.85,
+      "learning_rate": 4.464319029850747e-05,
+      "loss": 0.0001,
+      "step": 45940
+    },
+    {
+      "epoch": 42.86,
+      "learning_rate": 4.4642723880597016e-05,
+      "loss": 0.0001,
+      "step": 45944
+    },
+    {
+      "epoch": 42.86,
+      "learning_rate": 4.464225746268657e-05,
+      "loss": 0.0,
+      "step": 45948
+    },
+    {
+      "epoch": 42.87,
+      "learning_rate": 4.464179104477612e-05,
+      "loss": 0.0024,
+      "step": 45952
+    },
+    {
+      "epoch": 42.87,
+      "learning_rate": 4.4641324626865674e-05,
+      "loss": 0.0008,
+      "step": 45956
+    },
+    {
+      "epoch": 42.87,
+      "learning_rate": 4.464085820895523e-05,
+      "loss": 0.0011,
+      "step": 45960
+    },
+    {
+      "epoch": 42.88,
+      "learning_rate": 4.464039179104478e-05,
+      "loss": 0.0015,
+      "step": 45964
+    },
+    {
+      "epoch": 42.88,
+      "learning_rate": 4.4639925373134325e-05,
+      "loss": 0.0,
+      "step": 45968
+    },
+    {
+      "epoch": 42.88,
+      "learning_rate": 4.4639458955223886e-05,
+      "loss": 0.0001,
+      "step": 45972
+    },
+    {
+      "epoch": 42.89,
+      "learning_rate": 4.4638992537313435e-05,
+      "loss": 0.0009,
+      "step": 45976
+    },
+    {
+      "epoch": 42.89,
+      "learning_rate": 4.463852611940298e-05,
+      "loss": 0.0063,
+      "step": 45980
+    },
+    {
+      "epoch": 42.9,
+      "learning_rate": 4.463805970149254e-05,
+      "loss": 0.0,
+      "step": 45984
+    },
+    {
+      "epoch": 42.9,
+      "learning_rate": 4.463759328358209e-05,
+      "loss": 0.0006,
+      "step": 45988
+    },
+    {
+      "epoch": 42.9,
+      "learning_rate": 4.463712686567164e-05,
+      "loss": 0.0001,
+      "step": 45992
+    },
+    {
+      "epoch": 42.91,
+      "learning_rate": 4.4636660447761196e-05,
+      "loss": 0.0009,
+      "step": 45996
+    },
+    {
+      "epoch": 42.91,
+      "learning_rate": 4.463619402985075e-05,
+      "loss": 0.0002,
+      "step": 46000
+    },
+    {
+      "epoch": 42.91,
+      "eval_exact_match": 0.7340425531914894,
+      "eval_exec": 0.7678916827852998,
+      "eval_loss": 0.4125958979129791,
+      "eval_runtime": 1108.5153,
+      "eval_samples_per_second": 0.933,
+      "step": 46000
+    },
+    {
+      "epoch": 42.91,
+      "learning_rate": 4.4635727611940305e-05,
+      "loss": 0.0128,
+      "step": 46004
+    },
+    {
+      "epoch": 42.92,
+      "learning_rate": 4.4635261194029854e-05,
+      "loss": 0.0002,
+      "step": 46008
+    },
+    {
+      "epoch": 42.92,
+      "learning_rate": 4.46347947761194e-05,
+      "loss": 0.0003,
+      "step": 46012
+    },
+    {
+      "epoch": 42.93,
+      "learning_rate": 4.463432835820896e-05,
+      "loss": 0.0,
+      "step": 46016
+    },
+    {
+      "epoch": 42.93,
+      "learning_rate": 4.463386194029851e-05,
+      "loss": 0.0,
+      "step": 46020
+    },
+    {
+      "epoch": 42.93,
+      "learning_rate": 4.463339552238806e-05,
+      "loss": 0.0002,
+      "step": 46024
+    },
+    {
+      "epoch": 42.94,
+      "learning_rate": 4.4632929104477614e-05,
+      "loss": 0.0,
+      "step": 46028
+    },
+    {
+      "epoch": 42.94,
+      "learning_rate": 4.463246268656717e-05,
+      "loss": 0.0,
+      "step": 46032
+    },
+    {
+      "epoch": 42.94,
+      "learning_rate": 4.463199626865672e-05,
+      "loss": 0.0019,
+      "step": 46036
+    },
+    {
+      "epoch": 42.95,
+      "learning_rate": 4.463152985074627e-05,
+      "loss": 0.0003,
+      "step": 46040
+    },
+    {
+      "epoch": 42.95,
+      "learning_rate": 4.463106343283582e-05,
+      "loss": 0.0,
+      "step": 46044
+    },
+    {
+      "epoch": 42.96,
+      "learning_rate": 4.4630597014925375e-05,
+      "loss": 0.0,
+      "step": 46048
+    },
+    {
+      "epoch": 42.96,
+      "learning_rate": 4.463013059701493e-05,
+      "loss": 0.0001,
+      "step": 46052
+    },
+    {
+      "epoch": 42.96,
+      "learning_rate": 4.462966417910448e-05,
+      "loss": 0.0005,
+      "step": 46056
+    },
+    {
+      "epoch": 42.97,
+      "learning_rate": 4.4629197761194033e-05,
+      "loss": 0.0,
+      "step": 46060
+    },
+    {
+      "epoch": 42.97,
+      "learning_rate": 4.462873134328359e-05,
+      "loss": 0.0011,
+      "step": 46064
+    },
+    {
+      "epoch": 42.97,
+      "learning_rate": 4.4628264925373136e-05,
+      "loss": 0.0,
+      "step": 46068
+    },
+    {
+      "epoch": 42.98,
+      "learning_rate": 4.4627798507462685e-05,
+      "loss": 0.0,
+      "step": 46072
+    },
+    {
+      "epoch": 42.98,
+      "learning_rate": 4.462733208955224e-05,
+      "loss": 0.0008,
+      "step": 46076
+    },
+    {
+      "epoch": 42.98,
+      "learning_rate": 4.4626865671641794e-05,
+      "loss": 0.0051,
+      "step": 46080
+    },
+    {
+      "epoch": 42.99,
+      "learning_rate": 4.462639925373134e-05,
+      "loss": 0.0002,
+      "step": 46084
+    },
+    {
+      "epoch": 42.99,
+      "learning_rate": 4.46259328358209e-05,
+      "loss": 0.0,
+      "step": 46088
+    },
+    {
+      "epoch": 43.0,
+      "learning_rate": 4.462546641791045e-05,
+      "loss": 0.0,
+      "step": 46092
+    },
+    {
+      "epoch": 43.0,
+      "learning_rate": 4.4625e-05,
+      "loss": 0.0002,
+      "step": 46096
+    },
+    {
+      "epoch": 43.0,
+      "learning_rate": 4.4624533582089555e-05,
+      "loss": 0.0,
+      "step": 46100
+    },
+    {
+      "epoch": 43.01,
+      "learning_rate": 4.4624067164179104e-05,
+      "loss": 0.007,
+      "step": 46104
+    },
+    {
+      "epoch": 43.01,
+      "learning_rate": 4.462360074626866e-05,
+      "loss": 0.0003,
+      "step": 46108
+    },
+    {
+      "epoch": 43.01,
+      "learning_rate": 4.462313432835821e-05,
+      "loss": 0.0023,
+      "step": 46112
+    },
+    {
+      "epoch": 43.02,
+      "learning_rate": 4.462266791044776e-05,
+      "loss": 0.0001,
+      "step": 46116
+    },
+    {
+      "epoch": 43.02,
+      "learning_rate": 4.4622201492537316e-05,
+      "loss": 0.0003,
+      "step": 46120
+    },
+    {
+      "epoch": 43.03,
+      "learning_rate": 4.462173507462687e-05,
+      "loss": 0.0001,
+      "step": 46124
+    },
+    {
+      "epoch": 43.03,
+      "learning_rate": 4.462126865671642e-05,
+      "loss": 0.0001,
+      "step": 46128
+    },
+    {
+      "epoch": 43.03,
+      "learning_rate": 4.462080223880597e-05,
+      "loss": 0.0006,
+      "step": 46132
+    },
+    {
+      "epoch": 43.04,
+      "learning_rate": 4.462033582089552e-05,
+      "loss": 0.0,
+      "step": 46136
+    },
+    {
+      "epoch": 43.04,
+      "learning_rate": 4.461986940298508e-05,
+      "loss": 0.0001,
+      "step": 46140
+    },
+    {
+      "epoch": 43.04,
+      "learning_rate": 4.4619402985074626e-05,
+      "loss": 0.0,
+      "step": 46144
+    },
+    {
+      "epoch": 43.05,
+      "learning_rate": 4.461893656716418e-05,
+      "loss": 0.0,
+      "step": 46148
+    },
+    {
+      "epoch": 43.05,
+      "learning_rate": 4.4618470149253735e-05,
+      "loss": 0.0001,
+      "step": 46152
+    },
+    {
+      "epoch": 43.06,
+      "learning_rate": 4.4618003731343283e-05,
+      "loss": 0.0,
+      "step": 46156
+    },
+    {
+      "epoch": 43.06,
+      "learning_rate": 4.461753731343284e-05,
+      "loss": 0.0,
+      "step": 46160
+    },
+    {
+      "epoch": 43.06,
+      "learning_rate": 4.4617070895522386e-05,
+      "loss": 0.0,
+      "step": 46164
+    },
+    {
+      "epoch": 43.07,
+      "learning_rate": 4.461660447761195e-05,
+      "loss": 0.0,
+      "step": 46168
+    },
+    {
+      "epoch": 43.07,
+      "learning_rate": 4.4616138059701496e-05,
+      "loss": 0.0,
+      "step": 46172
+    },
+    {
+      "epoch": 43.07,
+      "learning_rate": 4.4615671641791044e-05,
+      "loss": 0.0001,
+      "step": 46176
+    },
+    {
+      "epoch": 43.08,
+      "learning_rate": 4.46152052238806e-05,
+      "loss": 0.0003,
+      "step": 46180
+    },
+    {
+      "epoch": 43.08,
+      "learning_rate": 4.4614738805970154e-05,
+      "loss": 0.0001,
+      "step": 46184
+    },
+    {
+      "epoch": 43.09,
+      "learning_rate": 4.46142723880597e-05,
+      "loss": 0.0,
+      "step": 46188
+    },
+    {
+      "epoch": 43.09,
+      "learning_rate": 4.461380597014926e-05,
+      "loss": 0.0,
+      "step": 46192
+    },
+    {
+      "epoch": 43.09,
+      "learning_rate": 4.4613339552238805e-05,
+      "loss": 0.0001,
+      "step": 46196
+    },
+    {
+      "epoch": 43.1,
+      "learning_rate": 4.461287313432836e-05,
+      "loss": 0.0,
+      "step": 46200
+    },
+    {
+      "epoch": 43.1,
+      "learning_rate": 4.4612406716417915e-05,
+      "loss": 0.0,
+      "step": 46204
+    },
+    {
+      "epoch": 43.1,
+      "learning_rate": 4.461194029850746e-05,
+      "loss": 0.0002,
+      "step": 46208
+    },
+    {
+      "epoch": 43.11,
+      "learning_rate": 4.461147388059702e-05,
+      "loss": 0.0,
+      "step": 46212
+    },
+    {
+      "epoch": 43.11,
+      "learning_rate": 4.461100746268657e-05,
+      "loss": 0.0012,
+      "step": 46216
+    },
+    {
+      "epoch": 43.12,
+      "learning_rate": 4.461054104477612e-05,
+      "loss": 0.0009,
+      "step": 46220
+    },
+    {
+      "epoch": 43.12,
+      "learning_rate": 4.461007462686567e-05,
+      "loss": 0.0003,
+      "step": 46224
+    },
+    {
+      "epoch": 43.12,
+      "learning_rate": 4.460960820895523e-05,
+      "loss": 0.0,
+      "step": 46228
+    },
+    {
+      "epoch": 43.13,
+      "learning_rate": 4.460914179104478e-05,
+      "loss": 0.0,
+      "step": 46232
+    },
+    {
+      "epoch": 43.13,
+      "learning_rate": 4.460867537313433e-05,
+      "loss": 0.0001,
+      "step": 46236
+    },
+    {
+      "epoch": 43.13,
+      "learning_rate": 4.460820895522388e-05,
+      "loss": 0.0,
+      "step": 46240
+    },
+    {
+      "epoch": 43.14,
+      "learning_rate": 4.460774253731344e-05,
+      "loss": 0.0,
+      "step": 46244
+    },
+    {
+      "epoch": 43.14,
+      "learning_rate": 4.4607276119402985e-05,
+      "loss": 0.0,
+      "step": 46248
+    },
+    {
+      "epoch": 43.15,
+      "learning_rate": 4.460680970149254e-05,
+      "loss": 0.0,
+      "step": 46252
+    },
+    {
+      "epoch": 43.15,
+      "learning_rate": 4.460634328358209e-05,
+      "loss": 0.0,
+      "step": 46256
+    },
+    {
+      "epoch": 43.15,
+      "learning_rate": 4.460587686567164e-05,
+      "loss": 0.0,
+      "step": 46260
+    },
+    {
+      "epoch": 43.16,
+      "learning_rate": 4.46054104477612e-05,
+      "loss": 0.0001,
+      "step": 46264
+    },
+    {
+      "epoch": 43.16,
+      "learning_rate": 4.4604944029850746e-05,
+      "loss": 0.0,
+      "step": 46268
+    },
+    {
+      "epoch": 43.16,
+      "learning_rate": 4.46044776119403e-05,
+      "loss": 0.0077,
+      "step": 46272
+    },
+    {
+      "epoch": 43.17,
+      "learning_rate": 4.4604011194029856e-05,
+      "loss": 0.0,
+      "step": 46276
+    },
+    {
+      "epoch": 43.17,
+      "learning_rate": 4.4603544776119404e-05,
+      "loss": 0.0,
+      "step": 46280
+    },
+    {
+      "epoch": 43.18,
+      "learning_rate": 4.460307835820895e-05,
+      "loss": 0.0,
+      "step": 46284
+    },
+    {
+      "epoch": 43.18,
+      "learning_rate": 4.4602611940298514e-05,
+      "loss": 0.0001,
+      "step": 46288
+    },
+    {
+      "epoch": 43.18,
+      "learning_rate": 4.460214552238806e-05,
+      "loss": 0.0015,
+      "step": 46292
+    },
+    {
+      "epoch": 43.19,
+      "learning_rate": 4.460167910447761e-05,
+      "loss": 0.0,
+      "step": 46296
+    },
+    {
+      "epoch": 43.19,
+      "learning_rate": 4.4601212686567165e-05,
+      "loss": 0.0,
+      "step": 46300
+    },
+    {
+      "epoch": 43.19,
+      "learning_rate": 4.460074626865672e-05,
+      "loss": 0.0015,
+      "step": 46304
+    },
+    {
+      "epoch": 43.2,
+      "learning_rate": 4.460027985074627e-05,
+      "loss": 0.0,
+      "step": 46308
+    },
+    {
+      "epoch": 43.2,
+      "learning_rate": 4.459981343283582e-05,
+      "loss": 0.0,
+      "step": 46312
+    },
+    {
+      "epoch": 43.21,
+      "learning_rate": 4.459934701492537e-05,
+      "loss": 0.0002,
+      "step": 46316
+    },
+    {
+      "epoch": 43.21,
+      "learning_rate": 4.4598880597014926e-05,
+      "loss": 0.0,
+      "step": 46320
+    },
+    {
+      "epoch": 43.21,
+      "learning_rate": 4.459841417910448e-05,
+      "loss": 0.0046,
+      "step": 46324
+    },
+    {
+      "epoch": 43.22,
+      "learning_rate": 4.459794776119403e-05,
+      "loss": 0.0,
+      "step": 46328
+    },
+    {
+      "epoch": 43.22,
+      "learning_rate": 4.4597481343283584e-05,
+      "loss": 0.0005,
+      "step": 46332
+    },
+    {
+      "epoch": 43.22,
+      "learning_rate": 4.459701492537314e-05,
+      "loss": 0.0001,
+      "step": 46336
+    },
+    {
+      "epoch": 43.23,
+      "learning_rate": 4.459654850746269e-05,
+      "loss": 0.0002,
+      "step": 46340
+    },
+    {
+      "epoch": 43.23,
+      "learning_rate": 4.459608208955224e-05,
+      "loss": 0.0006,
+      "step": 46344
+    },
+    {
+      "epoch": 43.24,
+      "learning_rate": 4.45956156716418e-05,
+      "loss": 0.0006,
+      "step": 46348
+    },
+    {
+      "epoch": 43.24,
+      "learning_rate": 4.4595149253731345e-05,
+      "loss": 0.0003,
+      "step": 46352
+    },
+    {
+      "epoch": 43.24,
+      "learning_rate": 4.45946828358209e-05,
+      "loss": 0.0,
+      "step": 46356
+    },
+    {
+      "epoch": 43.25,
+      "learning_rate": 4.459421641791045e-05,
+      "loss": 0.0003,
+      "step": 46360
+    },
+    {
+      "epoch": 43.25,
+      "learning_rate": 4.459375e-05,
+      "loss": 0.0005,
+      "step": 46364
+    },
+    {
+      "epoch": 43.25,
+      "learning_rate": 4.459328358208956e-05,
+      "loss": 0.0,
+      "step": 46368
+    },
+    {
+      "epoch": 43.26,
+      "learning_rate": 4.4592817164179106e-05,
+      "loss": 0.0005,
+      "step": 46372
+    },
+    {
+      "epoch": 43.26,
+      "learning_rate": 4.4592350746268654e-05,
+      "loss": 0.0,
+      "step": 46376
+    },
+    {
+      "epoch": 43.26,
+      "learning_rate": 4.4591884328358216e-05,
+      "loss": 0.0,
+      "step": 46380
+    },
+    {
+      "epoch": 43.27,
+      "learning_rate": 4.4591417910447764e-05,
+      "loss": 0.0,
+      "step": 46384
+    },
+    {
+      "epoch": 43.27,
+      "learning_rate": 4.459095149253731e-05,
+      "loss": 0.0001,
+      "step": 46388
+    },
+    {
+      "epoch": 43.28,
+      "learning_rate": 4.459048507462687e-05,
+      "loss": 0.0,
+      "step": 46392
+    },
+    {
+      "epoch": 43.28,
+      "learning_rate": 4.459001865671642e-05,
+      "loss": 0.001,
+      "step": 46396
+    },
+    {
+      "epoch": 43.28,
+      "learning_rate": 4.458955223880597e-05,
+      "loss": 0.0014,
+      "step": 46400
+    },
+    {
+      "epoch": 43.29,
+      "learning_rate": 4.4589085820895525e-05,
+      "loss": 0.0008,
+      "step": 46404
+    },
+    {
+      "epoch": 43.29,
+      "learning_rate": 4.458861940298508e-05,
+      "loss": 0.0,
+      "step": 46408
+    },
+    {
+      "epoch": 43.29,
+      "learning_rate": 4.458815298507463e-05,
+      "loss": 0.0,
+      "step": 46412
+    },
+    {
+      "epoch": 43.3,
+      "learning_rate": 4.458768656716418e-05,
+      "loss": 0.0011,
+      "step": 46416
+    },
+    {
+      "epoch": 43.3,
+      "learning_rate": 4.458722014925373e-05,
+      "loss": 0.0,
+      "step": 46420
+    },
+    {
+      "epoch": 43.31,
+      "learning_rate": 4.4586753731343286e-05,
+      "loss": 0.0,
+      "step": 46424
+    },
+    {
+      "epoch": 43.31,
+      "learning_rate": 4.458628731343284e-05,
+      "loss": 0.0001,
+      "step": 46428
+    },
+    {
+      "epoch": 43.31,
+      "learning_rate": 4.458582089552239e-05,
+      "loss": 0.0,
+      "step": 46432
+    },
+    {
+      "epoch": 43.32,
+      "learning_rate": 4.458535447761194e-05,
+      "loss": 0.0,
+      "step": 46436
+    },
+    {
+      "epoch": 43.32,
+      "learning_rate": 4.45848880597015e-05,
+      "loss": 0.0009,
+      "step": 46440
+    },
+    {
+      "epoch": 43.32,
+      "learning_rate": 4.458442164179105e-05,
+      "loss": 0.0002,
+      "step": 46444
+    },
+    {
+      "epoch": 43.33,
+      "learning_rate": 4.4583955223880595e-05,
+      "loss": 0.0,
+      "step": 46448
+    },
+    {
+      "epoch": 43.33,
+      "learning_rate": 4.458348880597015e-05,
+      "loss": 0.0021,
+      "step": 46452
+    },
+    {
+      "epoch": 43.34,
+      "learning_rate": 4.4583022388059705e-05,
+      "loss": 0.0,
+      "step": 46456
+    },
+    {
+      "epoch": 43.34,
+      "learning_rate": 4.458255597014925e-05,
+      "loss": 0.0012,
+      "step": 46460
+    },
+    {
+      "epoch": 43.34,
+      "learning_rate": 4.458208955223881e-05,
+      "loss": 0.0027,
+      "step": 46464
+    },
+    {
+      "epoch": 43.35,
+      "learning_rate": 4.458162313432836e-05,
+      "loss": 0.0,
+      "step": 46468
+    },
+    {
+      "epoch": 43.35,
+      "learning_rate": 4.458115671641791e-05,
+      "loss": 0.0001,
+      "step": 46472
+    },
+    {
+      "epoch": 43.35,
+      "learning_rate": 4.4580690298507466e-05,
+      "loss": 0.0002,
+      "step": 46476
+    },
+    {
+      "epoch": 43.36,
+      "learning_rate": 4.4580223880597014e-05,
+      "loss": 0.0002,
+      "step": 46480
+    },
+    {
+      "epoch": 43.36,
+      "learning_rate": 4.457975746268657e-05,
+      "loss": 0.0007,
+      "step": 46484
+    },
+    {
+      "epoch": 43.37,
+      "learning_rate": 4.4579291044776124e-05,
+      "loss": 0.0002,
+      "step": 46488
+    },
+    {
+      "epoch": 43.37,
+      "learning_rate": 4.457882462686567e-05,
+      "loss": 0.0009,
+      "step": 46492
+    },
+    {
+      "epoch": 43.37,
+      "learning_rate": 4.457835820895523e-05,
+      "loss": 0.0063,
+      "step": 46496
+    },
+    {
+      "epoch": 43.38,
+      "learning_rate": 4.457789179104478e-05,
+      "loss": 0.0004,
+      "step": 46500
+    },
+    {
+      "epoch": 43.38,
+      "eval_exact_match": 0.7263056092843327,
+      "eval_exec": 0.758220502901354,
+      "eval_loss": 0.437593549489975,
+      "eval_runtime": 1140.2178,
+      "eval_samples_per_second": 0.907,
+      "step": 46500
+    },
+    {
+      "epoch": 43.38,
+      "learning_rate": 4.457742537313433e-05,
+      "loss": 0.0002,
+      "step": 46504
+    },
+    {
+      "epoch": 43.38,
+      "learning_rate": 4.4576958955223885e-05,
+      "loss": 0.0,
+      "step": 46508
+    },
+    {
+      "epoch": 43.39,
+      "learning_rate": 4.457649253731343e-05,
+      "loss": 0.0,
+      "step": 46512
+    },
+    {
+      "epoch": 43.39,
+      "learning_rate": 4.457602611940299e-05,
+      "loss": 0.0004,
+      "step": 46516
+    },
+    {
+      "epoch": 43.4,
+      "learning_rate": 4.457555970149254e-05,
+      "loss": 0.0001,
+      "step": 46520
+    },
+    {
+      "epoch": 43.4,
+      "learning_rate": 4.457509328358209e-05,
+      "loss": 0.0053,
+      "step": 46524
+    },
+    {
+      "epoch": 43.4,
+      "learning_rate": 4.4574626865671646e-05,
+      "loss": 0.0002,
+      "step": 46528
+    },
+    {
+      "epoch": 43.41,
+      "learning_rate": 4.45741604477612e-05,
+      "loss": 0.0,
+      "step": 46532
+    },
+    {
+      "epoch": 43.41,
+      "learning_rate": 4.457369402985075e-05,
+      "loss": 0.0007,
+      "step": 46536
+    },
+    {
+      "epoch": 43.41,
+      "learning_rate": 4.45732276119403e-05,
+      "loss": 0.0001,
+      "step": 46540
+    },
+    {
+      "epoch": 43.42,
+      "learning_rate": 4.457276119402985e-05,
+      "loss": 0.0001,
+      "step": 46544
+    },
+    {
+      "epoch": 43.42,
+      "learning_rate": 4.457229477611941e-05,
+      "loss": 0.0,
+      "step": 46548
+    },
+    {
+      "epoch": 43.43,
+      "learning_rate": 4.4571828358208955e-05,
+      "loss": 0.0,
+      "step": 46552
+    },
+    {
+      "epoch": 43.43,
+      "learning_rate": 4.457136194029851e-05,
+      "loss": 0.0051,
+      "step": 46556
+    },
+    {
+      "epoch": 43.43,
+      "learning_rate": 4.4570895522388065e-05,
+      "loss": 0.0003,
+      "step": 46560
+    },
+    {
+      "epoch": 43.44,
+      "learning_rate": 4.457042910447761e-05,
+      "loss": 0.0001,
+      "step": 46564
+    },
+    {
+      "epoch": 43.44,
+      "learning_rate": 4.456996268656717e-05,
+      "loss": 0.0003,
+      "step": 46568
+    },
+    {
+      "epoch": 43.44,
+      "learning_rate": 4.4569496268656716e-05,
+      "loss": 0.0001,
+      "step": 46572
+    },
+    {
+      "epoch": 43.45,
+      "learning_rate": 4.456902985074627e-05,
+      "loss": 0.0,
+      "step": 46576
+    },
+    {
+      "epoch": 43.45,
+      "learning_rate": 4.4568563432835826e-05,
+      "loss": 0.0,
+      "step": 46580
+    },
+    {
+      "epoch": 43.46,
+      "learning_rate": 4.4568097014925374e-05,
+      "loss": 0.0002,
+      "step": 46584
+    },
+    {
+      "epoch": 43.46,
+      "learning_rate": 4.456763059701493e-05,
+      "loss": 0.0,
+      "step": 46588
+    },
+    {
+      "epoch": 43.46,
+      "learning_rate": 4.4567164179104484e-05,
+      "loss": 0.0,
+      "step": 46592
+    },
+    {
+      "epoch": 43.47,
+      "learning_rate": 4.456669776119403e-05,
+      "loss": 0.0,
+      "step": 46596
+    },
+    {
+      "epoch": 43.47,
+      "learning_rate": 4.456623134328358e-05,
+      "loss": 0.0,
+      "step": 46600
+    },
+    {
+      "epoch": 43.47,
+      "learning_rate": 4.4565764925373135e-05,
+      "loss": 0.0,
+      "step": 46604
+    },
+    {
+      "epoch": 43.48,
+      "learning_rate": 4.456529850746269e-05,
+      "loss": 0.0005,
+      "step": 46608
+    },
+    {
+      "epoch": 43.48,
+      "learning_rate": 4.456483208955224e-05,
+      "loss": 0.0,
+      "step": 46612
+    },
+    {
+      "epoch": 43.49,
+      "learning_rate": 4.456436567164179e-05,
+      "loss": 0.0,
+      "step": 46616
+    },
+    {
+      "epoch": 43.49,
+      "learning_rate": 4.456389925373135e-05,
+      "loss": 0.0001,
+      "step": 46620
+    },
+    {
+      "epoch": 43.49,
+      "learning_rate": 4.4563432835820896e-05,
+      "loss": 0.0,
+      "step": 46624
+    },
+    {
+      "epoch": 43.5,
+      "learning_rate": 4.456296641791045e-05,
+      "loss": 0.0015,
+      "step": 46628
+    },
+    {
+      "epoch": 43.5,
+      "learning_rate": 4.45625e-05,
+      "loss": 0.0,
+      "step": 46632
+    },
+    {
+      "epoch": 43.5,
+      "learning_rate": 4.4562033582089554e-05,
+      "loss": 0.0,
+      "step": 46636
+    },
+    {
+      "epoch": 43.51,
+      "learning_rate": 4.456156716417911e-05,
+      "loss": 0.0002,
+      "step": 46640
+    },
+    {
+      "epoch": 43.51,
+      "learning_rate": 4.456110074626866e-05,
+      "loss": 0.0,
+      "step": 46644
+    },
+    {
+      "epoch": 43.51,
+      "learning_rate": 4.4560634328358205e-05,
+      "loss": 0.0002,
+      "step": 46648
+    },
+    {
+      "epoch": 43.52,
+      "learning_rate": 4.456016791044777e-05,
+      "loss": 0.0001,
+      "step": 46652
+    },
+    {
+      "epoch": 43.52,
+      "learning_rate": 4.4559701492537315e-05,
+      "loss": 0.0001,
+      "step": 46656
+    },
+    {
+      "epoch": 43.53,
+      "learning_rate": 4.455923507462687e-05,
+      "loss": 0.0,
+      "step": 46660
+    },
+    {
+      "epoch": 43.53,
+      "learning_rate": 4.455876865671642e-05,
+      "loss": 0.0002,
+      "step": 46664
+    },
+    {
+      "epoch": 43.53,
+      "learning_rate": 4.455830223880597e-05,
+      "loss": 0.0,
+      "step": 46668
+    },
+    {
+      "epoch": 43.54,
+      "learning_rate": 4.455783582089553e-05,
+      "loss": 0.0003,
+      "step": 46672
+    },
+    {
+      "epoch": 43.54,
+      "learning_rate": 4.4557369402985076e-05,
+      "loss": 0.0001,
+      "step": 46676
+    },
+    {
+      "epoch": 43.54,
+      "learning_rate": 4.455690298507463e-05,
+      "loss": 0.0001,
+      "step": 46680
+    },
+    {
+      "epoch": 43.55,
+      "learning_rate": 4.4556436567164186e-05,
+      "loss": 0.0,
+      "step": 46684
+    },
+    {
+      "epoch": 43.55,
+      "learning_rate": 4.4555970149253734e-05,
+      "loss": 0.0,
+      "step": 46688
+    },
+    {
+      "epoch": 43.56,
+      "learning_rate": 4.455550373134328e-05,
+      "loss": 0.0007,
+      "step": 46692
+    },
+    {
+      "epoch": 43.56,
+      "learning_rate": 4.4555037313432844e-05,
+      "loss": 0.0042,
+      "step": 46696
+    },
+    {
+      "epoch": 43.56,
+      "learning_rate": 4.455457089552239e-05,
+      "loss": 0.0004,
+      "step": 46700
+    },
+    {
+      "epoch": 43.57,
+      "learning_rate": 4.455410447761194e-05,
+      "loss": 0.0001,
+      "step": 46704
+    },
+    {
+      "epoch": 43.57,
+      "learning_rate": 4.4553638059701495e-05,
+      "loss": 0.0,
+      "step": 46708
+    },
+    {
+      "epoch": 43.57,
+      "learning_rate": 4.455317164179105e-05,
+      "loss": 0.0002,
+      "step": 46712
+    },
+    {
+      "epoch": 43.58,
+      "learning_rate": 4.45527052238806e-05,
+      "loss": 0.0013,
+      "step": 46716
+    },
+    {
+      "epoch": 43.58,
+      "learning_rate": 4.455223880597015e-05,
+      "loss": 0.0001,
+      "step": 46720
+    },
+    {
+      "epoch": 43.59,
+      "learning_rate": 4.45517723880597e-05,
+      "loss": 0.0,
+      "step": 46724
+    },
+    {
+      "epoch": 43.59,
+      "learning_rate": 4.4551305970149256e-05,
+      "loss": 0.0,
+      "step": 46728
+    },
+    {
+      "epoch": 43.59,
+      "learning_rate": 4.455083955223881e-05,
+      "loss": 0.0,
+      "step": 46732
+    },
+    {
+      "epoch": 43.6,
+      "learning_rate": 4.455037313432836e-05,
+      "loss": 0.0001,
+      "step": 46736
+    },
+    {
+      "epoch": 43.6,
+      "learning_rate": 4.4549906716417914e-05,
+      "loss": 0.0001,
+      "step": 46740
+    },
+    {
+      "epoch": 43.6,
+      "learning_rate": 4.454944029850747e-05,
+      "loss": 0.0,
+      "step": 46744
+    },
+    {
+      "epoch": 43.61,
+      "learning_rate": 4.454897388059702e-05,
+      "loss": 0.0,
+      "step": 46748
+    },
+    {
+      "epoch": 43.61,
+      "learning_rate": 4.4548507462686565e-05,
+      "loss": 0.0,
+      "step": 46752
+    },
+    {
+      "epoch": 43.62,
+      "learning_rate": 4.454804104477612e-05,
+      "loss": 0.0001,
+      "step": 46756
+    },
+    {
+      "epoch": 43.62,
+      "learning_rate": 4.4547574626865675e-05,
+      "loss": 0.0,
+      "step": 46760
+    },
+    {
+      "epoch": 43.62,
+      "learning_rate": 4.454710820895522e-05,
+      "loss": 0.0,
+      "step": 46764
+    },
+    {
+      "epoch": 43.63,
+      "learning_rate": 4.454664179104478e-05,
+      "loss": 0.0,
+      "step": 46768
+    },
+    {
+      "epoch": 43.63,
+      "learning_rate": 4.454617537313433e-05,
+      "loss": 0.0,
+      "step": 46772
+    },
+    {
+      "epoch": 43.63,
+      "learning_rate": 4.454570895522388e-05,
+      "loss": 0.0004,
+      "step": 46776
+    },
+    {
+      "epoch": 43.64,
+      "learning_rate": 4.4545242537313436e-05,
+      "loss": 0.0009,
+      "step": 46780
+    },
+    {
+      "epoch": 43.64,
+      "learning_rate": 4.4544776119402984e-05,
+      "loss": 0.0,
+      "step": 46784
+    },
+    {
+      "epoch": 43.65,
+      "learning_rate": 4.454430970149254e-05,
+      "loss": 0.0,
+      "step": 46788
+    },
+    {
+      "epoch": 43.65,
+      "learning_rate": 4.4543843283582094e-05,
+      "loss": 0.0025,
+      "step": 46792
+    },
+    {
+      "epoch": 43.65,
+      "learning_rate": 4.454337686567164e-05,
+      "loss": 0.0,
+      "step": 46796
+    },
+    {
+      "epoch": 43.66,
+      "learning_rate": 4.45429104477612e-05,
+      "loss": 0.002,
+      "step": 46800
+    },
+    {
+      "epoch": 43.66,
+      "learning_rate": 4.454244402985075e-05,
+      "loss": 0.0,
+      "step": 46804
+    },
+    {
+      "epoch": 43.66,
+      "learning_rate": 4.45419776119403e-05,
+      "loss": 0.0002,
+      "step": 46808
+    },
+    {
+      "epoch": 43.67,
+      "learning_rate": 4.454151119402985e-05,
+      "loss": 0.0,
+      "step": 46812
+    },
+    {
+      "epoch": 43.67,
+      "learning_rate": 4.45410447761194e-05,
+      "loss": 0.0014,
+      "step": 46816
+    },
+    {
+      "epoch": 43.68,
+      "learning_rate": 4.454057835820896e-05,
+      "loss": 0.0003,
+      "step": 46820
+    },
+    {
+      "epoch": 43.68,
+      "learning_rate": 4.454011194029851e-05,
+      "loss": 0.0001,
+      "step": 46824
+    },
+    {
+      "epoch": 43.68,
+      "learning_rate": 4.453964552238806e-05,
+      "loss": 0.0,
+      "step": 46828
+    },
+    {
+      "epoch": 43.69,
+      "learning_rate": 4.4539179104477616e-05,
+      "loss": 0.0008,
+      "step": 46832
+    },
+    {
+      "epoch": 43.69,
+      "learning_rate": 4.453871268656717e-05,
+      "loss": 0.0144,
+      "step": 46836
+    },
+    {
+      "epoch": 43.69,
+      "learning_rate": 4.453824626865672e-05,
+      "loss": 0.0022,
+      "step": 46840
+    },
+    {
+      "epoch": 43.7,
+      "learning_rate": 4.453777985074627e-05,
+      "loss": 0.0,
+      "step": 46844
+    },
+    {
+      "epoch": 43.7,
+      "learning_rate": 4.453731343283583e-05,
+      "loss": 0.0001,
+      "step": 46848
+    },
+    {
+      "epoch": 43.71,
+      "learning_rate": 4.4536847014925377e-05,
+      "loss": 0.0,
+      "step": 46852
+    },
+    {
+      "epoch": 43.71,
+      "learning_rate": 4.4536380597014925e-05,
+      "loss": 0.0012,
+      "step": 46856
+    },
+    {
+      "epoch": 43.71,
+      "learning_rate": 4.453591417910448e-05,
+      "loss": 0.0,
+      "step": 46860
+    },
+    {
+      "epoch": 43.72,
+      "learning_rate": 4.4535447761194035e-05,
+      "loss": 0.0001,
+      "step": 46864
+    },
+    {
+      "epoch": 43.72,
+      "learning_rate": 4.453498134328358e-05,
+      "loss": 0.0003,
+      "step": 46868
+    },
+    {
+      "epoch": 43.72,
+      "learning_rate": 4.453451492537314e-05,
+      "loss": 0.0,
+      "step": 46872
+    },
+    {
+      "epoch": 43.73,
+      "learning_rate": 4.4534048507462686e-05,
+      "loss": 0.0012,
+      "step": 46876
+    },
+    {
+      "epoch": 43.73,
+      "learning_rate": 4.453358208955224e-05,
+      "loss": 0.0,
+      "step": 46880
+    },
+    {
+      "epoch": 43.73,
+      "learning_rate": 4.4533115671641796e-05,
+      "loss": 0.0,
+      "step": 46884
+    },
+    {
+      "epoch": 43.74,
+      "learning_rate": 4.4532649253731344e-05,
+      "loss": 0.0,
+      "step": 46888
+    },
+    {
+      "epoch": 43.74,
+      "learning_rate": 4.45321828358209e-05,
+      "loss": 0.0001,
+      "step": 46892
+    },
+    {
+      "epoch": 43.75,
+      "learning_rate": 4.4531716417910453e-05,
+      "loss": 0.0,
+      "step": 46896
+    },
+    {
+      "epoch": 43.75,
+      "learning_rate": 4.453125e-05,
+      "loss": 0.0226,
+      "step": 46900
+    },
+    {
+      "epoch": 43.75,
+      "learning_rate": 4.453078358208955e-05,
+      "loss": 0.0001,
+      "step": 46904
+    },
+    {
+      "epoch": 43.76,
+      "learning_rate": 4.453031716417911e-05,
+      "loss": 0.0018,
+      "step": 46908
+    },
+    {
+      "epoch": 43.76,
+      "learning_rate": 4.452985074626866e-05,
+      "loss": 0.0001,
+      "step": 46912
+    },
+    {
+      "epoch": 43.76,
+      "learning_rate": 4.452938432835821e-05,
+      "loss": 0.0017,
+      "step": 46916
+    },
+    {
+      "epoch": 43.77,
+      "learning_rate": 4.452891791044776e-05,
+      "loss": 0.0,
+      "step": 46920
+    },
+    {
+      "epoch": 43.77,
+      "learning_rate": 4.452845149253732e-05,
+      "loss": 0.0002,
+      "step": 46924
+    },
+    {
+      "epoch": 43.78,
+      "learning_rate": 4.4527985074626866e-05,
+      "loss": 0.0,
+      "step": 46928
+    },
+    {
+      "epoch": 43.78,
+      "learning_rate": 4.452751865671642e-05,
+      "loss": 0.0,
+      "step": 46932
+    },
+    {
+      "epoch": 43.78,
+      "learning_rate": 4.452705223880597e-05,
+      "loss": 0.0019,
+      "step": 46936
+    },
+    {
+      "epoch": 43.79,
+      "learning_rate": 4.4526585820895524e-05,
+      "loss": 0.0,
+      "step": 46940
+    },
+    {
+      "epoch": 43.79,
+      "learning_rate": 4.452611940298508e-05,
+      "loss": 0.0003,
+      "step": 46944
+    },
+    {
+      "epoch": 43.79,
+      "learning_rate": 4.452565298507463e-05,
+      "loss": 0.0004,
+      "step": 46948
+    },
+    {
+      "epoch": 43.8,
+      "learning_rate": 4.452518656716418e-05,
+      "loss": 0.0,
+      "step": 46952
+    },
+    {
+      "epoch": 43.8,
+      "learning_rate": 4.4524720149253736e-05,
+      "loss": 0.0001,
+      "step": 46956
+    },
+    {
+      "epoch": 43.81,
+      "learning_rate": 4.4524253731343285e-05,
+      "loss": 0.0,
+      "step": 46960
+    },
+    {
+      "epoch": 43.81,
+      "learning_rate": 4.452378731343283e-05,
+      "loss": 0.0004,
+      "step": 46964
+    },
+    {
+      "epoch": 43.81,
+      "learning_rate": 4.4523320895522394e-05,
+      "loss": 0.0001,
+      "step": 46968
+    },
+    {
+      "epoch": 43.82,
+      "learning_rate": 4.452285447761194e-05,
+      "loss": 0.0,
+      "step": 46972
+    },
+    {
+      "epoch": 43.82,
+      "learning_rate": 4.452238805970149e-05,
+      "loss": 0.0,
+      "step": 46976
+    },
+    {
+      "epoch": 43.82,
+      "learning_rate": 4.4521921641791046e-05,
+      "loss": 0.0001,
+      "step": 46980
+    },
+    {
+      "epoch": 43.83,
+      "learning_rate": 4.45214552238806e-05,
+      "loss": 0.0084,
+      "step": 46984
+    },
+    {
+      "epoch": 43.83,
+      "learning_rate": 4.4520988805970155e-05,
+      "loss": 0.0007,
+      "step": 46988
+    },
+    {
+      "epoch": 43.84,
+      "learning_rate": 4.4520522388059703e-05,
+      "loss": 0.0,
+      "step": 46992
+    },
+    {
+      "epoch": 43.84,
+      "learning_rate": 4.452005597014925e-05,
+      "loss": 0.0,
+      "step": 46996
+    },
+    {
+      "epoch": 43.84,
+      "learning_rate": 4.451958955223881e-05,
+      "loss": 0.0,
+      "step": 47000
+    },
+    {
+      "epoch": 43.84,
+      "eval_exact_match": 0.7321083172147002,
+      "eval_exec": 0.7688588007736944,
+      "eval_loss": 0.4131454825401306,
+      "eval_runtime": 1115.6019,
+      "eval_samples_per_second": 0.927,
+      "step": 47000
+    },
+    {
+      "epoch": 43.85,
+      "learning_rate": 4.451912313432836e-05,
+      "loss": 0.0,
+      "step": 47004
+    },
+    {
+      "epoch": 43.85,
+      "learning_rate": 4.451865671641791e-05,
+      "loss": 0.0021,
+      "step": 47008
+    },
+    {
+      "epoch": 43.85,
+      "learning_rate": 4.4518190298507464e-05,
+      "loss": 0.002,
+      "step": 47012
+    },
+    {
+      "epoch": 43.86,
+      "learning_rate": 4.451772388059702e-05,
+      "loss": 0.0084,
+      "step": 47016
+    },
+    {
+      "epoch": 43.86,
+      "learning_rate": 4.451725746268657e-05,
+      "loss": 0.0002,
+      "step": 47020
+    },
+    {
+      "epoch": 43.87,
+      "learning_rate": 4.451679104477612e-05,
+      "loss": 0.0007,
+      "step": 47024
+    },
+    {
+      "epoch": 43.87,
+      "learning_rate": 4.451632462686568e-05,
+      "loss": 0.0007,
+      "step": 47028
+    },
+    {
+      "epoch": 43.87,
+      "learning_rate": 4.4515858208955225e-05,
+      "loss": 0.0001,
+      "step": 47032
+    },
+    {
+      "epoch": 43.88,
+      "learning_rate": 4.451539179104478e-05,
+      "loss": 0.0001,
+      "step": 47036
+    },
+    {
+      "epoch": 43.88,
+      "learning_rate": 4.451492537313433e-05,
+      "loss": 0.0,
+      "step": 47040
+    },
+    {
+      "epoch": 43.88,
+      "learning_rate": 4.4514458955223883e-05,
+      "loss": 0.0002,
+      "step": 47044
+    },
+    {
+      "epoch": 43.89,
+      "learning_rate": 4.451399253731344e-05,
+      "loss": 0.0051,
+      "step": 47048
+    },
+    {
+      "epoch": 43.89,
+      "learning_rate": 4.4513526119402986e-05,
+      "loss": 0.0004,
+      "step": 47052
+    },
+    {
+      "epoch": 43.9,
+      "learning_rate": 4.4513059701492535e-05,
+      "loss": 0.0001,
+      "step": 47056
+    },
+    {
+      "epoch": 43.9,
+      "learning_rate": 4.4512593283582096e-05,
+      "loss": 0.0,
+      "step": 47060
+    },
+    {
+      "epoch": 43.9,
+      "learning_rate": 4.4512126865671644e-05,
+      "loss": 0.0005,
+      "step": 47064
+    },
+    {
+      "epoch": 43.91,
+      "learning_rate": 4.451166044776119e-05,
+      "loss": 0.0,
+      "step": 47068
+    },
+    {
+      "epoch": 43.91,
+      "learning_rate": 4.451119402985075e-05,
+      "loss": 0.0033,
+      "step": 47072
+    },
+    {
+      "epoch": 43.91,
+      "learning_rate": 4.45107276119403e-05,
+      "loss": 0.0003,
+      "step": 47076
+    },
+    {
+      "epoch": 43.92,
+      "learning_rate": 4.451026119402985e-05,
+      "loss": 0.0016,
+      "step": 47080
+    },
+    {
+      "epoch": 43.92,
+      "learning_rate": 4.4509794776119405e-05,
+      "loss": 0.0,
+      "step": 47084
+    },
+    {
+      "epoch": 43.93,
+      "learning_rate": 4.450932835820896e-05,
+      "loss": 0.001,
+      "step": 47088
+    },
+    {
+      "epoch": 43.93,
+      "learning_rate": 4.450886194029851e-05,
+      "loss": 0.0,
+      "step": 47092
+    },
+    {
+      "epoch": 43.93,
+      "learning_rate": 4.450839552238806e-05,
+      "loss": 0.0015,
+      "step": 47096
+    },
+    {
+      "epoch": 43.94,
+      "learning_rate": 4.450792910447761e-05,
+      "loss": 0.0,
+      "step": 47100
+    },
+    {
+      "epoch": 43.94,
+      "learning_rate": 4.4507462686567166e-05,
+      "loss": 0.001,
+      "step": 47104
+    },
+    {
+      "epoch": 43.94,
+      "learning_rate": 4.450699626865672e-05,
+      "loss": 0.0,
+      "step": 47108
+    },
+    {
+      "epoch": 43.95,
+      "learning_rate": 4.450652985074627e-05,
+      "loss": 0.0012,
+      "step": 47112
+    },
+    {
+      "epoch": 43.95,
+      "learning_rate": 4.450606343283582e-05,
+      "loss": 0.0,
+      "step": 47116
+    },
+    {
+      "epoch": 43.96,
+      "learning_rate": 4.450559701492538e-05,
+      "loss": 0.0,
+      "step": 47120
+    },
+    {
+      "epoch": 43.96,
+      "learning_rate": 4.450513059701493e-05,
+      "loss": 0.0018,
+      "step": 47124
+    },
+    {
+      "epoch": 43.96,
+      "learning_rate": 4.4504664179104475e-05,
+      "loss": 0.0002,
+      "step": 47128
+    },
+    {
+      "epoch": 43.97,
+      "learning_rate": 4.450419776119403e-05,
+      "loss": 0.0,
+      "step": 47132
+    },
+    {
+      "epoch": 43.97,
+      "learning_rate": 4.4503731343283585e-05,
+      "loss": 0.001,
+      "step": 47136
+    },
+    {
+      "epoch": 43.97,
+      "learning_rate": 4.4503264925373133e-05,
+      "loss": 0.0003,
+      "step": 47140
+    },
+    {
+      "epoch": 43.98,
+      "learning_rate": 4.450279850746269e-05,
+      "loss": 0.0081,
+      "step": 47144
+    },
+    {
+      "epoch": 43.98,
+      "learning_rate": 4.450233208955224e-05,
+      "loss": 0.0001,
+      "step": 47148
+    },
+    {
+      "epoch": 43.98,
+      "learning_rate": 4.45018656716418e-05,
+      "loss": 0.0001,
+      "step": 47152
+    },
+    {
+      "epoch": 43.99,
+      "learning_rate": 4.4501399253731346e-05,
+      "loss": 0.0,
+      "step": 47156
+    },
+    {
+      "epoch": 43.99,
+      "learning_rate": 4.4500932835820894e-05,
+      "loss": 0.0002,
+      "step": 47160
+    },
+    {
+      "epoch": 44.0,
+      "learning_rate": 4.450046641791045e-05,
+      "loss": 0.0,
+      "step": 47164
+    },
+    {
+      "epoch": 44.0,
+      "learning_rate": 4.4500000000000004e-05,
+      "loss": 0.0001,
+      "step": 47168
+    },
+    {
+      "epoch": 44.0,
+      "learning_rate": 4.449953358208955e-05,
+      "loss": 0.0001,
+      "step": 47172
+    },
+    {
+      "epoch": 44.01,
+      "learning_rate": 4.449906716417911e-05,
+      "loss": 0.0,
+      "step": 47176
+    },
+    {
+      "epoch": 44.01,
+      "learning_rate": 4.449860074626866e-05,
+      "loss": 0.0001,
+      "step": 47180
+    },
+    {
+      "epoch": 44.01,
+      "learning_rate": 4.449813432835821e-05,
+      "loss": 0.0017,
+      "step": 47184
+    },
+    {
+      "epoch": 44.02,
+      "learning_rate": 4.4497667910447765e-05,
+      "loss": 0.0003,
+      "step": 47188
+    },
+    {
+      "epoch": 44.02,
+      "learning_rate": 4.449720149253731e-05,
+      "loss": 0.0002,
+      "step": 47192
+    },
+    {
+      "epoch": 44.03,
+      "learning_rate": 4.449673507462687e-05,
+      "loss": 0.0,
+      "step": 47196
+    },
+    {
+      "epoch": 44.03,
+      "learning_rate": 4.449626865671642e-05,
+      "loss": 0.0003,
+      "step": 47200
+    },
+    {
+      "epoch": 44.03,
+      "learning_rate": 4.449580223880597e-05,
+      "loss": 0.0058,
+      "step": 47204
+    },
+    {
+      "epoch": 44.04,
+      "learning_rate": 4.4495335820895526e-05,
+      "loss": 0.0,
+      "step": 47208
+    },
+    {
+      "epoch": 44.04,
+      "learning_rate": 4.449486940298508e-05,
+      "loss": 0.0086,
+      "step": 47212
+    },
+    {
+      "epoch": 44.04,
+      "learning_rate": 4.449440298507463e-05,
+      "loss": 0.001,
+      "step": 47216
+    },
+    {
+      "epoch": 44.05,
+      "learning_rate": 4.449393656716418e-05,
+      "loss": 0.0016,
+      "step": 47220
+    },
+    {
+      "epoch": 44.05,
+      "learning_rate": 4.449347014925373e-05,
+      "loss": 0.0015,
+      "step": 47224
+    },
+    {
+      "epoch": 44.06,
+      "learning_rate": 4.449300373134329e-05,
+      "loss": 0.0006,
+      "step": 47228
+    },
+    {
+      "epoch": 44.06,
+      "learning_rate": 4.4492537313432835e-05,
+      "loss": 0.0002,
+      "step": 47232
+    },
+    {
+      "epoch": 44.06,
+      "learning_rate": 4.449207089552239e-05,
+      "loss": 0.0002,
+      "step": 47236
+    },
+    {
+      "epoch": 44.07,
+      "learning_rate": 4.4491604477611945e-05,
+      "loss": 0.0002,
+      "step": 47240
+    },
+    {
+      "epoch": 44.07,
+      "learning_rate": 4.449113805970149e-05,
+      "loss": 0.0001,
+      "step": 47244
+    },
+    {
+      "epoch": 44.07,
+      "learning_rate": 4.449067164179105e-05,
+      "loss": 0.0001,
+      "step": 47248
+    },
+    {
+      "epoch": 44.08,
+      "learning_rate": 4.4490205223880596e-05,
+      "loss": 0.002,
+      "step": 47252
+    },
+    {
+      "epoch": 44.08,
+      "learning_rate": 4.448973880597015e-05,
+      "loss": 0.0001,
+      "step": 47256
+    },
+    {
+      "epoch": 44.09,
+      "learning_rate": 4.4489272388059706e-05,
+      "loss": 0.0001,
+      "step": 47260
+    },
+    {
+      "epoch": 44.09,
+      "learning_rate": 4.4488805970149254e-05,
+      "loss": 0.0001,
+      "step": 47264
+    },
+    {
+      "epoch": 44.09,
+      "learning_rate": 4.448833955223881e-05,
+      "loss": 0.0,
+      "step": 47268
+    },
+    {
+      "epoch": 44.1,
+      "learning_rate": 4.4487873134328364e-05,
+      "loss": 0.0,
+      "step": 47272
+    },
+    {
+      "epoch": 44.1,
+      "learning_rate": 4.448740671641791e-05,
+      "loss": 0.0034,
+      "step": 47276
+    },
+    {
+      "epoch": 44.1,
+      "learning_rate": 4.448694029850746e-05,
+      "loss": 0.0004,
+      "step": 47280
+    },
+    {
+      "epoch": 44.11,
+      "learning_rate": 4.4486473880597015e-05,
+      "loss": 0.0,
+      "step": 47284
+    },
+    {
+      "epoch": 44.11,
+      "learning_rate": 4.448600746268657e-05,
+      "loss": 0.0001,
+      "step": 47288
+    },
+    {
+      "epoch": 44.12,
+      "learning_rate": 4.448554104477612e-05,
+      "loss": 0.0002,
+      "step": 47292
+    },
+    {
+      "epoch": 44.12,
+      "learning_rate": 4.448507462686567e-05,
+      "loss": 0.001,
+      "step": 47296
+    },
+    {
+      "epoch": 44.12,
+      "learning_rate": 4.448460820895523e-05,
+      "loss": 0.0006,
+      "step": 47300
+    },
+    {
+      "epoch": 44.13,
+      "learning_rate": 4.4484141791044776e-05,
+      "loss": 0.0003,
+      "step": 47304
+    },
+    {
+      "epoch": 44.13,
+      "learning_rate": 4.448367537313433e-05,
+      "loss": 0.0,
+      "step": 47308
+    },
+    {
+      "epoch": 44.13,
+      "learning_rate": 4.448320895522388e-05,
+      "loss": 0.0082,
+      "step": 47312
+    },
+    {
+      "epoch": 44.14,
+      "learning_rate": 4.448274253731344e-05,
+      "loss": 0.0,
+      "step": 47316
+    },
+    {
+      "epoch": 44.14,
+      "learning_rate": 4.448227611940299e-05,
+      "loss": 0.0001,
+      "step": 47320
+    },
+    {
+      "epoch": 44.15,
+      "learning_rate": 4.448180970149254e-05,
+      "loss": 0.0001,
+      "step": 47324
+    },
+    {
+      "epoch": 44.15,
+      "learning_rate": 4.448134328358209e-05,
+      "loss": 0.0,
+      "step": 47328
+    },
+    {
+      "epoch": 44.15,
+      "learning_rate": 4.448087686567165e-05,
+      "loss": 0.0004,
+      "step": 47332
+    },
+    {
+      "epoch": 44.16,
+      "learning_rate": 4.4480410447761195e-05,
+      "loss": 0.0002,
+      "step": 47336
+    },
+    {
+      "epoch": 44.16,
+      "learning_rate": 4.447994402985075e-05,
+      "loss": 0.0002,
+      "step": 47340
+    },
+    {
+      "epoch": 44.16,
+      "learning_rate": 4.44794776119403e-05,
+      "loss": 0.0,
+      "step": 47344
+    },
+    {
+      "epoch": 44.17,
+      "learning_rate": 4.447901119402985e-05,
+      "loss": 0.0,
+      "step": 47348
+    },
+    {
+      "epoch": 44.17,
+      "learning_rate": 4.447854477611941e-05,
+      "loss": 0.0047,
+      "step": 47352
+    },
+    {
+      "epoch": 44.18,
+      "learning_rate": 4.4478078358208956e-05,
+      "loss": 0.0003,
+      "step": 47356
+    },
+    {
+      "epoch": 44.18,
+      "learning_rate": 4.447761194029851e-05,
+      "loss": 0.0,
+      "step": 47360
+    },
+    {
+      "epoch": 44.18,
+      "learning_rate": 4.4477145522388066e-05,
+      "loss": 0.0006,
+      "step": 47364
+    },
+    {
+      "epoch": 44.19,
+      "learning_rate": 4.4476679104477614e-05,
+      "loss": 0.0,
+      "step": 47368
+    },
+    {
+      "epoch": 44.19,
+      "learning_rate": 4.447621268656716e-05,
+      "loss": 0.0005,
+      "step": 47372
+    },
+    {
+      "epoch": 44.19,
+      "learning_rate": 4.4475746268656724e-05,
+      "loss": 0.002,
+      "step": 47376
+    },
+    {
+      "epoch": 44.2,
+      "learning_rate": 4.447527985074627e-05,
+      "loss": 0.0,
+      "step": 47380
+    },
+    {
+      "epoch": 44.2,
+      "learning_rate": 4.447481343283582e-05,
+      "loss": 0.0141,
+      "step": 47384
+    },
+    {
+      "epoch": 44.21,
+      "learning_rate": 4.4474347014925375e-05,
+      "loss": 0.0001,
+      "step": 47388
+    },
+    {
+      "epoch": 44.21,
+      "learning_rate": 4.447388059701493e-05,
+      "loss": 0.0002,
+      "step": 47392
+    },
+    {
+      "epoch": 44.21,
+      "learning_rate": 4.447341417910448e-05,
+      "loss": 0.0001,
+      "step": 47396
+    },
+    {
+      "epoch": 44.22,
+      "learning_rate": 4.447294776119403e-05,
+      "loss": 0.0,
+      "step": 47400
+    },
+    {
+      "epoch": 44.22,
+      "learning_rate": 4.447248134328358e-05,
+      "loss": 0.0027,
+      "step": 47404
+    },
+    {
+      "epoch": 44.22,
+      "learning_rate": 4.4472014925373136e-05,
+      "loss": 0.0,
+      "step": 47408
+    },
+    {
+      "epoch": 44.23,
+      "learning_rate": 4.447154850746269e-05,
+      "loss": 0.0005,
+      "step": 47412
+    },
+    {
+      "epoch": 44.23,
+      "learning_rate": 4.447108208955224e-05,
+      "loss": 0.0,
+      "step": 47416
+    },
+    {
+      "epoch": 44.24,
+      "learning_rate": 4.4470615671641794e-05,
+      "loss": 0.0002,
+      "step": 47420
+    },
+    {
+      "epoch": 44.24,
+      "learning_rate": 4.447014925373135e-05,
+      "loss": 0.0,
+      "step": 47424
+    },
+    {
+      "epoch": 44.24,
+      "learning_rate": 4.44696828358209e-05,
+      "loss": 0.0004,
+      "step": 47428
+    },
+    {
+      "epoch": 44.25,
+      "learning_rate": 4.4469216417910445e-05,
+      "loss": 0.0,
+      "step": 47432
+    },
+    {
+      "epoch": 44.25,
+      "learning_rate": 4.446875e-05,
+      "loss": 0.001,
+      "step": 47436
+    },
+    {
+      "epoch": 44.25,
+      "learning_rate": 4.4468283582089555e-05,
+      "loss": 0.0,
+      "step": 47440
+    },
+    {
+      "epoch": 44.26,
+      "learning_rate": 4.44678171641791e-05,
+      "loss": 0.0,
+      "step": 47444
+    },
+    {
+      "epoch": 44.26,
+      "learning_rate": 4.446735074626866e-05,
+      "loss": 0.0,
+      "step": 47448
+    },
+    {
+      "epoch": 44.26,
+      "learning_rate": 4.446688432835821e-05,
+      "loss": 0.0012,
+      "step": 47452
+    },
+    {
+      "epoch": 44.27,
+      "learning_rate": 4.446641791044776e-05,
+      "loss": 0.0001,
+      "step": 47456
+    },
+    {
+      "epoch": 44.27,
+      "learning_rate": 4.4465951492537316e-05,
+      "loss": 0.0,
+      "step": 47460
+    },
+    {
+      "epoch": 44.28,
+      "learning_rate": 4.4465485074626864e-05,
+      "loss": 0.0,
+      "step": 47464
+    },
+    {
+      "epoch": 44.28,
+      "learning_rate": 4.446501865671642e-05,
+      "loss": 0.0001,
+      "step": 47468
+    },
+    {
+      "epoch": 44.28,
+      "learning_rate": 4.4464552238805974e-05,
+      "loss": 0.0012,
+      "step": 47472
+    },
+    {
+      "epoch": 44.29,
+      "learning_rate": 4.446408582089552e-05,
+      "loss": 0.0035,
+      "step": 47476
+    },
+    {
+      "epoch": 44.29,
+      "learning_rate": 4.446361940298508e-05,
+      "loss": 0.007,
+      "step": 47480
+    },
+    {
+      "epoch": 44.29,
+      "learning_rate": 4.446315298507463e-05,
+      "loss": 0.0001,
+      "step": 47484
+    },
+    {
+      "epoch": 44.3,
+      "learning_rate": 4.446268656716418e-05,
+      "loss": 0.0,
+      "step": 47488
+    },
+    {
+      "epoch": 44.3,
+      "learning_rate": 4.4462220149253735e-05,
+      "loss": 0.0,
+      "step": 47492
+    },
+    {
+      "epoch": 44.31,
+      "learning_rate": 4.446175373134328e-05,
+      "loss": 0.0,
+      "step": 47496
+    },
+    {
+      "epoch": 44.31,
+      "learning_rate": 4.446128731343284e-05,
+      "loss": 0.0004,
+      "step": 47500
+    },
+    {
+      "epoch": 44.31,
+      "eval_exact_match": 0.7330754352030948,
+      "eval_exec": 0.7707930367504836,
+      "eval_loss": 0.42296141386032104,
+      "eval_runtime": 1217.964,
+      "eval_samples_per_second": 0.849,
+      "step": 47500
+    },
+    {
+      "epoch": 44.31,
+      "learning_rate": 4.446082089552239e-05,
+      "loss": 0.0009,
+      "step": 47504
+    },
+    {
+      "epoch": 44.32,
+      "learning_rate": 4.446035447761194e-05,
+      "loss": 0.0001,
+      "step": 47508
+    },
+    {
+      "epoch": 44.32,
+      "learning_rate": 4.4459888059701496e-05,
+      "loss": 0.0001,
+      "step": 47512
+    },
+    {
+      "epoch": 44.32,
+      "learning_rate": 4.445942164179105e-05,
+      "loss": 0.0003,
+      "step": 47516
+    },
+    {
+      "epoch": 44.33,
+      "learning_rate": 4.44589552238806e-05,
+      "loss": 0.0,
+      "step": 47520
+    },
+    {
+      "epoch": 44.33,
+      "learning_rate": 4.445848880597015e-05,
+      "loss": 0.0,
+      "step": 47524
+    },
+    {
+      "epoch": 44.34,
+      "learning_rate": 4.445802238805971e-05,
+      "loss": 0.0,
+      "step": 47528
+    },
+    {
+      "epoch": 44.34,
+      "learning_rate": 4.445755597014926e-05,
+      "loss": 0.0001,
+      "step": 47532
+    },
+    {
+      "epoch": 44.34,
+      "learning_rate": 4.4457089552238805e-05,
+      "loss": 0.0001,
+      "step": 47536
+    },
+    {
+      "epoch": 44.35,
+      "learning_rate": 4.445662313432836e-05,
+      "loss": 0.0001,
+      "step": 47540
+    },
+    {
+      "epoch": 44.35,
+      "learning_rate": 4.4456156716417915e-05,
+      "loss": 0.0,
+      "step": 47544
+    },
+    {
+      "epoch": 44.35,
+      "learning_rate": 4.445569029850746e-05,
+      "loss": 0.0001,
+      "step": 47548
+    },
+    {
+      "epoch": 44.36,
+      "learning_rate": 4.445522388059702e-05,
+      "loss": 0.0,
+      "step": 47552
+    },
+    {
+      "epoch": 44.36,
+      "learning_rate": 4.4454757462686566e-05,
+      "loss": 0.0001,
+      "step": 47556
+    },
+    {
+      "epoch": 44.37,
+      "learning_rate": 4.445429104477612e-05,
+      "loss": 0.0001,
+      "step": 47560
+    },
+    {
+      "epoch": 44.37,
+      "learning_rate": 4.4453824626865676e-05,
+      "loss": 0.0001,
+      "step": 47564
+    },
+    {
+      "epoch": 44.37,
+      "learning_rate": 4.4453358208955224e-05,
+      "loss": 0.0001,
+      "step": 47568
+    },
+    {
+      "epoch": 44.38,
+      "learning_rate": 4.445289179104478e-05,
+      "loss": 0.0013,
+      "step": 47572
+    },
+    {
+      "epoch": 44.38,
+      "learning_rate": 4.4452425373134334e-05,
+      "loss": 0.0,
+      "step": 47576
+    },
+    {
+      "epoch": 44.38,
+      "learning_rate": 4.445195895522388e-05,
+      "loss": 0.0,
+      "step": 47580
+    },
+    {
+      "epoch": 44.39,
+      "learning_rate": 4.445149253731343e-05,
+      "loss": 0.0,
+      "step": 47584
+    },
+    {
+      "epoch": 44.39,
+      "learning_rate": 4.445102611940299e-05,
+      "loss": 0.0,
+      "step": 47588
+    },
+    {
+      "epoch": 44.4,
+      "learning_rate": 4.445055970149254e-05,
+      "loss": 0.0015,
+      "step": 47592
+    },
+    {
+      "epoch": 44.4,
+      "learning_rate": 4.445009328358209e-05,
+      "loss": 0.0003,
+      "step": 47596
+    },
+    {
+      "epoch": 44.4,
+      "learning_rate": 4.444962686567164e-05,
+      "loss": 0.0001,
+      "step": 47600
+    },
+    {
+      "epoch": 44.41,
+      "learning_rate": 4.44491604477612e-05,
+      "loss": 0.0001,
+      "step": 47604
+    },
+    {
+      "epoch": 44.41,
+      "learning_rate": 4.4448694029850746e-05,
+      "loss": 0.0,
+      "step": 47608
+    },
+    {
+      "epoch": 44.41,
+      "learning_rate": 4.44482276119403e-05,
+      "loss": 0.0,
+      "step": 47612
+    },
+    {
+      "epoch": 44.42,
+      "learning_rate": 4.444776119402985e-05,
+      "loss": 0.0001,
+      "step": 47616
+    },
+    {
+      "epoch": 44.42,
+      "learning_rate": 4.4447294776119404e-05,
+      "loss": 0.0001,
+      "step": 47620
+    },
+    {
+      "epoch": 44.43,
+      "learning_rate": 4.444682835820896e-05,
+      "loss": 0.0068,
+      "step": 47624
+    },
+    {
+      "epoch": 44.43,
+      "learning_rate": 4.444636194029851e-05,
+      "loss": 0.0001,
+      "step": 47628
+    },
+    {
+      "epoch": 44.43,
+      "learning_rate": 4.444589552238806e-05,
+      "loss": 0.0024,
+      "step": 47632
+    },
+    {
+      "epoch": 44.44,
+      "learning_rate": 4.444542910447762e-05,
+      "loss": 0.0,
+      "step": 47636
+    },
+    {
+      "epoch": 44.44,
+      "learning_rate": 4.4444962686567165e-05,
+      "loss": 0.0,
+      "step": 47640
+    },
+    {
+      "epoch": 44.44,
+      "learning_rate": 4.444449626865672e-05,
+      "loss": 0.0049,
+      "step": 47644
+    },
+    {
+      "epoch": 44.45,
+      "learning_rate": 4.4444029850746275e-05,
+      "loss": 0.0,
+      "step": 47648
+    },
+    {
+      "epoch": 44.45,
+      "learning_rate": 4.444356343283582e-05,
+      "loss": 0.0,
+      "step": 47652
+    },
+    {
+      "epoch": 44.46,
+      "learning_rate": 4.444309701492538e-05,
+      "loss": 0.0,
+      "step": 47656
+    },
+    {
+      "epoch": 44.46,
+      "learning_rate": 4.4442630597014926e-05,
+      "loss": 0.0002,
+      "step": 47660
+    },
+    {
+      "epoch": 44.46,
+      "learning_rate": 4.444216417910448e-05,
+      "loss": 0.0009,
+      "step": 47664
+    },
+    {
+      "epoch": 44.47,
+      "learning_rate": 4.4441697761194036e-05,
+      "loss": 0.0001,
+      "step": 47668
+    },
+    {
+      "epoch": 44.47,
+      "learning_rate": 4.4441231343283584e-05,
+      "loss": 0.0001,
+      "step": 47672
+    },
+    {
+      "epoch": 44.47,
+      "learning_rate": 4.444076492537313e-05,
+      "loss": 0.0007,
+      "step": 47676
+    },
+    {
+      "epoch": 44.48,
+      "learning_rate": 4.4440298507462694e-05,
+      "loss": 0.0,
+      "step": 47680
+    },
+    {
+      "epoch": 44.48,
+      "learning_rate": 4.443983208955224e-05,
+      "loss": 0.0,
+      "step": 47684
+    },
+    {
+      "epoch": 44.49,
+      "learning_rate": 4.443936567164179e-05,
+      "loss": 0.0003,
+      "step": 47688
+    },
+    {
+      "epoch": 44.49,
+      "learning_rate": 4.4438899253731345e-05,
+      "loss": 0.0021,
+      "step": 47692
+    },
+    {
+      "epoch": 44.49,
+      "learning_rate": 4.44384328358209e-05,
+      "loss": 0.0001,
+      "step": 47696
+    },
+    {
+      "epoch": 44.5,
+      "learning_rate": 4.443796641791045e-05,
+      "loss": 0.0,
+      "step": 47700
+    },
+    {
+      "epoch": 44.5,
+      "learning_rate": 4.44375e-05,
+      "loss": 0.0,
+      "step": 47704
+    },
+    {
+      "epoch": 44.5,
+      "learning_rate": 4.443703358208956e-05,
+      "loss": 0.0001,
+      "step": 47708
+    },
+    {
+      "epoch": 44.51,
+      "learning_rate": 4.4436567164179106e-05,
+      "loss": 0.0002,
+      "step": 47712
+    },
+    {
+      "epoch": 44.51,
+      "learning_rate": 4.443610074626866e-05,
+      "loss": 0.0,
+      "step": 47716
+    },
+    {
+      "epoch": 44.51,
+      "learning_rate": 4.443563432835821e-05,
+      "loss": 0.0001,
+      "step": 47720
+    },
+    {
+      "epoch": 44.52,
+      "learning_rate": 4.4435167910447764e-05,
+      "loss": 0.0001,
+      "step": 47724
+    },
+    {
+      "epoch": 44.52,
+      "learning_rate": 4.443470149253732e-05,
+      "loss": 0.0,
+      "step": 47728
+    },
+    {
+      "epoch": 44.53,
+      "learning_rate": 4.443423507462687e-05,
+      "loss": 0.0,
+      "step": 47732
+    },
+    {
+      "epoch": 44.53,
+      "learning_rate": 4.4433768656716415e-05,
+      "loss": 0.0,
+      "step": 47736
+    },
+    {
+      "epoch": 44.53,
+      "learning_rate": 4.4433302238805977e-05,
+      "loss": 0.0035,
+      "step": 47740
+    },
+    {
+      "epoch": 44.54,
+      "learning_rate": 4.4432835820895525e-05,
+      "loss": 0.0025,
+      "step": 47744
+    },
+    {
+      "epoch": 44.54,
+      "learning_rate": 4.443236940298507e-05,
+      "loss": 0.0,
+      "step": 47748
+    },
+    {
+      "epoch": 44.54,
+      "learning_rate": 4.443190298507463e-05,
+      "loss": 0.0,
+      "step": 47752
+    },
+    {
+      "epoch": 44.55,
+      "learning_rate": 4.443143656716418e-05,
+      "loss": 0.0031,
+      "step": 47756
+    },
+    {
+      "epoch": 44.55,
+      "learning_rate": 4.443097014925373e-05,
+      "loss": 0.0,
+      "step": 47760
+    },
+    {
+      "epoch": 44.56,
+      "learning_rate": 4.4430503731343286e-05,
+      "loss": 0.0004,
+      "step": 47764
+    },
+    {
+      "epoch": 44.56,
+      "learning_rate": 4.443003731343284e-05,
+      "loss": 0.0003,
+      "step": 47768
+    },
+    {
+      "epoch": 44.56,
+      "learning_rate": 4.442957089552239e-05,
+      "loss": 0.0,
+      "step": 47772
+    },
+    {
+      "epoch": 44.57,
+      "learning_rate": 4.4429104477611944e-05,
+      "loss": 0.0,
+      "step": 47776
+    },
+    {
+      "epoch": 44.57,
+      "learning_rate": 4.442863805970149e-05,
+      "loss": 0.0739,
+      "step": 47780
+    },
+    {
+      "epoch": 44.57,
+      "learning_rate": 4.442817164179105e-05,
+      "loss": 0.0,
+      "step": 47784
+    },
+    {
+      "epoch": 44.58,
+      "learning_rate": 4.44277052238806e-05,
+      "loss": 0.0,
+      "step": 47788
+    },
+    {
+      "epoch": 44.58,
+      "learning_rate": 4.442723880597015e-05,
+      "loss": 0.0,
+      "step": 47792
+    },
+    {
+      "epoch": 44.59,
+      "learning_rate": 4.44267723880597e-05,
+      "loss": 0.0002,
+      "step": 47796
+    },
+    {
+      "epoch": 44.59,
+      "learning_rate": 4.442630597014926e-05,
+      "loss": 0.0038,
+      "step": 47800
+    },
+    {
+      "epoch": 44.59,
+      "learning_rate": 4.442583955223881e-05,
+      "loss": 0.0001,
+      "step": 47804
+    },
+    {
+      "epoch": 44.6,
+      "learning_rate": 4.442537313432836e-05,
+      "loss": 0.0005,
+      "step": 47808
+    },
+    {
+      "epoch": 44.6,
+      "learning_rate": 4.442490671641791e-05,
+      "loss": 0.0001,
+      "step": 47812
+    },
+    {
+      "epoch": 44.6,
+      "learning_rate": 4.4424440298507466e-05,
+      "loss": 0.0005,
+      "step": 47816
+    },
+    {
+      "epoch": 44.61,
+      "learning_rate": 4.442397388059702e-05,
+      "loss": 0.0001,
+      "step": 47820
+    },
+    {
+      "epoch": 44.61,
+      "learning_rate": 4.442350746268657e-05,
+      "loss": 0.0047,
+      "step": 47824
+    },
+    {
+      "epoch": 44.62,
+      "learning_rate": 4.4423041044776124e-05,
+      "loss": 0.0,
+      "step": 47828
+    },
+    {
+      "epoch": 44.62,
+      "learning_rate": 4.442257462686568e-05,
+      "loss": 0.0,
+      "step": 47832
+    },
+    {
+      "epoch": 44.62,
+      "learning_rate": 4.4422108208955227e-05,
+      "loss": 0.0001,
+      "step": 47836
+    },
+    {
+      "epoch": 44.63,
+      "learning_rate": 4.4421641791044775e-05,
+      "loss": 0.0006,
+      "step": 47840
+    },
+    {
+      "epoch": 44.63,
+      "learning_rate": 4.442117537313433e-05,
+      "loss": 0.0001,
+      "step": 47844
+    },
+    {
+      "epoch": 44.63,
+      "learning_rate": 4.4420708955223885e-05,
+      "loss": 0.0003,
+      "step": 47848
+    },
+    {
+      "epoch": 44.64,
+      "learning_rate": 4.442024253731343e-05,
+      "loss": 0.0001,
+      "step": 47852
+    },
+    {
+      "epoch": 44.64,
+      "learning_rate": 4.441977611940299e-05,
+      "loss": 0.0005,
+      "step": 47856
+    },
+    {
+      "epoch": 44.65,
+      "learning_rate": 4.441930970149254e-05,
+      "loss": 0.0,
+      "step": 47860
+    },
+    {
+      "epoch": 44.65,
+      "learning_rate": 4.441884328358209e-05,
+      "loss": 0.0,
+      "step": 47864
+    },
+    {
+      "epoch": 44.65,
+      "learning_rate": 4.4418376865671646e-05,
+      "loss": 0.0,
+      "step": 47868
+    },
+    {
+      "epoch": 44.66,
+      "learning_rate": 4.4417910447761194e-05,
+      "loss": 0.0,
+      "step": 47872
+    },
+    {
+      "epoch": 44.66,
+      "learning_rate": 4.441744402985075e-05,
+      "loss": 0.0003,
+      "step": 47876
+    },
+    {
+      "epoch": 44.66,
+      "learning_rate": 4.4416977611940303e-05,
+      "loss": 0.0191,
+      "step": 47880
+    },
+    {
+      "epoch": 44.67,
+      "learning_rate": 4.441651119402985e-05,
+      "loss": 0.0,
+      "step": 47884
+    },
+    {
+      "epoch": 44.67,
+      "learning_rate": 4.4416044776119406e-05,
+      "loss": 0.0,
+      "step": 47888
+    },
+    {
+      "epoch": 44.68,
+      "learning_rate": 4.441557835820896e-05,
+      "loss": 0.0,
+      "step": 47892
+    },
+    {
+      "epoch": 44.68,
+      "learning_rate": 4.441511194029851e-05,
+      "loss": 0.0002,
+      "step": 47896
+    },
+    {
+      "epoch": 44.68,
+      "learning_rate": 4.441464552238806e-05,
+      "loss": 0.0,
+      "step": 47900
+    },
+    {
+      "epoch": 44.69,
+      "learning_rate": 4.441417910447761e-05,
+      "loss": 0.0,
+      "step": 47904
+    },
+    {
+      "epoch": 44.69,
+      "learning_rate": 4.441371268656717e-05,
+      "loss": 0.0,
+      "step": 47908
+    },
+    {
+      "epoch": 44.69,
+      "learning_rate": 4.4413246268656716e-05,
+      "loss": 0.0,
+      "step": 47912
+    },
+    {
+      "epoch": 44.7,
+      "learning_rate": 4.441277985074627e-05,
+      "loss": 0.0001,
+      "step": 47916
+    },
+    {
+      "epoch": 44.7,
+      "learning_rate": 4.4412313432835825e-05,
+      "loss": 0.0,
+      "step": 47920
+    },
+    {
+      "epoch": 44.71,
+      "learning_rate": 4.4411847014925374e-05,
+      "loss": 0.0049,
+      "step": 47924
+    },
+    {
+      "epoch": 44.71,
+      "learning_rate": 4.441138059701493e-05,
+      "loss": 0.0,
+      "step": 47928
+    },
+    {
+      "epoch": 44.71,
+      "learning_rate": 4.4410914179104477e-05,
+      "loss": 0.0,
+      "step": 47932
+    },
+    {
+      "epoch": 44.72,
+      "learning_rate": 4.441044776119403e-05,
+      "loss": 0.0001,
+      "step": 47936
+    },
+    {
+      "epoch": 44.72,
+      "learning_rate": 4.4409981343283586e-05,
+      "loss": 0.0,
+      "step": 47940
+    },
+    {
+      "epoch": 44.72,
+      "learning_rate": 4.4409514925373135e-05,
+      "loss": 0.0,
+      "step": 47944
+    },
+    {
+      "epoch": 44.73,
+      "learning_rate": 4.440904850746269e-05,
+      "loss": 0.0001,
+      "step": 47948
+    },
+    {
+      "epoch": 44.73,
+      "learning_rate": 4.4408582089552244e-05,
+      "loss": 0.0002,
+      "step": 47952
+    },
+    {
+      "epoch": 44.73,
+      "learning_rate": 4.440811567164179e-05,
+      "loss": 0.0001,
+      "step": 47956
+    },
+    {
+      "epoch": 44.74,
+      "learning_rate": 4.440764925373134e-05,
+      "loss": 0.0,
+      "step": 47960
+    },
+    {
+      "epoch": 44.74,
+      "learning_rate": 4.4407182835820896e-05,
+      "loss": 0.0001,
+      "step": 47964
+    },
+    {
+      "epoch": 44.75,
+      "learning_rate": 4.440671641791045e-05,
+      "loss": 0.0,
+      "step": 47968
+    },
+    {
+      "epoch": 44.75,
+      "learning_rate": 4.4406250000000005e-05,
+      "loss": 0.0001,
+      "step": 47972
+    },
+    {
+      "epoch": 44.75,
+      "learning_rate": 4.4405783582089553e-05,
+      "loss": 0.0003,
+      "step": 47976
+    },
+    {
+      "epoch": 44.76,
+      "learning_rate": 4.440531716417911e-05,
+      "loss": 0.0,
+      "step": 47980
+    },
+    {
+      "epoch": 44.76,
+      "learning_rate": 4.440485074626866e-05,
+      "loss": 0.0,
+      "step": 47984
+    },
+    {
+      "epoch": 44.76,
+      "learning_rate": 4.440438432835821e-05,
+      "loss": 0.0001,
+      "step": 47988
+    },
+    {
+      "epoch": 44.77,
+      "learning_rate": 4.440391791044776e-05,
+      "loss": 0.0,
+      "step": 47992
+    },
+    {
+      "epoch": 44.77,
+      "learning_rate": 4.440345149253732e-05,
+      "loss": 0.0,
+      "step": 47996
+    },
+    {
+      "epoch": 44.78,
+      "learning_rate": 4.440298507462687e-05,
+      "loss": 0.0,
+      "step": 48000
+    },
+    {
+      "epoch": 44.78,
+      "eval_exact_match": 0.7263056092843327,
+      "eval_exec": 0.7678916827852998,
+      "eval_loss": 0.43220871686935425,
+      "eval_runtime": 1142.6437,
+      "eval_samples_per_second": 0.905,
+      "step": 48000
+    },
+    {
+      "epoch": 44.78,
+      "learning_rate": 4.440251865671642e-05,
+      "loss": 0.0002,
+      "step": 48004
+    },
+    {
+      "epoch": 44.78,
+      "learning_rate": 4.440205223880597e-05,
+      "loss": 0.0,
+      "step": 48008
+    },
+    {
+      "epoch": 44.79,
+      "learning_rate": 4.440158582089553e-05,
+      "loss": 0.0,
+      "step": 48012
+    },
+    {
+      "epoch": 44.79,
+      "learning_rate": 4.4401119402985075e-05,
+      "loss": 0.0001,
+      "step": 48016
+    },
+    {
+      "epoch": 44.79,
+      "learning_rate": 4.440065298507463e-05,
+      "loss": 0.0003,
+      "step": 48020
+    },
+    {
+      "epoch": 44.8,
+      "learning_rate": 4.440018656716418e-05,
+      "loss": 0.0,
+      "step": 48024
+    },
+    {
+      "epoch": 44.8,
+      "learning_rate": 4.4399720149253733e-05,
+      "loss": 0.0001,
+      "step": 48028
+    },
+    {
+      "epoch": 44.81,
+      "learning_rate": 4.439925373134329e-05,
+      "loss": 0.0,
+      "step": 48032
+    },
+    {
+      "epoch": 44.81,
+      "learning_rate": 4.4398787313432836e-05,
+      "loss": 0.0,
+      "step": 48036
+    },
+    {
+      "epoch": 44.81,
+      "learning_rate": 4.439832089552239e-05,
+      "loss": 0.0001,
+      "step": 48040
+    },
+    {
+      "epoch": 44.82,
+      "learning_rate": 4.4397854477611946e-05,
+      "loss": 0.0,
+      "step": 48044
+    },
+    {
+      "epoch": 44.82,
+      "learning_rate": 4.4397388059701494e-05,
+      "loss": 0.0,
+      "step": 48048
+    },
+    {
+      "epoch": 44.82,
+      "learning_rate": 4.439692164179104e-05,
+      "loss": 0.0,
+      "step": 48052
+    },
+    {
+      "epoch": 44.83,
+      "learning_rate": 4.4396455223880604e-05,
+      "loss": 0.0,
+      "step": 48056
+    },
+    {
+      "epoch": 44.83,
+      "learning_rate": 4.439598880597015e-05,
+      "loss": 0.0043,
+      "step": 48060
+    },
+    {
+      "epoch": 44.84,
+      "learning_rate": 4.43955223880597e-05,
+      "loss": 0.0002,
+      "step": 48064
+    },
+    {
+      "epoch": 44.84,
+      "learning_rate": 4.4395055970149255e-05,
+      "loss": 0.0003,
+      "step": 48068
+    },
+    {
+      "epoch": 44.84,
+      "learning_rate": 4.439458955223881e-05,
+      "loss": 0.0001,
+      "step": 48072
+    },
+    {
+      "epoch": 44.85,
+      "learning_rate": 4.439412313432836e-05,
+      "loss": 0.0,
+      "step": 48076
+    },
+    {
+      "epoch": 44.85,
+      "learning_rate": 4.439365671641791e-05,
+      "loss": 0.0002,
+      "step": 48080
+    },
+    {
+      "epoch": 44.85,
+      "learning_rate": 4.439319029850746e-05,
+      "loss": 0.0075,
+      "step": 48084
+    },
+    {
+      "epoch": 44.86,
+      "learning_rate": 4.4392723880597016e-05,
+      "loss": 0.0,
+      "step": 48088
+    },
+    {
+      "epoch": 44.86,
+      "learning_rate": 4.439225746268657e-05,
+      "loss": 0.0001,
+      "step": 48092
+    },
+    {
+      "epoch": 44.87,
+      "learning_rate": 4.439179104477612e-05,
+      "loss": 0.0,
+      "step": 48096
+    },
+    {
+      "epoch": 44.87,
+      "learning_rate": 4.4391324626865674e-05,
+      "loss": 0.0,
+      "step": 48100
+    },
+    {
+      "epoch": 44.87,
+      "learning_rate": 4.439085820895523e-05,
+      "loss": 0.0038,
+      "step": 48104
+    },
+    {
+      "epoch": 44.88,
+      "learning_rate": 4.439039179104478e-05,
+      "loss": 0.0002,
+      "step": 48108
+    },
+    {
+      "epoch": 44.88,
+      "learning_rate": 4.4389925373134325e-05,
+      "loss": 0.0,
+      "step": 48112
+    },
+    {
+      "epoch": 44.88,
+      "learning_rate": 4.438945895522388e-05,
+      "loss": 0.0,
+      "step": 48116
+    },
+    {
+      "epoch": 44.89,
+      "learning_rate": 4.4388992537313435e-05,
+      "loss": 0.0055,
+      "step": 48120
+    },
+    {
+      "epoch": 44.89,
+      "learning_rate": 4.4388526119402983e-05,
+      "loss": 0.0001,
+      "step": 48124
+    },
+    {
+      "epoch": 44.9,
+      "learning_rate": 4.438805970149254e-05,
+      "loss": 0.0001,
+      "step": 48128
+    },
+    {
+      "epoch": 44.9,
+      "learning_rate": 4.438759328358209e-05,
+      "loss": 0.0003,
+      "step": 48132
+    },
+    {
+      "epoch": 44.9,
+      "learning_rate": 4.438712686567165e-05,
+      "loss": 0.0001,
+      "step": 48136
+    },
+    {
+      "epoch": 44.91,
+      "learning_rate": 4.4386660447761196e-05,
+      "loss": 0.0013,
+      "step": 48140
+    },
+    {
+      "epoch": 44.91,
+      "learning_rate": 4.4386194029850744e-05,
+      "loss": 0.0001,
+      "step": 48144
+    },
+    {
+      "epoch": 44.91,
+      "learning_rate": 4.4385727611940306e-05,
+      "loss": 0.0,
+      "step": 48148
+    },
+    {
+      "epoch": 44.92,
+      "learning_rate": 4.4385261194029854e-05,
+      "loss": 0.0,
+      "step": 48152
+    },
+    {
+      "epoch": 44.92,
+      "learning_rate": 4.43847947761194e-05,
+      "loss": 0.0,
+      "step": 48156
+    },
+    {
+      "epoch": 44.93,
+      "learning_rate": 4.438432835820896e-05,
+      "loss": 0.0001,
+      "step": 48160
+    },
+    {
+      "epoch": 44.93,
+      "learning_rate": 4.438386194029851e-05,
+      "loss": 0.0,
+      "step": 48164
+    },
+    {
+      "epoch": 44.93,
+      "learning_rate": 4.438339552238806e-05,
+      "loss": 0.0001,
+      "step": 48168
+    },
+    {
+      "epoch": 44.94,
+      "learning_rate": 4.4382929104477615e-05,
+      "loss": 0.0,
+      "step": 48172
+    },
+    {
+      "epoch": 44.94,
+      "learning_rate": 4.438246268656716e-05,
+      "loss": 0.0009,
+      "step": 48176
+    },
+    {
+      "epoch": 44.94,
+      "learning_rate": 4.438199626865672e-05,
+      "loss": 0.0,
+      "step": 48180
+    },
+    {
+      "epoch": 44.95,
+      "learning_rate": 4.438152985074627e-05,
+      "loss": 0.0001,
+      "step": 48184
+    },
+    {
+      "epoch": 44.95,
+      "learning_rate": 4.438106343283582e-05,
+      "loss": 0.0001,
+      "step": 48188
+    },
+    {
+      "epoch": 44.96,
+      "learning_rate": 4.4380597014925376e-05,
+      "loss": 0.0003,
+      "step": 48192
+    },
+    {
+      "epoch": 44.96,
+      "learning_rate": 4.438013059701493e-05,
+      "loss": 0.0,
+      "step": 48196
+    },
+    {
+      "epoch": 44.96,
+      "learning_rate": 4.437966417910448e-05,
+      "loss": 0.0,
+      "step": 48200
+    },
+    {
+      "epoch": 44.97,
+      "learning_rate": 4.437919776119403e-05,
+      "loss": 0.0,
+      "step": 48204
+    },
+    {
+      "epoch": 44.97,
+      "learning_rate": 4.437873134328359e-05,
+      "loss": 0.0,
+      "step": 48208
+    },
+    {
+      "epoch": 44.97,
+      "learning_rate": 4.437826492537314e-05,
+      "loss": 0.0002,
+      "step": 48212
+    },
+    {
+      "epoch": 44.98,
+      "learning_rate": 4.4377798507462685e-05,
+      "loss": 0.0,
+      "step": 48216
+    },
+    {
+      "epoch": 44.98,
+      "learning_rate": 4.437733208955224e-05,
+      "loss": 0.0,
+      "step": 48220
+    },
+    {
+      "epoch": 44.98,
+      "learning_rate": 4.4376865671641795e-05,
+      "loss": 0.0,
+      "step": 48224
+    },
+    {
+      "epoch": 44.99,
+      "learning_rate": 4.437639925373134e-05,
+      "loss": 0.0072,
+      "step": 48228
+    },
+    {
+      "epoch": 44.99,
+      "learning_rate": 4.43759328358209e-05,
+      "loss": 0.0,
+      "step": 48232
+    },
+    {
+      "epoch": 45.0,
+      "learning_rate": 4.4375466417910446e-05,
+      "loss": 0.0013,
+      "step": 48236
+    },
+    {
+      "epoch": 45.0,
+      "learning_rate": 4.4375e-05,
+      "loss": 0.0002,
+      "step": 48240
+    },
+    {
+      "epoch": 45.0,
+      "learning_rate": 4.4374533582089556e-05,
+      "loss": 0.0003,
+      "step": 48244
+    },
+    {
+      "epoch": 45.01,
+      "learning_rate": 4.4374067164179104e-05,
+      "loss": 0.0002,
+      "step": 48248
+    },
+    {
+      "epoch": 45.01,
+      "learning_rate": 4.437360074626866e-05,
+      "loss": 0.0009,
+      "step": 48252
+    },
+    {
+      "epoch": 45.01,
+      "learning_rate": 4.4373134328358214e-05,
+      "loss": 0.0,
+      "step": 48256
+    },
+    {
+      "epoch": 45.02,
+      "learning_rate": 4.437266791044776e-05,
+      "loss": 0.0001,
+      "step": 48260
+    },
+    {
+      "epoch": 45.02,
+      "learning_rate": 4.437220149253731e-05,
+      "loss": 0.0,
+      "step": 48264
+    },
+    {
+      "epoch": 45.03,
+      "learning_rate": 4.437173507462687e-05,
+      "loss": 0.0,
+      "step": 48268
+    },
+    {
+      "epoch": 45.03,
+      "learning_rate": 4.437126865671642e-05,
+      "loss": 0.0,
+      "step": 48272
+    },
+    {
+      "epoch": 45.03,
+      "learning_rate": 4.437080223880597e-05,
+      "loss": 0.0,
+      "step": 48276
+    },
+    {
+      "epoch": 45.04,
+      "learning_rate": 4.437033582089552e-05,
+      "loss": 0.0007,
+      "step": 48280
+    },
+    {
+      "epoch": 45.04,
+      "learning_rate": 4.436986940298508e-05,
+      "loss": 0.0005,
+      "step": 48284
+    },
+    {
+      "epoch": 45.04,
+      "learning_rate": 4.4369402985074626e-05,
+      "loss": 0.0,
+      "step": 48288
+    },
+    {
+      "epoch": 45.05,
+      "learning_rate": 4.436893656716418e-05,
+      "loss": 0.0002,
+      "step": 48292
+    },
+    {
+      "epoch": 45.05,
+      "learning_rate": 4.436847014925373e-05,
+      "loss": 0.0004,
+      "step": 48296
+    },
+    {
+      "epoch": 45.06,
+      "learning_rate": 4.436800373134329e-05,
+      "loss": 0.0005,
+      "step": 48300
+    },
+    {
+      "epoch": 45.06,
+      "learning_rate": 4.436753731343284e-05,
+      "loss": 0.0,
+      "step": 48304
+    },
+    {
+      "epoch": 45.06,
+      "learning_rate": 4.436707089552239e-05,
+      "loss": 0.0007,
+      "step": 48308
+    },
+    {
+      "epoch": 45.07,
+      "learning_rate": 4.436660447761194e-05,
+      "loss": 0.0,
+      "step": 48312
+    },
+    {
+      "epoch": 45.07,
+      "learning_rate": 4.43661380597015e-05,
+      "loss": 0.0,
+      "step": 48316
+    },
+    {
+      "epoch": 45.07,
+      "learning_rate": 4.4365671641791045e-05,
+      "loss": 0.0,
+      "step": 48320
+    },
+    {
+      "epoch": 45.08,
+      "learning_rate": 4.43652052238806e-05,
+      "loss": 0.001,
+      "step": 48324
+    },
+    {
+      "epoch": 45.08,
+      "learning_rate": 4.4364738805970155e-05,
+      "loss": 0.0001,
+      "step": 48328
+    },
+    {
+      "epoch": 45.09,
+      "learning_rate": 4.43642723880597e-05,
+      "loss": 0.001,
+      "step": 48332
+    },
+    {
+      "epoch": 45.09,
+      "learning_rate": 4.436380597014926e-05,
+      "loss": 0.0,
+      "step": 48336
+    },
+    {
+      "epoch": 45.09,
+      "learning_rate": 4.4363339552238806e-05,
+      "loss": 0.0,
+      "step": 48340
+    },
+    {
+      "epoch": 45.1,
+      "learning_rate": 4.436287313432836e-05,
+      "loss": 0.0044,
+      "step": 48344
+    },
+    {
+      "epoch": 45.1,
+      "learning_rate": 4.4362406716417916e-05,
+      "loss": 0.0,
+      "step": 48348
+    },
+    {
+      "epoch": 45.1,
+      "learning_rate": 4.4361940298507464e-05,
+      "loss": 0.0006,
+      "step": 48352
+    },
+    {
+      "epoch": 45.11,
+      "learning_rate": 4.436147388059701e-05,
+      "loss": 0.0,
+      "step": 48356
+    },
+    {
+      "epoch": 45.11,
+      "learning_rate": 4.4361007462686574e-05,
+      "loss": 0.0002,
+      "step": 48360
+    },
+    {
+      "epoch": 45.12,
+      "learning_rate": 4.436054104477612e-05,
+      "loss": 0.0002,
+      "step": 48364
+    },
+    {
+      "epoch": 45.12,
+      "learning_rate": 4.436007462686567e-05,
+      "loss": 0.0,
+      "step": 48368
+    },
+    {
+      "epoch": 45.12,
+      "learning_rate": 4.4359608208955225e-05,
+      "loss": 0.0002,
+      "step": 48372
+    },
+    {
+      "epoch": 45.13,
+      "learning_rate": 4.435914179104478e-05,
+      "loss": 0.0,
+      "step": 48376
+    },
+    {
+      "epoch": 45.13,
+      "learning_rate": 4.435867537313433e-05,
+      "loss": 0.0001,
+      "step": 48380
+    },
+    {
+      "epoch": 45.13,
+      "learning_rate": 4.435820895522388e-05,
+      "loss": 0.0,
+      "step": 48384
+    },
+    {
+      "epoch": 45.14,
+      "learning_rate": 4.435774253731344e-05,
+      "loss": 0.0,
+      "step": 48388
+    },
+    {
+      "epoch": 45.14,
+      "learning_rate": 4.4357276119402986e-05,
+      "loss": 0.0,
+      "step": 48392
+    },
+    {
+      "epoch": 45.15,
+      "learning_rate": 4.435680970149254e-05,
+      "loss": 0.0,
+      "step": 48396
+    },
+    {
+      "epoch": 45.15,
+      "learning_rate": 4.435634328358209e-05,
+      "loss": 0.0015,
+      "step": 48400
+    },
+    {
+      "epoch": 45.15,
+      "learning_rate": 4.4355876865671644e-05,
+      "loss": 0.0062,
+      "step": 48404
+    },
+    {
+      "epoch": 45.16,
+      "learning_rate": 4.43554104477612e-05,
+      "loss": 0.0,
+      "step": 48408
+    },
+    {
+      "epoch": 45.16,
+      "learning_rate": 4.435494402985075e-05,
+      "loss": 0.0,
+      "step": 48412
+    },
+    {
+      "epoch": 45.16,
+      "learning_rate": 4.4354477611940295e-05,
+      "loss": 0.0003,
+      "step": 48416
+    },
+    {
+      "epoch": 45.17,
+      "learning_rate": 4.435401119402986e-05,
+      "loss": 0.0006,
+      "step": 48420
+    },
+    {
+      "epoch": 45.17,
+      "learning_rate": 4.4353544776119405e-05,
+      "loss": 0.0,
+      "step": 48424
+    },
+    {
+      "epoch": 45.18,
+      "learning_rate": 4.435307835820895e-05,
+      "loss": 0.0,
+      "step": 48428
+    },
+    {
+      "epoch": 45.18,
+      "learning_rate": 4.435261194029851e-05,
+      "loss": 0.0,
+      "step": 48432
+    },
+    {
+      "epoch": 45.18,
+      "learning_rate": 4.435214552238806e-05,
+      "loss": 0.001,
+      "step": 48436
+    },
+    {
+      "epoch": 45.19,
+      "learning_rate": 4.435167910447761e-05,
+      "loss": 0.0001,
+      "step": 48440
+    },
+    {
+      "epoch": 45.19,
+      "learning_rate": 4.4351212686567166e-05,
+      "loss": 0.0,
+      "step": 48444
+    },
+    {
+      "epoch": 45.19,
+      "learning_rate": 4.435074626865672e-05,
+      "loss": 0.0001,
+      "step": 48448
+    },
+    {
+      "epoch": 45.2,
+      "learning_rate": 4.435027985074627e-05,
+      "loss": 0.0003,
+      "step": 48452
+    },
+    {
+      "epoch": 45.2,
+      "learning_rate": 4.4349813432835824e-05,
+      "loss": 0.0,
+      "step": 48456
+    },
+    {
+      "epoch": 45.21,
+      "learning_rate": 4.434934701492537e-05,
+      "loss": 0.0001,
+      "step": 48460
+    },
+    {
+      "epoch": 45.21,
+      "learning_rate": 4.434888059701493e-05,
+      "loss": 0.0004,
+      "step": 48464
+    },
+    {
+      "epoch": 45.21,
+      "learning_rate": 4.434841417910448e-05,
+      "loss": 0.0001,
+      "step": 48468
+    },
+    {
+      "epoch": 45.22,
+      "learning_rate": 4.434794776119403e-05,
+      "loss": 0.0,
+      "step": 48472
+    },
+    {
+      "epoch": 45.22,
+      "learning_rate": 4.4347481343283585e-05,
+      "loss": 0.0,
+      "step": 48476
+    },
+    {
+      "epoch": 45.22,
+      "learning_rate": 4.434701492537314e-05,
+      "loss": 0.0001,
+      "step": 48480
+    },
+    {
+      "epoch": 45.23,
+      "learning_rate": 4.434654850746269e-05,
+      "loss": 0.0009,
+      "step": 48484
+    },
+    {
+      "epoch": 45.23,
+      "learning_rate": 4.434608208955224e-05,
+      "loss": 0.0001,
+      "step": 48488
+    },
+    {
+      "epoch": 45.24,
+      "learning_rate": 4.434561567164179e-05,
+      "loss": 0.0,
+      "step": 48492
+    },
+    {
+      "epoch": 45.24,
+      "learning_rate": 4.4345149253731346e-05,
+      "loss": 0.0,
+      "step": 48496
+    },
+    {
+      "epoch": 45.24,
+      "learning_rate": 4.43446828358209e-05,
+      "loss": 0.0,
+      "step": 48500
+    },
+    {
+      "epoch": 45.24,
+      "eval_exact_match": 0.7272727272727273,
+      "eval_exec": 0.7659574468085106,
+      "eval_loss": 0.433497816324234,
+      "eval_runtime": 1151.4161,
+      "eval_samples_per_second": 0.898,
+      "step": 48500
+    },
+    {
+      "epoch": 45.25,
+      "learning_rate": 4.434421641791045e-05,
+      "loss": 0.0001,
+      "step": 48504
+    },
+    {
+      "epoch": 45.25,
+      "learning_rate": 4.4343750000000004e-05,
+      "loss": 0.0,
+      "step": 48508
+    },
+    {
+      "epoch": 45.25,
+      "learning_rate": 4.434328358208956e-05,
+      "loss": 0.0,
+      "step": 48512
+    },
+    {
+      "epoch": 45.26,
+      "learning_rate": 4.434281716417911e-05,
+      "loss": 0.0,
+      "step": 48516
+    },
+    {
+      "epoch": 45.26,
+      "learning_rate": 4.4342350746268655e-05,
+      "loss": 0.0,
+      "step": 48520
+    },
+    {
+      "epoch": 45.26,
+      "learning_rate": 4.434188432835821e-05,
+      "loss": 0.0,
+      "step": 48524
+    },
+    {
+      "epoch": 45.27,
+      "learning_rate": 4.4341417910447765e-05,
+      "loss": 0.0001,
+      "step": 48528
+    },
+    {
+      "epoch": 45.27,
+      "learning_rate": 4.434095149253731e-05,
+      "loss": 0.0,
+      "step": 48532
+    },
+    {
+      "epoch": 45.28,
+      "learning_rate": 4.434048507462687e-05,
+      "loss": 0.004,
+      "step": 48536
+    },
+    {
+      "epoch": 45.28,
+      "learning_rate": 4.434001865671642e-05,
+      "loss": 0.0,
+      "step": 48540
+    },
+    {
+      "epoch": 45.28,
+      "learning_rate": 4.433955223880597e-05,
+      "loss": 0.0,
+      "step": 48544
+    },
+    {
+      "epoch": 45.29,
+      "learning_rate": 4.4339085820895526e-05,
+      "loss": 0.0,
+      "step": 48548
+    },
+    {
+      "epoch": 45.29,
+      "learning_rate": 4.4338619402985074e-05,
+      "loss": 0.0,
+      "step": 48552
+    },
+    {
+      "epoch": 45.29,
+      "learning_rate": 4.433815298507463e-05,
+      "loss": 0.0,
+      "step": 48556
+    },
+    {
+      "epoch": 45.3,
+      "learning_rate": 4.4337686567164184e-05,
+      "loss": 0.0008,
+      "step": 48560
+    },
+    {
+      "epoch": 45.3,
+      "learning_rate": 4.433722014925373e-05,
+      "loss": 0.0,
+      "step": 48564
+    },
+    {
+      "epoch": 45.31,
+      "learning_rate": 4.433675373134329e-05,
+      "loss": 0.0003,
+      "step": 48568
+    },
+    {
+      "epoch": 45.31,
+      "learning_rate": 4.433628731343284e-05,
+      "loss": 0.0,
+      "step": 48572
+    },
+    {
+      "epoch": 45.31,
+      "learning_rate": 4.433582089552239e-05,
+      "loss": 0.0,
+      "step": 48576
+    },
+    {
+      "epoch": 45.32,
+      "learning_rate": 4.433535447761194e-05,
+      "loss": 0.0,
+      "step": 48580
+    },
+    {
+      "epoch": 45.32,
+      "learning_rate": 4.433488805970149e-05,
+      "loss": 0.0,
+      "step": 48584
+    },
+    {
+      "epoch": 45.32,
+      "learning_rate": 4.433442164179105e-05,
+      "loss": 0.0,
+      "step": 48588
+    },
+    {
+      "epoch": 45.33,
+      "learning_rate": 4.4333955223880596e-05,
+      "loss": 0.0001,
+      "step": 48592
+    },
+    {
+      "epoch": 45.33,
+      "learning_rate": 4.433348880597015e-05,
+      "loss": 0.0,
+      "step": 48596
+    },
+    {
+      "epoch": 45.34,
+      "learning_rate": 4.4333022388059706e-05,
+      "loss": 0.0,
+      "step": 48600
+    },
+    {
+      "epoch": 45.34,
+      "learning_rate": 4.4332555970149254e-05,
+      "loss": 0.0,
+      "step": 48604
+    },
+    {
+      "epoch": 45.34,
+      "learning_rate": 4.433208955223881e-05,
+      "loss": 0.0002,
+      "step": 48608
+    },
+    {
+      "epoch": 45.35,
+      "learning_rate": 4.433162313432836e-05,
+      "loss": 0.0,
+      "step": 48612
+    },
+    {
+      "epoch": 45.35,
+      "learning_rate": 4.433115671641792e-05,
+      "loss": 0.0001,
+      "step": 48616
+    },
+    {
+      "epoch": 45.35,
+      "learning_rate": 4.433069029850747e-05,
+      "loss": 0.0,
+      "step": 48620
+    },
+    {
+      "epoch": 45.36,
+      "learning_rate": 4.4330223880597015e-05,
+      "loss": 0.0027,
+      "step": 48624
+    },
+    {
+      "epoch": 45.36,
+      "learning_rate": 4.432975746268657e-05,
+      "loss": 0.0,
+      "step": 48628
+    },
+    {
+      "epoch": 45.37,
+      "learning_rate": 4.4329291044776125e-05,
+      "loss": 0.0002,
+      "step": 48632
+    },
+    {
+      "epoch": 45.37,
+      "learning_rate": 4.432882462686567e-05,
+      "loss": 0.0001,
+      "step": 48636
+    },
+    {
+      "epoch": 45.37,
+      "learning_rate": 4.432835820895523e-05,
+      "loss": 0.0,
+      "step": 48640
+    },
+    {
+      "epoch": 45.38,
+      "learning_rate": 4.4327891791044776e-05,
+      "loss": 0.0001,
+      "step": 48644
+    },
+    {
+      "epoch": 45.38,
+      "learning_rate": 4.432742537313433e-05,
+      "loss": 0.0001,
+      "step": 48648
+    },
+    {
+      "epoch": 45.38,
+      "learning_rate": 4.4326958955223886e-05,
+      "loss": 0.0,
+      "step": 48652
+    },
+    {
+      "epoch": 45.39,
+      "learning_rate": 4.4326492537313434e-05,
+      "loss": 0.0001,
+      "step": 48656
+    },
+    {
+      "epoch": 45.39,
+      "learning_rate": 4.432602611940299e-05,
+      "loss": 0.0,
+      "step": 48660
+    },
+    {
+      "epoch": 45.4,
+      "learning_rate": 4.4325559701492544e-05,
+      "loss": 0.0,
+      "step": 48664
+    },
+    {
+      "epoch": 45.4,
+      "learning_rate": 4.432509328358209e-05,
+      "loss": 0.0,
+      "step": 48668
+    },
+    {
+      "epoch": 45.4,
+      "learning_rate": 4.432462686567164e-05,
+      "loss": 0.0,
+      "step": 48672
+    },
+    {
+      "epoch": 45.41,
+      "learning_rate": 4.43241604477612e-05,
+      "loss": 0.0,
+      "step": 48676
+    },
+    {
+      "epoch": 45.41,
+      "learning_rate": 4.432369402985075e-05,
+      "loss": 0.0001,
+      "step": 48680
+    },
+    {
+      "epoch": 45.41,
+      "learning_rate": 4.43232276119403e-05,
+      "loss": 0.0,
+      "step": 48684
+    },
+    {
+      "epoch": 45.42,
+      "learning_rate": 4.432276119402985e-05,
+      "loss": 0.0006,
+      "step": 48688
+    },
+    {
+      "epoch": 45.42,
+      "learning_rate": 4.432229477611941e-05,
+      "loss": 0.0,
+      "step": 48692
+    },
+    {
+      "epoch": 45.43,
+      "learning_rate": 4.4321828358208956e-05,
+      "loss": 0.0005,
+      "step": 48696
+    },
+    {
+      "epoch": 45.43,
+      "learning_rate": 4.432136194029851e-05,
+      "loss": 0.0001,
+      "step": 48700
+    },
+    {
+      "epoch": 45.43,
+      "learning_rate": 4.432089552238806e-05,
+      "loss": 0.0,
+      "step": 48704
+    },
+    {
+      "epoch": 45.44,
+      "learning_rate": 4.4320429104477614e-05,
+      "loss": 0.0,
+      "step": 48708
+    },
+    {
+      "epoch": 45.44,
+      "learning_rate": 4.431996268656717e-05,
+      "loss": 0.0005,
+      "step": 48712
+    },
+    {
+      "epoch": 45.44,
+      "learning_rate": 4.431949626865672e-05,
+      "loss": 0.0,
+      "step": 48716
+    },
+    {
+      "epoch": 45.45,
+      "learning_rate": 4.431902985074627e-05,
+      "loss": 0.0,
+      "step": 48720
+    },
+    {
+      "epoch": 45.45,
+      "learning_rate": 4.4318563432835827e-05,
+      "loss": 0.0,
+      "step": 48724
+    },
+    {
+      "epoch": 45.46,
+      "learning_rate": 4.4318097014925375e-05,
+      "loss": 0.0,
+      "step": 48728
+    },
+    {
+      "epoch": 45.46,
+      "learning_rate": 4.431763059701492e-05,
+      "loss": 0.0001,
+      "step": 48732
+    },
+    {
+      "epoch": 45.46,
+      "learning_rate": 4.4317164179104484e-05,
+      "loss": 0.0,
+      "step": 48736
+    },
+    {
+      "epoch": 45.47,
+      "learning_rate": 4.431669776119403e-05,
+      "loss": 0.0,
+      "step": 48740
+    },
+    {
+      "epoch": 45.47,
+      "learning_rate": 4.431623134328358e-05,
+      "loss": 0.0001,
+      "step": 48744
+    },
+    {
+      "epoch": 45.47,
+      "learning_rate": 4.4315764925373136e-05,
+      "loss": 0.0,
+      "step": 48748
+    },
+    {
+      "epoch": 45.48,
+      "learning_rate": 4.431529850746269e-05,
+      "loss": 0.0,
+      "step": 48752
+    },
+    {
+      "epoch": 45.48,
+      "learning_rate": 4.431483208955224e-05,
+      "loss": 0.0,
+      "step": 48756
+    },
+    {
+      "epoch": 45.49,
+      "learning_rate": 4.4314365671641794e-05,
+      "loss": 0.0,
+      "step": 48760
+    },
+    {
+      "epoch": 45.49,
+      "learning_rate": 4.431389925373134e-05,
+      "loss": 0.0001,
+      "step": 48764
+    },
+    {
+      "epoch": 45.49,
+      "learning_rate": 4.43134328358209e-05,
+      "loss": 0.0,
+      "step": 48768
+    },
+    {
+      "epoch": 45.5,
+      "learning_rate": 4.431296641791045e-05,
+      "loss": 0.0,
+      "step": 48772
+    },
+    {
+      "epoch": 45.5,
+      "learning_rate": 4.43125e-05,
+      "loss": 0.0003,
+      "step": 48776
+    },
+    {
+      "epoch": 45.5,
+      "learning_rate": 4.4312033582089555e-05,
+      "loss": 0.0,
+      "step": 48780
+    },
+    {
+      "epoch": 45.51,
+      "learning_rate": 4.431156716417911e-05,
+      "loss": 0.0,
+      "step": 48784
+    },
+    {
+      "epoch": 45.51,
+      "learning_rate": 4.431110074626866e-05,
+      "loss": 0.0022,
+      "step": 48788
+    },
+    {
+      "epoch": 45.51,
+      "learning_rate": 4.431063432835821e-05,
+      "loss": 0.0019,
+      "step": 48792
+    },
+    {
+      "epoch": 45.52,
+      "learning_rate": 4.431016791044777e-05,
+      "loss": 0.0002,
+      "step": 48796
+    },
+    {
+      "epoch": 45.52,
+      "learning_rate": 4.4309701492537316e-05,
+      "loss": 0.0,
+      "step": 48800
+    },
+    {
+      "epoch": 45.53,
+      "learning_rate": 4.430923507462687e-05,
+      "loss": 0.0001,
+      "step": 48804
+    },
+    {
+      "epoch": 45.53,
+      "learning_rate": 4.430876865671642e-05,
+      "loss": 0.0,
+      "step": 48808
+    },
+    {
+      "epoch": 45.53,
+      "learning_rate": 4.4308302238805974e-05,
+      "loss": 0.0002,
+      "step": 48812
+    },
+    {
+      "epoch": 45.54,
+      "learning_rate": 4.430783582089553e-05,
+      "loss": 0.0,
+      "step": 48816
+    },
+    {
+      "epoch": 45.54,
+      "learning_rate": 4.4307369402985077e-05,
+      "loss": 0.0,
+      "step": 48820
+    },
+    {
+      "epoch": 45.54,
+      "learning_rate": 4.4306902985074625e-05,
+      "loss": 0.0,
+      "step": 48824
+    },
+    {
+      "epoch": 45.55,
+      "learning_rate": 4.4306436567164186e-05,
+      "loss": 0.0001,
+      "step": 48828
+    },
+    {
+      "epoch": 45.55,
+      "learning_rate": 4.4305970149253735e-05,
+      "loss": 0.005,
+      "step": 48832
+    },
+    {
+      "epoch": 45.56,
+      "learning_rate": 4.430550373134328e-05,
+      "loss": 0.0,
+      "step": 48836
+    },
+    {
+      "epoch": 45.56,
+      "learning_rate": 4.430503731343284e-05,
+      "loss": 0.0,
+      "step": 48840
+    },
+    {
+      "epoch": 45.56,
+      "learning_rate": 4.430457089552239e-05,
+      "loss": 0.0001,
+      "step": 48844
+    },
+    {
+      "epoch": 45.57,
+      "learning_rate": 4.430410447761194e-05,
+      "loss": 0.0,
+      "step": 48848
+    },
+    {
+      "epoch": 45.57,
+      "learning_rate": 4.4303638059701495e-05,
+      "loss": 0.0,
+      "step": 48852
+    },
+    {
+      "epoch": 45.57,
+      "learning_rate": 4.4303171641791044e-05,
+      "loss": 0.0,
+      "step": 48856
+    },
+    {
+      "epoch": 45.58,
+      "learning_rate": 4.43027052238806e-05,
+      "loss": 0.0,
+      "step": 48860
+    },
+    {
+      "epoch": 45.58,
+      "learning_rate": 4.4302238805970153e-05,
+      "loss": 0.0,
+      "step": 48864
+    },
+    {
+      "epoch": 45.59,
+      "learning_rate": 4.43017723880597e-05,
+      "loss": 0.0,
+      "step": 48868
+    },
+    {
+      "epoch": 45.59,
+      "learning_rate": 4.4301305970149256e-05,
+      "loss": 0.0001,
+      "step": 48872
+    },
+    {
+      "epoch": 45.59,
+      "learning_rate": 4.430083955223881e-05,
+      "loss": 0.0,
+      "step": 48876
+    },
+    {
+      "epoch": 45.6,
+      "learning_rate": 4.430037313432836e-05,
+      "loss": 0.0,
+      "step": 48880
+    },
+    {
+      "epoch": 45.6,
+      "learning_rate": 4.429990671641791e-05,
+      "loss": 0.0,
+      "step": 48884
+    },
+    {
+      "epoch": 45.6,
+      "learning_rate": 4.429944029850747e-05,
+      "loss": 0.0,
+      "step": 48888
+    },
+    {
+      "epoch": 45.61,
+      "learning_rate": 4.429897388059702e-05,
+      "loss": 0.0,
+      "step": 48892
+    },
+    {
+      "epoch": 45.61,
+      "learning_rate": 4.4298507462686566e-05,
+      "loss": 0.0002,
+      "step": 48896
+    },
+    {
+      "epoch": 45.62,
+      "learning_rate": 4.429804104477612e-05,
+      "loss": 0.0,
+      "step": 48900
+    },
+    {
+      "epoch": 45.62,
+      "learning_rate": 4.4297574626865675e-05,
+      "loss": 0.0,
+      "step": 48904
+    },
+    {
+      "epoch": 45.62,
+      "learning_rate": 4.4297108208955224e-05,
+      "loss": 0.0,
+      "step": 48908
+    },
+    {
+      "epoch": 45.63,
+      "learning_rate": 4.429664179104478e-05,
+      "loss": 0.0,
+      "step": 48912
+    },
+    {
+      "epoch": 45.63,
+      "learning_rate": 4.4296175373134327e-05,
+      "loss": 0.0057,
+      "step": 48916
+    },
+    {
+      "epoch": 45.63,
+      "learning_rate": 4.429570895522388e-05,
+      "loss": 0.0,
+      "step": 48920
+    },
+    {
+      "epoch": 45.64,
+      "learning_rate": 4.4295242537313436e-05,
+      "loss": 0.0002,
+      "step": 48924
+    },
+    {
+      "epoch": 45.64,
+      "learning_rate": 4.4294776119402985e-05,
+      "loss": 0.0003,
+      "step": 48928
+    },
+    {
+      "epoch": 45.65,
+      "learning_rate": 4.429430970149254e-05,
+      "loss": 0.0,
+      "step": 48932
+    },
+    {
+      "epoch": 45.65,
+      "learning_rate": 4.4293843283582094e-05,
+      "loss": 0.0003,
+      "step": 48936
+    },
+    {
+      "epoch": 45.65,
+      "learning_rate": 4.429337686567164e-05,
+      "loss": 0.0,
+      "step": 48940
+    },
+    {
+      "epoch": 45.66,
+      "learning_rate": 4.42929104477612e-05,
+      "loss": 0.0,
+      "step": 48944
+    },
+    {
+      "epoch": 45.66,
+      "learning_rate": 4.429244402985075e-05,
+      "loss": 0.0,
+      "step": 48948
+    },
+    {
+      "epoch": 45.66,
+      "learning_rate": 4.42919776119403e-05,
+      "loss": 0.0,
+      "step": 48952
+    },
+    {
+      "epoch": 45.67,
+      "learning_rate": 4.4291511194029855e-05,
+      "loss": 0.0,
+      "step": 48956
+    },
+    {
+      "epoch": 45.67,
+      "learning_rate": 4.4291044776119403e-05,
+      "loss": 0.0001,
+      "step": 48960
+    },
+    {
+      "epoch": 45.68,
+      "learning_rate": 4.429057835820896e-05,
+      "loss": 0.0,
+      "step": 48964
+    },
+    {
+      "epoch": 45.68,
+      "learning_rate": 4.429011194029851e-05,
+      "loss": 0.0,
+      "step": 48968
+    },
+    {
+      "epoch": 45.68,
+      "learning_rate": 4.428964552238806e-05,
+      "loss": 0.0004,
+      "step": 48972
+    },
+    {
+      "epoch": 45.69,
+      "learning_rate": 4.428917910447761e-05,
+      "loss": 0.0,
+      "step": 48976
+    },
+    {
+      "epoch": 45.69,
+      "learning_rate": 4.428871268656717e-05,
+      "loss": 0.0,
+      "step": 48980
+    },
+    {
+      "epoch": 45.69,
+      "learning_rate": 4.428824626865672e-05,
+      "loss": 0.0,
+      "step": 48984
+    },
+    {
+      "epoch": 45.7,
+      "learning_rate": 4.428777985074627e-05,
+      "loss": 0.0,
+      "step": 48988
+    },
+    {
+      "epoch": 45.7,
+      "learning_rate": 4.428731343283582e-05,
+      "loss": 0.0,
+      "step": 48992
+    },
+    {
+      "epoch": 45.71,
+      "learning_rate": 4.428684701492538e-05,
+      "loss": 0.0,
+      "step": 48996
+    },
+    {
+      "epoch": 45.71,
+      "learning_rate": 4.4286380597014925e-05,
+      "loss": 0.001,
+      "step": 49000
+    },
+    {
+      "epoch": 45.71,
+      "eval_exact_match": 0.7321083172147002,
+      "eval_exec": 0.7707930367504836,
+      "eval_loss": 0.4455174505710602,
+      "eval_runtime": 1081.8406,
+      "eval_samples_per_second": 0.956,
+      "step": 49000
+    },
+    {
+      "epoch": 45.71,
+      "learning_rate": 4.428591417910448e-05,
+      "loss": 0.0,
+      "step": 49004
+    },
+    {
+      "epoch": 45.72,
+      "learning_rate": 4.4285447761194035e-05,
+      "loss": 0.0001,
+      "step": 49008
+    },
+    {
+      "epoch": 45.72,
+      "learning_rate": 4.428498134328358e-05,
+      "loss": 0.0005,
+      "step": 49012
+    },
+    {
+      "epoch": 45.72,
+      "learning_rate": 4.428451492537314e-05,
+      "loss": 0.0,
+      "step": 49016
+    },
+    {
+      "epoch": 45.73,
+      "learning_rate": 4.4284048507462686e-05,
+      "loss": 0.0001,
+      "step": 49020
+    },
+    {
+      "epoch": 45.73,
+      "learning_rate": 4.428358208955224e-05,
+      "loss": 0.0,
+      "step": 49024
+    },
+    {
+      "epoch": 45.73,
+      "learning_rate": 4.4283115671641796e-05,
+      "loss": 0.0,
+      "step": 49028
+    },
+    {
+      "epoch": 45.74,
+      "learning_rate": 4.4282649253731344e-05,
+      "loss": 0.0002,
+      "step": 49032
+    },
+    {
+      "epoch": 45.74,
+      "learning_rate": 4.428218283582089e-05,
+      "loss": 0.0002,
+      "step": 49036
+    },
+    {
+      "epoch": 45.75,
+      "learning_rate": 4.4281716417910454e-05,
+      "loss": 0.0,
+      "step": 49040
+    },
+    {
+      "epoch": 45.75,
+      "learning_rate": 4.428125e-05,
+      "loss": 0.0,
+      "step": 49044
+    },
+    {
+      "epoch": 45.75,
+      "learning_rate": 4.428078358208955e-05,
+      "loss": 0.0,
+      "step": 49048
+    },
+    {
+      "epoch": 45.76,
+      "learning_rate": 4.4280317164179105e-05,
+      "loss": 0.0,
+      "step": 49052
+    },
+    {
+      "epoch": 45.76,
+      "learning_rate": 4.427985074626866e-05,
+      "loss": 0.0001,
+      "step": 49056
+    },
+    {
+      "epoch": 45.76,
+      "learning_rate": 4.427938432835821e-05,
+      "loss": 0.0003,
+      "step": 49060
+    },
+    {
+      "epoch": 45.77,
+      "learning_rate": 4.427891791044776e-05,
+      "loss": 0.0,
+      "step": 49064
+    },
+    {
+      "epoch": 45.77,
+      "learning_rate": 4.427845149253732e-05,
+      "loss": 0.0,
+      "step": 49068
+    },
+    {
+      "epoch": 45.78,
+      "learning_rate": 4.4277985074626866e-05,
+      "loss": 0.0001,
+      "step": 49072
+    },
+    {
+      "epoch": 45.78,
+      "learning_rate": 4.427751865671642e-05,
+      "loss": 0.0001,
+      "step": 49076
+    },
+    {
+      "epoch": 45.78,
+      "learning_rate": 4.427705223880597e-05,
+      "loss": 0.0,
+      "step": 49080
+    },
+    {
+      "epoch": 45.79,
+      "learning_rate": 4.4276585820895524e-05,
+      "loss": 0.0,
+      "step": 49084
+    },
+    {
+      "epoch": 45.79,
+      "learning_rate": 4.427611940298508e-05,
+      "loss": 0.0001,
+      "step": 49088
+    },
+    {
+      "epoch": 45.79,
+      "learning_rate": 4.427565298507463e-05,
+      "loss": 0.0011,
+      "step": 49092
+    },
+    {
+      "epoch": 45.8,
+      "learning_rate": 4.4275186567164175e-05,
+      "loss": 0.0001,
+      "step": 49096
+    },
+    {
+      "epoch": 45.8,
+      "learning_rate": 4.427472014925374e-05,
+      "loss": 0.0,
+      "step": 49100
+    },
+    {
+      "epoch": 45.81,
+      "learning_rate": 4.4274253731343285e-05,
+      "loss": 0.0,
+      "step": 49104
+    },
+    {
+      "epoch": 45.81,
+      "learning_rate": 4.427378731343284e-05,
+      "loss": 0.0,
+      "step": 49108
+    },
+    {
+      "epoch": 45.81,
+      "learning_rate": 4.427332089552239e-05,
+      "loss": 0.0,
+      "step": 49112
+    },
+    {
+      "epoch": 45.82,
+      "learning_rate": 4.427285447761194e-05,
+      "loss": 0.0001,
+      "step": 49116
+    },
+    {
+      "epoch": 45.82,
+      "learning_rate": 4.42723880597015e-05,
+      "loss": 0.0,
+      "step": 49120
+    },
+    {
+      "epoch": 45.82,
+      "learning_rate": 4.4271921641791046e-05,
+      "loss": 0.0005,
+      "step": 49124
+    },
+    {
+      "epoch": 45.83,
+      "learning_rate": 4.42714552238806e-05,
+      "loss": 0.0001,
+      "step": 49128
+    },
+    {
+      "epoch": 45.83,
+      "learning_rate": 4.4270988805970156e-05,
+      "loss": 0.0,
+      "step": 49132
+    },
+    {
+      "epoch": 45.84,
+      "learning_rate": 4.4270522388059704e-05,
+      "loss": 0.0,
+      "step": 49136
+    },
+    {
+      "epoch": 45.84,
+      "learning_rate": 4.427005597014925e-05,
+      "loss": 0.0,
+      "step": 49140
+    },
+    {
+      "epoch": 45.84,
+      "learning_rate": 4.426958955223881e-05,
+      "loss": 0.0005,
+      "step": 49144
+    },
+    {
+      "epoch": 45.85,
+      "learning_rate": 4.426912313432836e-05,
+      "loss": 0.0,
+      "step": 49148
+    },
+    {
+      "epoch": 45.85,
+      "learning_rate": 4.426865671641791e-05,
+      "loss": 0.0001,
+      "step": 49152
+    },
+    {
+      "epoch": 45.85,
+      "learning_rate": 4.4268190298507465e-05,
+      "loss": 0.0,
+      "step": 49156
+    },
+    {
+      "epoch": 45.86,
+      "learning_rate": 4.426772388059702e-05,
+      "loss": 0.0,
+      "step": 49160
+    },
+    {
+      "epoch": 45.86,
+      "learning_rate": 4.426725746268657e-05,
+      "loss": 0.0,
+      "step": 49164
+    },
+    {
+      "epoch": 45.87,
+      "learning_rate": 4.426679104477612e-05,
+      "loss": 0.0,
+      "step": 49168
+    },
+    {
+      "epoch": 45.87,
+      "learning_rate": 4.426632462686567e-05,
+      "loss": 0.0001,
+      "step": 49172
+    },
+    {
+      "epoch": 45.87,
+      "learning_rate": 4.4265858208955226e-05,
+      "loss": 0.0,
+      "step": 49176
+    },
+    {
+      "epoch": 45.88,
+      "learning_rate": 4.426539179104478e-05,
+      "loss": 0.0014,
+      "step": 49180
+    },
+    {
+      "epoch": 45.88,
+      "learning_rate": 4.426492537313433e-05,
+      "loss": 0.0,
+      "step": 49184
+    },
+    {
+      "epoch": 45.88,
+      "learning_rate": 4.4264458955223884e-05,
+      "loss": 0.0,
+      "step": 49188
+    },
+    {
+      "epoch": 45.89,
+      "learning_rate": 4.426399253731344e-05,
+      "loss": 0.0,
+      "step": 49192
+    },
+    {
+      "epoch": 45.89,
+      "learning_rate": 4.426352611940299e-05,
+      "loss": 0.0012,
+      "step": 49196
+    },
+    {
+      "epoch": 45.9,
+      "learning_rate": 4.4263059701492535e-05,
+      "loss": 0.0,
+      "step": 49200
+    },
+    {
+      "epoch": 45.9,
+      "learning_rate": 4.426259328358209e-05,
+      "loss": 0.0,
+      "step": 49204
+    },
+    {
+      "epoch": 45.9,
+      "learning_rate": 4.4262126865671645e-05,
+      "loss": 0.0,
+      "step": 49208
+    },
+    {
+      "epoch": 45.91,
+      "learning_rate": 4.426166044776119e-05,
+      "loss": 0.0,
+      "step": 49212
+    },
+    {
+      "epoch": 45.91,
+      "learning_rate": 4.426119402985075e-05,
+      "loss": 0.0,
+      "step": 49216
+    },
+    {
+      "epoch": 45.91,
+      "learning_rate": 4.42607276119403e-05,
+      "loss": 0.0003,
+      "step": 49220
+    },
+    {
+      "epoch": 45.92,
+      "learning_rate": 4.426026119402985e-05,
+      "loss": 0.0004,
+      "step": 49224
+    },
+    {
+      "epoch": 45.92,
+      "learning_rate": 4.4259794776119406e-05,
+      "loss": 0.0,
+      "step": 49228
+    },
+    {
+      "epoch": 45.93,
+      "learning_rate": 4.4259328358208954e-05,
+      "loss": 0.0,
+      "step": 49232
+    },
+    {
+      "epoch": 45.93,
+      "learning_rate": 4.425886194029851e-05,
+      "loss": 0.0021,
+      "step": 49236
+    },
+    {
+      "epoch": 45.93,
+      "learning_rate": 4.4258395522388064e-05,
+      "loss": 0.0017,
+      "step": 49240
+    },
+    {
+      "epoch": 45.94,
+      "learning_rate": 4.425792910447761e-05,
+      "loss": 0.0001,
+      "step": 49244
+    },
+    {
+      "epoch": 45.94,
+      "learning_rate": 4.425746268656717e-05,
+      "loss": 0.0,
+      "step": 49248
+    },
+    {
+      "epoch": 45.94,
+      "learning_rate": 4.425699626865672e-05,
+      "loss": 0.0017,
+      "step": 49252
+    },
+    {
+      "epoch": 45.95,
+      "learning_rate": 4.425652985074627e-05,
+      "loss": 0.0,
+      "step": 49256
+    },
+    {
+      "epoch": 45.95,
+      "learning_rate": 4.425606343283582e-05,
+      "loss": 0.0,
+      "step": 49260
+    },
+    {
+      "epoch": 45.96,
+      "learning_rate": 4.425559701492537e-05,
+      "loss": 0.0001,
+      "step": 49264
+    },
+    {
+      "epoch": 45.96,
+      "learning_rate": 4.425513059701493e-05,
+      "loss": 0.0,
+      "step": 49268
+    },
+    {
+      "epoch": 45.96,
+      "learning_rate": 4.425466417910448e-05,
+      "loss": 0.0034,
+      "step": 49272
+    },
+    {
+      "epoch": 45.97,
+      "learning_rate": 4.425419776119403e-05,
+      "loss": 0.0,
+      "step": 49276
+    },
+    {
+      "epoch": 45.97,
+      "learning_rate": 4.4253731343283586e-05,
+      "loss": 0.0001,
+      "step": 49280
+    },
+    {
+      "epoch": 45.97,
+      "learning_rate": 4.425326492537314e-05,
+      "loss": 0.0,
+      "step": 49284
+    },
+    {
+      "epoch": 45.98,
+      "learning_rate": 4.425279850746269e-05,
+      "loss": 0.0,
+      "step": 49288
+    },
+    {
+      "epoch": 45.98,
+      "learning_rate": 4.425233208955224e-05,
+      "loss": 0.0004,
+      "step": 49292
+    },
+    {
+      "epoch": 45.98,
+      "learning_rate": 4.42518656716418e-05,
+      "loss": 0.0001,
+      "step": 49296
+    },
+    {
+      "epoch": 45.99,
+      "learning_rate": 4.425139925373135e-05,
+      "loss": 0.0,
+      "step": 49300
+    },
+    {
+      "epoch": 45.99,
+      "learning_rate": 4.4250932835820895e-05,
+      "loss": 0.0,
+      "step": 49304
+    },
+    {
+      "epoch": 46.0,
+      "learning_rate": 4.425046641791045e-05,
+      "loss": 0.0039,
+      "step": 49308
+    },
+    {
+      "epoch": 46.0,
+      "learning_rate": 4.4250000000000005e-05,
+      "loss": 0.0,
+      "step": 49312
+    },
+    {
+      "epoch": 46.0,
+      "learning_rate": 4.424953358208955e-05,
+      "loss": 0.0001,
+      "step": 49316
+    },
+    {
+      "epoch": 46.01,
+      "learning_rate": 4.424906716417911e-05,
+      "loss": 0.0008,
+      "step": 49320
+    },
+    {
+      "epoch": 46.01,
+      "learning_rate": 4.4248600746268656e-05,
+      "loss": 0.0,
+      "step": 49324
+    },
+    {
+      "epoch": 46.01,
+      "learning_rate": 4.424813432835821e-05,
+      "loss": 0.0004,
+      "step": 49328
+    },
+    {
+      "epoch": 46.02,
+      "learning_rate": 4.4247667910447766e-05,
+      "loss": 0.0,
+      "step": 49332
+    },
+    {
+      "epoch": 46.02,
+      "learning_rate": 4.4247201492537314e-05,
+      "loss": 0.0,
+      "step": 49336
+    },
+    {
+      "epoch": 46.03,
+      "learning_rate": 4.424673507462687e-05,
+      "loss": 0.0,
+      "step": 49340
+    },
+    {
+      "epoch": 46.03,
+      "learning_rate": 4.4246268656716424e-05,
+      "loss": 0.0,
+      "step": 49344
+    },
+    {
+      "epoch": 46.03,
+      "learning_rate": 4.424580223880597e-05,
+      "loss": 0.0,
+      "step": 49348
+    },
+    {
+      "epoch": 46.04,
+      "learning_rate": 4.424533582089552e-05,
+      "loss": 0.0,
+      "step": 49352
+    },
+    {
+      "epoch": 46.04,
+      "learning_rate": 4.424486940298508e-05,
+      "loss": 0.0,
+      "step": 49356
+    },
+    {
+      "epoch": 46.04,
+      "learning_rate": 4.424440298507463e-05,
+      "loss": 0.0,
+      "step": 49360
+    },
+    {
+      "epoch": 46.05,
+      "learning_rate": 4.424393656716418e-05,
+      "loss": 0.0,
+      "step": 49364
+    },
+    {
+      "epoch": 46.05,
+      "learning_rate": 4.424347014925373e-05,
+      "loss": 0.0065,
+      "step": 49368
+    },
+    {
+      "epoch": 46.06,
+      "learning_rate": 4.424300373134329e-05,
+      "loss": 0.0,
+      "step": 49372
+    },
+    {
+      "epoch": 46.06,
+      "learning_rate": 4.4242537313432836e-05,
+      "loss": 0.0,
+      "step": 49376
+    },
+    {
+      "epoch": 46.06,
+      "learning_rate": 4.424207089552239e-05,
+      "loss": 0.0,
+      "step": 49380
+    },
+    {
+      "epoch": 46.07,
+      "learning_rate": 4.424160447761194e-05,
+      "loss": 0.0014,
+      "step": 49384
+    },
+    {
+      "epoch": 46.07,
+      "learning_rate": 4.4241138059701494e-05,
+      "loss": 0.0,
+      "step": 49388
+    },
+    {
+      "epoch": 46.07,
+      "learning_rate": 4.424067164179105e-05,
+      "loss": 0.0,
+      "step": 49392
+    },
+    {
+      "epoch": 46.08,
+      "learning_rate": 4.42402052238806e-05,
+      "loss": 0.0001,
+      "step": 49396
+    },
+    {
+      "epoch": 46.08,
+      "learning_rate": 4.423973880597015e-05,
+      "loss": 0.0001,
+      "step": 49400
+    },
+    {
+      "epoch": 46.09,
+      "learning_rate": 4.423927238805971e-05,
+      "loss": 0.0,
+      "step": 49404
+    },
+    {
+      "epoch": 46.09,
+      "learning_rate": 4.4238805970149255e-05,
+      "loss": 0.0,
+      "step": 49408
+    },
+    {
+      "epoch": 46.09,
+      "learning_rate": 4.42383395522388e-05,
+      "loss": 0.0,
+      "step": 49412
+    },
+    {
+      "epoch": 46.1,
+      "learning_rate": 4.4237873134328365e-05,
+      "loss": 0.0,
+      "step": 49416
+    },
+    {
+      "epoch": 46.1,
+      "learning_rate": 4.423740671641791e-05,
+      "loss": 0.0,
+      "step": 49420
+    },
+    {
+      "epoch": 46.1,
+      "learning_rate": 4.423694029850746e-05,
+      "loss": 0.0001,
+      "step": 49424
+    },
+    {
+      "epoch": 46.11,
+      "learning_rate": 4.4236473880597016e-05,
+      "loss": 0.0,
+      "step": 49428
+    },
+    {
+      "epoch": 46.11,
+      "learning_rate": 4.423600746268657e-05,
+      "loss": 0.0,
+      "step": 49432
+    },
+    {
+      "epoch": 46.12,
+      "learning_rate": 4.4235541044776126e-05,
+      "loss": 0.0,
+      "step": 49436
+    },
+    {
+      "epoch": 46.12,
+      "learning_rate": 4.4235074626865674e-05,
+      "loss": 0.0,
+      "step": 49440
+    },
+    {
+      "epoch": 46.12,
+      "learning_rate": 4.423460820895522e-05,
+      "loss": 0.0,
+      "step": 49444
+    },
+    {
+      "epoch": 46.13,
+      "learning_rate": 4.4234141791044784e-05,
+      "loss": 0.0001,
+      "step": 49448
+    },
+    {
+      "epoch": 46.13,
+      "learning_rate": 4.423367537313433e-05,
+      "loss": 0.0002,
+      "step": 49452
+    },
+    {
+      "epoch": 46.13,
+      "learning_rate": 4.423320895522388e-05,
+      "loss": 0.0,
+      "step": 49456
+    },
+    {
+      "epoch": 46.14,
+      "learning_rate": 4.4232742537313435e-05,
+      "loss": 0.0001,
+      "step": 49460
+    },
+    {
+      "epoch": 46.14,
+      "learning_rate": 4.423227611940299e-05,
+      "loss": 0.0061,
+      "step": 49464
+    },
+    {
+      "epoch": 46.15,
+      "learning_rate": 4.423180970149254e-05,
+      "loss": 0.0004,
+      "step": 49468
+    },
+    {
+      "epoch": 46.15,
+      "learning_rate": 4.423134328358209e-05,
+      "loss": 0.0,
+      "step": 49472
+    },
+    {
+      "epoch": 46.15,
+      "learning_rate": 4.423087686567165e-05,
+      "loss": 0.0001,
+      "step": 49476
+    },
+    {
+      "epoch": 46.16,
+      "learning_rate": 4.4230410447761196e-05,
+      "loss": 0.0,
+      "step": 49480
+    },
+    {
+      "epoch": 46.16,
+      "learning_rate": 4.422994402985075e-05,
+      "loss": 0.0,
+      "step": 49484
+    },
+    {
+      "epoch": 46.16,
+      "learning_rate": 4.42294776119403e-05,
+      "loss": 0.0002,
+      "step": 49488
+    },
+    {
+      "epoch": 46.17,
+      "learning_rate": 4.4229011194029854e-05,
+      "loss": 0.0,
+      "step": 49492
+    },
+    {
+      "epoch": 46.17,
+      "learning_rate": 4.422854477611941e-05,
+      "loss": 0.0,
+      "step": 49496
+    },
+    {
+      "epoch": 46.18,
+      "learning_rate": 4.422807835820896e-05,
+      "loss": 0.0006,
+      "step": 49500
+    },
+    {
+      "epoch": 46.18,
+      "eval_exact_match": 0.7214700193423598,
+      "eval_exec": 0.7572533849129593,
+      "eval_loss": 0.45067957043647766,
+      "eval_runtime": 1072.9162,
+      "eval_samples_per_second": 0.964,
+      "step": 49500
+    },
+    {
+      "epoch": 46.18,
+      "learning_rate": 4.4227611940298505e-05,
+      "loss": 0.0001,
+      "step": 49504
+    },
+    {
+      "epoch": 46.18,
+      "learning_rate": 4.422714552238807e-05,
+      "loss": 0.0,
+      "step": 49508
+    },
+    {
+      "epoch": 46.19,
+      "learning_rate": 4.4226679104477615e-05,
+      "loss": 0.0001,
+      "step": 49512
+    },
+    {
+      "epoch": 46.19,
+      "learning_rate": 4.422621268656716e-05,
+      "loss": 0.0,
+      "step": 49516
+    },
+    {
+      "epoch": 46.19,
+      "learning_rate": 4.422574626865672e-05,
+      "loss": 0.0001,
+      "step": 49520
+    },
+    {
+      "epoch": 46.2,
+      "learning_rate": 4.422527985074627e-05,
+      "loss": 0.0009,
+      "step": 49524
+    },
+    {
+      "epoch": 46.2,
+      "learning_rate": 4.422481343283582e-05,
+      "loss": 0.0,
+      "step": 49528
+    },
+    {
+      "epoch": 46.21,
+      "learning_rate": 4.4224347014925376e-05,
+      "loss": 0.0,
+      "step": 49532
+    },
+    {
+      "epoch": 46.21,
+      "learning_rate": 4.4223880597014924e-05,
+      "loss": 0.0,
+      "step": 49536
+    },
+    {
+      "epoch": 46.21,
+      "learning_rate": 4.422341417910448e-05,
+      "loss": 0.0001,
+      "step": 49540
+    },
+    {
+      "epoch": 46.22,
+      "learning_rate": 4.4222947761194034e-05,
+      "loss": 0.0,
+      "step": 49544
+    },
+    {
+      "epoch": 46.22,
+      "learning_rate": 4.422248134328358e-05,
+      "loss": 0.0009,
+      "step": 49548
+    },
+    {
+      "epoch": 46.22,
+      "learning_rate": 4.422201492537314e-05,
+      "loss": 0.0003,
+      "step": 49552
+    },
+    {
+      "epoch": 46.23,
+      "learning_rate": 4.422154850746269e-05,
+      "loss": 0.0022,
+      "step": 49556
+    },
+    {
+      "epoch": 46.23,
+      "learning_rate": 4.422108208955224e-05,
+      "loss": 0.0003,
+      "step": 49560
+    },
+    {
+      "epoch": 46.24,
+      "learning_rate": 4.422061567164179e-05,
+      "loss": 0.0001,
+      "step": 49564
+    },
+    {
+      "epoch": 46.24,
+      "learning_rate": 4.422014925373135e-05,
+      "loss": 0.0,
+      "step": 49568
+    },
+    {
+      "epoch": 46.24,
+      "learning_rate": 4.42196828358209e-05,
+      "loss": 0.0003,
+      "step": 49572
+    },
+    {
+      "epoch": 46.25,
+      "learning_rate": 4.4219216417910446e-05,
+      "loss": 0.0,
+      "step": 49576
+    },
+    {
+      "epoch": 46.25,
+      "learning_rate": 4.421875e-05,
+      "loss": 0.0,
+      "step": 49580
+    },
+    {
+      "epoch": 46.25,
+      "learning_rate": 4.4218283582089556e-05,
+      "loss": 0.0,
+      "step": 49584
+    },
+    {
+      "epoch": 46.26,
+      "learning_rate": 4.4217817164179104e-05,
+      "loss": 0.0004,
+      "step": 49588
+    },
+    {
+      "epoch": 46.26,
+      "learning_rate": 4.421735074626866e-05,
+      "loss": 0.0,
+      "step": 49592
+    },
+    {
+      "epoch": 46.26,
+      "learning_rate": 4.421688432835821e-05,
+      "loss": 0.0,
+      "step": 49596
+    },
+    {
+      "epoch": 46.27,
+      "learning_rate": 4.421641791044777e-05,
+      "loss": 0.0,
+      "step": 49600
+    },
+    {
+      "epoch": 46.27,
+      "learning_rate": 4.421595149253732e-05,
+      "loss": 0.0006,
+      "step": 49604
+    },
+    {
+      "epoch": 46.28,
+      "learning_rate": 4.4215485074626865e-05,
+      "loss": 0.0,
+      "step": 49608
+    },
+    {
+      "epoch": 46.28,
+      "learning_rate": 4.421501865671642e-05,
+      "loss": 0.0,
+      "step": 49612
+    },
+    {
+      "epoch": 46.28,
+      "learning_rate": 4.4214552238805975e-05,
+      "loss": 0.0008,
+      "step": 49616
+    },
+    {
+      "epoch": 46.29,
+      "learning_rate": 4.421408582089552e-05,
+      "loss": 0.0,
+      "step": 49620
+    },
+    {
+      "epoch": 46.29,
+      "learning_rate": 4.421361940298508e-05,
+      "loss": 0.0,
+      "step": 49624
+    },
+    {
+      "epoch": 46.29,
+      "learning_rate": 4.421315298507463e-05,
+      "loss": 0.0,
+      "step": 49628
+    },
+    {
+      "epoch": 46.3,
+      "learning_rate": 4.421268656716418e-05,
+      "loss": 0.0032,
+      "step": 49632
+    },
+    {
+      "epoch": 46.3,
+      "learning_rate": 4.4212220149253736e-05,
+      "loss": 0.0,
+      "step": 49636
+    },
+    {
+      "epoch": 46.31,
+      "learning_rate": 4.4211753731343284e-05,
+      "loss": 0.0001,
+      "step": 49640
+    },
+    {
+      "epoch": 46.31,
+      "learning_rate": 4.421128731343284e-05,
+      "loss": 0.0003,
+      "step": 49644
+    },
+    {
+      "epoch": 46.31,
+      "learning_rate": 4.4210820895522394e-05,
+      "loss": 0.0011,
+      "step": 49648
+    },
+    {
+      "epoch": 46.32,
+      "learning_rate": 4.421035447761194e-05,
+      "loss": 0.0,
+      "step": 49652
+    },
+    {
+      "epoch": 46.32,
+      "learning_rate": 4.420988805970149e-05,
+      "loss": 0.0,
+      "step": 49656
+    },
+    {
+      "epoch": 46.32,
+      "learning_rate": 4.420942164179105e-05,
+      "loss": 0.0,
+      "step": 49660
+    },
+    {
+      "epoch": 46.33,
+      "learning_rate": 4.42089552238806e-05,
+      "loss": 0.0,
+      "step": 49664
+    },
+    {
+      "epoch": 46.33,
+      "learning_rate": 4.420848880597015e-05,
+      "loss": 0.0004,
+      "step": 49668
+    },
+    {
+      "epoch": 46.34,
+      "learning_rate": 4.42080223880597e-05,
+      "loss": 0.0,
+      "step": 49672
+    },
+    {
+      "epoch": 46.34,
+      "learning_rate": 4.420755597014926e-05,
+      "loss": 0.0007,
+      "step": 49676
+    },
+    {
+      "epoch": 46.34,
+      "learning_rate": 4.4207089552238806e-05,
+      "loss": 0.0,
+      "step": 49680
+    },
+    {
+      "epoch": 46.35,
+      "learning_rate": 4.420662313432836e-05,
+      "loss": 0.0001,
+      "step": 49684
+    },
+    {
+      "epoch": 46.35,
+      "learning_rate": 4.4206156716417916e-05,
+      "loss": 0.0,
+      "step": 49688
+    },
+    {
+      "epoch": 46.35,
+      "learning_rate": 4.4205690298507464e-05,
+      "loss": 0.0006,
+      "step": 49692
+    },
+    {
+      "epoch": 46.36,
+      "learning_rate": 4.420522388059702e-05,
+      "loss": 0.0027,
+      "step": 49696
+    },
+    {
+      "epoch": 46.36,
+      "learning_rate": 4.420475746268657e-05,
+      "loss": 0.007,
+      "step": 49700
+    },
+    {
+      "epoch": 46.37,
+      "learning_rate": 4.420429104477612e-05,
+      "loss": 0.0,
+      "step": 49704
+    },
+    {
+      "epoch": 46.37,
+      "learning_rate": 4.4203824626865677e-05,
+      "loss": 0.0,
+      "step": 49708
+    },
+    {
+      "epoch": 46.37,
+      "learning_rate": 4.4203358208955225e-05,
+      "loss": 0.0001,
+      "step": 49712
+    },
+    {
+      "epoch": 46.38,
+      "learning_rate": 4.420289179104477e-05,
+      "loss": 0.0,
+      "step": 49716
+    },
+    {
+      "epoch": 46.38,
+      "learning_rate": 4.4202425373134334e-05,
+      "loss": 0.0003,
+      "step": 49720
+    },
+    {
+      "epoch": 46.38,
+      "learning_rate": 4.420195895522388e-05,
+      "loss": 0.0004,
+      "step": 49724
+    },
+    {
+      "epoch": 46.39,
+      "learning_rate": 4.420149253731343e-05,
+      "loss": 0.0,
+      "step": 49728
+    },
+    {
+      "epoch": 46.39,
+      "learning_rate": 4.4201026119402986e-05,
+      "loss": 0.0,
+      "step": 49732
+    },
+    {
+      "epoch": 46.4,
+      "learning_rate": 4.420055970149254e-05,
+      "loss": 0.0,
+      "step": 49736
+    },
+    {
+      "epoch": 46.4,
+      "learning_rate": 4.420009328358209e-05,
+      "loss": 0.0,
+      "step": 49740
+    },
+    {
+      "epoch": 46.4,
+      "learning_rate": 4.4199626865671644e-05,
+      "loss": 0.0,
+      "step": 49744
+    },
+    {
+      "epoch": 46.41,
+      "learning_rate": 4.41991604477612e-05,
+      "loss": 0.0,
+      "step": 49748
+    },
+    {
+      "epoch": 46.41,
+      "learning_rate": 4.419869402985075e-05,
+      "loss": 0.0,
+      "step": 49752
+    },
+    {
+      "epoch": 46.41,
+      "learning_rate": 4.41982276119403e-05,
+      "loss": 0.0,
+      "step": 49756
+    },
+    {
+      "epoch": 46.42,
+      "learning_rate": 4.419776119402985e-05,
+      "loss": 0.0004,
+      "step": 49760
+    },
+    {
+      "epoch": 46.42,
+      "learning_rate": 4.4197294776119405e-05,
+      "loss": 0.0001,
+      "step": 49764
+    },
+    {
+      "epoch": 46.43,
+      "learning_rate": 4.419682835820896e-05,
+      "loss": 0.0,
+      "step": 49768
+    },
+    {
+      "epoch": 46.43,
+      "learning_rate": 4.419636194029851e-05,
+      "loss": 0.0,
+      "step": 49772
+    },
+    {
+      "epoch": 46.43,
+      "learning_rate": 4.419589552238806e-05,
+      "loss": 0.0,
+      "step": 49776
+    },
+    {
+      "epoch": 46.44,
+      "learning_rate": 4.419542910447762e-05,
+      "loss": 0.0001,
+      "step": 49780
+    },
+    {
+      "epoch": 46.44,
+      "learning_rate": 4.4194962686567166e-05,
+      "loss": 0.0,
+      "step": 49784
+    },
+    {
+      "epoch": 46.44,
+      "learning_rate": 4.419449626865672e-05,
+      "loss": 0.0,
+      "step": 49788
+    },
+    {
+      "epoch": 46.45,
+      "learning_rate": 4.419402985074627e-05,
+      "loss": 0.0001,
+      "step": 49792
+    },
+    {
+      "epoch": 46.45,
+      "learning_rate": 4.4193563432835823e-05,
+      "loss": 0.0,
+      "step": 49796
+    },
+    {
+      "epoch": 46.46,
+      "learning_rate": 4.419309701492538e-05,
+      "loss": 0.0,
+      "step": 49800
+    },
+    {
+      "epoch": 46.46,
+      "learning_rate": 4.4192630597014927e-05,
+      "loss": 0.0,
+      "step": 49804
+    },
+    {
+      "epoch": 46.46,
+      "learning_rate": 4.419216417910448e-05,
+      "loss": 0.0,
+      "step": 49808
+    },
+    {
+      "epoch": 46.47,
+      "learning_rate": 4.4191697761194036e-05,
+      "loss": 0.0,
+      "step": 49812
+    },
+    {
+      "epoch": 46.47,
+      "learning_rate": 4.4191231343283584e-05,
+      "loss": 0.0,
+      "step": 49816
+    },
+    {
+      "epoch": 46.47,
+      "learning_rate": 4.419076492537313e-05,
+      "loss": 0.001,
+      "step": 49820
+    },
+    {
+      "epoch": 46.48,
+      "learning_rate": 4.419029850746269e-05,
+      "loss": 0.0,
+      "step": 49824
+    },
+    {
+      "epoch": 46.48,
+      "learning_rate": 4.418983208955224e-05,
+      "loss": 0.0006,
+      "step": 49828
+    },
+    {
+      "epoch": 46.49,
+      "learning_rate": 4.418936567164179e-05,
+      "loss": 0.0,
+      "step": 49832
+    },
+    {
+      "epoch": 46.49,
+      "learning_rate": 4.4188899253731345e-05,
+      "loss": 0.001,
+      "step": 49836
+    },
+    {
+      "epoch": 46.49,
+      "learning_rate": 4.41884328358209e-05,
+      "loss": 0.0,
+      "step": 49840
+    },
+    {
+      "epoch": 46.5,
+      "learning_rate": 4.418796641791045e-05,
+      "loss": 0.0,
+      "step": 49844
+    },
+    {
+      "epoch": 46.5,
+      "learning_rate": 4.4187500000000003e-05,
+      "loss": 0.0,
+      "step": 49848
+    },
+    {
+      "epoch": 46.5,
+      "learning_rate": 4.418703358208955e-05,
+      "loss": 0.0013,
+      "step": 49852
+    },
+    {
+      "epoch": 46.51,
+      "learning_rate": 4.4186567164179106e-05,
+      "loss": 0.0,
+      "step": 49856
+    },
+    {
+      "epoch": 46.51,
+      "learning_rate": 4.418610074626866e-05,
+      "loss": 0.0,
+      "step": 49860
+    },
+    {
+      "epoch": 46.51,
+      "learning_rate": 4.418563432835821e-05,
+      "loss": 0.0004,
+      "step": 49864
+    },
+    {
+      "epoch": 46.52,
+      "learning_rate": 4.4185167910447764e-05,
+      "loss": 0.0,
+      "step": 49868
+    },
+    {
+      "epoch": 46.52,
+      "learning_rate": 4.418470149253732e-05,
+      "loss": 0.0006,
+      "step": 49872
+    },
+    {
+      "epoch": 46.53,
+      "learning_rate": 4.418423507462687e-05,
+      "loss": 0.0,
+      "step": 49876
+    },
+    {
+      "epoch": 46.53,
+      "learning_rate": 4.4183768656716416e-05,
+      "loss": 0.0,
+      "step": 49880
+    },
+    {
+      "epoch": 46.53,
+      "learning_rate": 4.418330223880597e-05,
+      "loss": 0.0001,
+      "step": 49884
+    },
+    {
+      "epoch": 46.54,
+      "learning_rate": 4.4182835820895525e-05,
+      "loss": 0.0,
+      "step": 49888
+    },
+    {
+      "epoch": 46.54,
+      "learning_rate": 4.4182369402985074e-05,
+      "loss": 0.0,
+      "step": 49892
+    },
+    {
+      "epoch": 46.54,
+      "learning_rate": 4.418190298507463e-05,
+      "loss": 0.0013,
+      "step": 49896
+    },
+    {
+      "epoch": 46.55,
+      "learning_rate": 4.418143656716418e-05,
+      "loss": 0.0,
+      "step": 49900
+    },
+    {
+      "epoch": 46.55,
+      "learning_rate": 4.418097014925373e-05,
+      "loss": 0.0,
+      "step": 49904
+    },
+    {
+      "epoch": 46.56,
+      "learning_rate": 4.4180503731343286e-05,
+      "loss": 0.0012,
+      "step": 49908
+    },
+    {
+      "epoch": 46.56,
+      "learning_rate": 4.4180037313432835e-05,
+      "loss": 0.0005,
+      "step": 49912
+    },
+    {
+      "epoch": 46.56,
+      "learning_rate": 4.417957089552239e-05,
+      "loss": 0.0,
+      "step": 49916
+    },
+    {
+      "epoch": 46.57,
+      "learning_rate": 4.4179104477611944e-05,
+      "loss": 0.0027,
+      "step": 49920
+    },
+    {
+      "epoch": 46.57,
+      "learning_rate": 4.417863805970149e-05,
+      "loss": 0.0001,
+      "step": 49924
+    },
+    {
+      "epoch": 46.57,
+      "learning_rate": 4.417817164179105e-05,
+      "loss": 0.0,
+      "step": 49928
+    },
+    {
+      "epoch": 46.58,
+      "learning_rate": 4.41777052238806e-05,
+      "loss": 0.0,
+      "step": 49932
+    },
+    {
+      "epoch": 46.58,
+      "learning_rate": 4.417723880597015e-05,
+      "loss": 0.0002,
+      "step": 49936
+    },
+    {
+      "epoch": 46.59,
+      "learning_rate": 4.4176772388059705e-05,
+      "loss": 0.0,
+      "step": 49940
+    },
+    {
+      "epoch": 46.59,
+      "learning_rate": 4.4176305970149253e-05,
+      "loss": 0.0,
+      "step": 49944
+    },
+    {
+      "epoch": 46.59,
+      "learning_rate": 4.417583955223881e-05,
+      "loss": 0.0028,
+      "step": 49948
+    },
+    {
+      "epoch": 46.6,
+      "learning_rate": 4.417537313432836e-05,
+      "loss": 0.0,
+      "step": 49952
+    },
+    {
+      "epoch": 46.6,
+      "learning_rate": 4.417490671641791e-05,
+      "loss": 0.0002,
+      "step": 49956
+    },
+    {
+      "epoch": 46.6,
+      "learning_rate": 4.4174440298507466e-05,
+      "loss": 0.0001,
+      "step": 49960
+    },
+    {
+      "epoch": 46.61,
+      "learning_rate": 4.417397388059702e-05,
+      "loss": 0.0,
+      "step": 49964
+    },
+    {
+      "epoch": 46.61,
+      "learning_rate": 4.417350746268657e-05,
+      "loss": 0.0,
+      "step": 49968
+    },
+    {
+      "epoch": 46.62,
+      "learning_rate": 4.417304104477612e-05,
+      "loss": 0.0,
+      "step": 49972
+    },
+    {
+      "epoch": 46.62,
+      "learning_rate": 4.417257462686568e-05,
+      "loss": 0.0,
+      "step": 49976
+    },
+    {
+      "epoch": 46.62,
+      "learning_rate": 4.417210820895523e-05,
+      "loss": 0.0004,
+      "step": 49980
+    },
+    {
+      "epoch": 46.63,
+      "learning_rate": 4.4171641791044775e-05,
+      "loss": 0.0001,
+      "step": 49984
+    },
+    {
+      "epoch": 46.63,
+      "learning_rate": 4.417117537313433e-05,
+      "loss": 0.0002,
+      "step": 49988
+    },
+    {
+      "epoch": 46.63,
+      "learning_rate": 4.4170708955223885e-05,
+      "loss": 0.0,
+      "step": 49992
+    },
+    {
+      "epoch": 46.64,
+      "learning_rate": 4.417024253731343e-05,
+      "loss": 0.0,
+      "step": 49996
+    },
+    {
+      "epoch": 46.64,
+      "learning_rate": 4.416977611940299e-05,
+      "loss": 0.0001,
+      "step": 50000
+    },
+    {
+      "epoch": 46.64,
+      "eval_exact_match": 0.7156673114119922,
+      "eval_exec": 0.7485493230174082,
+      "eval_loss": 0.4576376974582672,
+      "eval_runtime": 1189.333,
+      "eval_samples_per_second": 0.869,
+      "step": 50000
+    },
+    {
+      "epoch": 46.65,
+      "learning_rate": 4.4169309701492536e-05,
+      "loss": 0.0001,
+      "step": 50004
+    },
+    {
+      "epoch": 46.65,
+      "learning_rate": 4.416884328358209e-05,
+      "loss": 0.0004,
+      "step": 50008
+    },
+    {
+      "epoch": 46.65,
+      "learning_rate": 4.4168376865671646e-05,
+      "loss": 0.0,
+      "step": 50012
+    },
+    {
+      "epoch": 46.66,
+      "learning_rate": 4.4167910447761194e-05,
+      "loss": 0.0,
+      "step": 50016
+    },
+    {
+      "epoch": 46.66,
+      "learning_rate": 4.416744402985075e-05,
+      "loss": 0.0001,
+      "step": 50020
+    },
+    {
+      "epoch": 46.66,
+      "learning_rate": 4.4166977611940304e-05,
+      "loss": 0.0,
+      "step": 50024
+    },
+    {
+      "epoch": 46.67,
+      "learning_rate": 4.416651119402985e-05,
+      "loss": 0.0,
+      "step": 50028
+    },
+    {
+      "epoch": 46.67,
+      "learning_rate": 4.41660447761194e-05,
+      "loss": 0.0,
+      "step": 50032
+    },
+    {
+      "epoch": 46.68,
+      "learning_rate": 4.416557835820896e-05,
+      "loss": 0.0002,
+      "step": 50036
+    },
+    {
+      "epoch": 46.68,
+      "learning_rate": 4.416511194029851e-05,
+      "loss": 0.0,
+      "step": 50040
+    },
+    {
+      "epoch": 46.68,
+      "learning_rate": 4.416464552238806e-05,
+      "loss": 0.0,
+      "step": 50044
+    },
+    {
+      "epoch": 46.69,
+      "learning_rate": 4.416417910447761e-05,
+      "loss": 0.0041,
+      "step": 50048
+    },
+    {
+      "epoch": 46.69,
+      "learning_rate": 4.416371268656717e-05,
+      "loss": 0.0001,
+      "step": 50052
+    },
+    {
+      "epoch": 46.69,
+      "learning_rate": 4.4163246268656716e-05,
+      "loss": 0.0001,
+      "step": 50056
+    },
+    {
+      "epoch": 46.7,
+      "learning_rate": 4.416277985074627e-05,
+      "loss": 0.0019,
+      "step": 50060
+    },
+    {
+      "epoch": 46.7,
+      "learning_rate": 4.416231343283582e-05,
+      "loss": 0.0,
+      "step": 50064
+    },
+    {
+      "epoch": 46.71,
+      "learning_rate": 4.4161847014925374e-05,
+      "loss": 0.0,
+      "step": 50068
+    },
+    {
+      "epoch": 46.71,
+      "learning_rate": 4.416138059701493e-05,
+      "loss": 0.0003,
+      "step": 50072
+    },
+    {
+      "epoch": 46.71,
+      "learning_rate": 4.416091417910448e-05,
+      "loss": 0.0,
+      "step": 50076
+    },
+    {
+      "epoch": 46.72,
+      "learning_rate": 4.416044776119403e-05,
+      "loss": 0.0,
+      "step": 50080
+    },
+    {
+      "epoch": 46.72,
+      "learning_rate": 4.415998134328359e-05,
+      "loss": 0.0,
+      "step": 50084
+    },
+    {
+      "epoch": 46.72,
+      "learning_rate": 4.4159514925373135e-05,
+      "loss": 0.0026,
+      "step": 50088
+    },
+    {
+      "epoch": 46.73,
+      "learning_rate": 4.415904850746269e-05,
+      "loss": 0.0001,
+      "step": 50092
+    },
+    {
+      "epoch": 46.73,
+      "learning_rate": 4.4158582089552245e-05,
+      "loss": 0.0,
+      "step": 50096
+    },
+    {
+      "epoch": 46.73,
+      "learning_rate": 4.415811567164179e-05,
+      "loss": 0.0,
+      "step": 50100
+    },
+    {
+      "epoch": 46.74,
+      "learning_rate": 4.415764925373135e-05,
+      "loss": 0.0,
+      "step": 50104
+    },
+    {
+      "epoch": 46.74,
+      "learning_rate": 4.4157182835820896e-05,
+      "loss": 0.0,
+      "step": 50108
+    },
+    {
+      "epoch": 46.75,
+      "learning_rate": 4.415671641791045e-05,
+      "loss": 0.0,
+      "step": 50112
+    },
+    {
+      "epoch": 46.75,
+      "learning_rate": 4.4156250000000006e-05,
+      "loss": 0.0001,
+      "step": 50116
+    },
+    {
+      "epoch": 46.75,
+      "learning_rate": 4.4155783582089554e-05,
+      "loss": 0.0018,
+      "step": 50120
+    },
+    {
+      "epoch": 46.76,
+      "learning_rate": 4.41553171641791e-05,
+      "loss": 0.0001,
+      "step": 50124
+    },
+    {
+      "epoch": 46.76,
+      "learning_rate": 4.4154850746268664e-05,
+      "loss": 0.0,
+      "step": 50128
+    },
+    {
+      "epoch": 46.76,
+      "learning_rate": 4.415438432835821e-05,
+      "loss": 0.0,
+      "step": 50132
+    },
+    {
+      "epoch": 46.77,
+      "learning_rate": 4.415391791044776e-05,
+      "loss": 0.0001,
+      "step": 50136
+    },
+    {
+      "epoch": 46.77,
+      "learning_rate": 4.4153451492537315e-05,
+      "loss": 0.0014,
+      "step": 50140
+    },
+    {
+      "epoch": 46.78,
+      "learning_rate": 4.415298507462687e-05,
+      "loss": 0.0,
+      "step": 50144
+    },
+    {
+      "epoch": 46.78,
+      "learning_rate": 4.415251865671642e-05,
+      "loss": 0.0001,
+      "step": 50148
+    },
+    {
+      "epoch": 46.78,
+      "learning_rate": 4.415205223880597e-05,
+      "loss": 0.0,
+      "step": 50152
+    },
+    {
+      "epoch": 46.79,
+      "learning_rate": 4.415158582089553e-05,
+      "loss": 0.0056,
+      "step": 50156
+    },
+    {
+      "epoch": 46.79,
+      "learning_rate": 4.4151119402985076e-05,
+      "loss": 0.0,
+      "step": 50160
+    },
+    {
+      "epoch": 46.79,
+      "learning_rate": 4.415065298507463e-05,
+      "loss": 0.0,
+      "step": 50164
+    },
+    {
+      "epoch": 46.8,
+      "learning_rate": 4.415018656716418e-05,
+      "loss": 0.0,
+      "step": 50168
+    },
+    {
+      "epoch": 46.8,
+      "learning_rate": 4.4149720149253734e-05,
+      "loss": 0.0001,
+      "step": 50172
+    },
+    {
+      "epoch": 46.81,
+      "learning_rate": 4.414925373134329e-05,
+      "loss": 0.0,
+      "step": 50176
+    },
+    {
+      "epoch": 46.81,
+      "learning_rate": 4.414878731343284e-05,
+      "loss": 0.0001,
+      "step": 50180
+    },
+    {
+      "epoch": 46.81,
+      "learning_rate": 4.4148320895522385e-05,
+      "loss": 0.0,
+      "step": 50184
+    },
+    {
+      "epoch": 46.82,
+      "learning_rate": 4.414785447761195e-05,
+      "loss": 0.0001,
+      "step": 50188
+    },
+    {
+      "epoch": 46.82,
+      "learning_rate": 4.4147388059701495e-05,
+      "loss": 0.0001,
+      "step": 50192
+    },
+    {
+      "epoch": 46.82,
+      "learning_rate": 4.414692164179104e-05,
+      "loss": 0.0,
+      "step": 50196
+    },
+    {
+      "epoch": 46.83,
+      "learning_rate": 4.41464552238806e-05,
+      "loss": 0.0,
+      "step": 50200
+    },
+    {
+      "epoch": 46.83,
+      "learning_rate": 4.414598880597015e-05,
+      "loss": 0.0,
+      "step": 50204
+    },
+    {
+      "epoch": 46.84,
+      "learning_rate": 4.41455223880597e-05,
+      "loss": 0.0,
+      "step": 50208
+    },
+    {
+      "epoch": 46.84,
+      "learning_rate": 4.4145055970149256e-05,
+      "loss": 0.0001,
+      "step": 50212
+    },
+    {
+      "epoch": 46.84,
+      "learning_rate": 4.4144589552238804e-05,
+      "loss": 0.0,
+      "step": 50216
+    },
+    {
+      "epoch": 46.85,
+      "learning_rate": 4.414412313432836e-05,
+      "loss": 0.0,
+      "step": 50220
+    },
+    {
+      "epoch": 46.85,
+      "learning_rate": 4.4143656716417914e-05,
+      "loss": 0.0001,
+      "step": 50224
+    },
+    {
+      "epoch": 46.85,
+      "learning_rate": 4.414319029850746e-05,
+      "loss": 0.0002,
+      "step": 50228
+    },
+    {
+      "epoch": 46.86,
+      "learning_rate": 4.414272388059702e-05,
+      "loss": 0.0,
+      "step": 50232
+    },
+    {
+      "epoch": 46.86,
+      "learning_rate": 4.414225746268657e-05,
+      "loss": 0.0001,
+      "step": 50236
+    },
+    {
+      "epoch": 46.87,
+      "learning_rate": 4.414179104477612e-05,
+      "loss": 0.0,
+      "step": 50240
+    },
+    {
+      "epoch": 46.87,
+      "learning_rate": 4.414132462686567e-05,
+      "loss": 0.0003,
+      "step": 50244
+    },
+    {
+      "epoch": 46.87,
+      "learning_rate": 4.414085820895523e-05,
+      "loss": 0.0,
+      "step": 50248
+    },
+    {
+      "epoch": 46.88,
+      "learning_rate": 4.414039179104478e-05,
+      "loss": 0.0001,
+      "step": 50252
+    },
+    {
+      "epoch": 46.88,
+      "learning_rate": 4.413992537313433e-05,
+      "loss": 0.0,
+      "step": 50256
+    },
+    {
+      "epoch": 46.88,
+      "learning_rate": 4.413945895522388e-05,
+      "loss": 0.0,
+      "step": 50260
+    },
+    {
+      "epoch": 46.89,
+      "learning_rate": 4.4138992537313436e-05,
+      "loss": 0.0,
+      "step": 50264
+    },
+    {
+      "epoch": 46.89,
+      "learning_rate": 4.413852611940299e-05,
+      "loss": 0.0001,
+      "step": 50268
+    },
+    {
+      "epoch": 46.9,
+      "learning_rate": 4.413805970149254e-05,
+      "loss": 0.0019,
+      "step": 50272
+    },
+    {
+      "epoch": 46.9,
+      "learning_rate": 4.413759328358209e-05,
+      "loss": 0.005,
+      "step": 50276
+    },
+    {
+      "epoch": 46.9,
+      "learning_rate": 4.413712686567165e-05,
+      "loss": 0.0,
+      "step": 50280
+    },
+    {
+      "epoch": 46.91,
+      "learning_rate": 4.41366604477612e-05,
+      "loss": 0.0008,
+      "step": 50284
+    },
+    {
+      "epoch": 46.91,
+      "learning_rate": 4.4136194029850745e-05,
+      "loss": 0.0,
+      "step": 50288
+    },
+    {
+      "epoch": 46.91,
+      "learning_rate": 4.41357276119403e-05,
+      "loss": 0.0001,
+      "step": 50292
+    },
+    {
+      "epoch": 46.92,
+      "learning_rate": 4.4135261194029855e-05,
+      "loss": 0.0001,
+      "step": 50296
+    },
+    {
+      "epoch": 46.92,
+      "learning_rate": 4.41347947761194e-05,
+      "loss": 0.0002,
+      "step": 50300
+    },
+    {
+      "epoch": 46.93,
+      "learning_rate": 4.413432835820896e-05,
+      "loss": 0.0,
+      "step": 50304
+    },
+    {
+      "epoch": 46.93,
+      "learning_rate": 4.413386194029851e-05,
+      "loss": 0.0022,
+      "step": 50308
+    },
+    {
+      "epoch": 46.93,
+      "learning_rate": 4.413339552238806e-05,
+      "loss": 0.0002,
+      "step": 50312
+    },
+    {
+      "epoch": 46.94,
+      "learning_rate": 4.4132929104477616e-05,
+      "loss": 0.0001,
+      "step": 50316
+    },
+    {
+      "epoch": 46.94,
+      "learning_rate": 4.4132462686567164e-05,
+      "loss": 0.009,
+      "step": 50320
+    },
+    {
+      "epoch": 46.94,
+      "learning_rate": 4.413199626865672e-05,
+      "loss": 0.0,
+      "step": 50324
+    },
+    {
+      "epoch": 46.95,
+      "learning_rate": 4.4131529850746274e-05,
+      "loss": 0.0002,
+      "step": 50328
+    },
+    {
+      "epoch": 46.95,
+      "learning_rate": 4.413106343283582e-05,
+      "loss": 0.0,
+      "step": 50332
+    },
+    {
+      "epoch": 46.96,
+      "learning_rate": 4.413059701492537e-05,
+      "loss": 0.0001,
+      "step": 50336
+    },
+    {
+      "epoch": 46.96,
+      "learning_rate": 4.413013059701493e-05,
+      "loss": 0.0023,
+      "step": 50340
+    },
+    {
+      "epoch": 46.96,
+      "learning_rate": 4.412966417910448e-05,
+      "loss": 0.0002,
+      "step": 50344
+    },
+    {
+      "epoch": 46.97,
+      "learning_rate": 4.412919776119403e-05,
+      "loss": 0.003,
+      "step": 50348
+    },
+    {
+      "epoch": 46.97,
+      "learning_rate": 4.412873134328358e-05,
+      "loss": 0.0,
+      "step": 50352
+    },
+    {
+      "epoch": 46.97,
+      "learning_rate": 4.412826492537314e-05,
+      "loss": 0.0,
+      "step": 50356
+    },
+    {
+      "epoch": 46.98,
+      "learning_rate": 4.4127798507462686e-05,
+      "loss": 0.0,
+      "step": 50360
+    },
+    {
+      "epoch": 46.98,
+      "learning_rate": 4.412733208955224e-05,
+      "loss": 0.0025,
+      "step": 50364
+    },
+    {
+      "epoch": 46.98,
+      "learning_rate": 4.4126865671641796e-05,
+      "loss": 0.0001,
+      "step": 50368
+    },
+    {
+      "epoch": 46.99,
+      "learning_rate": 4.4126399253731344e-05,
+      "loss": 0.0,
+      "step": 50372
+    },
+    {
+      "epoch": 46.99,
+      "learning_rate": 4.41259328358209e-05,
+      "loss": 0.0001,
+      "step": 50376
+    },
+    {
+      "epoch": 47.0,
+      "learning_rate": 4.412546641791045e-05,
+      "loss": 0.0002,
+      "step": 50380
+    },
+    {
+      "epoch": 47.0,
+      "learning_rate": 4.4125e-05,
+      "loss": 0.0,
+      "step": 50384
+    },
+    {
+      "epoch": 47.0,
+      "learning_rate": 4.412453358208956e-05,
+      "loss": 0.0026,
+      "step": 50388
+    },
+    {
+      "epoch": 47.01,
+      "learning_rate": 4.4124067164179105e-05,
+      "loss": 0.0,
+      "step": 50392
+    },
+    {
+      "epoch": 47.01,
+      "learning_rate": 4.412360074626865e-05,
+      "loss": 0.0049,
+      "step": 50396
+    },
+    {
+      "epoch": 47.01,
+      "learning_rate": 4.4123134328358215e-05,
+      "loss": 0.0,
+      "step": 50400
+    },
+    {
+      "epoch": 47.02,
+      "learning_rate": 4.412266791044776e-05,
+      "loss": 0.0,
+      "step": 50404
+    },
+    {
+      "epoch": 47.02,
+      "learning_rate": 4.412220149253731e-05,
+      "loss": 0.0013,
+      "step": 50408
+    },
+    {
+      "epoch": 47.03,
+      "learning_rate": 4.4121735074626866e-05,
+      "loss": 0.0011,
+      "step": 50412
+    },
+    {
+      "epoch": 47.03,
+      "learning_rate": 4.412126865671642e-05,
+      "loss": 0.0001,
+      "step": 50416
+    },
+    {
+      "epoch": 47.03,
+      "learning_rate": 4.4120802238805976e-05,
+      "loss": 0.0005,
+      "step": 50420
+    },
+    {
+      "epoch": 47.04,
+      "learning_rate": 4.4120335820895524e-05,
+      "loss": 0.0,
+      "step": 50424
+    },
+    {
+      "epoch": 47.04,
+      "learning_rate": 4.411986940298508e-05,
+      "loss": 0.0,
+      "step": 50428
+    },
+    {
+      "epoch": 47.04,
+      "learning_rate": 4.4119402985074634e-05,
+      "loss": 0.0,
+      "step": 50432
+    },
+    {
+      "epoch": 47.05,
+      "learning_rate": 4.411893656716418e-05,
+      "loss": 0.0,
+      "step": 50436
+    },
+    {
+      "epoch": 47.05,
+      "learning_rate": 4.411847014925373e-05,
+      "loss": 0.0,
+      "step": 50440
+    },
+    {
+      "epoch": 47.06,
+      "learning_rate": 4.4118003731343285e-05,
+      "loss": 0.0,
+      "step": 50444
+    },
+    {
+      "epoch": 47.06,
+      "learning_rate": 4.411753731343284e-05,
+      "loss": 0.0002,
+      "step": 50448
+    },
+    {
+      "epoch": 47.06,
+      "learning_rate": 4.411707089552239e-05,
+      "loss": 0.0,
+      "step": 50452
+    },
+    {
+      "epoch": 47.07,
+      "learning_rate": 4.411660447761194e-05,
+      "loss": 0.0,
+      "step": 50456
+    },
+    {
+      "epoch": 47.07,
+      "learning_rate": 4.41161380597015e-05,
+      "loss": 0.0,
+      "step": 50460
+    },
+    {
+      "epoch": 47.07,
+      "learning_rate": 4.4115671641791046e-05,
+      "loss": 0.0,
+      "step": 50464
+    },
+    {
+      "epoch": 47.08,
+      "learning_rate": 4.41152052238806e-05,
+      "loss": 0.0001,
+      "step": 50468
+    },
+    {
+      "epoch": 47.08,
+      "learning_rate": 4.411473880597015e-05,
+      "loss": 0.0002,
+      "step": 50472
+    },
+    {
+      "epoch": 47.09,
+      "learning_rate": 4.4114272388059704e-05,
+      "loss": 0.0018,
+      "step": 50476
+    },
+    {
+      "epoch": 47.09,
+      "learning_rate": 4.411380597014926e-05,
+      "loss": 0.0001,
+      "step": 50480
+    },
+    {
+      "epoch": 47.09,
+      "learning_rate": 4.411333955223881e-05,
+      "loss": 0.0001,
+      "step": 50484
+    },
+    {
+      "epoch": 47.1,
+      "learning_rate": 4.411287313432836e-05,
+      "loss": 0.0,
+      "step": 50488
+    },
+    {
+      "epoch": 47.1,
+      "learning_rate": 4.411240671641792e-05,
+      "loss": 0.0,
+      "step": 50492
+    },
+    {
+      "epoch": 47.1,
+      "learning_rate": 4.4111940298507465e-05,
+      "loss": 0.0,
+      "step": 50496
+    },
+    {
+      "epoch": 47.11,
+      "learning_rate": 4.411147388059701e-05,
+      "loss": 0.0001,
+      "step": 50500
+    },
+    {
+      "epoch": 47.11,
+      "eval_exact_match": 0.7321083172147002,
+      "eval_exec": 0.7649903288201161,
+      "eval_loss": 0.4581345319747925,
+      "eval_runtime": 1124.8565,
+      "eval_samples_per_second": 0.919,
+      "step": 50500
+    },
+    {
+      "epoch": 47.11,
+      "learning_rate": 4.411100746268657e-05,
+      "loss": 0.0001,
+      "step": 50504
+    },
+    {
+      "epoch": 47.12,
+      "learning_rate": 4.411054104477612e-05,
+      "loss": 0.0,
+      "step": 50508
+    },
+    {
+      "epoch": 47.12,
+      "learning_rate": 4.411007462686567e-05,
+      "loss": 0.0,
+      "step": 50512
+    },
+    {
+      "epoch": 47.12,
+      "learning_rate": 4.4109608208955226e-05,
+      "loss": 0.0,
+      "step": 50516
+    },
+    {
+      "epoch": 47.13,
+      "learning_rate": 4.410914179104478e-05,
+      "loss": 0.0001,
+      "step": 50520
+    },
+    {
+      "epoch": 47.13,
+      "learning_rate": 4.410867537313433e-05,
+      "loss": 0.0001,
+      "step": 50524
+    },
+    {
+      "epoch": 47.13,
+      "learning_rate": 4.4108208955223884e-05,
+      "loss": 0.0002,
+      "step": 50528
+    },
+    {
+      "epoch": 47.14,
+      "learning_rate": 4.410774253731343e-05,
+      "loss": 0.0001,
+      "step": 50532
+    },
+    {
+      "epoch": 47.14,
+      "learning_rate": 4.410727611940299e-05,
+      "loss": 0.0,
+      "step": 50536
+    },
+    {
+      "epoch": 47.15,
+      "learning_rate": 4.410680970149254e-05,
+      "loss": 0.0006,
+      "step": 50540
+    },
+    {
+      "epoch": 47.15,
+      "learning_rate": 4.410634328358209e-05,
+      "loss": 0.0,
+      "step": 50544
+    },
+    {
+      "epoch": 47.15,
+      "learning_rate": 4.4105876865671645e-05,
+      "loss": 0.0,
+      "step": 50548
+    },
+    {
+      "epoch": 47.16,
+      "learning_rate": 4.41054104477612e-05,
+      "loss": 0.0001,
+      "step": 50552
+    },
+    {
+      "epoch": 47.16,
+      "learning_rate": 4.410494402985075e-05,
+      "loss": 0.0,
+      "step": 50556
+    },
+    {
+      "epoch": 47.16,
+      "learning_rate": 4.4104477611940296e-05,
+      "loss": 0.0,
+      "step": 50560
+    },
+    {
+      "epoch": 47.17,
+      "learning_rate": 4.410401119402985e-05,
+      "loss": 0.0001,
+      "step": 50564
+    },
+    {
+      "epoch": 47.17,
+      "learning_rate": 4.4103544776119406e-05,
+      "loss": 0.0,
+      "step": 50568
+    },
+    {
+      "epoch": 47.18,
+      "learning_rate": 4.4103078358208954e-05,
+      "loss": 0.0,
+      "step": 50572
+    },
+    {
+      "epoch": 47.18,
+      "learning_rate": 4.410261194029851e-05,
+      "loss": 0.0001,
+      "step": 50576
+    },
+    {
+      "epoch": 47.18,
+      "learning_rate": 4.4102145522388064e-05,
+      "loss": 0.0,
+      "step": 50580
+    },
+    {
+      "epoch": 47.19,
+      "learning_rate": 4.410167910447762e-05,
+      "loss": 0.0,
+      "step": 50584
+    },
+    {
+      "epoch": 47.19,
+      "learning_rate": 4.410121268656717e-05,
+      "loss": 0.0001,
+      "step": 50588
+    },
+    {
+      "epoch": 47.19,
+      "learning_rate": 4.4100746268656715e-05,
+      "loss": 0.0,
+      "step": 50592
+    },
+    {
+      "epoch": 47.2,
+      "learning_rate": 4.4100279850746276e-05,
+      "loss": 0.0,
+      "step": 50596
+    },
+    {
+      "epoch": 47.2,
+      "learning_rate": 4.4099813432835825e-05,
+      "loss": 0.0,
+      "step": 50600
+    },
+    {
+      "epoch": 47.21,
+      "learning_rate": 4.409934701492537e-05,
+      "loss": 0.0008,
+      "step": 50604
+    },
+    {
+      "epoch": 47.21,
+      "learning_rate": 4.409888059701493e-05,
+      "loss": 0.0,
+      "step": 50608
+    },
+    {
+      "epoch": 47.21,
+      "learning_rate": 4.409841417910448e-05,
+      "loss": 0.0,
+      "step": 50612
+    },
+    {
+      "epoch": 47.22,
+      "learning_rate": 4.409794776119403e-05,
+      "loss": 0.0,
+      "step": 50616
+    },
+    {
+      "epoch": 47.22,
+      "learning_rate": 4.4097481343283586e-05,
+      "loss": 0.0,
+      "step": 50620
+    },
+    {
+      "epoch": 47.22,
+      "learning_rate": 4.4097014925373134e-05,
+      "loss": 0.0,
+      "step": 50624
+    },
+    {
+      "epoch": 47.23,
+      "learning_rate": 4.409654850746269e-05,
+      "loss": 0.0001,
+      "step": 50628
+    },
+    {
+      "epoch": 47.23,
+      "learning_rate": 4.4096082089552244e-05,
+      "loss": 0.0011,
+      "step": 50632
+    },
+    {
+      "epoch": 47.24,
+      "learning_rate": 4.409561567164179e-05,
+      "loss": 0.0024,
+      "step": 50636
+    },
+    {
+      "epoch": 47.24,
+      "learning_rate": 4.4095149253731347e-05,
+      "loss": 0.0,
+      "step": 50640
+    },
+    {
+      "epoch": 47.24,
+      "learning_rate": 4.40946828358209e-05,
+      "loss": 0.0002,
+      "step": 50644
+    },
+    {
+      "epoch": 47.25,
+      "learning_rate": 4.409421641791045e-05,
+      "loss": 0.0,
+      "step": 50648
+    },
+    {
+      "epoch": 47.25,
+      "learning_rate": 4.409375e-05,
+      "loss": 0.0,
+      "step": 50652
+    },
+    {
+      "epoch": 47.25,
+      "learning_rate": 4.409328358208956e-05,
+      "loss": 0.0,
+      "step": 50656
+    },
+    {
+      "epoch": 47.26,
+      "learning_rate": 4.409281716417911e-05,
+      "loss": 0.0,
+      "step": 50660
+    },
+    {
+      "epoch": 47.26,
+      "learning_rate": 4.4092350746268656e-05,
+      "loss": 0.0,
+      "step": 50664
+    },
+    {
+      "epoch": 47.26,
+      "learning_rate": 4.409188432835821e-05,
+      "loss": 0.0,
+      "step": 50668
+    },
+    {
+      "epoch": 47.27,
+      "learning_rate": 4.4091417910447766e-05,
+      "loss": 0.0,
+      "step": 50672
+    },
+    {
+      "epoch": 47.27,
+      "learning_rate": 4.4090951492537314e-05,
+      "loss": 0.0,
+      "step": 50676
+    },
+    {
+      "epoch": 47.28,
+      "learning_rate": 4.409048507462687e-05,
+      "loss": 0.0001,
+      "step": 50680
+    },
+    {
+      "epoch": 47.28,
+      "learning_rate": 4.409001865671642e-05,
+      "loss": 0.0002,
+      "step": 50684
+    },
+    {
+      "epoch": 47.28,
+      "learning_rate": 4.408955223880597e-05,
+      "loss": 0.0001,
+      "step": 50688
+    },
+    {
+      "epoch": 47.29,
+      "learning_rate": 4.4089085820895526e-05,
+      "loss": 0.0,
+      "step": 50692
+    },
+    {
+      "epoch": 47.29,
+      "learning_rate": 4.4088619402985075e-05,
+      "loss": 0.0003,
+      "step": 50696
+    },
+    {
+      "epoch": 47.29,
+      "learning_rate": 4.408815298507463e-05,
+      "loss": 0.0,
+      "step": 50700
+    },
+    {
+      "epoch": 47.3,
+      "learning_rate": 4.4087686567164184e-05,
+      "loss": 0.0,
+      "step": 50704
+    },
+    {
+      "epoch": 47.3,
+      "learning_rate": 4.408722014925373e-05,
+      "loss": 0.0,
+      "step": 50708
+    },
+    {
+      "epoch": 47.31,
+      "learning_rate": 4.408675373134328e-05,
+      "loss": 0.0,
+      "step": 50712
+    },
+    {
+      "epoch": 47.31,
+      "learning_rate": 4.408628731343284e-05,
+      "loss": 0.0003,
+      "step": 50716
+    },
+    {
+      "epoch": 47.31,
+      "learning_rate": 4.408582089552239e-05,
+      "loss": 0.0,
+      "step": 50720
+    },
+    {
+      "epoch": 47.32,
+      "learning_rate": 4.408535447761194e-05,
+      "loss": 0.0,
+      "step": 50724
+    },
+    {
+      "epoch": 47.32,
+      "learning_rate": 4.4084888059701494e-05,
+      "loss": 0.0,
+      "step": 50728
+    },
+    {
+      "epoch": 47.32,
+      "learning_rate": 4.408442164179105e-05,
+      "loss": 0.0,
+      "step": 50732
+    },
+    {
+      "epoch": 47.33,
+      "learning_rate": 4.4083955223880597e-05,
+      "loss": 0.0,
+      "step": 50736
+    },
+    {
+      "epoch": 47.33,
+      "learning_rate": 4.408348880597015e-05,
+      "loss": 0.0,
+      "step": 50740
+    },
+    {
+      "epoch": 47.34,
+      "learning_rate": 4.40830223880597e-05,
+      "loss": 0.0,
+      "step": 50744
+    },
+    {
+      "epoch": 47.34,
+      "learning_rate": 4.408255597014926e-05,
+      "loss": 0.0,
+      "step": 50748
+    },
+    {
+      "epoch": 47.34,
+      "learning_rate": 4.408208955223881e-05,
+      "loss": 0.0001,
+      "step": 50752
+    },
+    {
+      "epoch": 47.35,
+      "learning_rate": 4.408162313432836e-05,
+      "loss": 0.0,
+      "step": 50756
+    },
+    {
+      "epoch": 47.35,
+      "learning_rate": 4.408115671641791e-05,
+      "loss": 0.0,
+      "step": 50760
+    },
+    {
+      "epoch": 47.35,
+      "learning_rate": 4.408069029850747e-05,
+      "loss": 0.0,
+      "step": 50764
+    },
+    {
+      "epoch": 47.36,
+      "learning_rate": 4.4080223880597016e-05,
+      "loss": 0.0001,
+      "step": 50768
+    },
+    {
+      "epoch": 47.36,
+      "learning_rate": 4.407975746268657e-05,
+      "loss": 0.0,
+      "step": 50772
+    },
+    {
+      "epoch": 47.37,
+      "learning_rate": 4.4079291044776125e-05,
+      "loss": 0.0,
+      "step": 50776
+    },
+    {
+      "epoch": 47.37,
+      "learning_rate": 4.4078824626865673e-05,
+      "loss": 0.0,
+      "step": 50780
+    },
+    {
+      "epoch": 47.37,
+      "learning_rate": 4.407835820895523e-05,
+      "loss": 0.0004,
+      "step": 50784
+    },
+    {
+      "epoch": 47.38,
+      "learning_rate": 4.4077891791044777e-05,
+      "loss": 0.0,
+      "step": 50788
+    },
+    {
+      "epoch": 47.38,
+      "learning_rate": 4.407742537313433e-05,
+      "loss": 0.0,
+      "step": 50792
+    },
+    {
+      "epoch": 47.38,
+      "learning_rate": 4.4076958955223886e-05,
+      "loss": 0.0,
+      "step": 50796
+    },
+    {
+      "epoch": 47.39,
+      "learning_rate": 4.4076492537313434e-05,
+      "loss": 0.0,
+      "step": 50800
+    },
+    {
+      "epoch": 47.39,
+      "learning_rate": 4.407602611940298e-05,
+      "loss": 0.0002,
+      "step": 50804
+    },
+    {
+      "epoch": 47.4,
+      "learning_rate": 4.4075559701492544e-05,
+      "loss": 0.0,
+      "step": 50808
+    },
+    {
+      "epoch": 47.4,
+      "learning_rate": 4.407509328358209e-05,
+      "loss": 0.0012,
+      "step": 50812
+    },
+    {
+      "epoch": 47.4,
+      "learning_rate": 4.407462686567164e-05,
+      "loss": 0.0003,
+      "step": 50816
+    },
+    {
+      "epoch": 47.41,
+      "learning_rate": 4.4074160447761195e-05,
+      "loss": 0.0,
+      "step": 50820
+    },
+    {
+      "epoch": 47.41,
+      "learning_rate": 4.407369402985075e-05,
+      "loss": 0.0004,
+      "step": 50824
+    },
+    {
+      "epoch": 47.41,
+      "learning_rate": 4.40732276119403e-05,
+      "loss": 0.0024,
+      "step": 50828
+    },
+    {
+      "epoch": 47.42,
+      "learning_rate": 4.4072761194029853e-05,
+      "loss": 0.0017,
+      "step": 50832
+    },
+    {
+      "epoch": 47.42,
+      "learning_rate": 4.407229477611941e-05,
+      "loss": 0.0005,
+      "step": 50836
+    },
+    {
+      "epoch": 47.43,
+      "learning_rate": 4.4071828358208956e-05,
+      "loss": 0.0,
+      "step": 50840
+    },
+    {
+      "epoch": 47.43,
+      "learning_rate": 4.407136194029851e-05,
+      "loss": 0.0,
+      "step": 50844
+    },
+    {
+      "epoch": 47.43,
+      "learning_rate": 4.407089552238806e-05,
+      "loss": 0.0,
+      "step": 50848
+    },
+    {
+      "epoch": 47.44,
+      "learning_rate": 4.4070429104477614e-05,
+      "loss": 0.0003,
+      "step": 50852
+    },
+    {
+      "epoch": 47.44,
+      "learning_rate": 4.406996268656717e-05,
+      "loss": 0.0,
+      "step": 50856
+    },
+    {
+      "epoch": 47.44,
+      "learning_rate": 4.406949626865672e-05,
+      "loss": 0.0,
+      "step": 50860
+    },
+    {
+      "epoch": 47.45,
+      "learning_rate": 4.4069029850746266e-05,
+      "loss": 0.0001,
+      "step": 50864
+    },
+    {
+      "epoch": 47.45,
+      "learning_rate": 4.406856343283583e-05,
+      "loss": 0.0,
+      "step": 50868
+    },
+    {
+      "epoch": 47.46,
+      "learning_rate": 4.4068097014925375e-05,
+      "loss": 0.0009,
+      "step": 50872
+    },
+    {
+      "epoch": 47.46,
+      "learning_rate": 4.4067630597014923e-05,
+      "loss": 0.0001,
+      "step": 50876
+    },
+    {
+      "epoch": 47.46,
+      "learning_rate": 4.406716417910448e-05,
+      "loss": 0.0,
+      "step": 50880
+    },
+    {
+      "epoch": 47.47,
+      "learning_rate": 4.406669776119403e-05,
+      "loss": 0.0003,
+      "step": 50884
+    },
+    {
+      "epoch": 47.47,
+      "learning_rate": 4.406623134328358e-05,
+      "loss": 0.0002,
+      "step": 50888
+    },
+    {
+      "epoch": 47.47,
+      "learning_rate": 4.4065764925373136e-05,
+      "loss": 0.0,
+      "step": 50892
+    },
+    {
+      "epoch": 47.48,
+      "learning_rate": 4.4065298507462684e-05,
+      "loss": 0.0001,
+      "step": 50896
+    },
+    {
+      "epoch": 47.48,
+      "learning_rate": 4.406483208955224e-05,
+      "loss": 0.0,
+      "step": 50900
+    },
+    {
+      "epoch": 47.49,
+      "learning_rate": 4.4064365671641794e-05,
+      "loss": 0.0,
+      "step": 50904
+    },
+    {
+      "epoch": 47.49,
+      "learning_rate": 4.406389925373134e-05,
+      "loss": 0.0015,
+      "step": 50908
+    },
+    {
+      "epoch": 47.49,
+      "learning_rate": 4.40634328358209e-05,
+      "loss": 0.0,
+      "step": 50912
+    },
+    {
+      "epoch": 47.5,
+      "learning_rate": 4.406296641791045e-05,
+      "loss": 0.0011,
+      "step": 50916
+    },
+    {
+      "epoch": 47.5,
+      "learning_rate": 4.40625e-05,
+      "loss": 0.0001,
+      "step": 50920
+    },
+    {
+      "epoch": 47.5,
+      "learning_rate": 4.4062033582089555e-05,
+      "loss": 0.0,
+      "step": 50924
+    },
+    {
+      "epoch": 47.51,
+      "learning_rate": 4.406156716417911e-05,
+      "loss": 0.0006,
+      "step": 50928
+    },
+    {
+      "epoch": 47.51,
+      "learning_rate": 4.406110074626866e-05,
+      "loss": 0.0,
+      "step": 50932
+    },
+    {
+      "epoch": 47.51,
+      "learning_rate": 4.406063432835821e-05,
+      "loss": 0.0,
+      "step": 50936
+    },
+    {
+      "epoch": 47.52,
+      "learning_rate": 4.406016791044776e-05,
+      "loss": 0.0003,
+      "step": 50940
+    },
+    {
+      "epoch": 47.52,
+      "learning_rate": 4.4059701492537316e-05,
+      "loss": 0.0,
+      "step": 50944
+    },
+    {
+      "epoch": 47.53,
+      "learning_rate": 4.405923507462687e-05,
+      "loss": 0.0001,
+      "step": 50948
+    },
+    {
+      "epoch": 47.53,
+      "learning_rate": 4.405876865671642e-05,
+      "loss": 0.0,
+      "step": 50952
+    },
+    {
+      "epoch": 47.53,
+      "learning_rate": 4.405830223880597e-05,
+      "loss": 0.0001,
+      "step": 50956
+    },
+    {
+      "epoch": 47.54,
+      "learning_rate": 4.405783582089553e-05,
+      "loss": 0.0,
+      "step": 50960
+    },
+    {
+      "epoch": 47.54,
+      "learning_rate": 4.405736940298508e-05,
+      "loss": 0.0,
+      "step": 50964
+    },
+    {
+      "epoch": 47.54,
+      "learning_rate": 4.4056902985074625e-05,
+      "loss": 0.0,
+      "step": 50968
+    },
+    {
+      "epoch": 47.55,
+      "learning_rate": 4.405643656716418e-05,
+      "loss": 0.0001,
+      "step": 50972
+    },
+    {
+      "epoch": 47.55,
+      "learning_rate": 4.4055970149253735e-05,
+      "loss": 0.0,
+      "step": 50976
+    },
+    {
+      "epoch": 47.56,
+      "learning_rate": 4.405550373134328e-05,
+      "loss": 0.0016,
+      "step": 50980
+    },
+    {
+      "epoch": 47.56,
+      "learning_rate": 4.405503731343284e-05,
+      "loss": 0.0,
+      "step": 50984
+    },
+    {
+      "epoch": 47.56,
+      "learning_rate": 4.405457089552239e-05,
+      "loss": 0.0,
+      "step": 50988
+    },
+    {
+      "epoch": 47.57,
+      "learning_rate": 4.405410447761194e-05,
+      "loss": 0.0022,
+      "step": 50992
+    },
+    {
+      "epoch": 47.57,
+      "learning_rate": 4.4053638059701496e-05,
+      "loss": 0.0,
+      "step": 50996
+    },
+    {
+      "epoch": 47.57,
+      "learning_rate": 4.4053171641791044e-05,
+      "loss": 0.0,
+      "step": 51000
+    },
+    {
+      "epoch": 47.57,
+      "eval_exact_match": 0.7369439071566731,
+      "eval_exec": 0.7649903288201161,
+      "eval_loss": 0.47057628631591797,
+      "eval_runtime": 1132.9681,
+      "eval_samples_per_second": 0.913,
+      "step": 51000
+    },
+    {
+      "epoch": 47.58,
+      "learning_rate": 4.40527052238806e-05,
+      "loss": 0.0,
+      "step": 51004
+    },
+    {
+      "epoch": 47.58,
+      "learning_rate": 4.4052238805970154e-05,
+      "loss": 0.0002,
+      "step": 51008
+    },
+    {
+      "epoch": 47.59,
+      "learning_rate": 4.40517723880597e-05,
+      "loss": 0.0029,
+      "step": 51012
+    },
+    {
+      "epoch": 47.59,
+      "learning_rate": 4.405130597014925e-05,
+      "loss": 0.0,
+      "step": 51016
+    },
+    {
+      "epoch": 47.59,
+      "learning_rate": 4.405083955223881e-05,
+      "loss": 0.0055,
+      "step": 51020
+    },
+    {
+      "epoch": 47.6,
+      "learning_rate": 4.405037313432836e-05,
+      "loss": 0.0,
+      "step": 51024
+    },
+    {
+      "epoch": 47.6,
+      "learning_rate": 4.404990671641791e-05,
+      "loss": 0.0,
+      "step": 51028
+    },
+    {
+      "epoch": 47.6,
+      "learning_rate": 4.404944029850746e-05,
+      "loss": 0.0,
+      "step": 51032
+    },
+    {
+      "epoch": 47.61,
+      "learning_rate": 4.404897388059702e-05,
+      "loss": 0.0,
+      "step": 51036
+    },
+    {
+      "epoch": 47.61,
+      "learning_rate": 4.4048507462686566e-05,
+      "loss": 0.0,
+      "step": 51040
+    },
+    {
+      "epoch": 47.62,
+      "learning_rate": 4.404804104477612e-05,
+      "loss": 0.0,
+      "step": 51044
+    },
+    {
+      "epoch": 47.62,
+      "learning_rate": 4.4047574626865676e-05,
+      "loss": 0.0007,
+      "step": 51048
+    },
+    {
+      "epoch": 47.62,
+      "learning_rate": 4.4047108208955224e-05,
+      "loss": 0.0,
+      "step": 51052
+    },
+    {
+      "epoch": 47.63,
+      "learning_rate": 4.404664179104478e-05,
+      "loss": 0.0,
+      "step": 51056
+    },
+    {
+      "epoch": 47.63,
+      "learning_rate": 4.404617537313433e-05,
+      "loss": 0.0001,
+      "step": 51060
+    },
+    {
+      "epoch": 47.63,
+      "learning_rate": 4.404570895522388e-05,
+      "loss": 0.0011,
+      "step": 51064
+    },
+    {
+      "epoch": 47.64,
+      "learning_rate": 4.404524253731344e-05,
+      "loss": 0.0,
+      "step": 51068
+    },
+    {
+      "epoch": 47.64,
+      "learning_rate": 4.4044776119402985e-05,
+      "loss": 0.0027,
+      "step": 51072
+    },
+    {
+      "epoch": 47.65,
+      "learning_rate": 4.404430970149254e-05,
+      "loss": 0.0003,
+      "step": 51076
+    },
+    {
+      "epoch": 47.65,
+      "learning_rate": 4.4043843283582095e-05,
+      "loss": 0.0,
+      "step": 51080
+    },
+    {
+      "epoch": 47.65,
+      "learning_rate": 4.404337686567164e-05,
+      "loss": 0.0002,
+      "step": 51084
+    },
+    {
+      "epoch": 47.66,
+      "learning_rate": 4.40429104477612e-05,
+      "loss": 0.0011,
+      "step": 51088
+    },
+    {
+      "epoch": 47.66,
+      "learning_rate": 4.4042444029850746e-05,
+      "loss": 0.0,
+      "step": 51092
+    },
+    {
+      "epoch": 47.66,
+      "learning_rate": 4.40419776119403e-05,
+      "loss": 0.0048,
+      "step": 51096
+    },
+    {
+      "epoch": 47.67,
+      "learning_rate": 4.4041511194029856e-05,
+      "loss": 0.0,
+      "step": 51100
+    },
+    {
+      "epoch": 47.67,
+      "learning_rate": 4.4041044776119404e-05,
+      "loss": 0.0,
+      "step": 51104
+    },
+    {
+      "epoch": 47.68,
+      "learning_rate": 4.404057835820896e-05,
+      "loss": 0.0,
+      "step": 51108
+    },
+    {
+      "epoch": 47.68,
+      "learning_rate": 4.4040111940298514e-05,
+      "loss": 0.0,
+      "step": 51112
+    },
+    {
+      "epoch": 47.68,
+      "learning_rate": 4.403964552238806e-05,
+      "loss": 0.0026,
+      "step": 51116
+    },
+    {
+      "epoch": 47.69,
+      "learning_rate": 4.403917910447761e-05,
+      "loss": 0.0002,
+      "step": 51120
+    },
+    {
+      "epoch": 47.69,
+      "learning_rate": 4.4038712686567165e-05,
+      "loss": 0.0,
+      "step": 51124
+    },
+    {
+      "epoch": 47.69,
+      "learning_rate": 4.403824626865672e-05,
+      "loss": 0.0021,
+      "step": 51128
+    },
+    {
+      "epoch": 47.7,
+      "learning_rate": 4.403777985074627e-05,
+      "loss": 0.0008,
+      "step": 51132
+    },
+    {
+      "epoch": 47.7,
+      "learning_rate": 4.403731343283582e-05,
+      "loss": 0.0,
+      "step": 51136
+    },
+    {
+      "epoch": 47.71,
+      "learning_rate": 4.403684701492538e-05,
+      "loss": 0.0001,
+      "step": 51140
+    },
+    {
+      "epoch": 47.71,
+      "learning_rate": 4.4036380597014926e-05,
+      "loss": 0.0002,
+      "step": 51144
+    },
+    {
+      "epoch": 47.71,
+      "learning_rate": 4.403591417910448e-05,
+      "loss": 0.0,
+      "step": 51148
+    },
+    {
+      "epoch": 47.72,
+      "learning_rate": 4.403544776119403e-05,
+      "loss": 0.0,
+      "step": 51152
+    },
+    {
+      "epoch": 47.72,
+      "learning_rate": 4.4034981343283584e-05,
+      "loss": 0.0,
+      "step": 51156
+    },
+    {
+      "epoch": 47.72,
+      "learning_rate": 4.403451492537314e-05,
+      "loss": 0.0003,
+      "step": 51160
+    },
+    {
+      "epoch": 47.73,
+      "learning_rate": 4.403404850746269e-05,
+      "loss": 0.0004,
+      "step": 51164
+    },
+    {
+      "epoch": 47.73,
+      "learning_rate": 4.403358208955224e-05,
+      "loss": 0.0,
+      "step": 51168
+    },
+    {
+      "epoch": 47.73,
+      "learning_rate": 4.40331156716418e-05,
+      "loss": 0.0,
+      "step": 51172
+    },
+    {
+      "epoch": 47.74,
+      "learning_rate": 4.4032649253731345e-05,
+      "loss": 0.0021,
+      "step": 51176
+    },
+    {
+      "epoch": 47.74,
+      "learning_rate": 4.403218283582089e-05,
+      "loss": 0.0,
+      "step": 51180
+    },
+    {
+      "epoch": 47.75,
+      "learning_rate": 4.403171641791045e-05,
+      "loss": 0.0,
+      "step": 51184
+    },
+    {
+      "epoch": 47.75,
+      "learning_rate": 4.403125e-05,
+      "loss": 0.0001,
+      "step": 51188
+    },
+    {
+      "epoch": 47.75,
+      "learning_rate": 4.403078358208955e-05,
+      "loss": 0.0001,
+      "step": 51192
+    },
+    {
+      "epoch": 47.76,
+      "learning_rate": 4.4030317164179106e-05,
+      "loss": 0.0041,
+      "step": 51196
+    },
+    {
+      "epoch": 47.76,
+      "learning_rate": 4.402985074626866e-05,
+      "loss": 0.0,
+      "step": 51200
+    },
+    {
+      "epoch": 47.76,
+      "learning_rate": 4.402938432835821e-05,
+      "loss": 0.0,
+      "step": 51204
+    },
+    {
+      "epoch": 47.77,
+      "learning_rate": 4.4028917910447764e-05,
+      "loss": 0.0,
+      "step": 51208
+    },
+    {
+      "epoch": 47.77,
+      "learning_rate": 4.402845149253731e-05,
+      "loss": 0.0,
+      "step": 51212
+    },
+    {
+      "epoch": 47.78,
+      "learning_rate": 4.402798507462687e-05,
+      "loss": 0.0,
+      "step": 51216
+    },
+    {
+      "epoch": 47.78,
+      "learning_rate": 4.402751865671642e-05,
+      "loss": 0.0018,
+      "step": 51220
+    },
+    {
+      "epoch": 47.78,
+      "learning_rate": 4.402705223880597e-05,
+      "loss": 0.0,
+      "step": 51224
+    },
+    {
+      "epoch": 47.79,
+      "learning_rate": 4.4026585820895525e-05,
+      "loss": 0.0,
+      "step": 51228
+    },
+    {
+      "epoch": 47.79,
+      "learning_rate": 4.402611940298508e-05,
+      "loss": 0.0001,
+      "step": 51232
+    },
+    {
+      "epoch": 47.79,
+      "learning_rate": 4.402565298507463e-05,
+      "loss": 0.0,
+      "step": 51236
+    },
+    {
+      "epoch": 47.8,
+      "learning_rate": 4.402518656716418e-05,
+      "loss": 0.0001,
+      "step": 51240
+    },
+    {
+      "epoch": 47.8,
+      "learning_rate": 4.402472014925373e-05,
+      "loss": 0.0001,
+      "step": 51244
+    },
+    {
+      "epoch": 47.81,
+      "learning_rate": 4.4024253731343286e-05,
+      "loss": 0.0,
+      "step": 51248
+    },
+    {
+      "epoch": 47.81,
+      "learning_rate": 4.402378731343284e-05,
+      "loss": 0.0005,
+      "step": 51252
+    },
+    {
+      "epoch": 47.81,
+      "learning_rate": 4.402332089552239e-05,
+      "loss": 0.0004,
+      "step": 51256
+    },
+    {
+      "epoch": 47.82,
+      "learning_rate": 4.4022854477611944e-05,
+      "loss": 0.0,
+      "step": 51260
+    },
+    {
+      "epoch": 47.82,
+      "learning_rate": 4.40223880597015e-05,
+      "loss": 0.0,
+      "step": 51264
+    },
+    {
+      "epoch": 47.82,
+      "learning_rate": 4.402192164179105e-05,
+      "loss": 0.0,
+      "step": 51268
+    },
+    {
+      "epoch": 47.83,
+      "learning_rate": 4.4021455223880595e-05,
+      "loss": 0.0,
+      "step": 51272
+    },
+    {
+      "epoch": 47.83,
+      "learning_rate": 4.402098880597016e-05,
+      "loss": 0.0008,
+      "step": 51276
+    },
+    {
+      "epoch": 47.84,
+      "learning_rate": 4.4020522388059705e-05,
+      "loss": 0.0001,
+      "step": 51280
+    },
+    {
+      "epoch": 47.84,
+      "learning_rate": 4.402005597014925e-05,
+      "loss": 0.0001,
+      "step": 51284
+    },
+    {
+      "epoch": 47.84,
+      "learning_rate": 4.401958955223881e-05,
+      "loss": 0.0,
+      "step": 51288
+    },
+    {
+      "epoch": 47.85,
+      "learning_rate": 4.401912313432836e-05,
+      "loss": 0.0,
+      "step": 51292
+    },
+    {
+      "epoch": 47.85,
+      "learning_rate": 4.401865671641791e-05,
+      "loss": 0.0,
+      "step": 51296
+    },
+    {
+      "epoch": 47.85,
+      "learning_rate": 4.4018190298507466e-05,
+      "loss": 0.0001,
+      "step": 51300
+    },
+    {
+      "epoch": 47.86,
+      "learning_rate": 4.4017723880597014e-05,
+      "loss": 0.0,
+      "step": 51304
+    },
+    {
+      "epoch": 47.86,
+      "learning_rate": 4.401725746268657e-05,
+      "loss": 0.0001,
+      "step": 51308
+    },
+    {
+      "epoch": 47.87,
+      "learning_rate": 4.4016791044776124e-05,
+      "loss": 0.0,
+      "step": 51312
+    },
+    {
+      "epoch": 47.87,
+      "learning_rate": 4.401632462686567e-05,
+      "loss": 0.0,
+      "step": 51316
+    },
+    {
+      "epoch": 47.87,
+      "learning_rate": 4.401585820895523e-05,
+      "loss": 0.0,
+      "step": 51320
+    },
+    {
+      "epoch": 47.88,
+      "learning_rate": 4.401539179104478e-05,
+      "loss": 0.0,
+      "step": 51324
+    },
+    {
+      "epoch": 47.88,
+      "learning_rate": 4.401492537313433e-05,
+      "loss": 0.0,
+      "step": 51328
+    },
+    {
+      "epoch": 47.88,
+      "learning_rate": 4.401445895522388e-05,
+      "loss": 0.0,
+      "step": 51332
+    },
+    {
+      "epoch": 47.89,
+      "learning_rate": 4.401399253731344e-05,
+      "loss": 0.0,
+      "step": 51336
+    },
+    {
+      "epoch": 47.89,
+      "learning_rate": 4.401352611940299e-05,
+      "loss": 0.0,
+      "step": 51340
+    },
+    {
+      "epoch": 47.9,
+      "learning_rate": 4.4013059701492536e-05,
+      "loss": 0.0001,
+      "step": 51344
+    },
+    {
+      "epoch": 47.9,
+      "learning_rate": 4.401259328358209e-05,
+      "loss": 0.0,
+      "step": 51348
+    },
+    {
+      "epoch": 47.9,
+      "learning_rate": 4.4012126865671646e-05,
+      "loss": 0.0,
+      "step": 51352
+    },
+    {
+      "epoch": 47.91,
+      "learning_rate": 4.4011660447761194e-05,
+      "loss": 0.0048,
+      "step": 51356
+    },
+    {
+      "epoch": 47.91,
+      "learning_rate": 4.401119402985075e-05,
+      "loss": 0.0,
+      "step": 51360
+    },
+    {
+      "epoch": 47.91,
+      "learning_rate": 4.40107276119403e-05,
+      "loss": 0.0,
+      "step": 51364
+    },
+    {
+      "epoch": 47.92,
+      "learning_rate": 4.401026119402985e-05,
+      "loss": 0.0,
+      "step": 51368
+    },
+    {
+      "epoch": 47.92,
+      "learning_rate": 4.400979477611941e-05,
+      "loss": 0.0014,
+      "step": 51372
+    },
+    {
+      "epoch": 47.93,
+      "learning_rate": 4.4009328358208955e-05,
+      "loss": 0.0012,
+      "step": 51376
+    },
+    {
+      "epoch": 47.93,
+      "learning_rate": 4.400886194029851e-05,
+      "loss": 0.0,
+      "step": 51380
+    },
+    {
+      "epoch": 47.93,
+      "learning_rate": 4.4008395522388065e-05,
+      "loss": 0.0,
+      "step": 51384
+    },
+    {
+      "epoch": 47.94,
+      "learning_rate": 4.400792910447761e-05,
+      "loss": 0.0,
+      "step": 51388
+    },
+    {
+      "epoch": 47.94,
+      "learning_rate": 4.400746268656716e-05,
+      "loss": 0.0,
+      "step": 51392
+    },
+    {
+      "epoch": 47.94,
+      "learning_rate": 4.400699626865672e-05,
+      "loss": 0.0,
+      "step": 51396
+    },
+    {
+      "epoch": 47.95,
+      "learning_rate": 4.400652985074627e-05,
+      "loss": 0.0002,
+      "step": 51400
+    },
+    {
+      "epoch": 47.95,
+      "learning_rate": 4.4006063432835826e-05,
+      "loss": 0.0001,
+      "step": 51404
+    },
+    {
+      "epoch": 47.96,
+      "learning_rate": 4.4005597014925374e-05,
+      "loss": 0.0003,
+      "step": 51408
+    },
+    {
+      "epoch": 47.96,
+      "learning_rate": 4.400513059701493e-05,
+      "loss": 0.0002,
+      "step": 51412
+    },
+    {
+      "epoch": 47.96,
+      "learning_rate": 4.4004664179104484e-05,
+      "loss": 0.0001,
+      "step": 51416
+    },
+    {
+      "epoch": 47.97,
+      "learning_rate": 4.400419776119403e-05,
+      "loss": 0.0,
+      "step": 51420
+    },
+    {
+      "epoch": 47.97,
+      "learning_rate": 4.400373134328358e-05,
+      "loss": 0.0,
+      "step": 51424
+    },
+    {
+      "epoch": 47.97,
+      "learning_rate": 4.400326492537314e-05,
+      "loss": 0.0,
+      "step": 51428
+    },
+    {
+      "epoch": 47.98,
+      "learning_rate": 4.400279850746269e-05,
+      "loss": 0.0,
+      "step": 51432
+    },
+    {
+      "epoch": 47.98,
+      "learning_rate": 4.400233208955224e-05,
+      "loss": 0.0,
+      "step": 51436
+    },
+    {
+      "epoch": 47.98,
+      "learning_rate": 4.400186567164179e-05,
+      "loss": 0.0,
+      "step": 51440
+    },
+    {
+      "epoch": 47.99,
+      "learning_rate": 4.400139925373135e-05,
+      "loss": 0.0,
+      "step": 51444
+    },
+    {
+      "epoch": 47.99,
+      "learning_rate": 4.4000932835820896e-05,
+      "loss": 0.0001,
+      "step": 51448
+    },
+    {
+      "epoch": 48.0,
+      "learning_rate": 4.400046641791045e-05,
+      "loss": 0.0008,
+      "step": 51452
+    },
+    {
+      "epoch": 48.0,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.0,
+      "step": 51456
+    },
+    {
+      "epoch": 48.0,
+      "learning_rate": 4.3999533582089554e-05,
+      "loss": 0.0001,
+      "step": 51460
+    },
+    {
+      "epoch": 48.01,
+      "learning_rate": 4.399906716417911e-05,
+      "loss": 0.0,
+      "step": 51464
+    },
+    {
+      "epoch": 48.01,
+      "learning_rate": 4.399860074626866e-05,
+      "loss": 0.0,
+      "step": 51468
+    },
+    {
+      "epoch": 48.01,
+      "learning_rate": 4.399813432835821e-05,
+      "loss": 0.0,
+      "step": 51472
+    },
+    {
+      "epoch": 48.02,
+      "learning_rate": 4.399766791044777e-05,
+      "loss": 0.0,
+      "step": 51476
+    },
+    {
+      "epoch": 48.02,
+      "learning_rate": 4.3997201492537315e-05,
+      "loss": 0.0,
+      "step": 51480
+    },
+    {
+      "epoch": 48.03,
+      "learning_rate": 4.399673507462686e-05,
+      "loss": 0.0005,
+      "step": 51484
+    },
+    {
+      "epoch": 48.03,
+      "learning_rate": 4.3996268656716425e-05,
+      "loss": 0.0012,
+      "step": 51488
+    },
+    {
+      "epoch": 48.03,
+      "learning_rate": 4.399580223880597e-05,
+      "loss": 0.0146,
+      "step": 51492
+    },
+    {
+      "epoch": 48.04,
+      "learning_rate": 4.399533582089552e-05,
+      "loss": 0.0004,
+      "step": 51496
+    },
+    {
+      "epoch": 48.04,
+      "learning_rate": 4.3994869402985076e-05,
+      "loss": 0.0,
+      "step": 51500
+    },
+    {
+      "epoch": 48.04,
+      "eval_exact_match": 0.741779497098646,
+      "eval_exec": 0.776595744680851,
+      "eval_loss": 0.44945940375328064,
+      "eval_runtime": 1251.2204,
+      "eval_samples_per_second": 0.826,
+      "step": 51500
+    },
+    {
+      "epoch": 48.04,
+      "learning_rate": 4.399440298507463e-05,
+      "loss": 0.0,
+      "step": 51504
+    },
+    {
+      "epoch": 48.05,
+      "learning_rate": 4.399393656716418e-05,
+      "loss": 0.0001,
+      "step": 51508
+    },
+    {
+      "epoch": 48.05,
+      "learning_rate": 4.3993470149253734e-05,
+      "loss": 0.0,
+      "step": 51512
+    },
+    {
+      "epoch": 48.06,
+      "learning_rate": 4.399300373134329e-05,
+      "loss": 0.0,
+      "step": 51516
+    },
+    {
+      "epoch": 48.06,
+      "learning_rate": 4.399253731343284e-05,
+      "loss": 0.0,
+      "step": 51520
+    },
+    {
+      "epoch": 48.06,
+      "learning_rate": 4.399207089552239e-05,
+      "loss": 0.0,
+      "step": 51524
+    },
+    {
+      "epoch": 48.07,
+      "learning_rate": 4.399160447761194e-05,
+      "loss": 0.0004,
+      "step": 51528
+    },
+    {
+      "epoch": 48.07,
+      "learning_rate": 4.3991138059701495e-05,
+      "loss": 0.0,
+      "step": 51532
+    },
+    {
+      "epoch": 48.07,
+      "learning_rate": 4.399067164179105e-05,
+      "loss": 0.0001,
+      "step": 51536
+    },
+    {
+      "epoch": 48.08,
+      "learning_rate": 4.39902052238806e-05,
+      "loss": 0.0022,
+      "step": 51540
+    },
+    {
+      "epoch": 48.08,
+      "learning_rate": 4.3989738805970146e-05,
+      "loss": 0.0001,
+      "step": 51544
+    },
+    {
+      "epoch": 48.09,
+      "learning_rate": 4.398927238805971e-05,
+      "loss": 0.0,
+      "step": 51548
+    },
+    {
+      "epoch": 48.09,
+      "learning_rate": 4.3988805970149256e-05,
+      "loss": 0.0,
+      "step": 51552
+    },
+    {
+      "epoch": 48.09,
+      "learning_rate": 4.3988339552238804e-05,
+      "loss": 0.0001,
+      "step": 51556
+    },
+    {
+      "epoch": 48.1,
+      "learning_rate": 4.398787313432836e-05,
+      "loss": 0.0,
+      "step": 51560
+    },
+    {
+      "epoch": 48.1,
+      "learning_rate": 4.3987406716417914e-05,
+      "loss": 0.0003,
+      "step": 51564
+    },
+    {
+      "epoch": 48.1,
+      "learning_rate": 4.398694029850747e-05,
+      "loss": 0.0,
+      "step": 51568
+    },
+    {
+      "epoch": 48.11,
+      "learning_rate": 4.398647388059702e-05,
+      "loss": 0.0001,
+      "step": 51572
+    },
+    {
+      "epoch": 48.11,
+      "learning_rate": 4.3986007462686565e-05,
+      "loss": 0.0,
+      "step": 51576
+    },
+    {
+      "epoch": 48.12,
+      "learning_rate": 4.3985541044776126e-05,
+      "loss": 0.0,
+      "step": 51580
+    },
+    {
+      "epoch": 48.12,
+      "learning_rate": 4.3985074626865675e-05,
+      "loss": 0.0002,
+      "step": 51584
+    },
+    {
+      "epoch": 48.12,
+      "learning_rate": 4.398460820895522e-05,
+      "loss": 0.0009,
+      "step": 51588
+    },
+    {
+      "epoch": 48.13,
+      "learning_rate": 4.398414179104478e-05,
+      "loss": 0.0,
+      "step": 51592
+    },
+    {
+      "epoch": 48.13,
+      "learning_rate": 4.398367537313433e-05,
+      "loss": 0.0001,
+      "step": 51596
+    },
+    {
+      "epoch": 48.13,
+      "learning_rate": 4.398320895522388e-05,
+      "loss": 0.0,
+      "step": 51600
+    },
+    {
+      "epoch": 48.14,
+      "learning_rate": 4.3982742537313436e-05,
+      "loss": 0.0,
+      "step": 51604
+    },
+    {
+      "epoch": 48.14,
+      "learning_rate": 4.398227611940299e-05,
+      "loss": 0.0,
+      "step": 51608
+    },
+    {
+      "epoch": 48.15,
+      "learning_rate": 4.398180970149254e-05,
+      "loss": 0.0007,
+      "step": 51612
+    },
+    {
+      "epoch": 48.15,
+      "learning_rate": 4.3981343283582094e-05,
+      "loss": 0.0,
+      "step": 51616
+    },
+    {
+      "epoch": 48.15,
+      "learning_rate": 4.398087686567164e-05,
+      "loss": 0.0001,
+      "step": 51620
+    },
+    {
+      "epoch": 48.16,
+      "learning_rate": 4.3980410447761197e-05,
+      "loss": 0.0,
+      "step": 51624
+    },
+    {
+      "epoch": 48.16,
+      "learning_rate": 4.397994402985075e-05,
+      "loss": 0.0,
+      "step": 51628
+    },
+    {
+      "epoch": 48.16,
+      "learning_rate": 4.39794776119403e-05,
+      "loss": 0.0082,
+      "step": 51632
+    },
+    {
+      "epoch": 48.17,
+      "learning_rate": 4.397901119402985e-05,
+      "loss": 0.0001,
+      "step": 51636
+    },
+    {
+      "epoch": 48.17,
+      "learning_rate": 4.397854477611941e-05,
+      "loss": 0.0,
+      "step": 51640
+    },
+    {
+      "epoch": 48.18,
+      "learning_rate": 4.397807835820896e-05,
+      "loss": 0.0001,
+      "step": 51644
+    },
+    {
+      "epoch": 48.18,
+      "learning_rate": 4.3977611940298506e-05,
+      "loss": 0.0,
+      "step": 51648
+    },
+    {
+      "epoch": 48.18,
+      "learning_rate": 4.397714552238806e-05,
+      "loss": 0.0,
+      "step": 51652
+    },
+    {
+      "epoch": 48.19,
+      "learning_rate": 4.3976679104477615e-05,
+      "loss": 0.0001,
+      "step": 51656
+    },
+    {
+      "epoch": 48.19,
+      "learning_rate": 4.3976212686567164e-05,
+      "loss": 0.0,
+      "step": 51660
+    },
+    {
+      "epoch": 48.19,
+      "learning_rate": 4.397574626865672e-05,
+      "loss": 0.0139,
+      "step": 51664
+    },
+    {
+      "epoch": 48.2,
+      "learning_rate": 4.3975279850746273e-05,
+      "loss": 0.0,
+      "step": 51668
+    },
+    {
+      "epoch": 48.2,
+      "learning_rate": 4.397481343283582e-05,
+      "loss": 0.0007,
+      "step": 51672
+    },
+    {
+      "epoch": 48.21,
+      "learning_rate": 4.3974347014925376e-05,
+      "loss": 0.0,
+      "step": 51676
+    },
+    {
+      "epoch": 48.21,
+      "learning_rate": 4.3973880597014925e-05,
+      "loss": 0.0,
+      "step": 51680
+    },
+    {
+      "epoch": 48.21,
+      "learning_rate": 4.397341417910448e-05,
+      "loss": 0.0,
+      "step": 51684
+    },
+    {
+      "epoch": 48.22,
+      "learning_rate": 4.3972947761194034e-05,
+      "loss": 0.0001,
+      "step": 51688
+    },
+    {
+      "epoch": 48.22,
+      "learning_rate": 4.397248134328358e-05,
+      "loss": 0.0,
+      "step": 51692
+    },
+    {
+      "epoch": 48.22,
+      "learning_rate": 4.397201492537313e-05,
+      "loss": 0.0,
+      "step": 51696
+    },
+    {
+      "epoch": 48.23,
+      "learning_rate": 4.397154850746269e-05,
+      "loss": 0.001,
+      "step": 51700
+    },
+    {
+      "epoch": 48.23,
+      "learning_rate": 4.397108208955224e-05,
+      "loss": 0.001,
+      "step": 51704
+    },
+    {
+      "epoch": 48.24,
+      "learning_rate": 4.397061567164179e-05,
+      "loss": 0.0,
+      "step": 51708
+    },
+    {
+      "epoch": 48.24,
+      "learning_rate": 4.3970149253731344e-05,
+      "loss": 0.0001,
+      "step": 51712
+    },
+    {
+      "epoch": 48.24,
+      "learning_rate": 4.39696828358209e-05,
+      "loss": 0.0001,
+      "step": 51716
+    },
+    {
+      "epoch": 48.25,
+      "learning_rate": 4.3969216417910447e-05,
+      "loss": 0.0027,
+      "step": 51720
+    },
+    {
+      "epoch": 48.25,
+      "learning_rate": 4.396875e-05,
+      "loss": 0.0,
+      "step": 51724
+    },
+    {
+      "epoch": 48.25,
+      "learning_rate": 4.3968283582089556e-05,
+      "loss": 0.0,
+      "step": 51728
+    },
+    {
+      "epoch": 48.26,
+      "learning_rate": 4.396781716417911e-05,
+      "loss": 0.0004,
+      "step": 51732
+    },
+    {
+      "epoch": 48.26,
+      "learning_rate": 4.396735074626866e-05,
+      "loss": 0.0001,
+      "step": 51736
+    },
+    {
+      "epoch": 48.26,
+      "learning_rate": 4.396688432835821e-05,
+      "loss": 0.0,
+      "step": 51740
+    },
+    {
+      "epoch": 48.27,
+      "learning_rate": 4.396641791044776e-05,
+      "loss": 0.0,
+      "step": 51744
+    },
+    {
+      "epoch": 48.27,
+      "learning_rate": 4.396595149253732e-05,
+      "loss": 0.0,
+      "step": 51748
+    },
+    {
+      "epoch": 48.28,
+      "learning_rate": 4.3965485074626866e-05,
+      "loss": 0.0001,
+      "step": 51752
+    },
+    {
+      "epoch": 48.28,
+      "learning_rate": 4.396501865671642e-05,
+      "loss": 0.0,
+      "step": 51756
+    },
+    {
+      "epoch": 48.28,
+      "learning_rate": 4.3964552238805975e-05,
+      "loss": 0.0001,
+      "step": 51760
+    },
+    {
+      "epoch": 48.29,
+      "learning_rate": 4.3964085820895523e-05,
+      "loss": 0.0001,
+      "step": 51764
+    },
+    {
+      "epoch": 48.29,
+      "learning_rate": 4.396361940298508e-05,
+      "loss": 0.0,
+      "step": 51768
+    },
+    {
+      "epoch": 48.29,
+      "learning_rate": 4.3963152985074626e-05,
+      "loss": 0.0001,
+      "step": 51772
+    },
+    {
+      "epoch": 48.3,
+      "learning_rate": 4.396268656716418e-05,
+      "loss": 0.0004,
+      "step": 51776
+    },
+    {
+      "epoch": 48.3,
+      "learning_rate": 4.3962220149253736e-05,
+      "loss": 0.0,
+      "step": 51780
+    },
+    {
+      "epoch": 48.31,
+      "learning_rate": 4.3961753731343284e-05,
+      "loss": 0.0,
+      "step": 51784
+    },
+    {
+      "epoch": 48.31,
+      "learning_rate": 4.396128731343284e-05,
+      "loss": 0.0022,
+      "step": 51788
+    },
+    {
+      "epoch": 48.31,
+      "learning_rate": 4.3960820895522394e-05,
+      "loss": 0.0,
+      "step": 51792
+    },
+    {
+      "epoch": 48.32,
+      "learning_rate": 4.396035447761194e-05,
+      "loss": 0.0007,
+      "step": 51796
+    },
+    {
+      "epoch": 48.32,
+      "learning_rate": 4.395988805970149e-05,
+      "loss": 0.0,
+      "step": 51800
+    },
+    {
+      "epoch": 48.32,
+      "learning_rate": 4.3959421641791045e-05,
+      "loss": 0.0,
+      "step": 51804
+    },
+    {
+      "epoch": 48.33,
+      "learning_rate": 4.39589552238806e-05,
+      "loss": 0.0,
+      "step": 51808
+    },
+    {
+      "epoch": 48.33,
+      "learning_rate": 4.395848880597015e-05,
+      "loss": 0.0,
+      "step": 51812
+    },
+    {
+      "epoch": 48.34,
+      "learning_rate": 4.39580223880597e-05,
+      "loss": 0.0,
+      "step": 51816
+    },
+    {
+      "epoch": 48.34,
+      "learning_rate": 4.395755597014926e-05,
+      "loss": 0.0,
+      "step": 51820
+    },
+    {
+      "epoch": 48.34,
+      "learning_rate": 4.3957089552238806e-05,
+      "loss": 0.0,
+      "step": 51824
+    },
+    {
+      "epoch": 48.35,
+      "learning_rate": 4.395662313432836e-05,
+      "loss": 0.0,
+      "step": 51828
+    },
+    {
+      "epoch": 48.35,
+      "learning_rate": 4.395615671641791e-05,
+      "loss": 0.0001,
+      "step": 51832
+    },
+    {
+      "epoch": 48.35,
+      "learning_rate": 4.3955690298507464e-05,
+      "loss": 0.0001,
+      "step": 51836
+    },
+    {
+      "epoch": 48.36,
+      "learning_rate": 4.395522388059702e-05,
+      "loss": 0.0001,
+      "step": 51840
+    },
+    {
+      "epoch": 48.36,
+      "learning_rate": 4.395475746268657e-05,
+      "loss": 0.0,
+      "step": 51844
+    },
+    {
+      "epoch": 48.37,
+      "learning_rate": 4.395429104477612e-05,
+      "loss": 0.0028,
+      "step": 51848
+    },
+    {
+      "epoch": 48.37,
+      "learning_rate": 4.395382462686568e-05,
+      "loss": 0.0024,
+      "step": 51852
+    },
+    {
+      "epoch": 48.37,
+      "learning_rate": 4.3953358208955225e-05,
+      "loss": 0.0002,
+      "step": 51856
+    },
+    {
+      "epoch": 48.38,
+      "learning_rate": 4.3952891791044773e-05,
+      "loss": 0.0001,
+      "step": 51860
+    },
+    {
+      "epoch": 48.38,
+      "learning_rate": 4.395242537313433e-05,
+      "loss": 0.0001,
+      "step": 51864
+    },
+    {
+      "epoch": 48.38,
+      "learning_rate": 4.395195895522388e-05,
+      "loss": 0.0,
+      "step": 51868
+    },
+    {
+      "epoch": 48.39,
+      "learning_rate": 4.395149253731343e-05,
+      "loss": 0.0002,
+      "step": 51872
+    },
+    {
+      "epoch": 48.39,
+      "learning_rate": 4.3951026119402986e-05,
+      "loss": 0.0,
+      "step": 51876
+    },
+    {
+      "epoch": 48.4,
+      "learning_rate": 4.395055970149254e-05,
+      "loss": 0.0,
+      "step": 51880
+    },
+    {
+      "epoch": 48.4,
+      "learning_rate": 4.395009328358209e-05,
+      "loss": 0.0,
+      "step": 51884
+    },
+    {
+      "epoch": 48.4,
+      "learning_rate": 4.3949626865671644e-05,
+      "loss": 0.0,
+      "step": 51888
+    },
+    {
+      "epoch": 48.41,
+      "learning_rate": 4.394916044776119e-05,
+      "loss": 0.0,
+      "step": 51892
+    },
+    {
+      "epoch": 48.41,
+      "learning_rate": 4.3948694029850754e-05,
+      "loss": 0.0,
+      "step": 51896
+    },
+    {
+      "epoch": 48.41,
+      "learning_rate": 4.39482276119403e-05,
+      "loss": 0.0,
+      "step": 51900
+    },
+    {
+      "epoch": 48.42,
+      "learning_rate": 4.394776119402985e-05,
+      "loss": 0.0,
+      "step": 51904
+    },
+    {
+      "epoch": 48.42,
+      "learning_rate": 4.3947294776119405e-05,
+      "loss": 0.0,
+      "step": 51908
+    },
+    {
+      "epoch": 48.43,
+      "learning_rate": 4.394682835820896e-05,
+      "loss": 0.0003,
+      "step": 51912
+    },
+    {
+      "epoch": 48.43,
+      "learning_rate": 4.394636194029851e-05,
+      "loss": 0.0024,
+      "step": 51916
+    },
+    {
+      "epoch": 48.43,
+      "learning_rate": 4.394589552238806e-05,
+      "loss": 0.0,
+      "step": 51920
+    },
+    {
+      "epoch": 48.44,
+      "learning_rate": 4.394542910447761e-05,
+      "loss": 0.0001,
+      "step": 51924
+    },
+    {
+      "epoch": 48.44,
+      "learning_rate": 4.3944962686567166e-05,
+      "loss": 0.0,
+      "step": 51928
+    },
+    {
+      "epoch": 48.44,
+      "learning_rate": 4.394449626865672e-05,
+      "loss": 0.0,
+      "step": 51932
+    },
+    {
+      "epoch": 48.45,
+      "learning_rate": 4.394402985074627e-05,
+      "loss": 0.0,
+      "step": 51936
+    },
+    {
+      "epoch": 48.45,
+      "learning_rate": 4.3943563432835824e-05,
+      "loss": 0.0001,
+      "step": 51940
+    },
+    {
+      "epoch": 48.46,
+      "learning_rate": 4.394309701492538e-05,
+      "loss": 0.0,
+      "step": 51944
+    },
+    {
+      "epoch": 48.46,
+      "learning_rate": 4.394263059701493e-05,
+      "loss": 0.0029,
+      "step": 51948
+    },
+    {
+      "epoch": 48.46,
+      "learning_rate": 4.3942164179104475e-05,
+      "loss": 0.0,
+      "step": 51952
+    },
+    {
+      "epoch": 48.47,
+      "learning_rate": 4.394169776119404e-05,
+      "loss": 0.0,
+      "step": 51956
+    },
+    {
+      "epoch": 48.47,
+      "learning_rate": 4.3941231343283585e-05,
+      "loss": 0.0,
+      "step": 51960
+    },
+    {
+      "epoch": 48.47,
+      "learning_rate": 4.394076492537313e-05,
+      "loss": 0.0002,
+      "step": 51964
+    },
+    {
+      "epoch": 48.48,
+      "learning_rate": 4.394029850746269e-05,
+      "loss": 0.0,
+      "step": 51968
+    },
+    {
+      "epoch": 48.48,
+      "learning_rate": 4.393983208955224e-05,
+      "loss": 0.0,
+      "step": 51972
+    },
+    {
+      "epoch": 48.49,
+      "learning_rate": 4.393936567164179e-05,
+      "loss": 0.002,
+      "step": 51976
+    },
+    {
+      "epoch": 48.49,
+      "learning_rate": 4.3938899253731346e-05,
+      "loss": 0.0007,
+      "step": 51980
+    },
+    {
+      "epoch": 48.49,
+      "learning_rate": 4.3938432835820894e-05,
+      "loss": 0.0,
+      "step": 51984
+    },
+    {
+      "epoch": 48.5,
+      "learning_rate": 4.393796641791045e-05,
+      "loss": 0.001,
+      "step": 51988
+    },
+    {
+      "epoch": 48.5,
+      "learning_rate": 4.3937500000000004e-05,
+      "loss": 0.0,
+      "step": 51992
+    },
+    {
+      "epoch": 48.5,
+      "learning_rate": 4.393703358208955e-05,
+      "loss": 0.0,
+      "step": 51996
+    },
+    {
+      "epoch": 48.51,
+      "learning_rate": 4.393656716417911e-05,
+      "loss": 0.0018,
+      "step": 52000
+    },
+    {
+      "epoch": 48.51,
+      "eval_exact_match": 0.7446808510638298,
+      "eval_exec": 0.7746615087040619,
+      "eval_loss": 0.44412875175476074,
+      "eval_runtime": 1559.1098,
+      "eval_samples_per_second": 0.663,
+      "step": 52000
+    },
+    {
+      "epoch": 48.51,
+      "learning_rate": 4.393610074626866e-05,
+      "loss": 0.0,
+      "step": 52004
+    },
+    {
+      "epoch": 48.51,
+      "learning_rate": 4.393563432835821e-05,
+      "loss": 0.0,
+      "step": 52008
+    },
+    {
+      "epoch": 48.52,
+      "learning_rate": 4.393516791044776e-05,
+      "loss": 0.0,
+      "step": 52012
+    },
+    {
+      "epoch": 48.52,
+      "learning_rate": 4.393470149253732e-05,
+      "loss": 0.0,
+      "step": 52016
+    },
+    {
+      "epoch": 48.53,
+      "learning_rate": 4.393423507462687e-05,
+      "loss": 0.0,
+      "step": 52020
+    },
+    {
+      "epoch": 48.53,
+      "learning_rate": 4.3933768656716416e-05,
+      "loss": 0.0,
+      "step": 52024
+    },
+    {
+      "epoch": 48.53,
+      "learning_rate": 4.393330223880597e-05,
+      "loss": 0.0,
+      "step": 52028
+    },
+    {
+      "epoch": 48.54,
+      "learning_rate": 4.3932835820895526e-05,
+      "loss": 0.0003,
+      "step": 52032
+    },
+    {
+      "epoch": 48.54,
+      "learning_rate": 4.3932369402985074e-05,
+      "loss": 0.0018,
+      "step": 52036
+    },
+    {
+      "epoch": 48.54,
+      "learning_rate": 4.393190298507463e-05,
+      "loss": 0.0,
+      "step": 52040
+    },
+    {
+      "epoch": 48.55,
+      "learning_rate": 4.393143656716418e-05,
+      "loss": 0.0,
+      "step": 52044
+    },
+    {
+      "epoch": 48.55,
+      "learning_rate": 4.393097014925374e-05,
+      "loss": 0.0,
+      "step": 52048
+    },
+    {
+      "epoch": 48.56,
+      "learning_rate": 4.393050373134329e-05,
+      "loss": 0.0,
+      "step": 52052
+    },
+    {
+      "epoch": 48.56,
+      "learning_rate": 4.3930037313432835e-05,
+      "loss": 0.0,
+      "step": 52056
+    },
+    {
+      "epoch": 48.56,
+      "learning_rate": 4.392957089552239e-05,
+      "loss": 0.0022,
+      "step": 52060
+    },
+    {
+      "epoch": 48.57,
+      "learning_rate": 4.3929104477611945e-05,
+      "loss": 0.0001,
+      "step": 52064
+    },
+    {
+      "epoch": 48.57,
+      "learning_rate": 4.392863805970149e-05,
+      "loss": 0.0056,
+      "step": 52068
+    },
+    {
+      "epoch": 48.57,
+      "learning_rate": 4.392817164179105e-05,
+      "loss": 0.0,
+      "step": 52072
+    },
+    {
+      "epoch": 48.58,
+      "learning_rate": 4.39277052238806e-05,
+      "loss": 0.0,
+      "step": 52076
+    },
+    {
+      "epoch": 48.58,
+      "learning_rate": 4.392723880597015e-05,
+      "loss": 0.0,
+      "step": 52080
+    },
+    {
+      "epoch": 48.59,
+      "learning_rate": 4.3926772388059706e-05,
+      "loss": 0.0,
+      "step": 52084
+    },
+    {
+      "epoch": 48.59,
+      "learning_rate": 4.3926305970149254e-05,
+      "loss": 0.0,
+      "step": 52088
+    },
+    {
+      "epoch": 48.59,
+      "learning_rate": 4.392583955223881e-05,
+      "loss": 0.0,
+      "step": 52092
+    },
+    {
+      "epoch": 48.6,
+      "learning_rate": 4.3925373134328364e-05,
+      "loss": 0.0,
+      "step": 52096
+    },
+    {
+      "epoch": 48.6,
+      "learning_rate": 4.392490671641791e-05,
+      "loss": 0.0003,
+      "step": 52100
+    },
+    {
+      "epoch": 48.6,
+      "learning_rate": 4.392444029850746e-05,
+      "loss": 0.0,
+      "step": 52104
+    },
+    {
+      "epoch": 48.61,
+      "learning_rate": 4.392397388059702e-05,
+      "loss": 0.0,
+      "step": 52108
+    },
+    {
+      "epoch": 48.61,
+      "learning_rate": 4.392350746268657e-05,
+      "loss": 0.0,
+      "step": 52112
+    },
+    {
+      "epoch": 48.62,
+      "learning_rate": 4.392304104477612e-05,
+      "loss": 0.0019,
+      "step": 52116
+    },
+    {
+      "epoch": 48.62,
+      "learning_rate": 4.392257462686567e-05,
+      "loss": 0.0,
+      "step": 52120
+    },
+    {
+      "epoch": 48.62,
+      "learning_rate": 4.392210820895523e-05,
+      "loss": 0.0,
+      "step": 52124
+    },
+    {
+      "epoch": 48.63,
+      "learning_rate": 4.3921641791044776e-05,
+      "loss": 0.0,
+      "step": 52128
+    },
+    {
+      "epoch": 48.63,
+      "learning_rate": 4.392117537313433e-05,
+      "loss": 0.0005,
+      "step": 52132
+    },
+    {
+      "epoch": 48.63,
+      "learning_rate": 4.3920708955223886e-05,
+      "loss": 0.0001,
+      "step": 52136
+    },
+    {
+      "epoch": 48.64,
+      "learning_rate": 4.3920242537313434e-05,
+      "loss": 0.0001,
+      "step": 52140
+    },
+    {
+      "epoch": 48.64,
+      "learning_rate": 4.391977611940299e-05,
+      "loss": 0.0,
+      "step": 52144
+    },
+    {
+      "epoch": 48.65,
+      "learning_rate": 4.391930970149254e-05,
+      "loss": 0.0002,
+      "step": 52148
+    },
+    {
+      "epoch": 48.65,
+      "learning_rate": 4.391884328358209e-05,
+      "loss": 0.0001,
+      "step": 52152
+    },
+    {
+      "epoch": 48.65,
+      "learning_rate": 4.391837686567165e-05,
+      "loss": 0.0,
+      "step": 52156
+    },
+    {
+      "epoch": 48.66,
+      "learning_rate": 4.3917910447761195e-05,
+      "loss": 0.0,
+      "step": 52160
+    },
+    {
+      "epoch": 48.66,
+      "learning_rate": 4.391744402985074e-05,
+      "loss": 0.0,
+      "step": 52164
+    },
+    {
+      "epoch": 48.66,
+      "learning_rate": 4.3916977611940305e-05,
+      "loss": 0.0,
+      "step": 52168
+    },
+    {
+      "epoch": 48.67,
+      "learning_rate": 4.391651119402985e-05,
+      "loss": 0.0,
+      "step": 52172
+    },
+    {
+      "epoch": 48.67,
+      "learning_rate": 4.39160447761194e-05,
+      "loss": 0.0,
+      "step": 52176
+    },
+    {
+      "epoch": 48.68,
+      "learning_rate": 4.3915578358208956e-05,
+      "loss": 0.0001,
+      "step": 52180
+    },
+    {
+      "epoch": 48.68,
+      "learning_rate": 4.391511194029851e-05,
+      "loss": 0.0024,
+      "step": 52184
+    },
+    {
+      "epoch": 48.68,
+      "learning_rate": 4.391464552238806e-05,
+      "loss": 0.0001,
+      "step": 52188
+    },
+    {
+      "epoch": 48.69,
+      "learning_rate": 4.3914179104477614e-05,
+      "loss": 0.0,
+      "step": 52192
+    },
+    {
+      "epoch": 48.69,
+      "learning_rate": 4.391371268656717e-05,
+      "loss": 0.0,
+      "step": 52196
+    },
+    {
+      "epoch": 48.69,
+      "learning_rate": 4.391324626865672e-05,
+      "loss": 0.0,
+      "step": 52200
+    },
+    {
+      "epoch": 48.7,
+      "learning_rate": 4.391277985074627e-05,
+      "loss": 0.0,
+      "step": 52204
+    },
+    {
+      "epoch": 48.7,
+      "learning_rate": 4.391231343283582e-05,
+      "loss": 0.0001,
+      "step": 52208
+    },
+    {
+      "epoch": 48.71,
+      "learning_rate": 4.3911847014925375e-05,
+      "loss": 0.0,
+      "step": 52212
+    },
+    {
+      "epoch": 48.71,
+      "learning_rate": 4.391138059701493e-05,
+      "loss": 0.0,
+      "step": 52216
+    },
+    {
+      "epoch": 48.71,
+      "learning_rate": 4.391091417910448e-05,
+      "loss": 0.0016,
+      "step": 52220
+    },
+    {
+      "epoch": 48.72,
+      "learning_rate": 4.391044776119403e-05,
+      "loss": 0.0001,
+      "step": 52224
+    },
+    {
+      "epoch": 48.72,
+      "learning_rate": 4.390998134328359e-05,
+      "loss": 0.0,
+      "step": 52228
+    },
+    {
+      "epoch": 48.72,
+      "learning_rate": 4.3909514925373136e-05,
+      "loss": 0.0009,
+      "step": 52232
+    },
+    {
+      "epoch": 48.73,
+      "learning_rate": 4.390904850746269e-05,
+      "loss": 0.0,
+      "step": 52236
+    },
+    {
+      "epoch": 48.73,
+      "learning_rate": 4.390858208955224e-05,
+      "loss": 0.0,
+      "step": 52240
+    },
+    {
+      "epoch": 48.73,
+      "learning_rate": 4.3908115671641794e-05,
+      "loss": 0.0001,
+      "step": 52244
+    },
+    {
+      "epoch": 48.74,
+      "learning_rate": 4.390764925373135e-05,
+      "loss": 0.0,
+      "step": 52248
+    },
+    {
+      "epoch": 48.74,
+      "learning_rate": 4.39071828358209e-05,
+      "loss": 0.0005,
+      "step": 52252
+    },
+    {
+      "epoch": 48.75,
+      "learning_rate": 4.3906716417910445e-05,
+      "loss": 0.0001,
+      "step": 52256
+    },
+    {
+      "epoch": 48.75,
+      "learning_rate": 4.390625000000001e-05,
+      "loss": 0.0,
+      "step": 52260
+    },
+    {
+      "epoch": 48.75,
+      "learning_rate": 4.3905783582089555e-05,
+      "loss": 0.0,
+      "step": 52264
+    },
+    {
+      "epoch": 48.76,
+      "learning_rate": 4.39053171641791e-05,
+      "loss": 0.0,
+      "step": 52268
+    },
+    {
+      "epoch": 48.76,
+      "learning_rate": 4.390485074626866e-05,
+      "loss": 0.0,
+      "step": 52272
+    },
+    {
+      "epoch": 48.76,
+      "learning_rate": 4.390438432835821e-05,
+      "loss": 0.0,
+      "step": 52276
+    },
+    {
+      "epoch": 48.77,
+      "learning_rate": 4.390391791044776e-05,
+      "loss": 0.0,
+      "step": 52280
+    },
+    {
+      "epoch": 48.77,
+      "learning_rate": 4.3903451492537316e-05,
+      "loss": 0.0001,
+      "step": 52284
+    },
+    {
+      "epoch": 48.78,
+      "learning_rate": 4.390298507462687e-05,
+      "loss": 0.0,
+      "step": 52288
+    },
+    {
+      "epoch": 48.78,
+      "learning_rate": 4.390251865671642e-05,
+      "loss": 0.0,
+      "step": 52292
+    },
+    {
+      "epoch": 48.78,
+      "learning_rate": 4.3902052238805974e-05,
+      "loss": 0.0,
+      "step": 52296
+    },
+    {
+      "epoch": 48.79,
+      "learning_rate": 4.390158582089552e-05,
+      "loss": 0.0,
+      "step": 52300
+    },
+    {
+      "epoch": 48.79,
+      "learning_rate": 4.390111940298508e-05,
+      "loss": 0.0113,
+      "step": 52304
+    },
+    {
+      "epoch": 48.79,
+      "learning_rate": 4.390065298507463e-05,
+      "loss": 0.0003,
+      "step": 52308
+    },
+    {
+      "epoch": 48.8,
+      "learning_rate": 4.390018656716418e-05,
+      "loss": 0.0004,
+      "step": 52312
+    },
+    {
+      "epoch": 48.8,
+      "learning_rate": 4.389972014925373e-05,
+      "loss": 0.0001,
+      "step": 52316
+    },
+    {
+      "epoch": 48.81,
+      "learning_rate": 4.389925373134329e-05,
+      "loss": 0.0,
+      "step": 52320
+    },
+    {
+      "epoch": 48.81,
+      "learning_rate": 4.389878731343284e-05,
+      "loss": 0.0,
+      "step": 52324
+    },
+    {
+      "epoch": 48.81,
+      "learning_rate": 4.3898320895522386e-05,
+      "loss": 0.0001,
+      "step": 52328
+    },
+    {
+      "epoch": 48.82,
+      "learning_rate": 4.389785447761194e-05,
+      "loss": 0.0,
+      "step": 52332
+    },
+    {
+      "epoch": 48.82,
+      "learning_rate": 4.3897388059701496e-05,
+      "loss": 0.0,
+      "step": 52336
+    },
+    {
+      "epoch": 48.82,
+      "learning_rate": 4.3896921641791044e-05,
+      "loss": 0.0,
+      "step": 52340
+    },
+    {
+      "epoch": 48.83,
+      "learning_rate": 4.38964552238806e-05,
+      "loss": 0.0,
+      "step": 52344
+    },
+    {
+      "epoch": 48.83,
+      "learning_rate": 4.3895988805970154e-05,
+      "loss": 0.0,
+      "step": 52348
+    },
+    {
+      "epoch": 48.84,
+      "learning_rate": 4.38955223880597e-05,
+      "loss": 0.0,
+      "step": 52352
+    },
+    {
+      "epoch": 48.84,
+      "learning_rate": 4.389505597014926e-05,
+      "loss": 0.0008,
+      "step": 52356
+    },
+    {
+      "epoch": 48.84,
+      "learning_rate": 4.3894589552238805e-05,
+      "loss": 0.0001,
+      "step": 52360
+    },
+    {
+      "epoch": 48.85,
+      "learning_rate": 4.389412313432836e-05,
+      "loss": 0.0,
+      "step": 52364
+    },
+    {
+      "epoch": 48.85,
+      "learning_rate": 4.3893656716417915e-05,
+      "loss": 0.0,
+      "step": 52368
+    },
+    {
+      "epoch": 48.85,
+      "learning_rate": 4.389319029850746e-05,
+      "loss": 0.0001,
+      "step": 52372
+    },
+    {
+      "epoch": 48.86,
+      "learning_rate": 4.389272388059702e-05,
+      "loss": 0.0,
+      "step": 52376
+    },
+    {
+      "epoch": 48.86,
+      "learning_rate": 4.389225746268657e-05,
+      "loss": 0.0,
+      "step": 52380
+    },
+    {
+      "epoch": 48.87,
+      "learning_rate": 4.389179104477612e-05,
+      "loss": 0.0,
+      "step": 52384
+    },
+    {
+      "epoch": 48.87,
+      "learning_rate": 4.3891324626865676e-05,
+      "loss": 0.0003,
+      "step": 52388
+    },
+    {
+      "epoch": 48.87,
+      "learning_rate": 4.3890858208955224e-05,
+      "loss": 0.0,
+      "step": 52392
+    },
+    {
+      "epoch": 48.88,
+      "learning_rate": 4.389039179104478e-05,
+      "loss": 0.0,
+      "step": 52396
+    },
+    {
+      "epoch": 48.88,
+      "learning_rate": 4.3889925373134334e-05,
+      "loss": 0.0,
+      "step": 52400
+    },
+    {
+      "epoch": 48.88,
+      "learning_rate": 4.388945895522388e-05,
+      "loss": 0.0,
+      "step": 52404
+    },
+    {
+      "epoch": 48.89,
+      "learning_rate": 4.388899253731344e-05,
+      "loss": 0.0,
+      "step": 52408
+    },
+    {
+      "epoch": 48.89,
+      "learning_rate": 4.388852611940299e-05,
+      "loss": 0.0003,
+      "step": 52412
+    },
+    {
+      "epoch": 48.9,
+      "learning_rate": 4.388805970149254e-05,
+      "loss": 0.0,
+      "step": 52416
+    },
+    {
+      "epoch": 48.9,
+      "learning_rate": 4.388759328358209e-05,
+      "loss": 0.0,
+      "step": 52420
+    },
+    {
+      "epoch": 48.9,
+      "learning_rate": 4.388712686567164e-05,
+      "loss": 0.0,
+      "step": 52424
+    },
+    {
+      "epoch": 48.91,
+      "learning_rate": 4.38866604477612e-05,
+      "loss": 0.0,
+      "step": 52428
+    },
+    {
+      "epoch": 48.91,
+      "learning_rate": 4.3886194029850746e-05,
+      "loss": 0.0011,
+      "step": 52432
+    },
+    {
+      "epoch": 48.91,
+      "learning_rate": 4.38857276119403e-05,
+      "loss": 0.0,
+      "step": 52436
+    },
+    {
+      "epoch": 48.92,
+      "learning_rate": 4.3885261194029856e-05,
+      "loss": 0.0003,
+      "step": 52440
+    },
+    {
+      "epoch": 48.92,
+      "learning_rate": 4.3884794776119404e-05,
+      "loss": 0.0,
+      "step": 52444
+    },
+    {
+      "epoch": 48.93,
+      "learning_rate": 4.388432835820896e-05,
+      "loss": 0.0,
+      "step": 52448
+    },
+    {
+      "epoch": 48.93,
+      "learning_rate": 4.388386194029851e-05,
+      "loss": 0.0,
+      "step": 52452
+    },
+    {
+      "epoch": 48.93,
+      "learning_rate": 4.388339552238806e-05,
+      "loss": 0.0001,
+      "step": 52456
+    },
+    {
+      "epoch": 48.94,
+      "learning_rate": 4.3882929104477617e-05,
+      "loss": 0.0,
+      "step": 52460
+    },
+    {
+      "epoch": 48.94,
+      "learning_rate": 4.3882462686567165e-05,
+      "loss": 0.0,
+      "step": 52464
+    },
+    {
+      "epoch": 48.94,
+      "learning_rate": 4.388199626865672e-05,
+      "loss": 0.0,
+      "step": 52468
+    },
+    {
+      "epoch": 48.95,
+      "learning_rate": 4.3881529850746275e-05,
+      "loss": 0.0005,
+      "step": 52472
+    },
+    {
+      "epoch": 48.95,
+      "learning_rate": 4.388106343283582e-05,
+      "loss": 0.003,
+      "step": 52476
+    },
+    {
+      "epoch": 48.96,
+      "learning_rate": 4.388059701492537e-05,
+      "loss": 0.0,
+      "step": 52480
+    },
+    {
+      "epoch": 48.96,
+      "learning_rate": 4.3880130597014926e-05,
+      "loss": 0.0029,
+      "step": 52484
+    },
+    {
+      "epoch": 48.96,
+      "learning_rate": 4.387966417910448e-05,
+      "loss": 0.002,
+      "step": 52488
+    },
+    {
+      "epoch": 48.97,
+      "learning_rate": 4.387919776119403e-05,
+      "loss": 0.0,
+      "step": 52492
+    },
+    {
+      "epoch": 48.97,
+      "learning_rate": 4.3878731343283584e-05,
+      "loss": 0.0,
+      "step": 52496
+    },
+    {
+      "epoch": 48.97,
+      "learning_rate": 4.387826492537314e-05,
+      "loss": 0.0016,
+      "step": 52500
+    },
+    {
+      "epoch": 48.97,
+      "eval_exact_match": 0.741779497098646,
+      "eval_exec": 0.769825918762089,
+      "eval_loss": 0.4436146914958954,
+      "eval_runtime": 1145.0497,
+      "eval_samples_per_second": 0.903,
+      "step": 52500
+    },
+    {
+      "epoch": 48.98,
+      "learning_rate": 4.387779850746269e-05,
+      "loss": 0.0001,
+      "step": 52504
+    },
+    {
+      "epoch": 48.98,
+      "learning_rate": 4.387733208955224e-05,
+      "loss": 0.0001,
+      "step": 52508
+    },
+    {
+      "epoch": 48.98,
+      "learning_rate": 4.387686567164179e-05,
+      "loss": 0.0,
+      "step": 52512
+    },
+    {
+      "epoch": 48.99,
+      "learning_rate": 4.3876399253731345e-05,
+      "loss": 0.0019,
+      "step": 52516
+    },
+    {
+      "epoch": 48.99,
+      "learning_rate": 4.38759328358209e-05,
+      "loss": 0.0001,
+      "step": 52520
+    },
+    {
+      "epoch": 49.0,
+      "learning_rate": 4.387546641791045e-05,
+      "loss": 0.0,
+      "step": 52524
+    },
+    {
+      "epoch": 49.0,
+      "learning_rate": 4.3875e-05,
+      "loss": 0.0,
+      "step": 52528
+    },
+    {
+      "epoch": 49.0,
+      "learning_rate": 4.387453358208956e-05,
+      "loss": 0.0,
+      "step": 52532
+    },
+    {
+      "epoch": 49.01,
+      "learning_rate": 4.3874067164179106e-05,
+      "loss": 0.0,
+      "step": 52536
+    },
+    {
+      "epoch": 49.01,
+      "learning_rate": 4.387360074626866e-05,
+      "loss": 0.001,
+      "step": 52540
+    },
+    {
+      "epoch": 49.01,
+      "learning_rate": 4.387313432835821e-05,
+      "loss": 0.0,
+      "step": 52544
+    },
+    {
+      "epoch": 49.02,
+      "learning_rate": 4.3872667910447764e-05,
+      "loss": 0.0,
+      "step": 52548
+    },
+    {
+      "epoch": 49.02,
+      "learning_rate": 4.387220149253732e-05,
+      "loss": 0.0006,
+      "step": 52552
+    },
+    {
+      "epoch": 49.03,
+      "learning_rate": 4.387173507462687e-05,
+      "loss": 0.0,
+      "step": 52556
+    },
+    {
+      "epoch": 49.03,
+      "learning_rate": 4.387126865671642e-05,
+      "loss": 0.0029,
+      "step": 52560
+    },
+    {
+      "epoch": 49.03,
+      "learning_rate": 4.3870802238805976e-05,
+      "loss": 0.0,
+      "step": 52564
+    },
+    {
+      "epoch": 49.04,
+      "learning_rate": 4.3870335820895525e-05,
+      "loss": 0.0,
+      "step": 52568
+    },
+    {
+      "epoch": 49.04,
+      "learning_rate": 4.386986940298507e-05,
+      "loss": 0.0001,
+      "step": 52572
+    },
+    {
+      "epoch": 49.04,
+      "learning_rate": 4.3869402985074634e-05,
+      "loss": 0.0005,
+      "step": 52576
+    },
+    {
+      "epoch": 49.05,
+      "learning_rate": 4.386893656716418e-05,
+      "loss": 0.0,
+      "step": 52580
+    },
+    {
+      "epoch": 49.05,
+      "learning_rate": 4.386847014925373e-05,
+      "loss": 0.0,
+      "step": 52584
+    },
+    {
+      "epoch": 49.06,
+      "learning_rate": 4.3868003731343286e-05,
+      "loss": 0.0001,
+      "step": 52588
+    },
+    {
+      "epoch": 49.06,
+      "learning_rate": 4.386753731343284e-05,
+      "loss": 0.0,
+      "step": 52592
+    },
+    {
+      "epoch": 49.06,
+      "learning_rate": 4.386707089552239e-05,
+      "loss": 0.0001,
+      "step": 52596
+    },
+    {
+      "epoch": 49.07,
+      "learning_rate": 4.3866604477611943e-05,
+      "loss": 0.0009,
+      "step": 52600
+    },
+    {
+      "epoch": 49.07,
+      "learning_rate": 4.386613805970149e-05,
+      "loss": 0.0,
+      "step": 52604
+    },
+    {
+      "epoch": 49.07,
+      "learning_rate": 4.3865671641791047e-05,
+      "loss": 0.0001,
+      "step": 52608
+    },
+    {
+      "epoch": 49.08,
+      "learning_rate": 4.38652052238806e-05,
+      "loss": 0.0,
+      "step": 52612
+    },
+    {
+      "epoch": 49.08,
+      "learning_rate": 4.386473880597015e-05,
+      "loss": 0.0001,
+      "step": 52616
+    },
+    {
+      "epoch": 49.09,
+      "learning_rate": 4.3864272388059704e-05,
+      "loss": 0.0,
+      "step": 52620
+    },
+    {
+      "epoch": 49.09,
+      "learning_rate": 4.386380597014926e-05,
+      "loss": 0.0006,
+      "step": 52624
+    },
+    {
+      "epoch": 49.09,
+      "learning_rate": 4.386333955223881e-05,
+      "loss": 0.0008,
+      "step": 52628
+    },
+    {
+      "epoch": 49.1,
+      "learning_rate": 4.3862873134328356e-05,
+      "loss": 0.0001,
+      "step": 52632
+    },
+    {
+      "epoch": 49.1,
+      "learning_rate": 4.386240671641792e-05,
+      "loss": 0.0001,
+      "step": 52636
+    },
+    {
+      "epoch": 49.1,
+      "learning_rate": 4.3861940298507465e-05,
+      "loss": 0.0,
+      "step": 52640
+    },
+    {
+      "epoch": 49.11,
+      "learning_rate": 4.3861473880597014e-05,
+      "loss": 0.0003,
+      "step": 52644
+    },
+    {
+      "epoch": 49.11,
+      "learning_rate": 4.386100746268657e-05,
+      "loss": 0.0111,
+      "step": 52648
+    },
+    {
+      "epoch": 49.12,
+      "learning_rate": 4.3860541044776123e-05,
+      "loss": 0.0,
+      "step": 52652
+    },
+    {
+      "epoch": 49.12,
+      "learning_rate": 4.386007462686567e-05,
+      "loss": 0.0003,
+      "step": 52656
+    },
+    {
+      "epoch": 49.12,
+      "learning_rate": 4.3859608208955226e-05,
+      "loss": 0.0016,
+      "step": 52660
+    },
+    {
+      "epoch": 49.13,
+      "learning_rate": 4.3859141791044775e-05,
+      "loss": 0.0,
+      "step": 52664
+    },
+    {
+      "epoch": 49.13,
+      "learning_rate": 4.385867537313433e-05,
+      "loss": 0.0,
+      "step": 52668
+    },
+    {
+      "epoch": 49.13,
+      "learning_rate": 4.3858208955223884e-05,
+      "loss": 0.0,
+      "step": 52672
+    },
+    {
+      "epoch": 49.14,
+      "learning_rate": 4.385774253731343e-05,
+      "loss": 0.0005,
+      "step": 52676
+    },
+    {
+      "epoch": 49.14,
+      "learning_rate": 4.385727611940299e-05,
+      "loss": 0.0,
+      "step": 52680
+    },
+    {
+      "epoch": 49.15,
+      "learning_rate": 4.385680970149254e-05,
+      "loss": 0.0,
+      "step": 52684
+    },
+    {
+      "epoch": 49.15,
+      "learning_rate": 4.385634328358209e-05,
+      "loss": 0.0001,
+      "step": 52688
+    },
+    {
+      "epoch": 49.15,
+      "learning_rate": 4.385587686567164e-05,
+      "loss": 0.0,
+      "step": 52692
+    },
+    {
+      "epoch": 49.16,
+      "learning_rate": 4.38554104477612e-05,
+      "loss": 0.0001,
+      "step": 52696
+    },
+    {
+      "epoch": 49.16,
+      "learning_rate": 4.385494402985075e-05,
+      "loss": 0.0,
+      "step": 52700
+    },
+    {
+      "epoch": 49.16,
+      "learning_rate": 4.38544776119403e-05,
+      "loss": 0.0,
+      "step": 52704
+    },
+    {
+      "epoch": 49.17,
+      "learning_rate": 4.385401119402985e-05,
+      "loss": 0.0,
+      "step": 52708
+    },
+    {
+      "epoch": 49.17,
+      "learning_rate": 4.3853544776119406e-05,
+      "loss": 0.0003,
+      "step": 52712
+    },
+    {
+      "epoch": 49.18,
+      "learning_rate": 4.385307835820896e-05,
+      "loss": 0.0009,
+      "step": 52716
+    },
+    {
+      "epoch": 49.18,
+      "learning_rate": 4.385261194029851e-05,
+      "loss": 0.0001,
+      "step": 52720
+    },
+    {
+      "epoch": 49.18,
+      "learning_rate": 4.385214552238806e-05,
+      "loss": 0.0008,
+      "step": 52724
+    },
+    {
+      "epoch": 49.19,
+      "learning_rate": 4.385167910447762e-05,
+      "loss": 0.005,
+      "step": 52728
+    },
+    {
+      "epoch": 49.19,
+      "learning_rate": 4.385121268656717e-05,
+      "loss": 0.0,
+      "step": 52732
+    },
+    {
+      "epoch": 49.19,
+      "learning_rate": 4.3850746268656715e-05,
+      "loss": 0.0001,
+      "step": 52736
+    },
+    {
+      "epoch": 49.2,
+      "learning_rate": 4.385027985074627e-05,
+      "loss": 0.0,
+      "step": 52740
+    },
+    {
+      "epoch": 49.2,
+      "learning_rate": 4.3849813432835825e-05,
+      "loss": 0.0,
+      "step": 52744
+    },
+    {
+      "epoch": 49.21,
+      "learning_rate": 4.3849347014925373e-05,
+      "loss": 0.0,
+      "step": 52748
+    },
+    {
+      "epoch": 49.21,
+      "learning_rate": 4.384888059701493e-05,
+      "loss": 0.0002,
+      "step": 52752
+    },
+    {
+      "epoch": 49.21,
+      "learning_rate": 4.384841417910448e-05,
+      "loss": 0.0001,
+      "step": 52756
+    },
+    {
+      "epoch": 49.22,
+      "learning_rate": 4.384794776119403e-05,
+      "loss": 0.0,
+      "step": 52760
+    },
+    {
+      "epoch": 49.22,
+      "learning_rate": 4.3847481343283586e-05,
+      "loss": 0.0006,
+      "step": 52764
+    },
+    {
+      "epoch": 49.22,
+      "learning_rate": 4.3847014925373134e-05,
+      "loss": 0.0009,
+      "step": 52768
+    },
+    {
+      "epoch": 49.23,
+      "learning_rate": 4.384654850746269e-05,
+      "loss": 0.0,
+      "step": 52772
+    },
+    {
+      "epoch": 49.23,
+      "learning_rate": 4.3846082089552244e-05,
+      "loss": 0.0003,
+      "step": 52776
+    },
+    {
+      "epoch": 49.24,
+      "learning_rate": 4.384561567164179e-05,
+      "loss": 0.0001,
+      "step": 52780
+    },
+    {
+      "epoch": 49.24,
+      "learning_rate": 4.384514925373134e-05,
+      "loss": 0.0,
+      "step": 52784
+    },
+    {
+      "epoch": 49.24,
+      "learning_rate": 4.38446828358209e-05,
+      "loss": 0.0,
+      "step": 52788
+    },
+    {
+      "epoch": 49.25,
+      "learning_rate": 4.384421641791045e-05,
+      "loss": 0.0,
+      "step": 52792
+    },
+    {
+      "epoch": 49.25,
+      "learning_rate": 4.384375e-05,
+      "loss": 0.0008,
+      "step": 52796
+    },
+    {
+      "epoch": 49.25,
+      "learning_rate": 4.384328358208955e-05,
+      "loss": 0.0001,
+      "step": 52800
+    },
+    {
+      "epoch": 49.26,
+      "learning_rate": 4.384281716417911e-05,
+      "loss": 0.0,
+      "step": 52804
+    },
+    {
+      "epoch": 49.26,
+      "learning_rate": 4.3842350746268656e-05,
+      "loss": 0.0,
+      "step": 52808
+    },
+    {
+      "epoch": 49.26,
+      "learning_rate": 4.384188432835821e-05,
+      "loss": 0.0127,
+      "step": 52812
+    },
+    {
+      "epoch": 49.27,
+      "learning_rate": 4.3841417910447766e-05,
+      "loss": 0.0,
+      "step": 52816
+    },
+    {
+      "epoch": 49.27,
+      "learning_rate": 4.3840951492537314e-05,
+      "loss": 0.0001,
+      "step": 52820
+    },
+    {
+      "epoch": 49.28,
+      "learning_rate": 4.384048507462687e-05,
+      "loss": 0.0,
+      "step": 52824
+    },
+    {
+      "epoch": 49.28,
+      "learning_rate": 4.384001865671642e-05,
+      "loss": 0.0,
+      "step": 52828
+    },
+    {
+      "epoch": 49.28,
+      "learning_rate": 4.383955223880597e-05,
+      "loss": 0.0002,
+      "step": 52832
+    },
+    {
+      "epoch": 49.29,
+      "learning_rate": 4.383908582089553e-05,
+      "loss": 0.0,
+      "step": 52836
+    },
+    {
+      "epoch": 49.29,
+      "learning_rate": 4.3838619402985075e-05,
+      "loss": 0.0007,
+      "step": 52840
+    },
+    {
+      "epoch": 49.29,
+      "learning_rate": 4.3838152985074623e-05,
+      "loss": 0.0006,
+      "step": 52844
+    },
+    {
+      "epoch": 49.3,
+      "learning_rate": 4.3837686567164185e-05,
+      "loss": 0.0001,
+      "step": 52848
+    },
+    {
+      "epoch": 49.3,
+      "learning_rate": 4.383722014925373e-05,
+      "loss": 0.0,
+      "step": 52852
+    },
+    {
+      "epoch": 49.31,
+      "learning_rate": 4.383675373134328e-05,
+      "loss": 0.0001,
+      "step": 52856
+    },
+    {
+      "epoch": 49.31,
+      "learning_rate": 4.3836287313432836e-05,
+      "loss": 0.0004,
+      "step": 52860
+    },
+    {
+      "epoch": 49.31,
+      "learning_rate": 4.383582089552239e-05,
+      "loss": 0.0001,
+      "step": 52864
+    },
+    {
+      "epoch": 49.32,
+      "learning_rate": 4.3835354477611946e-05,
+      "loss": 0.0017,
+      "step": 52868
+    },
+    {
+      "epoch": 49.32,
+      "learning_rate": 4.3834888059701494e-05,
+      "loss": 0.0015,
+      "step": 52872
+    },
+    {
+      "epoch": 49.32,
+      "learning_rate": 4.383442164179105e-05,
+      "loss": 0.0,
+      "step": 52876
+    },
+    {
+      "epoch": 49.33,
+      "learning_rate": 4.3833955223880604e-05,
+      "loss": 0.0,
+      "step": 52880
+    },
+    {
+      "epoch": 49.33,
+      "learning_rate": 4.383348880597015e-05,
+      "loss": 0.0,
+      "step": 52884
+    },
+    {
+      "epoch": 49.34,
+      "learning_rate": 4.38330223880597e-05,
+      "loss": 0.0006,
+      "step": 52888
+    },
+    {
+      "epoch": 49.34,
+      "learning_rate": 4.3832555970149255e-05,
+      "loss": 0.0,
+      "step": 52892
+    },
+    {
+      "epoch": 49.34,
+      "learning_rate": 4.383208955223881e-05,
+      "loss": 0.0,
+      "step": 52896
+    },
+    {
+      "epoch": 49.35,
+      "learning_rate": 4.383162313432836e-05,
+      "loss": 0.0005,
+      "step": 52900
+    },
+    {
+      "epoch": 49.35,
+      "learning_rate": 4.383115671641791e-05,
+      "loss": 0.0,
+      "step": 52904
+    },
+    {
+      "epoch": 49.35,
+      "learning_rate": 4.383069029850747e-05,
+      "loss": 0.0005,
+      "step": 52908
+    },
+    {
+      "epoch": 49.36,
+      "learning_rate": 4.3830223880597016e-05,
+      "loss": 0.0,
+      "step": 52912
+    },
+    {
+      "epoch": 49.36,
+      "learning_rate": 4.382975746268657e-05,
+      "loss": 0.0001,
+      "step": 52916
+    },
+    {
+      "epoch": 49.37,
+      "learning_rate": 4.382929104477612e-05,
+      "loss": 0.0003,
+      "step": 52920
+    },
+    {
+      "epoch": 49.37,
+      "learning_rate": 4.3828824626865674e-05,
+      "loss": 0.0,
+      "step": 52924
+    },
+    {
+      "epoch": 49.37,
+      "learning_rate": 4.382835820895523e-05,
+      "loss": 0.0,
+      "step": 52928
+    },
+    {
+      "epoch": 49.38,
+      "learning_rate": 4.382789179104478e-05,
+      "loss": 0.0,
+      "step": 52932
+    },
+    {
+      "epoch": 49.38,
+      "learning_rate": 4.3827425373134325e-05,
+      "loss": 0.0002,
+      "step": 52936
+    },
+    {
+      "epoch": 49.38,
+      "learning_rate": 4.382695895522389e-05,
+      "loss": 0.0002,
+      "step": 52940
+    },
+    {
+      "epoch": 49.39,
+      "learning_rate": 4.3826492537313435e-05,
+      "loss": 0.0,
+      "step": 52944
+    },
+    {
+      "epoch": 49.39,
+      "learning_rate": 4.382602611940298e-05,
+      "loss": 0.0,
+      "step": 52948
+    },
+    {
+      "epoch": 49.4,
+      "learning_rate": 4.382555970149254e-05,
+      "loss": 0.0007,
+      "step": 52952
+    },
+    {
+      "epoch": 49.4,
+      "learning_rate": 4.382509328358209e-05,
+      "loss": 0.0001,
+      "step": 52956
+    },
+    {
+      "epoch": 49.4,
+      "learning_rate": 4.382462686567164e-05,
+      "loss": 0.0,
+      "step": 52960
+    },
+    {
+      "epoch": 49.41,
+      "learning_rate": 4.3824160447761196e-05,
+      "loss": 0.0001,
+      "step": 52964
+    },
+    {
+      "epoch": 49.41,
+      "learning_rate": 4.382369402985075e-05,
+      "loss": 0.0003,
+      "step": 52968
+    },
+    {
+      "epoch": 49.41,
+      "learning_rate": 4.38232276119403e-05,
+      "loss": 0.0006,
+      "step": 52972
+    },
+    {
+      "epoch": 49.42,
+      "learning_rate": 4.3822761194029854e-05,
+      "loss": 0.0,
+      "step": 52976
+    },
+    {
+      "epoch": 49.42,
+      "learning_rate": 4.38222947761194e-05,
+      "loss": 0.0,
+      "step": 52980
+    },
+    {
+      "epoch": 49.43,
+      "learning_rate": 4.382182835820896e-05,
+      "loss": 0.0001,
+      "step": 52984
+    },
+    {
+      "epoch": 49.43,
+      "learning_rate": 4.382136194029851e-05,
+      "loss": 0.0,
+      "step": 52988
+    },
+    {
+      "epoch": 49.43,
+      "learning_rate": 4.382089552238806e-05,
+      "loss": 0.0,
+      "step": 52992
+    },
+    {
+      "epoch": 49.44,
+      "learning_rate": 4.382042910447761e-05,
+      "loss": 0.0,
+      "step": 52996
+    },
+    {
+      "epoch": 49.44,
+      "learning_rate": 4.381996268656717e-05,
+      "loss": 0.0,
+      "step": 53000
+    },
+    {
+      "epoch": 49.44,
+      "eval_exact_match": 0.7485493230174082,
+      "eval_exec": 0.7804642166344294,
+      "eval_loss": 0.4629165828227997,
+      "eval_runtime": 1125.3708,
+      "eval_samples_per_second": 0.919,
+      "step": 53000
+    },
+    {
+      "epoch": 49.44,
+      "learning_rate": 4.381949626865672e-05,
+      "loss": 0.0061,
+      "step": 53004
+    },
+    {
+      "epoch": 49.45,
+      "learning_rate": 4.3819029850746266e-05,
+      "loss": 0.0,
+      "step": 53008
+    },
+    {
+      "epoch": 49.45,
+      "learning_rate": 4.381856343283582e-05,
+      "loss": 0.0001,
+      "step": 53012
+    },
+    {
+      "epoch": 49.46,
+      "learning_rate": 4.3818097014925376e-05,
+      "loss": 0.0,
+      "step": 53016
+    },
+    {
+      "epoch": 49.46,
+      "learning_rate": 4.3817630597014924e-05,
+      "loss": 0.0,
+      "step": 53020
+    },
+    {
+      "epoch": 49.46,
+      "learning_rate": 4.381716417910448e-05,
+      "loss": 0.0,
+      "step": 53024
+    },
+    {
+      "epoch": 49.47,
+      "learning_rate": 4.3816697761194034e-05,
+      "loss": 0.0013,
+      "step": 53028
+    },
+    {
+      "epoch": 49.47,
+      "learning_rate": 4.381623134328359e-05,
+      "loss": 0.0001,
+      "step": 53032
+    },
+    {
+      "epoch": 49.47,
+      "learning_rate": 4.381576492537314e-05,
+      "loss": 0.0,
+      "step": 53036
+    },
+    {
+      "epoch": 49.48,
+      "learning_rate": 4.3815298507462685e-05,
+      "loss": 0.0,
+      "step": 53040
+    },
+    {
+      "epoch": 49.48,
+      "learning_rate": 4.381483208955225e-05,
+      "loss": 0.0003,
+      "step": 53044
+    },
+    {
+      "epoch": 49.49,
+      "learning_rate": 4.3814365671641795e-05,
+      "loss": 0.0003,
+      "step": 53048
+    },
+    {
+      "epoch": 49.49,
+      "learning_rate": 4.381389925373134e-05,
+      "loss": 0.0,
+      "step": 53052
+    },
+    {
+      "epoch": 49.49,
+      "learning_rate": 4.38134328358209e-05,
+      "loss": 0.0,
+      "step": 53056
+    },
+    {
+      "epoch": 49.5,
+      "learning_rate": 4.381296641791045e-05,
+      "loss": 0.0001,
+      "step": 53060
+    },
+    {
+      "epoch": 49.5,
+      "learning_rate": 4.38125e-05,
+      "loss": 0.0,
+      "step": 53064
+    },
+    {
+      "epoch": 49.5,
+      "learning_rate": 4.3812033582089556e-05,
+      "loss": 0.0,
+      "step": 53068
+    },
+    {
+      "epoch": 49.51,
+      "learning_rate": 4.3811567164179104e-05,
+      "loss": 0.0002,
+      "step": 53072
+    },
+    {
+      "epoch": 49.51,
+      "learning_rate": 4.381110074626866e-05,
+      "loss": 0.0001,
+      "step": 53076
+    },
+    {
+      "epoch": 49.51,
+      "learning_rate": 4.3810634328358214e-05,
+      "loss": 0.0,
+      "step": 53080
+    },
+    {
+      "epoch": 49.52,
+      "learning_rate": 4.381016791044776e-05,
+      "loss": 0.0001,
+      "step": 53084
+    },
+    {
+      "epoch": 49.52,
+      "learning_rate": 4.380970149253732e-05,
+      "loss": 0.0,
+      "step": 53088
+    },
+    {
+      "epoch": 49.53,
+      "learning_rate": 4.380923507462687e-05,
+      "loss": 0.0006,
+      "step": 53092
+    },
+    {
+      "epoch": 49.53,
+      "learning_rate": 4.380876865671642e-05,
+      "loss": 0.0001,
+      "step": 53096
+    },
+    {
+      "epoch": 49.53,
+      "learning_rate": 4.380830223880597e-05,
+      "loss": 0.0,
+      "step": 53100
+    },
+    {
+      "epoch": 49.54,
+      "learning_rate": 4.380783582089552e-05,
+      "loss": 0.0028,
+      "step": 53104
+    },
+    {
+      "epoch": 49.54,
+      "learning_rate": 4.380736940298508e-05,
+      "loss": 0.0,
+      "step": 53108
+    },
+    {
+      "epoch": 49.54,
+      "learning_rate": 4.3806902985074626e-05,
+      "loss": 0.0001,
+      "step": 53112
+    },
+    {
+      "epoch": 49.55,
+      "learning_rate": 4.380643656716418e-05,
+      "loss": 0.0,
+      "step": 53116
+    },
+    {
+      "epoch": 49.55,
+      "learning_rate": 4.3805970149253736e-05,
+      "loss": 0.0,
+      "step": 53120
+    },
+    {
+      "epoch": 49.56,
+      "learning_rate": 4.3805503731343284e-05,
+      "loss": 0.0001,
+      "step": 53124
+    },
+    {
+      "epoch": 49.56,
+      "learning_rate": 4.380503731343284e-05,
+      "loss": 0.0,
+      "step": 53128
+    },
+    {
+      "epoch": 49.56,
+      "learning_rate": 4.380457089552239e-05,
+      "loss": 0.0,
+      "step": 53132
+    },
+    {
+      "epoch": 49.57,
+      "learning_rate": 4.380410447761194e-05,
+      "loss": 0.0,
+      "step": 53136
+    },
+    {
+      "epoch": 49.57,
+      "learning_rate": 4.38036380597015e-05,
+      "loss": 0.0001,
+      "step": 53140
+    },
+    {
+      "epoch": 49.57,
+      "learning_rate": 4.3803171641791045e-05,
+      "loss": 0.0001,
+      "step": 53144
+    },
+    {
+      "epoch": 49.58,
+      "learning_rate": 4.38027052238806e-05,
+      "loss": 0.0001,
+      "step": 53148
+    },
+    {
+      "epoch": 49.58,
+      "learning_rate": 4.3802238805970155e-05,
+      "loss": 0.0,
+      "step": 53152
+    },
+    {
+      "epoch": 49.59,
+      "learning_rate": 4.38017723880597e-05,
+      "loss": 0.0,
+      "step": 53156
+    },
+    {
+      "epoch": 49.59,
+      "learning_rate": 4.380130597014925e-05,
+      "loss": 0.0,
+      "step": 53160
+    },
+    {
+      "epoch": 49.59,
+      "learning_rate": 4.3800839552238806e-05,
+      "loss": 0.0001,
+      "step": 53164
+    },
+    {
+      "epoch": 49.6,
+      "learning_rate": 4.380037313432836e-05,
+      "loss": 0.0,
+      "step": 53168
+    },
+    {
+      "epoch": 49.6,
+      "learning_rate": 4.379990671641791e-05,
+      "loss": 0.0001,
+      "step": 53172
+    },
+    {
+      "epoch": 49.6,
+      "learning_rate": 4.3799440298507464e-05,
+      "loss": 0.0015,
+      "step": 53176
+    },
+    {
+      "epoch": 49.61,
+      "learning_rate": 4.379897388059702e-05,
+      "loss": 0.0,
+      "step": 53180
+    },
+    {
+      "epoch": 49.61,
+      "learning_rate": 4.379850746268657e-05,
+      "loss": 0.0,
+      "step": 53184
+    },
+    {
+      "epoch": 49.62,
+      "learning_rate": 4.379804104477612e-05,
+      "loss": 0.0,
+      "step": 53188
+    },
+    {
+      "epoch": 49.62,
+      "learning_rate": 4.379757462686567e-05,
+      "loss": 0.0007,
+      "step": 53192
+    },
+    {
+      "epoch": 49.62,
+      "learning_rate": 4.379710820895523e-05,
+      "loss": 0.0007,
+      "step": 53196
+    },
+    {
+      "epoch": 49.63,
+      "learning_rate": 4.379664179104478e-05,
+      "loss": 0.0,
+      "step": 53200
+    },
+    {
+      "epoch": 49.63,
+      "learning_rate": 4.379617537313433e-05,
+      "loss": 0.0002,
+      "step": 53204
+    },
+    {
+      "epoch": 49.63,
+      "learning_rate": 4.379570895522388e-05,
+      "loss": 0.0001,
+      "step": 53208
+    },
+    {
+      "epoch": 49.64,
+      "learning_rate": 4.379524253731344e-05,
+      "loss": 0.0003,
+      "step": 53212
+    },
+    {
+      "epoch": 49.64,
+      "learning_rate": 4.3794776119402986e-05,
+      "loss": 0.0,
+      "step": 53216
+    },
+    {
+      "epoch": 49.65,
+      "learning_rate": 4.379430970149254e-05,
+      "loss": 0.0,
+      "step": 53220
+    },
+    {
+      "epoch": 49.65,
+      "learning_rate": 4.379384328358209e-05,
+      "loss": 0.0,
+      "step": 53224
+    },
+    {
+      "epoch": 49.65,
+      "learning_rate": 4.3793376865671644e-05,
+      "loss": 0.0,
+      "step": 53228
+    },
+    {
+      "epoch": 49.66,
+      "learning_rate": 4.37929104477612e-05,
+      "loss": 0.0,
+      "step": 53232
+    },
+    {
+      "epoch": 49.66,
+      "learning_rate": 4.379244402985075e-05,
+      "loss": 0.0001,
+      "step": 53236
+    },
+    {
+      "epoch": 49.66,
+      "learning_rate": 4.37919776119403e-05,
+      "loss": 0.001,
+      "step": 53240
+    },
+    {
+      "epoch": 49.67,
+      "learning_rate": 4.379151119402986e-05,
+      "loss": 0.0,
+      "step": 53244
+    },
+    {
+      "epoch": 49.67,
+      "learning_rate": 4.3791044776119405e-05,
+      "loss": 0.0,
+      "step": 53248
+    },
+    {
+      "epoch": 49.68,
+      "learning_rate": 4.379057835820895e-05,
+      "loss": 0.0005,
+      "step": 53252
+    },
+    {
+      "epoch": 49.68,
+      "learning_rate": 4.3790111940298515e-05,
+      "loss": 0.0,
+      "step": 53256
+    },
+    {
+      "epoch": 49.68,
+      "learning_rate": 4.378964552238806e-05,
+      "loss": 0.0001,
+      "step": 53260
+    },
+    {
+      "epoch": 49.69,
+      "learning_rate": 4.378917910447761e-05,
+      "loss": 0.0,
+      "step": 53264
+    },
+    {
+      "epoch": 49.69,
+      "learning_rate": 4.3788712686567166e-05,
+      "loss": 0.0,
+      "step": 53268
+    },
+    {
+      "epoch": 49.69,
+      "learning_rate": 4.378824626865672e-05,
+      "loss": 0.0,
+      "step": 53272
+    },
+    {
+      "epoch": 49.7,
+      "learning_rate": 4.378777985074627e-05,
+      "loss": 0.0021,
+      "step": 53276
+    },
+    {
+      "epoch": 49.7,
+      "learning_rate": 4.3787313432835824e-05,
+      "loss": 0.0,
+      "step": 53280
+    },
+    {
+      "epoch": 49.71,
+      "learning_rate": 4.378684701492537e-05,
+      "loss": 0.0,
+      "step": 53284
+    },
+    {
+      "epoch": 49.71,
+      "learning_rate": 4.378638059701493e-05,
+      "loss": 0.0001,
+      "step": 53288
+    },
+    {
+      "epoch": 49.71,
+      "learning_rate": 4.378591417910448e-05,
+      "loss": 0.0,
+      "step": 53292
+    },
+    {
+      "epoch": 49.72,
+      "learning_rate": 4.378544776119403e-05,
+      "loss": 0.0,
+      "step": 53296
+    },
+    {
+      "epoch": 49.72,
+      "learning_rate": 4.3784981343283585e-05,
+      "loss": 0.0,
+      "step": 53300
+    },
+    {
+      "epoch": 49.72,
+      "learning_rate": 4.378451492537314e-05,
+      "loss": 0.0,
+      "step": 53304
+    },
+    {
+      "epoch": 49.73,
+      "learning_rate": 4.378404850746269e-05,
+      "loss": 0.0,
+      "step": 53308
+    },
+    {
+      "epoch": 49.73,
+      "learning_rate": 4.3783582089552236e-05,
+      "loss": 0.0,
+      "step": 53312
+    },
+    {
+      "epoch": 49.73,
+      "learning_rate": 4.37831156716418e-05,
+      "loss": 0.0,
+      "step": 53316
+    },
+    {
+      "epoch": 49.74,
+      "learning_rate": 4.3782649253731346e-05,
+      "loss": 0.0015,
+      "step": 53320
+    },
+    {
+      "epoch": 49.74,
+      "learning_rate": 4.3782182835820894e-05,
+      "loss": 0.0,
+      "step": 53324
+    },
+    {
+      "epoch": 49.75,
+      "learning_rate": 4.378171641791045e-05,
+      "loss": 0.0002,
+      "step": 53328
+    },
+    {
+      "epoch": 49.75,
+      "learning_rate": 4.3781250000000004e-05,
+      "loss": 0.0,
+      "step": 53332
+    },
+    {
+      "epoch": 49.75,
+      "learning_rate": 4.378078358208955e-05,
+      "loss": 0.001,
+      "step": 53336
+    },
+    {
+      "epoch": 49.76,
+      "learning_rate": 4.378031716417911e-05,
+      "loss": 0.0,
+      "step": 53340
+    },
+    {
+      "epoch": 49.76,
+      "learning_rate": 4.3779850746268655e-05,
+      "loss": 0.0007,
+      "step": 53344
+    },
+    {
+      "epoch": 49.76,
+      "learning_rate": 4.377938432835821e-05,
+      "loss": 0.0,
+      "step": 53348
+    },
+    {
+      "epoch": 49.77,
+      "learning_rate": 4.3778917910447765e-05,
+      "loss": 0.0,
+      "step": 53352
+    },
+    {
+      "epoch": 49.77,
+      "learning_rate": 4.377845149253731e-05,
+      "loss": 0.013,
+      "step": 53356
+    },
+    {
+      "epoch": 49.78,
+      "learning_rate": 4.377798507462687e-05,
+      "loss": 0.0007,
+      "step": 53360
+    },
+    {
+      "epoch": 49.78,
+      "learning_rate": 4.377751865671642e-05,
+      "loss": 0.0,
+      "step": 53364
+    },
+    {
+      "epoch": 49.78,
+      "learning_rate": 4.377705223880597e-05,
+      "loss": 0.0,
+      "step": 53368
+    },
+    {
+      "epoch": 49.79,
+      "learning_rate": 4.3776585820895526e-05,
+      "loss": 0.0,
+      "step": 53372
+    },
+    {
+      "epoch": 49.79,
+      "learning_rate": 4.377611940298508e-05,
+      "loss": 0.0,
+      "step": 53376
+    },
+    {
+      "epoch": 49.79,
+      "learning_rate": 4.377565298507463e-05,
+      "loss": 0.0,
+      "step": 53380
+    },
+    {
+      "epoch": 49.8,
+      "learning_rate": 4.3775186567164184e-05,
+      "loss": 0.0,
+      "step": 53384
+    },
+    {
+      "epoch": 49.8,
+      "learning_rate": 4.377472014925373e-05,
+      "loss": 0.0006,
+      "step": 53388
+    },
+    {
+      "epoch": 49.81,
+      "learning_rate": 4.377425373134329e-05,
+      "loss": 0.0032,
+      "step": 53392
+    },
+    {
+      "epoch": 49.81,
+      "learning_rate": 4.377378731343284e-05,
+      "loss": 0.0001,
+      "step": 53396
+    },
+    {
+      "epoch": 49.81,
+      "learning_rate": 4.377332089552239e-05,
+      "loss": 0.0,
+      "step": 53400
+    },
+    {
+      "epoch": 49.82,
+      "learning_rate": 4.377285447761194e-05,
+      "loss": 0.0,
+      "step": 53404
+    },
+    {
+      "epoch": 49.82,
+      "learning_rate": 4.37723880597015e-05,
+      "loss": 0.0,
+      "step": 53408
+    },
+    {
+      "epoch": 49.82,
+      "learning_rate": 4.377192164179105e-05,
+      "loss": 0.0001,
+      "step": 53412
+    },
+    {
+      "epoch": 49.83,
+      "learning_rate": 4.3771455223880596e-05,
+      "loss": 0.0,
+      "step": 53416
+    },
+    {
+      "epoch": 49.83,
+      "learning_rate": 4.377098880597015e-05,
+      "loss": 0.0,
+      "step": 53420
+    },
+    {
+      "epoch": 49.84,
+      "learning_rate": 4.3770522388059706e-05,
+      "loss": 0.0,
+      "step": 53424
+    },
+    {
+      "epoch": 49.84,
+      "learning_rate": 4.3770055970149254e-05,
+      "loss": 0.0,
+      "step": 53428
+    },
+    {
+      "epoch": 49.84,
+      "learning_rate": 4.376958955223881e-05,
+      "loss": 0.0,
+      "step": 53432
+    },
+    {
+      "epoch": 49.85,
+      "learning_rate": 4.3769123134328364e-05,
+      "loss": 0.0,
+      "step": 53436
+    },
+    {
+      "epoch": 49.85,
+      "learning_rate": 4.376865671641791e-05,
+      "loss": 0.0,
+      "step": 53440
+    },
+    {
+      "epoch": 49.85,
+      "learning_rate": 4.3768190298507467e-05,
+      "loss": 0.0021,
+      "step": 53444
+    },
+    {
+      "epoch": 49.86,
+      "learning_rate": 4.3767723880597015e-05,
+      "loss": 0.0,
+      "step": 53448
+    },
+    {
+      "epoch": 49.86,
+      "learning_rate": 4.376725746268657e-05,
+      "loss": 0.0002,
+      "step": 53452
+    },
+    {
+      "epoch": 49.87,
+      "learning_rate": 4.3766791044776125e-05,
+      "loss": 0.0021,
+      "step": 53456
+    },
+    {
+      "epoch": 49.87,
+      "learning_rate": 4.376632462686567e-05,
+      "loss": 0.0,
+      "step": 53460
+    },
+    {
+      "epoch": 49.87,
+      "learning_rate": 4.376585820895522e-05,
+      "loss": 0.0,
+      "step": 53464
+    },
+    {
+      "epoch": 49.88,
+      "learning_rate": 4.376539179104478e-05,
+      "loss": 0.0001,
+      "step": 53468
+    },
+    {
+      "epoch": 49.88,
+      "learning_rate": 4.376492537313433e-05,
+      "loss": 0.0,
+      "step": 53472
+    },
+    {
+      "epoch": 49.88,
+      "learning_rate": 4.376445895522388e-05,
+      "loss": 0.0,
+      "step": 53476
+    },
+    {
+      "epoch": 49.89,
+      "learning_rate": 4.3763992537313434e-05,
+      "loss": 0.0,
+      "step": 53480
+    },
+    {
+      "epoch": 49.89,
+      "learning_rate": 4.376352611940299e-05,
+      "loss": 0.0005,
+      "step": 53484
+    },
+    {
+      "epoch": 49.9,
+      "learning_rate": 4.376305970149254e-05,
+      "loss": 0.0,
+      "step": 53488
+    },
+    {
+      "epoch": 49.9,
+      "learning_rate": 4.376259328358209e-05,
+      "loss": 0.0,
+      "step": 53492
+    },
+    {
+      "epoch": 49.9,
+      "learning_rate": 4.3762126865671646e-05,
+      "loss": 0.0,
+      "step": 53496
+    },
+    {
+      "epoch": 49.91,
+      "learning_rate": 4.3761660447761195e-05,
+      "loss": 0.0,
+      "step": 53500
+    },
+    {
+      "epoch": 49.91,
+      "eval_exact_match": 0.7495164410058027,
+      "eval_exec": 0.7833655705996132,
+      "eval_loss": 0.4703017771244049,
+      "eval_runtime": 1159.4214,
+      "eval_samples_per_second": 0.892,
+      "step": 53500
+    },
+    {
+      "epoch": 49.91,
+      "learning_rate": 4.376119402985075e-05,
+      "loss": 0.001,
+      "step": 53504
+    },
+    {
+      "epoch": 49.91,
+      "learning_rate": 4.37607276119403e-05,
+      "loss": 0.0,
+      "step": 53508
+    },
+    {
+      "epoch": 49.92,
+      "learning_rate": 4.376026119402985e-05,
+      "loss": 0.0,
+      "step": 53512
+    },
+    {
+      "epoch": 49.92,
+      "learning_rate": 4.375979477611941e-05,
+      "loss": 0.0004,
+      "step": 53516
+    },
+    {
+      "epoch": 49.93,
+      "learning_rate": 4.3759328358208956e-05,
+      "loss": 0.0,
+      "step": 53520
+    },
+    {
+      "epoch": 49.93,
+      "learning_rate": 4.375886194029851e-05,
+      "loss": 0.0,
+      "step": 53524
+    },
+    {
+      "epoch": 49.93,
+      "learning_rate": 4.3758395522388065e-05,
+      "loss": 0.0,
+      "step": 53528
+    },
+    {
+      "epoch": 49.94,
+      "learning_rate": 4.3757929104477614e-05,
+      "loss": 0.0,
+      "step": 53532
+    },
+    {
+      "epoch": 49.94,
+      "learning_rate": 4.375746268656717e-05,
+      "loss": 0.0,
+      "step": 53536
+    },
+    {
+      "epoch": 49.94,
+      "learning_rate": 4.3756996268656717e-05,
+      "loss": 0.0023,
+      "step": 53540
+    },
+    {
+      "epoch": 49.95,
+      "learning_rate": 4.375652985074627e-05,
+      "loss": 0.0001,
+      "step": 53544
+    },
+    {
+      "epoch": 49.95,
+      "learning_rate": 4.3756063432835826e-05,
+      "loss": 0.0,
+      "step": 53548
+    },
+    {
+      "epoch": 49.96,
+      "learning_rate": 4.3755597014925375e-05,
+      "loss": 0.0,
+      "step": 53552
+    },
+    {
+      "epoch": 49.96,
+      "learning_rate": 4.375513059701493e-05,
+      "loss": 0.0,
+      "step": 53556
+    },
+    {
+      "epoch": 49.96,
+      "learning_rate": 4.3754664179104484e-05,
+      "loss": 0.0001,
+      "step": 53560
+    },
+    {
+      "epoch": 49.97,
+      "learning_rate": 4.375419776119403e-05,
+      "loss": 0.0001,
+      "step": 53564
+    },
+    {
+      "epoch": 49.97,
+      "learning_rate": 4.375373134328358e-05,
+      "loss": 0.0,
+      "step": 53568
+    },
+    {
+      "epoch": 49.97,
+      "learning_rate": 4.3753264925373136e-05,
+      "loss": 0.0001,
+      "step": 53572
+    },
+    {
+      "epoch": 49.98,
+      "learning_rate": 4.375279850746269e-05,
+      "loss": 0.0002,
+      "step": 53576
+    },
+    {
+      "epoch": 49.98,
+      "learning_rate": 4.375233208955224e-05,
+      "loss": 0.0003,
+      "step": 53580
+    },
+    {
+      "epoch": 49.98,
+      "learning_rate": 4.3751865671641793e-05,
+      "loss": 0.0001,
+      "step": 53584
+    },
+    {
+      "epoch": 49.99,
+      "learning_rate": 4.375139925373135e-05,
+      "loss": 0.0024,
+      "step": 53588
+    },
+    {
+      "epoch": 49.99,
+      "learning_rate": 4.3750932835820897e-05,
+      "loss": 0.0,
+      "step": 53592
+    },
+    {
+      "epoch": 50.0,
+      "learning_rate": 4.375046641791045e-05,
+      "loss": 0.0,
+      "step": 53596
+    },
+    {
+      "epoch": 50.0,
+      "learning_rate": 4.375e-05,
+      "loss": 0.0,
+      "step": 53600
+    },
+    {
+      "epoch": 50.0,
+      "learning_rate": 4.3749533582089554e-05,
+      "loss": 0.0,
+      "step": 53604
+    },
+    {
+      "epoch": 50.01,
+      "learning_rate": 4.374906716417911e-05,
+      "loss": 0.0,
+      "step": 53608
+    },
+    {
+      "epoch": 50.01,
+      "learning_rate": 4.374860074626866e-05,
+      "loss": 0.0008,
+      "step": 53612
+    },
+    {
+      "epoch": 50.01,
+      "learning_rate": 4.3748134328358206e-05,
+      "loss": 0.0,
+      "step": 53616
+    },
+    {
+      "epoch": 50.02,
+      "learning_rate": 4.374766791044777e-05,
+      "loss": 0.0003,
+      "step": 53620
+    },
+    {
+      "epoch": 50.02,
+      "learning_rate": 4.3747201492537315e-05,
+      "loss": 0.0001,
+      "step": 53624
+    },
+    {
+      "epoch": 50.03,
+      "learning_rate": 4.3746735074626864e-05,
+      "loss": 0.003,
+      "step": 53628
+    },
+    {
+      "epoch": 50.03,
+      "learning_rate": 4.374626865671642e-05,
+      "loss": 0.0002,
+      "step": 53632
+    },
+    {
+      "epoch": 50.03,
+      "learning_rate": 4.3745802238805973e-05,
+      "loss": 0.0001,
+      "step": 53636
+    },
+    {
+      "epoch": 50.04,
+      "learning_rate": 4.374533582089552e-05,
+      "loss": 0.0,
+      "step": 53640
+    },
+    {
+      "epoch": 50.04,
+      "learning_rate": 4.3744869402985076e-05,
+      "loss": 0.0003,
+      "step": 53644
+    },
+    {
+      "epoch": 50.04,
+      "learning_rate": 4.374440298507463e-05,
+      "loss": 0.0,
+      "step": 53648
+    },
+    {
+      "epoch": 50.05,
+      "learning_rate": 4.374393656716418e-05,
+      "loss": 0.0,
+      "step": 53652
+    },
+    {
+      "epoch": 50.05,
+      "learning_rate": 4.3743470149253734e-05,
+      "loss": 0.0,
+      "step": 53656
+    },
+    {
+      "epoch": 50.06,
+      "learning_rate": 4.374300373134328e-05,
+      "loss": 0.0002,
+      "step": 53660
+    },
+    {
+      "epoch": 50.06,
+      "learning_rate": 4.374253731343284e-05,
+      "loss": 0.0001,
+      "step": 53664
+    },
+    {
+      "epoch": 50.06,
+      "learning_rate": 4.374207089552239e-05,
+      "loss": 0.0005,
+      "step": 53668
+    },
+    {
+      "epoch": 50.07,
+      "learning_rate": 4.374160447761194e-05,
+      "loss": 0.0001,
+      "step": 53672
+    },
+    {
+      "epoch": 50.07,
+      "learning_rate": 4.374113805970149e-05,
+      "loss": 0.0001,
+      "step": 53676
+    },
+    {
+      "epoch": 50.07,
+      "learning_rate": 4.374067164179105e-05,
+      "loss": 0.0,
+      "step": 53680
+    },
+    {
+      "epoch": 50.08,
+      "learning_rate": 4.37402052238806e-05,
+      "loss": 0.006,
+      "step": 53684
+    },
+    {
+      "epoch": 50.08,
+      "learning_rate": 4.373973880597015e-05,
+      "loss": 0.0,
+      "step": 53688
+    },
+    {
+      "epoch": 50.09,
+      "learning_rate": 4.37392723880597e-05,
+      "loss": 0.0,
+      "step": 53692
+    },
+    {
+      "epoch": 50.09,
+      "learning_rate": 4.3738805970149256e-05,
+      "loss": 0.0001,
+      "step": 53696
+    },
+    {
+      "epoch": 50.09,
+      "learning_rate": 4.373833955223881e-05,
+      "loss": 0.0004,
+      "step": 53700
+    },
+    {
+      "epoch": 50.1,
+      "learning_rate": 4.373787313432836e-05,
+      "loss": 0.0,
+      "step": 53704
+    },
+    {
+      "epoch": 50.1,
+      "learning_rate": 4.3737406716417914e-05,
+      "loss": 0.0,
+      "step": 53708
+    },
+    {
+      "epoch": 50.1,
+      "learning_rate": 4.373694029850747e-05,
+      "loss": 0.0,
+      "step": 53712
+    },
+    {
+      "epoch": 50.11,
+      "learning_rate": 4.373647388059702e-05,
+      "loss": 0.0001,
+      "step": 53716
+    },
+    {
+      "epoch": 50.11,
+      "learning_rate": 4.3736007462686565e-05,
+      "loss": 0.0,
+      "step": 53720
+    },
+    {
+      "epoch": 50.12,
+      "learning_rate": 4.373554104477613e-05,
+      "loss": 0.0,
+      "step": 53724
+    },
+    {
+      "epoch": 50.12,
+      "learning_rate": 4.3735074626865675e-05,
+      "loss": 0.0,
+      "step": 53728
+    },
+    {
+      "epoch": 50.12,
+      "learning_rate": 4.3734608208955223e-05,
+      "loss": 0.0001,
+      "step": 53732
+    },
+    {
+      "epoch": 50.13,
+      "learning_rate": 4.373414179104478e-05,
+      "loss": 0.0022,
+      "step": 53736
+    },
+    {
+      "epoch": 50.13,
+      "learning_rate": 4.373367537313433e-05,
+      "loss": 0.0,
+      "step": 53740
+    },
+    {
+      "epoch": 50.13,
+      "learning_rate": 4.373320895522388e-05,
+      "loss": 0.0,
+      "step": 53744
+    },
+    {
+      "epoch": 50.14,
+      "learning_rate": 4.3732742537313436e-05,
+      "loss": 0.0,
+      "step": 53748
+    },
+    {
+      "epoch": 50.14,
+      "learning_rate": 4.3732276119402984e-05,
+      "loss": 0.0001,
+      "step": 53752
+    },
+    {
+      "epoch": 50.15,
+      "learning_rate": 4.373180970149254e-05,
+      "loss": 0.0,
+      "step": 53756
+    },
+    {
+      "epoch": 50.15,
+      "learning_rate": 4.3731343283582094e-05,
+      "loss": 0.0001,
+      "step": 53760
+    },
+    {
+      "epoch": 50.15,
+      "learning_rate": 4.373087686567164e-05,
+      "loss": 0.0001,
+      "step": 53764
+    },
+    {
+      "epoch": 50.16,
+      "learning_rate": 4.37304104477612e-05,
+      "loss": 0.0,
+      "step": 53768
+    },
+    {
+      "epoch": 50.16,
+      "learning_rate": 4.372994402985075e-05,
+      "loss": 0.0002,
+      "step": 53772
+    },
+    {
+      "epoch": 50.16,
+      "learning_rate": 4.37294776119403e-05,
+      "loss": 0.0,
+      "step": 53776
+    },
+    {
+      "epoch": 50.17,
+      "learning_rate": 4.372901119402985e-05,
+      "loss": 0.0,
+      "step": 53780
+    },
+    {
+      "epoch": 50.17,
+      "learning_rate": 4.37285447761194e-05,
+      "loss": 0.0,
+      "step": 53784
+    },
+    {
+      "epoch": 50.18,
+      "learning_rate": 4.372807835820896e-05,
+      "loss": 0.0,
+      "step": 53788
+    },
+    {
+      "epoch": 50.18,
+      "learning_rate": 4.3727611940298506e-05,
+      "loss": 0.0001,
+      "step": 53792
+    },
+    {
+      "epoch": 50.18,
+      "learning_rate": 4.372714552238806e-05,
+      "loss": 0.0002,
+      "step": 53796
+    },
+    {
+      "epoch": 50.19,
+      "learning_rate": 4.3726679104477616e-05,
+      "loss": 0.0003,
+      "step": 53800
+    },
+    {
+      "epoch": 50.19,
+      "learning_rate": 4.3726212686567164e-05,
+      "loss": 0.0,
+      "step": 53804
+    },
+    {
+      "epoch": 50.19,
+      "learning_rate": 4.372574626865672e-05,
+      "loss": 0.0006,
+      "step": 53808
+    },
+    {
+      "epoch": 50.2,
+      "learning_rate": 4.372527985074627e-05,
+      "loss": 0.0,
+      "step": 53812
+    },
+    {
+      "epoch": 50.2,
+      "learning_rate": 4.372481343283582e-05,
+      "loss": 0.0,
+      "step": 53816
+    },
+    {
+      "epoch": 50.21,
+      "learning_rate": 4.372434701492538e-05,
+      "loss": 0.0003,
+      "step": 53820
+    },
+    {
+      "epoch": 50.21,
+      "learning_rate": 4.3723880597014925e-05,
+      "loss": 0.0001,
+      "step": 53824
+    },
+    {
+      "epoch": 50.21,
+      "learning_rate": 4.372341417910448e-05,
+      "loss": 0.0001,
+      "step": 53828
+    },
+    {
+      "epoch": 50.22,
+      "learning_rate": 4.3722947761194035e-05,
+      "loss": 0.0,
+      "step": 53832
+    },
+    {
+      "epoch": 50.22,
+      "learning_rate": 4.372248134328358e-05,
+      "loss": 0.0003,
+      "step": 53836
+    },
+    {
+      "epoch": 50.22,
+      "learning_rate": 4.372201492537313e-05,
+      "loss": 0.0001,
+      "step": 53840
+    },
+    {
+      "epoch": 50.23,
+      "learning_rate": 4.3721548507462686e-05,
+      "loss": 0.0,
+      "step": 53844
+    },
+    {
+      "epoch": 50.23,
+      "learning_rate": 4.372108208955224e-05,
+      "loss": 0.0026,
+      "step": 53848
+    },
+    {
+      "epoch": 50.24,
+      "learning_rate": 4.3720615671641796e-05,
+      "loss": 0.0,
+      "step": 53852
+    },
+    {
+      "epoch": 50.24,
+      "learning_rate": 4.3720149253731344e-05,
+      "loss": 0.0017,
+      "step": 53856
+    },
+    {
+      "epoch": 50.24,
+      "learning_rate": 4.37196828358209e-05,
+      "loss": 0.0,
+      "step": 53860
+    },
+    {
+      "epoch": 50.25,
+      "learning_rate": 4.3719216417910454e-05,
+      "loss": 0.0,
+      "step": 53864
+    },
+    {
+      "epoch": 50.25,
+      "learning_rate": 4.371875e-05,
+      "loss": 0.0001,
+      "step": 53868
+    },
+    {
+      "epoch": 50.25,
+      "learning_rate": 4.371828358208955e-05,
+      "loss": 0.0,
+      "step": 53872
+    },
+    {
+      "epoch": 50.26,
+      "learning_rate": 4.371781716417911e-05,
+      "loss": 0.0003,
+      "step": 53876
+    },
+    {
+      "epoch": 50.26,
+      "learning_rate": 4.371735074626866e-05,
+      "loss": 0.0,
+      "step": 53880
+    },
+    {
+      "epoch": 50.26,
+      "learning_rate": 4.371688432835821e-05,
+      "loss": 0.0001,
+      "step": 53884
+    },
+    {
+      "epoch": 50.27,
+      "learning_rate": 4.371641791044776e-05,
+      "loss": 0.0,
+      "step": 53888
+    },
+    {
+      "epoch": 50.27,
+      "learning_rate": 4.371595149253732e-05,
+      "loss": 0.0,
+      "step": 53892
+    },
+    {
+      "epoch": 50.28,
+      "learning_rate": 4.3715485074626866e-05,
+      "loss": 0.0001,
+      "step": 53896
+    },
+    {
+      "epoch": 50.28,
+      "learning_rate": 4.371501865671642e-05,
+      "loss": 0.0001,
+      "step": 53900
+    },
+    {
+      "epoch": 50.28,
+      "learning_rate": 4.371455223880597e-05,
+      "loss": 0.0001,
+      "step": 53904
+    },
+    {
+      "epoch": 50.29,
+      "learning_rate": 4.3714085820895524e-05,
+      "loss": 0.0001,
+      "step": 53908
+    },
+    {
+      "epoch": 50.29,
+      "learning_rate": 4.371361940298508e-05,
+      "loss": 0.0,
+      "step": 53912
+    },
+    {
+      "epoch": 50.29,
+      "learning_rate": 4.371315298507463e-05,
+      "loss": 0.0006,
+      "step": 53916
+    },
+    {
+      "epoch": 50.3,
+      "learning_rate": 4.371268656716418e-05,
+      "loss": 0.0,
+      "step": 53920
+    },
+    {
+      "epoch": 50.3,
+      "learning_rate": 4.371222014925374e-05,
+      "loss": 0.0003,
+      "step": 53924
+    },
+    {
+      "epoch": 50.31,
+      "learning_rate": 4.3711753731343285e-05,
+      "loss": 0.0,
+      "step": 53928
+    },
+    {
+      "epoch": 50.31,
+      "learning_rate": 4.371128731343283e-05,
+      "loss": 0.0,
+      "step": 53932
+    },
+    {
+      "epoch": 50.31,
+      "learning_rate": 4.3710820895522395e-05,
+      "loss": 0.0071,
+      "step": 53936
+    },
+    {
+      "epoch": 50.32,
+      "learning_rate": 4.371035447761194e-05,
+      "loss": 0.0003,
+      "step": 53940
+    },
+    {
+      "epoch": 50.32,
+      "learning_rate": 4.370988805970149e-05,
+      "loss": 0.0,
+      "step": 53944
+    },
+    {
+      "epoch": 50.32,
+      "learning_rate": 4.3709421641791046e-05,
+      "loss": 0.0,
+      "step": 53948
+    },
+    {
+      "epoch": 50.33,
+      "learning_rate": 4.37089552238806e-05,
+      "loss": 0.0017,
+      "step": 53952
+    },
+    {
+      "epoch": 50.33,
+      "learning_rate": 4.370848880597015e-05,
+      "loss": 0.0029,
+      "step": 53956
+    },
+    {
+      "epoch": 50.34,
+      "learning_rate": 4.3708022388059704e-05,
+      "loss": 0.0,
+      "step": 53960
+    },
+    {
+      "epoch": 50.34,
+      "learning_rate": 4.370755597014925e-05,
+      "loss": 0.0,
+      "step": 53964
+    },
+    {
+      "epoch": 50.34,
+      "learning_rate": 4.370708955223881e-05,
+      "loss": 0.0,
+      "step": 53968
+    },
+    {
+      "epoch": 50.35,
+      "learning_rate": 4.370662313432836e-05,
+      "loss": 0.0,
+      "step": 53972
+    },
+    {
+      "epoch": 50.35,
+      "learning_rate": 4.370615671641791e-05,
+      "loss": 0.0,
+      "step": 53976
+    },
+    {
+      "epoch": 50.35,
+      "learning_rate": 4.3705690298507465e-05,
+      "loss": 0.0004,
+      "step": 53980
+    },
+    {
+      "epoch": 50.36,
+      "learning_rate": 4.370522388059702e-05,
+      "loss": 0.0,
+      "step": 53984
+    },
+    {
+      "epoch": 50.36,
+      "learning_rate": 4.370475746268657e-05,
+      "loss": 0.001,
+      "step": 53988
+    },
+    {
+      "epoch": 50.37,
+      "learning_rate": 4.3704291044776116e-05,
+      "loss": 0.0,
+      "step": 53992
+    },
+    {
+      "epoch": 50.37,
+      "learning_rate": 4.370382462686568e-05,
+      "loss": 0.0,
+      "step": 53996
+    },
+    {
+      "epoch": 50.37,
+      "learning_rate": 4.3703358208955226e-05,
+      "loss": 0.0,
+      "step": 54000
+    },
+    {
+      "epoch": 50.37,
+      "eval_exact_match": 0.746615087040619,
+      "eval_exec": 0.7669245647969052,
+      "eval_loss": 0.46396273374557495,
+      "eval_runtime": 1103.7037,
+      "eval_samples_per_second": 0.937,
+      "step": 54000
+    },
+    {
+      "epoch": 50.38,
+      "learning_rate": 4.3702891791044774e-05,
+      "loss": 0.0,
+      "step": 54004
+    },
+    {
+      "epoch": 50.38,
+      "learning_rate": 4.370242537313433e-05,
+      "loss": 0.0,
+      "step": 54008
+    },
+    {
+      "epoch": 50.38,
+      "learning_rate": 4.3701958955223884e-05,
+      "loss": 0.0002,
+      "step": 54012
+    },
+    {
+      "epoch": 50.39,
+      "learning_rate": 4.370149253731344e-05,
+      "loss": 0.0,
+      "step": 54016
+    },
+    {
+      "epoch": 50.39,
+      "learning_rate": 4.370102611940299e-05,
+      "loss": 0.0011,
+      "step": 54020
+    },
+    {
+      "epoch": 50.4,
+      "learning_rate": 4.3700559701492535e-05,
+      "loss": 0.0,
+      "step": 54024
+    },
+    {
+      "epoch": 50.4,
+      "learning_rate": 4.37000932835821e-05,
+      "loss": 0.0,
+      "step": 54028
+    },
+    {
+      "epoch": 50.4,
+      "learning_rate": 4.3699626865671645e-05,
+      "loss": 0.0,
+      "step": 54032
+    },
+    {
+      "epoch": 50.41,
+      "learning_rate": 4.369916044776119e-05,
+      "loss": 0.0,
+      "step": 54036
+    },
+    {
+      "epoch": 50.41,
+      "learning_rate": 4.369869402985075e-05,
+      "loss": 0.0,
+      "step": 54040
+    },
+    {
+      "epoch": 50.41,
+      "learning_rate": 4.36982276119403e-05,
+      "loss": 0.0001,
+      "step": 54044
+    },
+    {
+      "epoch": 50.42,
+      "learning_rate": 4.369776119402985e-05,
+      "loss": 0.0,
+      "step": 54048
+    },
+    {
+      "epoch": 50.42,
+      "learning_rate": 4.3697294776119406e-05,
+      "loss": 0.0,
+      "step": 54052
+    },
+    {
+      "epoch": 50.43,
+      "learning_rate": 4.369682835820896e-05,
+      "loss": 0.0,
+      "step": 54056
+    },
+    {
+      "epoch": 50.43,
+      "learning_rate": 4.369636194029851e-05,
+      "loss": 0.0004,
+      "step": 54060
+    },
+    {
+      "epoch": 50.43,
+      "learning_rate": 4.3695895522388064e-05,
+      "loss": 0.0,
+      "step": 54064
+    },
+    {
+      "epoch": 50.44,
+      "learning_rate": 4.369542910447761e-05,
+      "loss": 0.0,
+      "step": 54068
+    },
+    {
+      "epoch": 50.44,
+      "learning_rate": 4.369496268656717e-05,
+      "loss": 0.0004,
+      "step": 54072
+    },
+    {
+      "epoch": 50.44,
+      "learning_rate": 4.369449626865672e-05,
+      "loss": 0.0001,
+      "step": 54076
+    },
+    {
+      "epoch": 50.45,
+      "learning_rate": 4.369402985074627e-05,
+      "loss": 0.0,
+      "step": 54080
+    },
+    {
+      "epoch": 50.45,
+      "learning_rate": 4.369356343283582e-05,
+      "loss": 0.0,
+      "step": 54084
+    },
+    {
+      "epoch": 50.46,
+      "learning_rate": 4.369309701492538e-05,
+      "loss": 0.0,
+      "step": 54088
+    },
+    {
+      "epoch": 50.46,
+      "learning_rate": 4.369263059701493e-05,
+      "loss": 0.0,
+      "step": 54092
+    },
+    {
+      "epoch": 50.46,
+      "learning_rate": 4.3692164179104476e-05,
+      "loss": 0.0001,
+      "step": 54096
+    },
+    {
+      "epoch": 50.47,
+      "learning_rate": 4.369169776119403e-05,
+      "loss": 0.0,
+      "step": 54100
+    },
+    {
+      "epoch": 50.47,
+      "learning_rate": 4.3691231343283586e-05,
+      "loss": 0.0,
+      "step": 54104
+    },
+    {
+      "epoch": 50.47,
+      "learning_rate": 4.3690764925373134e-05,
+      "loss": 0.0,
+      "step": 54108
+    },
+    {
+      "epoch": 50.48,
+      "learning_rate": 4.369029850746269e-05,
+      "loss": 0.0003,
+      "step": 54112
+    },
+    {
+      "epoch": 50.48,
+      "learning_rate": 4.3689832089552244e-05,
+      "loss": 0.0,
+      "step": 54116
+    },
+    {
+      "epoch": 50.49,
+      "learning_rate": 4.368936567164179e-05,
+      "loss": 0.0,
+      "step": 54120
+    },
+    {
+      "epoch": 50.49,
+      "learning_rate": 4.368889925373135e-05,
+      "loss": 0.0009,
+      "step": 54124
+    },
+    {
+      "epoch": 50.49,
+      "learning_rate": 4.3688432835820895e-05,
+      "loss": 0.0,
+      "step": 54128
+    },
+    {
+      "epoch": 50.5,
+      "learning_rate": 4.368796641791045e-05,
+      "loss": 0.0001,
+      "step": 54132
+    },
+    {
+      "epoch": 50.5,
+      "learning_rate": 4.3687500000000005e-05,
+      "loss": 0.0,
+      "step": 54136
+    },
+    {
+      "epoch": 50.5,
+      "learning_rate": 4.368703358208955e-05,
+      "loss": 0.0003,
+      "step": 54140
+    },
+    {
+      "epoch": 50.51,
+      "learning_rate": 4.36865671641791e-05,
+      "loss": 0.0001,
+      "step": 54144
+    },
+    {
+      "epoch": 50.51,
+      "learning_rate": 4.368610074626866e-05,
+      "loss": 0.0003,
+      "step": 54148
+    },
+    {
+      "epoch": 50.51,
+      "learning_rate": 4.368563432835821e-05,
+      "loss": 0.0005,
+      "step": 54152
+    },
+    {
+      "epoch": 50.52,
+      "learning_rate": 4.368516791044776e-05,
+      "loss": 0.0,
+      "step": 54156
+    },
+    {
+      "epoch": 50.52,
+      "learning_rate": 4.3684701492537314e-05,
+      "loss": 0.0,
+      "step": 54160
+    },
+    {
+      "epoch": 50.53,
+      "learning_rate": 4.368423507462687e-05,
+      "loss": 0.0,
+      "step": 54164
+    },
+    {
+      "epoch": 50.53,
+      "learning_rate": 4.368376865671642e-05,
+      "loss": 0.0004,
+      "step": 54168
+    },
+    {
+      "epoch": 50.53,
+      "learning_rate": 4.368330223880597e-05,
+      "loss": 0.0002,
+      "step": 54172
+    },
+    {
+      "epoch": 50.54,
+      "learning_rate": 4.368283582089553e-05,
+      "loss": 0.0,
+      "step": 54176
+    },
+    {
+      "epoch": 50.54,
+      "learning_rate": 4.368236940298508e-05,
+      "loss": 0.0003,
+      "step": 54180
+    },
+    {
+      "epoch": 50.54,
+      "learning_rate": 4.368190298507463e-05,
+      "loss": 0.0,
+      "step": 54184
+    },
+    {
+      "epoch": 50.55,
+      "learning_rate": 4.368143656716418e-05,
+      "loss": 0.0,
+      "step": 54188
+    },
+    {
+      "epoch": 50.55,
+      "learning_rate": 4.368097014925373e-05,
+      "loss": 0.0,
+      "step": 54192
+    },
+    {
+      "epoch": 50.56,
+      "learning_rate": 4.368050373134329e-05,
+      "loss": 0.0,
+      "step": 54196
+    },
+    {
+      "epoch": 50.56,
+      "learning_rate": 4.3680037313432836e-05,
+      "loss": 0.0,
+      "step": 54200
+    },
+    {
+      "epoch": 50.56,
+      "learning_rate": 4.367957089552239e-05,
+      "loss": 0.0141,
+      "step": 54204
+    },
+    {
+      "epoch": 50.57,
+      "learning_rate": 4.3679104477611946e-05,
+      "loss": 0.0001,
+      "step": 54208
+    },
+    {
+      "epoch": 50.57,
+      "learning_rate": 4.3678638059701494e-05,
+      "loss": 0.0006,
+      "step": 54212
+    },
+    {
+      "epoch": 50.57,
+      "learning_rate": 4.367817164179105e-05,
+      "loss": 0.0,
+      "step": 54216
+    },
+    {
+      "epoch": 50.58,
+      "learning_rate": 4.36777052238806e-05,
+      "loss": 0.0,
+      "step": 54220
+    },
+    {
+      "epoch": 50.58,
+      "learning_rate": 4.367723880597015e-05,
+      "loss": 0.0,
+      "step": 54224
+    },
+    {
+      "epoch": 50.59,
+      "learning_rate": 4.367677238805971e-05,
+      "loss": 0.0,
+      "step": 54228
+    },
+    {
+      "epoch": 50.59,
+      "learning_rate": 4.3676305970149255e-05,
+      "loss": 0.0,
+      "step": 54232
+    },
+    {
+      "epoch": 50.59,
+      "learning_rate": 4.367583955223881e-05,
+      "loss": 0.0,
+      "step": 54236
+    },
+    {
+      "epoch": 50.6,
+      "learning_rate": 4.3675373134328365e-05,
+      "loss": 0.0006,
+      "step": 54240
+    },
+    {
+      "epoch": 50.6,
+      "learning_rate": 4.367490671641791e-05,
+      "loss": 0.0,
+      "step": 54244
+    },
+    {
+      "epoch": 50.6,
+      "learning_rate": 4.367444029850746e-05,
+      "loss": 0.0001,
+      "step": 54248
+    },
+    {
+      "epoch": 50.61,
+      "learning_rate": 4.3673973880597016e-05,
+      "loss": 0.0,
+      "step": 54252
+    },
+    {
+      "epoch": 50.61,
+      "learning_rate": 4.367350746268657e-05,
+      "loss": 0.0002,
+      "step": 54256
+    },
+    {
+      "epoch": 50.62,
+      "learning_rate": 4.367304104477612e-05,
+      "loss": 0.0,
+      "step": 54260
+    },
+    {
+      "epoch": 50.62,
+      "learning_rate": 4.3672574626865674e-05,
+      "loss": 0.0,
+      "step": 54264
+    },
+    {
+      "epoch": 50.62,
+      "learning_rate": 4.367210820895523e-05,
+      "loss": 0.0,
+      "step": 54268
+    },
+    {
+      "epoch": 50.63,
+      "learning_rate": 4.367164179104478e-05,
+      "loss": 0.0001,
+      "step": 54272
+    },
+    {
+      "epoch": 50.63,
+      "learning_rate": 4.367117537313433e-05,
+      "loss": 0.0057,
+      "step": 54276
+    },
+    {
+      "epoch": 50.63,
+      "learning_rate": 4.367070895522388e-05,
+      "loss": 0.0004,
+      "step": 54280
+    },
+    {
+      "epoch": 50.64,
+      "learning_rate": 4.3670242537313435e-05,
+      "loss": 0.0,
+      "step": 54284
+    },
+    {
+      "epoch": 50.64,
+      "learning_rate": 4.366977611940299e-05,
+      "loss": 0.0,
+      "step": 54288
+    },
+    {
+      "epoch": 50.65,
+      "learning_rate": 4.366930970149254e-05,
+      "loss": 0.0006,
+      "step": 54292
+    },
+    {
+      "epoch": 50.65,
+      "learning_rate": 4.366884328358209e-05,
+      "loss": 0.0,
+      "step": 54296
+    },
+    {
+      "epoch": 50.65,
+      "learning_rate": 4.366837686567165e-05,
+      "loss": 0.0,
+      "step": 54300
+    },
+    {
+      "epoch": 50.66,
+      "learning_rate": 4.3667910447761196e-05,
+      "loss": 0.0,
+      "step": 54304
+    },
+    {
+      "epoch": 50.66,
+      "learning_rate": 4.3667444029850744e-05,
+      "loss": 0.0001,
+      "step": 54308
+    },
+    {
+      "epoch": 50.66,
+      "learning_rate": 4.36669776119403e-05,
+      "loss": 0.0012,
+      "step": 54312
+    },
+    {
+      "epoch": 50.67,
+      "learning_rate": 4.3666511194029854e-05,
+      "loss": 0.0,
+      "step": 54316
+    },
+    {
+      "epoch": 50.67,
+      "learning_rate": 4.36660447761194e-05,
+      "loss": 0.0,
+      "step": 54320
+    },
+    {
+      "epoch": 50.68,
+      "learning_rate": 4.366557835820896e-05,
+      "loss": 0.0009,
+      "step": 54324
+    },
+    {
+      "epoch": 50.68,
+      "learning_rate": 4.366511194029851e-05,
+      "loss": 0.0001,
+      "step": 54328
+    },
+    {
+      "epoch": 50.68,
+      "learning_rate": 4.366464552238806e-05,
+      "loss": 0.0,
+      "step": 54332
+    },
+    {
+      "epoch": 50.69,
+      "learning_rate": 4.3664179104477615e-05,
+      "loss": 0.0,
+      "step": 54336
+    },
+    {
+      "epoch": 50.69,
+      "learning_rate": 4.366371268656716e-05,
+      "loss": 0.0,
+      "step": 54340
+    },
+    {
+      "epoch": 50.69,
+      "learning_rate": 4.3663246268656724e-05,
+      "loss": 0.0,
+      "step": 54344
+    },
+    {
+      "epoch": 50.7,
+      "learning_rate": 4.366277985074627e-05,
+      "loss": 0.0,
+      "step": 54348
+    },
+    {
+      "epoch": 50.7,
+      "learning_rate": 4.366231343283582e-05,
+      "loss": 0.0002,
+      "step": 54352
+    },
+    {
+      "epoch": 50.71,
+      "learning_rate": 4.3661847014925376e-05,
+      "loss": 0.0,
+      "step": 54356
+    },
+    {
+      "epoch": 50.71,
+      "learning_rate": 4.366138059701493e-05,
+      "loss": 0.0,
+      "step": 54360
+    },
+    {
+      "epoch": 50.71,
+      "learning_rate": 4.366091417910448e-05,
+      "loss": 0.0,
+      "step": 54364
+    },
+    {
+      "epoch": 50.72,
+      "learning_rate": 4.3660447761194034e-05,
+      "loss": 0.0,
+      "step": 54368
+    },
+    {
+      "epoch": 50.72,
+      "learning_rate": 4.365998134328358e-05,
+      "loss": 0.0,
+      "step": 54372
+    },
+    {
+      "epoch": 50.72,
+      "learning_rate": 4.365951492537314e-05,
+      "loss": 0.0,
+      "step": 54376
+    },
+    {
+      "epoch": 50.73,
+      "learning_rate": 4.365904850746269e-05,
+      "loss": 0.0001,
+      "step": 54380
+    },
+    {
+      "epoch": 50.73,
+      "learning_rate": 4.365858208955224e-05,
+      "loss": 0.0034,
+      "step": 54384
+    },
+    {
+      "epoch": 50.73,
+      "learning_rate": 4.3658115671641795e-05,
+      "loss": 0.0045,
+      "step": 54388
+    },
+    {
+      "epoch": 50.74,
+      "learning_rate": 4.365764925373135e-05,
+      "loss": 0.0,
+      "step": 54392
+    },
+    {
+      "epoch": 50.74,
+      "learning_rate": 4.36571828358209e-05,
+      "loss": 0.0,
+      "step": 54396
+    },
+    {
+      "epoch": 50.75,
+      "learning_rate": 4.3656716417910446e-05,
+      "loss": 0.0016,
+      "step": 54400
+    },
+    {
+      "epoch": 50.75,
+      "learning_rate": 4.365625000000001e-05,
+      "loss": 0.0011,
+      "step": 54404
+    },
+    {
+      "epoch": 50.75,
+      "learning_rate": 4.3655783582089556e-05,
+      "loss": 0.0026,
+      "step": 54408
+    },
+    {
+      "epoch": 50.76,
+      "learning_rate": 4.3655317164179104e-05,
+      "loss": 0.0,
+      "step": 54412
+    },
+    {
+      "epoch": 50.76,
+      "learning_rate": 4.365485074626866e-05,
+      "loss": 0.0,
+      "step": 54416
+    },
+    {
+      "epoch": 50.76,
+      "learning_rate": 4.3654384328358214e-05,
+      "loss": 0.0004,
+      "step": 54420
+    },
+    {
+      "epoch": 50.77,
+      "learning_rate": 4.365391791044776e-05,
+      "loss": 0.0006,
+      "step": 54424
+    },
+    {
+      "epoch": 50.77,
+      "learning_rate": 4.3653451492537317e-05,
+      "loss": 0.0,
+      "step": 54428
+    },
+    {
+      "epoch": 50.78,
+      "learning_rate": 4.3652985074626865e-05,
+      "loss": 0.0001,
+      "step": 54432
+    },
+    {
+      "epoch": 50.78,
+      "learning_rate": 4.365251865671642e-05,
+      "loss": 0.0,
+      "step": 54436
+    },
+    {
+      "epoch": 50.78,
+      "learning_rate": 4.3652052238805975e-05,
+      "loss": 0.0,
+      "step": 54440
+    },
+    {
+      "epoch": 50.79,
+      "learning_rate": 4.365158582089552e-05,
+      "loss": 0.0,
+      "step": 54444
+    },
+    {
+      "epoch": 50.79,
+      "learning_rate": 4.365111940298508e-05,
+      "loss": 0.0013,
+      "step": 54448
+    },
+    {
+      "epoch": 50.79,
+      "learning_rate": 4.365065298507463e-05,
+      "loss": 0.0,
+      "step": 54452
+    },
+    {
+      "epoch": 50.8,
+      "learning_rate": 4.365018656716418e-05,
+      "loss": 0.0,
+      "step": 54456
+    },
+    {
+      "epoch": 50.8,
+      "learning_rate": 4.364972014925373e-05,
+      "loss": 0.0,
+      "step": 54460
+    },
+    {
+      "epoch": 50.81,
+      "learning_rate": 4.3649253731343284e-05,
+      "loss": 0.0,
+      "step": 54464
+    },
+    {
+      "epoch": 50.81,
+      "learning_rate": 4.364878731343284e-05,
+      "loss": 0.0014,
+      "step": 54468
+    },
+    {
+      "epoch": 50.81,
+      "learning_rate": 4.364832089552239e-05,
+      "loss": 0.0028,
+      "step": 54472
+    },
+    {
+      "epoch": 50.82,
+      "learning_rate": 4.364785447761194e-05,
+      "loss": 0.0,
+      "step": 54476
+    },
+    {
+      "epoch": 50.82,
+      "learning_rate": 4.3647388059701496e-05,
+      "loss": 0.0,
+      "step": 54480
+    },
+    {
+      "epoch": 50.82,
+      "learning_rate": 4.3646921641791045e-05,
+      "loss": 0.0002,
+      "step": 54484
+    },
+    {
+      "epoch": 50.83,
+      "learning_rate": 4.36464552238806e-05,
+      "loss": 0.0,
+      "step": 54488
+    },
+    {
+      "epoch": 50.83,
+      "learning_rate": 4.364598880597015e-05,
+      "loss": 0.0,
+      "step": 54492
+    },
+    {
+      "epoch": 50.84,
+      "learning_rate": 4.36455223880597e-05,
+      "loss": 0.0017,
+      "step": 54496
+    },
+    {
+      "epoch": 50.84,
+      "learning_rate": 4.364505597014926e-05,
+      "loss": 0.0037,
+      "step": 54500
+    },
+    {
+      "epoch": 50.84,
+      "eval_exact_match": 0.7446808510638298,
+      "eval_exec": 0.7756286266924565,
+      "eval_loss": 0.45397448539733887,
+      "eval_runtime": 1155.302,
+      "eval_samples_per_second": 0.895,
+      "step": 54500
+    },
+    {
+      "epoch": 50.84,
+      "learning_rate": 4.3644589552238806e-05,
+      "loss": 0.0,
+      "step": 54504
+    },
+    {
+      "epoch": 50.85,
+      "learning_rate": 4.364412313432836e-05,
+      "loss": 0.0,
+      "step": 54508
+    },
+    {
+      "epoch": 50.85,
+      "learning_rate": 4.3643656716417915e-05,
+      "loss": 0.0,
+      "step": 54512
+    },
+    {
+      "epoch": 50.85,
+      "learning_rate": 4.3643190298507464e-05,
+      "loss": 0.0,
+      "step": 54516
+    },
+    {
+      "epoch": 50.86,
+      "learning_rate": 4.364272388059702e-05,
+      "loss": 0.0,
+      "step": 54520
+    },
+    {
+      "epoch": 50.86,
+      "learning_rate": 4.3642257462686567e-05,
+      "loss": 0.0,
+      "step": 54524
+    },
+    {
+      "epoch": 50.87,
+      "learning_rate": 4.364179104477612e-05,
+      "loss": 0.0001,
+      "step": 54528
+    },
+    {
+      "epoch": 50.87,
+      "learning_rate": 4.3641324626865676e-05,
+      "loss": 0.0,
+      "step": 54532
+    },
+    {
+      "epoch": 50.87,
+      "learning_rate": 4.3640858208955225e-05,
+      "loss": 0.0,
+      "step": 54536
+    },
+    {
+      "epoch": 50.88,
+      "learning_rate": 4.364039179104478e-05,
+      "loss": 0.0023,
+      "step": 54540
+    },
+    {
+      "epoch": 50.88,
+      "learning_rate": 4.3639925373134334e-05,
+      "loss": 0.0,
+      "step": 54544
+    },
+    {
+      "epoch": 50.88,
+      "learning_rate": 4.363945895522388e-05,
+      "loss": 0.0,
+      "step": 54548
+    },
+    {
+      "epoch": 50.89,
+      "learning_rate": 4.363899253731343e-05,
+      "loss": 0.0,
+      "step": 54552
+    },
+    {
+      "epoch": 50.89,
+      "learning_rate": 4.363852611940299e-05,
+      "loss": 0.0,
+      "step": 54556
+    },
+    {
+      "epoch": 50.9,
+      "learning_rate": 4.363805970149254e-05,
+      "loss": 0.0066,
+      "step": 54560
+    },
+    {
+      "epoch": 50.9,
+      "learning_rate": 4.363759328358209e-05,
+      "loss": 0.0,
+      "step": 54564
+    },
+    {
+      "epoch": 50.9,
+      "learning_rate": 4.3637126865671643e-05,
+      "loss": 0.0,
+      "step": 54568
+    },
+    {
+      "epoch": 50.91,
+      "learning_rate": 4.36366604477612e-05,
+      "loss": 0.0029,
+      "step": 54572
+    },
+    {
+      "epoch": 50.91,
+      "learning_rate": 4.3636194029850746e-05,
+      "loss": 0.0005,
+      "step": 54576
+    },
+    {
+      "epoch": 50.91,
+      "learning_rate": 4.36357276119403e-05,
+      "loss": 0.0,
+      "step": 54580
+    },
+    {
+      "epoch": 50.92,
+      "learning_rate": 4.363526119402985e-05,
+      "loss": 0.0,
+      "step": 54584
+    },
+    {
+      "epoch": 50.92,
+      "learning_rate": 4.3634794776119404e-05,
+      "loss": 0.0003,
+      "step": 54588
+    },
+    {
+      "epoch": 50.93,
+      "learning_rate": 4.363432835820896e-05,
+      "loss": 0.0001,
+      "step": 54592
+    },
+    {
+      "epoch": 50.93,
+      "learning_rate": 4.363386194029851e-05,
+      "loss": 0.0,
+      "step": 54596
+    },
+    {
+      "epoch": 50.93,
+      "learning_rate": 4.363339552238806e-05,
+      "loss": 0.0,
+      "step": 54600
+    },
+    {
+      "epoch": 50.94,
+      "learning_rate": 4.363292910447762e-05,
+      "loss": 0.0,
+      "step": 54604
+    },
+    {
+      "epoch": 50.94,
+      "learning_rate": 4.3632462686567165e-05,
+      "loss": 0.0032,
+      "step": 54608
+    },
+    {
+      "epoch": 50.94,
+      "learning_rate": 4.3631996268656714e-05,
+      "loss": 0.0,
+      "step": 54612
+    },
+    {
+      "epoch": 50.95,
+      "learning_rate": 4.3631529850746275e-05,
+      "loss": 0.0,
+      "step": 54616
+    },
+    {
+      "epoch": 50.95,
+      "learning_rate": 4.363106343283582e-05,
+      "loss": 0.0,
+      "step": 54620
+    },
+    {
+      "epoch": 50.96,
+      "learning_rate": 4.363059701492537e-05,
+      "loss": 0.0057,
+      "step": 54624
+    },
+    {
+      "epoch": 50.96,
+      "learning_rate": 4.3630130597014926e-05,
+      "loss": 0.0001,
+      "step": 54628
+    },
+    {
+      "epoch": 50.96,
+      "learning_rate": 4.362966417910448e-05,
+      "loss": 0.0,
+      "step": 54632
+    },
+    {
+      "epoch": 50.97,
+      "learning_rate": 4.362919776119403e-05,
+      "loss": 0.0001,
+      "step": 54636
+    },
+    {
+      "epoch": 50.97,
+      "learning_rate": 4.3628731343283584e-05,
+      "loss": 0.0,
+      "step": 54640
+    },
+    {
+      "epoch": 50.97,
+      "learning_rate": 4.362826492537313e-05,
+      "loss": 0.0,
+      "step": 54644
+    },
+    {
+      "epoch": 50.98,
+      "learning_rate": 4.362779850746269e-05,
+      "loss": 0.0001,
+      "step": 54648
+    },
+    {
+      "epoch": 50.98,
+      "learning_rate": 4.362733208955224e-05,
+      "loss": 0.0001,
+      "step": 54652
+    },
+    {
+      "epoch": 50.98,
+      "learning_rate": 4.362686567164179e-05,
+      "loss": 0.0,
+      "step": 54656
+    },
+    {
+      "epoch": 50.99,
+      "learning_rate": 4.3626399253731345e-05,
+      "loss": 0.0003,
+      "step": 54660
+    },
+    {
+      "epoch": 50.99,
+      "learning_rate": 4.36259328358209e-05,
+      "loss": 0.0,
+      "step": 54664
+    },
+    {
+      "epoch": 51.0,
+      "learning_rate": 4.362546641791045e-05,
+      "loss": 0.0,
+      "step": 54668
+    },
+    {
+      "epoch": 51.0,
+      "learning_rate": 4.3625e-05,
+      "loss": 0.0013,
+      "step": 54672
+    },
+    {
+      "epoch": 51.0,
+      "learning_rate": 4.362453358208956e-05,
+      "loss": 0.0006,
+      "step": 54676
+    },
+    {
+      "epoch": 51.01,
+      "learning_rate": 4.3624067164179106e-05,
+      "loss": 0.0006,
+      "step": 54680
+    },
+    {
+      "epoch": 51.01,
+      "learning_rate": 4.362360074626866e-05,
+      "loss": 0.0001,
+      "step": 54684
+    },
+    {
+      "epoch": 51.01,
+      "learning_rate": 4.362313432835821e-05,
+      "loss": 0.0001,
+      "step": 54688
+    },
+    {
+      "epoch": 51.02,
+      "learning_rate": 4.3622667910447764e-05,
+      "loss": 0.0001,
+      "step": 54692
+    },
+    {
+      "epoch": 51.02,
+      "learning_rate": 4.362220149253732e-05,
+      "loss": 0.0001,
+      "step": 54696
+    },
+    {
+      "epoch": 51.03,
+      "learning_rate": 4.362173507462687e-05,
+      "loss": 0.0,
+      "step": 54700
+    },
+    {
+      "epoch": 51.03,
+      "learning_rate": 4.3621268656716415e-05,
+      "loss": 0.0,
+      "step": 54704
+    },
+    {
+      "epoch": 51.03,
+      "learning_rate": 4.362080223880598e-05,
+      "loss": 0.0,
+      "step": 54708
+    },
+    {
+      "epoch": 51.04,
+      "learning_rate": 4.3620335820895525e-05,
+      "loss": 0.0001,
+      "step": 54712
+    },
+    {
+      "epoch": 51.04,
+      "learning_rate": 4.3619869402985073e-05,
+      "loss": 0.0023,
+      "step": 54716
+    },
+    {
+      "epoch": 51.04,
+      "learning_rate": 4.361940298507463e-05,
+      "loss": 0.0,
+      "step": 54720
+    },
+    {
+      "epoch": 51.05,
+      "learning_rate": 4.361893656716418e-05,
+      "loss": 0.0,
+      "step": 54724
+    },
+    {
+      "epoch": 51.05,
+      "learning_rate": 4.361847014925373e-05,
+      "loss": 0.0,
+      "step": 54728
+    },
+    {
+      "epoch": 51.06,
+      "learning_rate": 4.3618003731343286e-05,
+      "loss": 0.0,
+      "step": 54732
+    },
+    {
+      "epoch": 51.06,
+      "learning_rate": 4.361753731343284e-05,
+      "loss": 0.0,
+      "step": 54736
+    },
+    {
+      "epoch": 51.06,
+      "learning_rate": 4.361707089552239e-05,
+      "loss": 0.0002,
+      "step": 54740
+    },
+    {
+      "epoch": 51.07,
+      "learning_rate": 4.3616604477611944e-05,
+      "loss": 0.0007,
+      "step": 54744
+    },
+    {
+      "epoch": 51.07,
+      "learning_rate": 4.361613805970149e-05,
+      "loss": 0.0002,
+      "step": 54748
+    },
+    {
+      "epoch": 51.07,
+      "learning_rate": 4.361567164179105e-05,
+      "loss": 0.0005,
+      "step": 54752
+    },
+    {
+      "epoch": 51.08,
+      "learning_rate": 4.36152052238806e-05,
+      "loss": 0.0,
+      "step": 54756
+    },
+    {
+      "epoch": 51.08,
+      "learning_rate": 4.361473880597015e-05,
+      "loss": 0.0,
+      "step": 54760
+    },
+    {
+      "epoch": 51.09,
+      "learning_rate": 4.36142723880597e-05,
+      "loss": 0.0,
+      "step": 54764
+    },
+    {
+      "epoch": 51.09,
+      "learning_rate": 4.361380597014926e-05,
+      "loss": 0.0,
+      "step": 54768
+    },
+    {
+      "epoch": 51.09,
+      "learning_rate": 4.361333955223881e-05,
+      "loss": 0.0092,
+      "step": 54772
+    },
+    {
+      "epoch": 51.1,
+      "learning_rate": 4.3612873134328356e-05,
+      "loss": 0.0002,
+      "step": 54776
+    },
+    {
+      "epoch": 51.1,
+      "learning_rate": 4.361240671641791e-05,
+      "loss": 0.0003,
+      "step": 54780
+    },
+    {
+      "epoch": 51.1,
+      "learning_rate": 4.3611940298507466e-05,
+      "loss": 0.0001,
+      "step": 54784
+    },
+    {
+      "epoch": 51.11,
+      "learning_rate": 4.3611473880597014e-05,
+      "loss": 0.0,
+      "step": 54788
+    },
+    {
+      "epoch": 51.11,
+      "learning_rate": 4.361100746268657e-05,
+      "loss": 0.0,
+      "step": 54792
+    },
+    {
+      "epoch": 51.12,
+      "learning_rate": 4.3610541044776124e-05,
+      "loss": 0.0007,
+      "step": 54796
+    },
+    {
+      "epoch": 51.12,
+      "learning_rate": 4.361007462686567e-05,
+      "loss": 0.0,
+      "step": 54800
+    },
+    {
+      "epoch": 51.12,
+      "learning_rate": 4.360960820895523e-05,
+      "loss": 0.0001,
+      "step": 54804
+    },
+    {
+      "epoch": 51.13,
+      "learning_rate": 4.3609141791044775e-05,
+      "loss": 0.0,
+      "step": 54808
+    },
+    {
+      "epoch": 51.13,
+      "learning_rate": 4.360867537313433e-05,
+      "loss": 0.0,
+      "step": 54812
+    },
+    {
+      "epoch": 51.13,
+      "learning_rate": 4.3608208955223885e-05,
+      "loss": 0.0,
+      "step": 54816
+    },
+    {
+      "epoch": 51.14,
+      "learning_rate": 4.360774253731343e-05,
+      "loss": 0.0,
+      "step": 54820
+    },
+    {
+      "epoch": 51.14,
+      "learning_rate": 4.360727611940298e-05,
+      "loss": 0.0,
+      "step": 54824
+    },
+    {
+      "epoch": 51.15,
+      "learning_rate": 4.360680970149254e-05,
+      "loss": 0.0,
+      "step": 54828
+    },
+    {
+      "epoch": 51.15,
+      "learning_rate": 4.360634328358209e-05,
+      "loss": 0.0,
+      "step": 54832
+    },
+    {
+      "epoch": 51.15,
+      "learning_rate": 4.3605876865671646e-05,
+      "loss": 0.0012,
+      "step": 54836
+    },
+    {
+      "epoch": 51.16,
+      "learning_rate": 4.3605410447761194e-05,
+      "loss": 0.0001,
+      "step": 54840
+    },
+    {
+      "epoch": 51.16,
+      "learning_rate": 4.360494402985075e-05,
+      "loss": 0.0,
+      "step": 54844
+    },
+    {
+      "epoch": 51.16,
+      "learning_rate": 4.3604477611940304e-05,
+      "loss": 0.0,
+      "step": 54848
+    },
+    {
+      "epoch": 51.17,
+      "learning_rate": 4.360401119402985e-05,
+      "loss": 0.0,
+      "step": 54852
+    },
+    {
+      "epoch": 51.17,
+      "learning_rate": 4.360354477611941e-05,
+      "loss": 0.0,
+      "step": 54856
+    },
+    {
+      "epoch": 51.18,
+      "learning_rate": 4.360307835820896e-05,
+      "loss": 0.0,
+      "step": 54860
+    },
+    {
+      "epoch": 51.18,
+      "learning_rate": 4.360261194029851e-05,
+      "loss": 0.0003,
+      "step": 54864
+    },
+    {
+      "epoch": 51.18,
+      "learning_rate": 4.360214552238806e-05,
+      "loss": 0.0001,
+      "step": 54868
+    },
+    {
+      "epoch": 51.19,
+      "learning_rate": 4.360167910447761e-05,
+      "loss": 0.0,
+      "step": 54872
+    },
+    {
+      "epoch": 51.19,
+      "learning_rate": 4.360121268656717e-05,
+      "loss": 0.0001,
+      "step": 54876
+    },
+    {
+      "epoch": 51.19,
+      "learning_rate": 4.3600746268656716e-05,
+      "loss": 0.0009,
+      "step": 54880
+    },
+    {
+      "epoch": 51.2,
+      "learning_rate": 4.360027985074627e-05,
+      "loss": 0.0003,
+      "step": 54884
+    },
+    {
+      "epoch": 51.2,
+      "learning_rate": 4.3599813432835826e-05,
+      "loss": 0.0,
+      "step": 54888
+    },
+    {
+      "epoch": 51.21,
+      "learning_rate": 4.3599347014925374e-05,
+      "loss": 0.0001,
+      "step": 54892
+    },
+    {
+      "epoch": 51.21,
+      "learning_rate": 4.359888059701493e-05,
+      "loss": 0.0,
+      "step": 54896
+    },
+    {
+      "epoch": 51.21,
+      "learning_rate": 4.359841417910448e-05,
+      "loss": 0.0002,
+      "step": 54900
+    },
+    {
+      "epoch": 51.22,
+      "learning_rate": 4.359794776119403e-05,
+      "loss": 0.0,
+      "step": 54904
+    },
+    {
+      "epoch": 51.22,
+      "learning_rate": 4.359748134328359e-05,
+      "loss": 0.0,
+      "step": 54908
+    },
+    {
+      "epoch": 51.22,
+      "learning_rate": 4.3597014925373135e-05,
+      "loss": 0.0,
+      "step": 54912
+    },
+    {
+      "epoch": 51.23,
+      "learning_rate": 4.359654850746269e-05,
+      "loss": 0.0,
+      "step": 54916
+    },
+    {
+      "epoch": 51.23,
+      "learning_rate": 4.3596082089552245e-05,
+      "loss": 0.0,
+      "step": 54920
+    },
+    {
+      "epoch": 51.24,
+      "learning_rate": 4.359561567164179e-05,
+      "loss": 0.004,
+      "step": 54924
+    },
+    {
+      "epoch": 51.24,
+      "learning_rate": 4.359514925373134e-05,
+      "loss": 0.0,
+      "step": 54928
+    },
+    {
+      "epoch": 51.24,
+      "learning_rate": 4.3594682835820896e-05,
+      "loss": 0.0001,
+      "step": 54932
+    },
+    {
+      "epoch": 51.25,
+      "learning_rate": 4.359421641791045e-05,
+      "loss": 0.0001,
+      "step": 54936
+    },
+    {
+      "epoch": 51.25,
+      "learning_rate": 4.359375e-05,
+      "loss": 0.0,
+      "step": 54940
+    },
+    {
+      "epoch": 51.25,
+      "learning_rate": 4.3593283582089554e-05,
+      "loss": 0.0,
+      "step": 54944
+    },
+    {
+      "epoch": 51.26,
+      "learning_rate": 4.359281716417911e-05,
+      "loss": 0.0,
+      "step": 54948
+    },
+    {
+      "epoch": 51.26,
+      "learning_rate": 4.359235074626866e-05,
+      "loss": 0.0,
+      "step": 54952
+    },
+    {
+      "epoch": 51.26,
+      "learning_rate": 4.359188432835821e-05,
+      "loss": 0.0,
+      "step": 54956
+    },
+    {
+      "epoch": 51.27,
+      "learning_rate": 4.359141791044776e-05,
+      "loss": 0.0,
+      "step": 54960
+    },
+    {
+      "epoch": 51.27,
+      "learning_rate": 4.3590951492537315e-05,
+      "loss": 0.0,
+      "step": 54964
+    },
+    {
+      "epoch": 51.28,
+      "learning_rate": 4.359048507462687e-05,
+      "loss": 0.0001,
+      "step": 54968
+    },
+    {
+      "epoch": 51.28,
+      "learning_rate": 4.359001865671642e-05,
+      "loss": 0.0,
+      "step": 54972
+    },
+    {
+      "epoch": 51.28,
+      "learning_rate": 4.358955223880597e-05,
+      "loss": 0.0006,
+      "step": 54976
+    },
+    {
+      "epoch": 51.29,
+      "learning_rate": 4.358908582089553e-05,
+      "loss": 0.0003,
+      "step": 54980
+    },
+    {
+      "epoch": 51.29,
+      "learning_rate": 4.3588619402985076e-05,
+      "loss": 0.0,
+      "step": 54984
+    },
+    {
+      "epoch": 51.29,
+      "learning_rate": 4.3588152985074624e-05,
+      "loss": 0.0,
+      "step": 54988
+    },
+    {
+      "epoch": 51.3,
+      "learning_rate": 4.358768656716418e-05,
+      "loss": 0.0001,
+      "step": 54992
+    },
+    {
+      "epoch": 51.3,
+      "learning_rate": 4.3587220149253734e-05,
+      "loss": 0.0,
+      "step": 54996
+    },
+    {
+      "epoch": 51.31,
+      "learning_rate": 4.358675373134329e-05,
+      "loss": 0.0,
+      "step": 55000
+    },
+    {
+      "epoch": 51.31,
+      "eval_exact_match": 0.7379110251450677,
+      "eval_exec": 0.758220502901354,
+      "eval_loss": 0.4803534746170044,
+      "eval_runtime": 1126.6002,
+      "eval_samples_per_second": 0.918,
+      "step": 55000
+    },
+    {
+      "epoch": 51.31,
+      "learning_rate": 4.358628731343284e-05,
+      "loss": 0.0,
+      "step": 55004
+    },
+    {
+      "epoch": 51.31,
+      "learning_rate": 4.358582089552239e-05,
+      "loss": 0.0,
+      "step": 55008
+    },
+    {
+      "epoch": 51.32,
+      "learning_rate": 4.358535447761195e-05,
+      "loss": 0.0004,
+      "step": 55012
+    },
+    {
+      "epoch": 51.32,
+      "learning_rate": 4.3584888059701495e-05,
+      "loss": 0.001,
+      "step": 55016
+    },
+    {
+      "epoch": 51.32,
+      "learning_rate": 4.358442164179104e-05,
+      "loss": 0.0,
+      "step": 55020
+    },
+    {
+      "epoch": 51.33,
+      "learning_rate": 4.3583955223880605e-05,
+      "loss": 0.0,
+      "step": 55024
+    },
+    {
+      "epoch": 51.33,
+      "learning_rate": 4.358348880597015e-05,
+      "loss": 0.001,
+      "step": 55028
+    },
+    {
+      "epoch": 51.34,
+      "learning_rate": 4.35830223880597e-05,
+      "loss": 0.0003,
+      "step": 55032
+    },
+    {
+      "epoch": 51.34,
+      "learning_rate": 4.3582555970149256e-05,
+      "loss": 0.0001,
+      "step": 55036
+    },
+    {
+      "epoch": 51.34,
+      "learning_rate": 4.358208955223881e-05,
+      "loss": 0.0013,
+      "step": 55040
+    },
+    {
+      "epoch": 51.35,
+      "learning_rate": 4.358162313432836e-05,
+      "loss": 0.0,
+      "step": 55044
+    },
+    {
+      "epoch": 51.35,
+      "learning_rate": 4.3581156716417914e-05,
+      "loss": 0.0,
+      "step": 55048
+    },
+    {
+      "epoch": 51.35,
+      "learning_rate": 4.358069029850746e-05,
+      "loss": 0.0,
+      "step": 55052
+    },
+    {
+      "epoch": 51.36,
+      "learning_rate": 4.358022388059702e-05,
+      "loss": 0.0,
+      "step": 55056
+    },
+    {
+      "epoch": 51.36,
+      "learning_rate": 4.357975746268657e-05,
+      "loss": 0.0,
+      "step": 55060
+    },
+    {
+      "epoch": 51.37,
+      "learning_rate": 4.357929104477612e-05,
+      "loss": 0.0,
+      "step": 55064
+    },
+    {
+      "epoch": 51.37,
+      "learning_rate": 4.3578824626865675e-05,
+      "loss": 0.0,
+      "step": 55068
+    },
+    {
+      "epoch": 51.37,
+      "learning_rate": 4.357835820895523e-05,
+      "loss": 0.0053,
+      "step": 55072
+    },
+    {
+      "epoch": 51.38,
+      "learning_rate": 4.357789179104478e-05,
+      "loss": 0.0002,
+      "step": 55076
+    },
+    {
+      "epoch": 51.38,
+      "learning_rate": 4.3577425373134326e-05,
+      "loss": 0.0,
+      "step": 55080
+    },
+    {
+      "epoch": 51.38,
+      "learning_rate": 4.357695895522389e-05,
+      "loss": 0.0,
+      "step": 55084
+    },
+    {
+      "epoch": 51.39,
+      "learning_rate": 4.3576492537313436e-05,
+      "loss": 0.0,
+      "step": 55088
+    },
+    {
+      "epoch": 51.39,
+      "learning_rate": 4.3576026119402984e-05,
+      "loss": 0.0001,
+      "step": 55092
+    },
+    {
+      "epoch": 51.4,
+      "learning_rate": 4.357555970149254e-05,
+      "loss": 0.0,
+      "step": 55096
+    },
+    {
+      "epoch": 51.4,
+      "learning_rate": 4.3575093283582094e-05,
+      "loss": 0.0,
+      "step": 55100
+    },
+    {
+      "epoch": 51.4,
+      "learning_rate": 4.357462686567164e-05,
+      "loss": 0.0,
+      "step": 55104
+    },
+    {
+      "epoch": 51.41,
+      "learning_rate": 4.35741604477612e-05,
+      "loss": 0.0,
+      "step": 55108
+    },
+    {
+      "epoch": 51.41,
+      "learning_rate": 4.3573694029850745e-05,
+      "loss": 0.0002,
+      "step": 55112
+    },
+    {
+      "epoch": 51.41,
+      "learning_rate": 4.35732276119403e-05,
+      "loss": 0.0002,
+      "step": 55116
+    },
+    {
+      "epoch": 51.42,
+      "learning_rate": 4.3572761194029855e-05,
+      "loss": 0.0,
+      "step": 55120
+    },
+    {
+      "epoch": 51.42,
+      "learning_rate": 4.35722947761194e-05,
+      "loss": 0.0,
+      "step": 55124
+    },
+    {
+      "epoch": 51.43,
+      "learning_rate": 4.357182835820896e-05,
+      "loss": 0.0,
+      "step": 55128
+    },
+    {
+      "epoch": 51.43,
+      "learning_rate": 4.357136194029851e-05,
+      "loss": 0.0,
+      "step": 55132
+    },
+    {
+      "epoch": 51.43,
+      "learning_rate": 4.357089552238806e-05,
+      "loss": 0.0,
+      "step": 55136
+    },
+    {
+      "epoch": 51.44,
+      "learning_rate": 4.357042910447761e-05,
+      "loss": 0.0001,
+      "step": 55140
+    },
+    {
+      "epoch": 51.44,
+      "learning_rate": 4.3569962686567164e-05,
+      "loss": 0.0001,
+      "step": 55144
+    },
+    {
+      "epoch": 51.44,
+      "learning_rate": 4.356949626865672e-05,
+      "loss": 0.0001,
+      "step": 55148
+    },
+    {
+      "epoch": 51.45,
+      "learning_rate": 4.356902985074627e-05,
+      "loss": 0.0001,
+      "step": 55152
+    },
+    {
+      "epoch": 51.45,
+      "learning_rate": 4.356856343283582e-05,
+      "loss": 0.0039,
+      "step": 55156
+    },
+    {
+      "epoch": 51.46,
+      "learning_rate": 4.356809701492538e-05,
+      "loss": 0.0003,
+      "step": 55160
+    },
+    {
+      "epoch": 51.46,
+      "learning_rate": 4.356763059701493e-05,
+      "loss": 0.0,
+      "step": 55164
+    },
+    {
+      "epoch": 51.46,
+      "learning_rate": 4.356716417910448e-05,
+      "loss": 0.0,
+      "step": 55168
+    },
+    {
+      "epoch": 51.47,
+      "learning_rate": 4.356669776119403e-05,
+      "loss": 0.0,
+      "step": 55172
+    },
+    {
+      "epoch": 51.47,
+      "learning_rate": 4.356623134328359e-05,
+      "loss": 0.0,
+      "step": 55176
+    },
+    {
+      "epoch": 51.47,
+      "learning_rate": 4.356576492537314e-05,
+      "loss": 0.0004,
+      "step": 55180
+    },
+    {
+      "epoch": 51.48,
+      "learning_rate": 4.3565298507462686e-05,
+      "loss": 0.0001,
+      "step": 55184
+    },
+    {
+      "epoch": 51.48,
+      "learning_rate": 4.356483208955224e-05,
+      "loss": 0.0,
+      "step": 55188
+    },
+    {
+      "epoch": 51.49,
+      "learning_rate": 4.3564365671641796e-05,
+      "loss": 0.0001,
+      "step": 55192
+    },
+    {
+      "epoch": 51.49,
+      "learning_rate": 4.3563899253731344e-05,
+      "loss": 0.0001,
+      "step": 55196
+    },
+    {
+      "epoch": 51.49,
+      "learning_rate": 4.35634328358209e-05,
+      "loss": 0.0,
+      "step": 55200
+    },
+    {
+      "epoch": 51.5,
+      "learning_rate": 4.356296641791045e-05,
+      "loss": 0.0,
+      "step": 55204
+    },
+    {
+      "epoch": 51.5,
+      "learning_rate": 4.35625e-05,
+      "loss": 0.0002,
+      "step": 55208
+    },
+    {
+      "epoch": 51.5,
+      "learning_rate": 4.356203358208956e-05,
+      "loss": 0.0,
+      "step": 55212
+    },
+    {
+      "epoch": 51.51,
+      "learning_rate": 4.3561567164179105e-05,
+      "loss": 0.0001,
+      "step": 55216
+    },
+    {
+      "epoch": 51.51,
+      "learning_rate": 4.356110074626866e-05,
+      "loss": 0.0,
+      "step": 55220
+    },
+    {
+      "epoch": 51.51,
+      "learning_rate": 4.3560634328358215e-05,
+      "loss": 0.0,
+      "step": 55224
+    },
+    {
+      "epoch": 51.52,
+      "learning_rate": 4.356016791044776e-05,
+      "loss": 0.0001,
+      "step": 55228
+    },
+    {
+      "epoch": 51.52,
+      "learning_rate": 4.355970149253731e-05,
+      "loss": 0.0017,
+      "step": 55232
+    },
+    {
+      "epoch": 51.53,
+      "learning_rate": 4.355923507462687e-05,
+      "loss": 0.0003,
+      "step": 55236
+    },
+    {
+      "epoch": 51.53,
+      "learning_rate": 4.355876865671642e-05,
+      "loss": 0.0,
+      "step": 55240
+    },
+    {
+      "epoch": 51.53,
+      "learning_rate": 4.355830223880597e-05,
+      "loss": 0.0,
+      "step": 55244
+    },
+    {
+      "epoch": 51.54,
+      "learning_rate": 4.3557835820895524e-05,
+      "loss": 0.0,
+      "step": 55248
+    },
+    {
+      "epoch": 51.54,
+      "learning_rate": 4.355736940298508e-05,
+      "loss": 0.0,
+      "step": 55252
+    },
+    {
+      "epoch": 51.54,
+      "learning_rate": 4.355690298507463e-05,
+      "loss": 0.0005,
+      "step": 55256
+    },
+    {
+      "epoch": 51.55,
+      "learning_rate": 4.355643656716418e-05,
+      "loss": 0.0,
+      "step": 55260
+    },
+    {
+      "epoch": 51.55,
+      "learning_rate": 4.355597014925373e-05,
+      "loss": 0.0003,
+      "step": 55264
+    },
+    {
+      "epoch": 51.56,
+      "learning_rate": 4.3555503731343285e-05,
+      "loss": 0.0001,
+      "step": 55268
+    },
+    {
+      "epoch": 51.56,
+      "learning_rate": 4.355503731343284e-05,
+      "loss": 0.0,
+      "step": 55272
+    },
+    {
+      "epoch": 51.56,
+      "learning_rate": 4.355457089552239e-05,
+      "loss": 0.0,
+      "step": 55276
+    },
+    {
+      "epoch": 51.57,
+      "learning_rate": 4.355410447761194e-05,
+      "loss": 0.0,
+      "step": 55280
+    },
+    {
+      "epoch": 51.57,
+      "learning_rate": 4.35536380597015e-05,
+      "loss": 0.0009,
+      "step": 55284
+    },
+    {
+      "epoch": 51.57,
+      "learning_rate": 4.3553171641791046e-05,
+      "loss": 0.0,
+      "step": 55288
+    },
+    {
+      "epoch": 51.58,
+      "learning_rate": 4.3552705223880594e-05,
+      "loss": 0.0,
+      "step": 55292
+    },
+    {
+      "epoch": 51.58,
+      "learning_rate": 4.3552238805970156e-05,
+      "loss": 0.0,
+      "step": 55296
+    },
+    {
+      "epoch": 51.59,
+      "learning_rate": 4.3551772388059704e-05,
+      "loss": 0.0,
+      "step": 55300
+    },
+    {
+      "epoch": 51.59,
+      "learning_rate": 4.355130597014925e-05,
+      "loss": 0.004,
+      "step": 55304
+    },
+    {
+      "epoch": 51.59,
+      "learning_rate": 4.355083955223881e-05,
+      "loss": 0.0,
+      "step": 55308
+    },
+    {
+      "epoch": 51.6,
+      "learning_rate": 4.355037313432836e-05,
+      "loss": 0.0,
+      "step": 55312
+    },
+    {
+      "epoch": 51.6,
+      "learning_rate": 4.3549906716417917e-05,
+      "loss": 0.004,
+      "step": 55316
+    },
+    {
+      "epoch": 51.6,
+      "learning_rate": 4.3549440298507465e-05,
+      "loss": 0.0,
+      "step": 55320
+    },
+    {
+      "epoch": 51.61,
+      "learning_rate": 4.354897388059701e-05,
+      "loss": 0.0002,
+      "step": 55324
+    },
+    {
+      "epoch": 51.61,
+      "learning_rate": 4.3548507462686574e-05,
+      "loss": 0.0,
+      "step": 55328
+    },
+    {
+      "epoch": 51.62,
+      "learning_rate": 4.354804104477612e-05,
+      "loss": 0.0001,
+      "step": 55332
+    },
+    {
+      "epoch": 51.62,
+      "learning_rate": 4.354757462686567e-05,
+      "loss": 0.0017,
+      "step": 55336
+    },
+    {
+      "epoch": 51.62,
+      "learning_rate": 4.3547108208955226e-05,
+      "loss": 0.0,
+      "step": 55340
+    },
+    {
+      "epoch": 51.63,
+      "learning_rate": 4.354664179104478e-05,
+      "loss": 0.0002,
+      "step": 55344
+    },
+    {
+      "epoch": 51.63,
+      "learning_rate": 4.354617537313433e-05,
+      "loss": 0.0,
+      "step": 55348
+    },
+    {
+      "epoch": 51.63,
+      "learning_rate": 4.3545708955223884e-05,
+      "loss": 0.0,
+      "step": 55352
+    },
+    {
+      "epoch": 51.64,
+      "learning_rate": 4.354524253731344e-05,
+      "loss": 0.0001,
+      "step": 55356
+    },
+    {
+      "epoch": 51.64,
+      "learning_rate": 4.354477611940299e-05,
+      "loss": 0.0027,
+      "step": 55360
+    },
+    {
+      "epoch": 51.65,
+      "learning_rate": 4.354430970149254e-05,
+      "loss": 0.0,
+      "step": 55364
+    },
+    {
+      "epoch": 51.65,
+      "learning_rate": 4.354384328358209e-05,
+      "loss": 0.0,
+      "step": 55368
+    },
+    {
+      "epoch": 51.65,
+      "learning_rate": 4.3543376865671645e-05,
+      "loss": 0.0043,
+      "step": 55372
+    },
+    {
+      "epoch": 51.66,
+      "learning_rate": 4.35429104477612e-05,
+      "loss": 0.0015,
+      "step": 55376
+    },
+    {
+      "epoch": 51.66,
+      "learning_rate": 4.354244402985075e-05,
+      "loss": 0.0,
+      "step": 55380
+    },
+    {
+      "epoch": 51.66,
+      "learning_rate": 4.3541977611940296e-05,
+      "loss": 0.0005,
+      "step": 55384
+    },
+    {
+      "epoch": 51.67,
+      "learning_rate": 4.354151119402986e-05,
+      "loss": 0.0001,
+      "step": 55388
+    },
+    {
+      "epoch": 51.67,
+      "learning_rate": 4.3541044776119406e-05,
+      "loss": 0.0,
+      "step": 55392
+    },
+    {
+      "epoch": 51.68,
+      "learning_rate": 4.3540578358208954e-05,
+      "loss": 0.0,
+      "step": 55396
+    },
+    {
+      "epoch": 51.68,
+      "learning_rate": 4.354011194029851e-05,
+      "loss": 0.0001,
+      "step": 55400
+    },
+    {
+      "epoch": 51.68,
+      "learning_rate": 4.3539645522388063e-05,
+      "loss": 0.0,
+      "step": 55404
+    },
+    {
+      "epoch": 51.69,
+      "learning_rate": 4.353917910447761e-05,
+      "loss": 0.0,
+      "step": 55408
+    },
+    {
+      "epoch": 51.69,
+      "learning_rate": 4.3538712686567167e-05,
+      "loss": 0.0001,
+      "step": 55412
+    },
+    {
+      "epoch": 51.69,
+      "learning_rate": 4.353824626865672e-05,
+      "loss": 0.0001,
+      "step": 55416
+    },
+    {
+      "epoch": 51.7,
+      "learning_rate": 4.353777985074627e-05,
+      "loss": 0.0029,
+      "step": 55420
+    },
+    {
+      "epoch": 51.7,
+      "learning_rate": 4.3537313432835824e-05,
+      "loss": 0.0116,
+      "step": 55424
+    },
+    {
+      "epoch": 51.71,
+      "learning_rate": 4.353684701492537e-05,
+      "loss": 0.0006,
+      "step": 55428
+    },
+    {
+      "epoch": 51.71,
+      "learning_rate": 4.353638059701493e-05,
+      "loss": 0.0003,
+      "step": 55432
+    },
+    {
+      "epoch": 51.71,
+      "learning_rate": 4.353591417910448e-05,
+      "loss": 0.0,
+      "step": 55436
+    },
+    {
+      "epoch": 51.72,
+      "learning_rate": 4.353544776119403e-05,
+      "loss": 0.0006,
+      "step": 55440
+    },
+    {
+      "epoch": 51.72,
+      "learning_rate": 4.353498134328358e-05,
+      "loss": 0.0,
+      "step": 55444
+    },
+    {
+      "epoch": 51.72,
+      "learning_rate": 4.353451492537314e-05,
+      "loss": 0.0004,
+      "step": 55448
+    },
+    {
+      "epoch": 51.73,
+      "learning_rate": 4.353404850746269e-05,
+      "loss": 0.001,
+      "step": 55452
+    },
+    {
+      "epoch": 51.73,
+      "learning_rate": 4.353358208955224e-05,
+      "loss": 0.0003,
+      "step": 55456
+    },
+    {
+      "epoch": 51.73,
+      "learning_rate": 4.353311567164179e-05,
+      "loss": 0.001,
+      "step": 55460
+    },
+    {
+      "epoch": 51.74,
+      "learning_rate": 4.3532649253731346e-05,
+      "loss": 0.0,
+      "step": 55464
+    },
+    {
+      "epoch": 51.74,
+      "learning_rate": 4.3532182835820895e-05,
+      "loss": 0.0005,
+      "step": 55468
+    },
+    {
+      "epoch": 51.75,
+      "learning_rate": 4.353171641791045e-05,
+      "loss": 0.0009,
+      "step": 55472
+    },
+    {
+      "epoch": 51.75,
+      "learning_rate": 4.3531250000000004e-05,
+      "loss": 0.0001,
+      "step": 55476
+    },
+    {
+      "epoch": 51.75,
+      "learning_rate": 4.353078358208956e-05,
+      "loss": 0.0,
+      "step": 55480
+    },
+    {
+      "epoch": 51.76,
+      "learning_rate": 4.353031716417911e-05,
+      "loss": 0.0028,
+      "step": 55484
+    },
+    {
+      "epoch": 51.76,
+      "learning_rate": 4.3529850746268656e-05,
+      "loss": 0.0,
+      "step": 55488
+    },
+    {
+      "epoch": 51.76,
+      "learning_rate": 4.352938432835821e-05,
+      "loss": 0.0,
+      "step": 55492
+    },
+    {
+      "epoch": 51.77,
+      "learning_rate": 4.3528917910447765e-05,
+      "loss": 0.0,
+      "step": 55496
+    },
+    {
+      "epoch": 51.77,
+      "learning_rate": 4.3528451492537314e-05,
+      "loss": 0.0,
+      "step": 55500
+    },
+    {
+      "epoch": 51.77,
+      "eval_exact_match": 0.741779497098646,
+      "eval_exec": 0.758220502901354,
+      "eval_loss": 0.46916627883911133,
+      "eval_runtime": 1122.7652,
+      "eval_samples_per_second": 0.921,
+      "step": 55500
+    },
+    {
+      "epoch": 51.78,
+      "learning_rate": 4.352798507462687e-05,
+      "loss": 0.0,
+      "step": 55504
+    },
+    {
+      "epoch": 51.78,
+      "learning_rate": 4.352751865671642e-05,
+      "loss": 0.0,
+      "step": 55508
+    },
+    {
+      "epoch": 51.78,
+      "learning_rate": 4.352705223880597e-05,
+      "loss": 0.0034,
+      "step": 55512
+    },
+    {
+      "epoch": 51.79,
+      "learning_rate": 4.3526585820895526e-05,
+      "loss": 0.0009,
+      "step": 55516
+    },
+    {
+      "epoch": 51.79,
+      "learning_rate": 4.3526119402985075e-05,
+      "loss": 0.002,
+      "step": 55520
+    },
+    {
+      "epoch": 51.79,
+      "learning_rate": 4.352565298507463e-05,
+      "loss": 0.0001,
+      "step": 55524
+    },
+    {
+      "epoch": 51.8,
+      "learning_rate": 4.3525186567164184e-05,
+      "loss": 0.0008,
+      "step": 55528
+    },
+    {
+      "epoch": 51.8,
+      "learning_rate": 4.352472014925373e-05,
+      "loss": 0.0002,
+      "step": 55532
+    },
+    {
+      "epoch": 51.81,
+      "learning_rate": 4.352425373134329e-05,
+      "loss": 0.0,
+      "step": 55536
+    },
+    {
+      "epoch": 51.81,
+      "learning_rate": 4.352378731343284e-05,
+      "loss": 0.0005,
+      "step": 55540
+    },
+    {
+      "epoch": 51.81,
+      "learning_rate": 4.352332089552239e-05,
+      "loss": 0.0001,
+      "step": 55544
+    },
+    {
+      "epoch": 51.82,
+      "learning_rate": 4.352285447761194e-05,
+      "loss": 0.0,
+      "step": 55548
+    },
+    {
+      "epoch": 51.82,
+      "learning_rate": 4.3522388059701493e-05,
+      "loss": 0.0001,
+      "step": 55552
+    },
+    {
+      "epoch": 51.82,
+      "learning_rate": 4.352192164179105e-05,
+      "loss": 0.0,
+      "step": 55556
+    },
+    {
+      "epoch": 51.83,
+      "learning_rate": 4.3521455223880596e-05,
+      "loss": 0.0,
+      "step": 55560
+    },
+    {
+      "epoch": 51.83,
+      "learning_rate": 4.352098880597015e-05,
+      "loss": 0.0004,
+      "step": 55564
+    },
+    {
+      "epoch": 51.84,
+      "learning_rate": 4.3520522388059706e-05,
+      "loss": 0.0001,
+      "step": 55568
+    },
+    {
+      "epoch": 51.84,
+      "learning_rate": 4.3520055970149254e-05,
+      "loss": 0.0,
+      "step": 55572
+    },
+    {
+      "epoch": 51.84,
+      "learning_rate": 4.351958955223881e-05,
+      "loss": 0.0005,
+      "step": 55576
+    },
+    {
+      "epoch": 51.85,
+      "learning_rate": 4.351912313432836e-05,
+      "loss": 0.0001,
+      "step": 55580
+    },
+    {
+      "epoch": 51.85,
+      "learning_rate": 4.351865671641791e-05,
+      "loss": 0.0,
+      "step": 55584
+    },
+    {
+      "epoch": 51.85,
+      "learning_rate": 4.351819029850747e-05,
+      "loss": 0.0,
+      "step": 55588
+    },
+    {
+      "epoch": 51.86,
+      "learning_rate": 4.3517723880597015e-05,
+      "loss": 0.0,
+      "step": 55592
+    },
+    {
+      "epoch": 51.86,
+      "learning_rate": 4.351725746268657e-05,
+      "loss": 0.0001,
+      "step": 55596
+    },
+    {
+      "epoch": 51.87,
+      "learning_rate": 4.3516791044776125e-05,
+      "loss": 0.0001,
+      "step": 55600
+    },
+    {
+      "epoch": 51.87,
+      "learning_rate": 4.351632462686567e-05,
+      "loss": 0.0001,
+      "step": 55604
+    },
+    {
+      "epoch": 51.87,
+      "learning_rate": 4.351585820895522e-05,
+      "loss": 0.0,
+      "step": 55608
+    },
+    {
+      "epoch": 51.88,
+      "learning_rate": 4.3515391791044776e-05,
+      "loss": 0.0,
+      "step": 55612
+    },
+    {
+      "epoch": 51.88,
+      "learning_rate": 4.351492537313433e-05,
+      "loss": 0.0,
+      "step": 55616
+    },
+    {
+      "epoch": 51.88,
+      "learning_rate": 4.351445895522388e-05,
+      "loss": 0.0021,
+      "step": 55620
+    },
+    {
+      "epoch": 51.89,
+      "learning_rate": 4.3513992537313434e-05,
+      "loss": 0.0,
+      "step": 55624
+    },
+    {
+      "epoch": 51.89,
+      "learning_rate": 4.351352611940299e-05,
+      "loss": 0.0003,
+      "step": 55628
+    },
+    {
+      "epoch": 51.9,
+      "learning_rate": 4.351305970149254e-05,
+      "loss": 0.0013,
+      "step": 55632
+    },
+    {
+      "epoch": 51.9,
+      "learning_rate": 4.351259328358209e-05,
+      "loss": 0.0,
+      "step": 55636
+    },
+    {
+      "epoch": 51.9,
+      "learning_rate": 4.351212686567164e-05,
+      "loss": 0.0,
+      "step": 55640
+    },
+    {
+      "epoch": 51.91,
+      "learning_rate": 4.35116604477612e-05,
+      "loss": 0.0,
+      "step": 55644
+    },
+    {
+      "epoch": 51.91,
+      "learning_rate": 4.351119402985075e-05,
+      "loss": 0.0001,
+      "step": 55648
+    },
+    {
+      "epoch": 51.91,
+      "learning_rate": 4.35107276119403e-05,
+      "loss": 0.0001,
+      "step": 55652
+    },
+    {
+      "epoch": 51.92,
+      "learning_rate": 4.351026119402985e-05,
+      "loss": 0.0002,
+      "step": 55656
+    },
+    {
+      "epoch": 51.92,
+      "learning_rate": 4.350979477611941e-05,
+      "loss": 0.0008,
+      "step": 55660
+    },
+    {
+      "epoch": 51.93,
+      "learning_rate": 4.3509328358208956e-05,
+      "loss": 0.0,
+      "step": 55664
+    },
+    {
+      "epoch": 51.93,
+      "learning_rate": 4.350886194029851e-05,
+      "loss": 0.0,
+      "step": 55668
+    },
+    {
+      "epoch": 51.93,
+      "learning_rate": 4.350839552238806e-05,
+      "loss": 0.0,
+      "step": 55672
+    },
+    {
+      "epoch": 51.94,
+      "learning_rate": 4.3507929104477614e-05,
+      "loss": 0.0001,
+      "step": 55676
+    },
+    {
+      "epoch": 51.94,
+      "learning_rate": 4.350746268656717e-05,
+      "loss": 0.0,
+      "step": 55680
+    },
+    {
+      "epoch": 51.94,
+      "learning_rate": 4.350699626865672e-05,
+      "loss": 0.0,
+      "step": 55684
+    },
+    {
+      "epoch": 51.95,
+      "learning_rate": 4.350652985074627e-05,
+      "loss": 0.0004,
+      "step": 55688
+    },
+    {
+      "epoch": 51.95,
+      "learning_rate": 4.350606343283583e-05,
+      "loss": 0.0,
+      "step": 55692
+    },
+    {
+      "epoch": 51.96,
+      "learning_rate": 4.3505597014925375e-05,
+      "loss": 0.0001,
+      "step": 55696
+    },
+    {
+      "epoch": 51.96,
+      "learning_rate": 4.350513059701492e-05,
+      "loss": 0.0,
+      "step": 55700
+    },
+    {
+      "epoch": 51.96,
+      "learning_rate": 4.3504664179104485e-05,
+      "loss": 0.0001,
+      "step": 55704
+    },
+    {
+      "epoch": 51.97,
+      "learning_rate": 4.350419776119403e-05,
+      "loss": 0.0001,
+      "step": 55708
+    },
+    {
+      "epoch": 51.97,
+      "learning_rate": 4.350373134328358e-05,
+      "loss": 0.0003,
+      "step": 55712
+    },
+    {
+      "epoch": 51.97,
+      "learning_rate": 4.3503264925373136e-05,
+      "loss": 0.0,
+      "step": 55716
+    },
+    {
+      "epoch": 51.98,
+      "learning_rate": 4.350279850746269e-05,
+      "loss": 0.0027,
+      "step": 55720
+    },
+    {
+      "epoch": 51.98,
+      "learning_rate": 4.350233208955224e-05,
+      "loss": 0.0,
+      "step": 55724
+    },
+    {
+      "epoch": 51.98,
+      "learning_rate": 4.3501865671641794e-05,
+      "loss": 0.0002,
+      "step": 55728
+    },
+    {
+      "epoch": 51.99,
+      "learning_rate": 4.350139925373134e-05,
+      "loss": 0.0,
+      "step": 55732
+    },
+    {
+      "epoch": 51.99,
+      "learning_rate": 4.35009328358209e-05,
+      "loss": 0.0001,
+      "step": 55736
+    },
+    {
+      "epoch": 52.0,
+      "learning_rate": 4.350046641791045e-05,
+      "loss": 0.0003,
+      "step": 55740
+    },
+    {
+      "epoch": 52.0,
+      "learning_rate": 4.35e-05,
+      "loss": 0.0,
+      "step": 55744
+    },
+    {
+      "epoch": 52.0,
+      "learning_rate": 4.3499533582089555e-05,
+      "loss": 0.0,
+      "step": 55748
+    },
+    {
+      "epoch": 52.01,
+      "learning_rate": 4.349906716417911e-05,
+      "loss": 0.0,
+      "step": 55752
+    },
+    {
+      "epoch": 52.01,
+      "learning_rate": 4.349860074626866e-05,
+      "loss": 0.0,
+      "step": 55756
+    },
+    {
+      "epoch": 52.01,
+      "learning_rate": 4.3498134328358206e-05,
+      "loss": 0.0,
+      "step": 55760
+    },
+    {
+      "epoch": 52.02,
+      "learning_rate": 4.349766791044777e-05,
+      "loss": 0.0,
+      "step": 55764
+    },
+    {
+      "epoch": 52.02,
+      "learning_rate": 4.3497201492537316e-05,
+      "loss": 0.0,
+      "step": 55768
+    },
+    {
+      "epoch": 52.03,
+      "learning_rate": 4.3496735074626864e-05,
+      "loss": 0.0001,
+      "step": 55772
+    },
+    {
+      "epoch": 52.03,
+      "learning_rate": 4.349626865671642e-05,
+      "loss": 0.0,
+      "step": 55776
+    },
+    {
+      "epoch": 52.03,
+      "learning_rate": 4.3495802238805974e-05,
+      "loss": 0.0006,
+      "step": 55780
+    },
+    {
+      "epoch": 52.04,
+      "learning_rate": 4.349533582089552e-05,
+      "loss": 0.0044,
+      "step": 55784
+    },
+    {
+      "epoch": 52.04,
+      "learning_rate": 4.349486940298508e-05,
+      "loss": 0.0,
+      "step": 55788
+    },
+    {
+      "epoch": 52.04,
+      "learning_rate": 4.3494402985074625e-05,
+      "loss": 0.0,
+      "step": 55792
+    },
+    {
+      "epoch": 52.05,
+      "learning_rate": 4.349393656716418e-05,
+      "loss": 0.0,
+      "step": 55796
+    },
+    {
+      "epoch": 52.05,
+      "learning_rate": 4.3493470149253735e-05,
+      "loss": 0.0005,
+      "step": 55800
+    },
+    {
+      "epoch": 52.06,
+      "learning_rate": 4.349300373134328e-05,
+      "loss": 0.0,
+      "step": 55804
+    },
+    {
+      "epoch": 52.06,
+      "learning_rate": 4.349253731343284e-05,
+      "loss": 0.0,
+      "step": 55808
+    },
+    {
+      "epoch": 52.06,
+      "learning_rate": 4.349207089552239e-05,
+      "loss": 0.0001,
+      "step": 55812
+    },
+    {
+      "epoch": 52.07,
+      "learning_rate": 4.349160447761194e-05,
+      "loss": 0.0024,
+      "step": 55816
+    },
+    {
+      "epoch": 52.07,
+      "learning_rate": 4.3491138059701496e-05,
+      "loss": 0.0015,
+      "step": 55820
+    },
+    {
+      "epoch": 52.07,
+      "learning_rate": 4.3490671641791044e-05,
+      "loss": 0.0001,
+      "step": 55824
+    },
+    {
+      "epoch": 52.08,
+      "learning_rate": 4.34902052238806e-05,
+      "loss": 0.0,
+      "step": 55828
+    },
+    {
+      "epoch": 52.08,
+      "learning_rate": 4.3489738805970154e-05,
+      "loss": 0.0,
+      "step": 55832
+    },
+    {
+      "epoch": 52.09,
+      "learning_rate": 4.34892723880597e-05,
+      "loss": 0.003,
+      "step": 55836
+    },
+    {
+      "epoch": 52.09,
+      "learning_rate": 4.348880597014926e-05,
+      "loss": 0.0,
+      "step": 55840
+    },
+    {
+      "epoch": 52.09,
+      "learning_rate": 4.348833955223881e-05,
+      "loss": 0.0,
+      "step": 55844
+    },
+    {
+      "epoch": 52.1,
+      "learning_rate": 4.348787313432836e-05,
+      "loss": 0.0008,
+      "step": 55848
+    },
+    {
+      "epoch": 52.1,
+      "learning_rate": 4.348740671641791e-05,
+      "loss": 0.0,
+      "step": 55852
+    },
+    {
+      "epoch": 52.1,
+      "learning_rate": 4.348694029850747e-05,
+      "loss": 0.0002,
+      "step": 55856
+    },
+    {
+      "epoch": 52.11,
+      "learning_rate": 4.348647388059702e-05,
+      "loss": 0.0001,
+      "step": 55860
+    },
+    {
+      "epoch": 52.11,
+      "learning_rate": 4.3486007462686566e-05,
+      "loss": 0.0001,
+      "step": 55864
+    },
+    {
+      "epoch": 52.12,
+      "learning_rate": 4.348554104477612e-05,
+      "loss": 0.0,
+      "step": 55868
+    },
+    {
+      "epoch": 52.12,
+      "learning_rate": 4.3485074626865676e-05,
+      "loss": 0.0004,
+      "step": 55872
+    },
+    {
+      "epoch": 52.12,
+      "learning_rate": 4.3484608208955224e-05,
+      "loss": 0.0,
+      "step": 55876
+    },
+    {
+      "epoch": 52.13,
+      "learning_rate": 4.348414179104478e-05,
+      "loss": 0.0001,
+      "step": 55880
+    },
+    {
+      "epoch": 52.13,
+      "learning_rate": 4.348367537313433e-05,
+      "loss": 0.0,
+      "step": 55884
+    },
+    {
+      "epoch": 52.13,
+      "learning_rate": 4.348320895522388e-05,
+      "loss": 0.0,
+      "step": 55888
+    },
+    {
+      "epoch": 52.14,
+      "learning_rate": 4.348274253731344e-05,
+      "loss": 0.0,
+      "step": 55892
+    },
+    {
+      "epoch": 52.14,
+      "learning_rate": 4.3482276119402985e-05,
+      "loss": 0.0,
+      "step": 55896
+    },
+    {
+      "epoch": 52.15,
+      "learning_rate": 4.348180970149254e-05,
+      "loss": 0.0,
+      "step": 55900
+    },
+    {
+      "epoch": 52.15,
+      "learning_rate": 4.3481343283582095e-05,
+      "loss": 0.0,
+      "step": 55904
+    },
+    {
+      "epoch": 52.15,
+      "learning_rate": 4.348087686567164e-05,
+      "loss": 0.0001,
+      "step": 55908
+    },
+    {
+      "epoch": 52.16,
+      "learning_rate": 4.348041044776119e-05,
+      "loss": 0.0003,
+      "step": 55912
+    },
+    {
+      "epoch": 52.16,
+      "learning_rate": 4.347994402985075e-05,
+      "loss": 0.0,
+      "step": 55916
+    },
+    {
+      "epoch": 52.16,
+      "learning_rate": 4.34794776119403e-05,
+      "loss": 0.0014,
+      "step": 55920
+    },
+    {
+      "epoch": 52.17,
+      "learning_rate": 4.347901119402985e-05,
+      "loss": 0.0005,
+      "step": 55924
+    },
+    {
+      "epoch": 52.17,
+      "learning_rate": 4.3478544776119404e-05,
+      "loss": 0.0,
+      "step": 55928
+    },
+    {
+      "epoch": 52.18,
+      "learning_rate": 4.347807835820896e-05,
+      "loss": 0.0,
+      "step": 55932
+    },
+    {
+      "epoch": 52.18,
+      "learning_rate": 4.347761194029851e-05,
+      "loss": 0.0,
+      "step": 55936
+    },
+    {
+      "epoch": 52.18,
+      "learning_rate": 4.347714552238806e-05,
+      "loss": 0.0,
+      "step": 55940
+    },
+    {
+      "epoch": 52.19,
+      "learning_rate": 4.347667910447761e-05,
+      "loss": 0.0,
+      "step": 55944
+    },
+    {
+      "epoch": 52.19,
+      "learning_rate": 4.3476212686567165e-05,
+      "loss": 0.0,
+      "step": 55948
+    },
+    {
+      "epoch": 52.19,
+      "learning_rate": 4.347574626865672e-05,
+      "loss": 0.0,
+      "step": 55952
+    },
+    {
+      "epoch": 52.2,
+      "learning_rate": 4.347527985074627e-05,
+      "loss": 0.0001,
+      "step": 55956
+    },
+    {
+      "epoch": 52.2,
+      "learning_rate": 4.347481343283582e-05,
+      "loss": 0.0,
+      "step": 55960
+    },
+    {
+      "epoch": 52.21,
+      "learning_rate": 4.347434701492538e-05,
+      "loss": 0.0,
+      "step": 55964
+    },
+    {
+      "epoch": 52.21,
+      "learning_rate": 4.3473880597014926e-05,
+      "loss": 0.0,
+      "step": 55968
+    },
+    {
+      "epoch": 52.21,
+      "learning_rate": 4.347341417910448e-05,
+      "loss": 0.0,
+      "step": 55972
+    },
+    {
+      "epoch": 52.22,
+      "learning_rate": 4.3472947761194036e-05,
+      "loss": 0.0001,
+      "step": 55976
+    },
+    {
+      "epoch": 52.22,
+      "learning_rate": 4.3472481343283584e-05,
+      "loss": 0.0,
+      "step": 55980
+    },
+    {
+      "epoch": 52.22,
+      "learning_rate": 4.347201492537314e-05,
+      "loss": 0.0,
+      "step": 55984
+    },
+    {
+      "epoch": 52.23,
+      "learning_rate": 4.347154850746269e-05,
+      "loss": 0.0005,
+      "step": 55988
+    },
+    {
+      "epoch": 52.23,
+      "learning_rate": 4.347108208955224e-05,
+      "loss": 0.0,
+      "step": 55992
+    },
+    {
+      "epoch": 52.24,
+      "learning_rate": 4.34706156716418e-05,
+      "loss": 0.0,
+      "step": 55996
+    },
+    {
+      "epoch": 52.24,
+      "learning_rate": 4.3470149253731345e-05,
+      "loss": 0.0,
+      "step": 56000
+    },
+    {
+      "epoch": 52.24,
+      "eval_exact_match": 0.7398452611218569,
+      "eval_exec": 0.7630560928433269,
+      "eval_loss": 0.48453760147094727,
+      "eval_runtime": 1139.6139,
+      "eval_samples_per_second": 0.907,
+      "step": 56000
+    },
+    {
+      "epoch": 52.24,
+      "learning_rate": 4.346968283582089e-05,
+      "loss": 0.0,
+      "step": 56004
+    },
+    {
+      "epoch": 52.25,
+      "learning_rate": 4.3469216417910455e-05,
+      "loss": 0.0,
+      "step": 56008
+    },
+    {
+      "epoch": 52.25,
+      "learning_rate": 4.346875e-05,
+      "loss": 0.0,
+      "step": 56012
+    },
+    {
+      "epoch": 52.25,
+      "learning_rate": 4.346828358208955e-05,
+      "loss": 0.0,
+      "step": 56016
+    },
+    {
+      "epoch": 52.26,
+      "learning_rate": 4.3467817164179106e-05,
+      "loss": 0.0,
+      "step": 56020
+    },
+    {
+      "epoch": 52.26,
+      "learning_rate": 4.346735074626866e-05,
+      "loss": 0.0,
+      "step": 56024
+    },
+    {
+      "epoch": 52.26,
+      "learning_rate": 4.346688432835821e-05,
+      "loss": 0.0,
+      "step": 56028
+    },
+    {
+      "epoch": 52.27,
+      "learning_rate": 4.3466417910447764e-05,
+      "loss": 0.0,
+      "step": 56032
+    },
+    {
+      "epoch": 52.27,
+      "learning_rate": 4.346595149253732e-05,
+      "loss": 0.0001,
+      "step": 56036
+    },
+    {
+      "epoch": 52.28,
+      "learning_rate": 4.346548507462687e-05,
+      "loss": 0.0,
+      "step": 56040
+    },
+    {
+      "epoch": 52.28,
+      "learning_rate": 4.346501865671642e-05,
+      "loss": 0.0,
+      "step": 56044
+    },
+    {
+      "epoch": 52.28,
+      "learning_rate": 4.346455223880597e-05,
+      "loss": 0.0032,
+      "step": 56048
+    },
+    {
+      "epoch": 52.29,
+      "learning_rate": 4.3464085820895525e-05,
+      "loss": 0.0,
+      "step": 56052
+    },
+    {
+      "epoch": 52.29,
+      "learning_rate": 4.346361940298508e-05,
+      "loss": 0.0,
+      "step": 56056
+    },
+    {
+      "epoch": 52.29,
+      "learning_rate": 4.346315298507463e-05,
+      "loss": 0.0021,
+      "step": 56060
+    },
+    {
+      "epoch": 52.3,
+      "learning_rate": 4.3462686567164176e-05,
+      "loss": 0.0,
+      "step": 56064
+    },
+    {
+      "epoch": 52.3,
+      "learning_rate": 4.346222014925374e-05,
+      "loss": 0.0,
+      "step": 56068
+    },
+    {
+      "epoch": 52.31,
+      "learning_rate": 4.3461753731343286e-05,
+      "loss": 0.0,
+      "step": 56072
+    },
+    {
+      "epoch": 52.31,
+      "learning_rate": 4.3461287313432834e-05,
+      "loss": 0.0,
+      "step": 56076
+    },
+    {
+      "epoch": 52.31,
+      "learning_rate": 4.346082089552239e-05,
+      "loss": 0.0,
+      "step": 56080
+    },
+    {
+      "epoch": 52.32,
+      "learning_rate": 4.3460354477611944e-05,
+      "loss": 0.0,
+      "step": 56084
+    },
+    {
+      "epoch": 52.32,
+      "learning_rate": 4.345988805970149e-05,
+      "loss": 0.0,
+      "step": 56088
+    },
+    {
+      "epoch": 52.32,
+      "learning_rate": 4.345942164179105e-05,
+      "loss": 0.0001,
+      "step": 56092
+    },
+    {
+      "epoch": 52.33,
+      "learning_rate": 4.34589552238806e-05,
+      "loss": 0.0,
+      "step": 56096
+    },
+    {
+      "epoch": 52.33,
+      "learning_rate": 4.345848880597015e-05,
+      "loss": 0.0003,
+      "step": 56100
+    },
+    {
+      "epoch": 52.34,
+      "learning_rate": 4.3458022388059705e-05,
+      "loss": 0.0,
+      "step": 56104
+    },
+    {
+      "epoch": 52.34,
+      "learning_rate": 4.345755597014925e-05,
+      "loss": 0.0,
+      "step": 56108
+    },
+    {
+      "epoch": 52.34,
+      "learning_rate": 4.345708955223881e-05,
+      "loss": 0.0002,
+      "step": 56112
+    },
+    {
+      "epoch": 52.35,
+      "learning_rate": 4.345662313432836e-05,
+      "loss": 0.001,
+      "step": 56116
+    },
+    {
+      "epoch": 52.35,
+      "learning_rate": 4.345615671641791e-05,
+      "loss": 0.0005,
+      "step": 56120
+    },
+    {
+      "epoch": 52.35,
+      "learning_rate": 4.345569029850746e-05,
+      "loss": 0.0041,
+      "step": 56124
+    },
+    {
+      "epoch": 52.36,
+      "learning_rate": 4.345522388059702e-05,
+      "loss": 0.0,
+      "step": 56128
+    },
+    {
+      "epoch": 52.36,
+      "learning_rate": 4.345475746268657e-05,
+      "loss": 0.0,
+      "step": 56132
+    },
+    {
+      "epoch": 52.37,
+      "learning_rate": 4.3454291044776124e-05,
+      "loss": 0.0,
+      "step": 56136
+    },
+    {
+      "epoch": 52.37,
+      "learning_rate": 4.345382462686567e-05,
+      "loss": 0.0,
+      "step": 56140
+    },
+    {
+      "epoch": 52.37,
+      "learning_rate": 4.345335820895523e-05,
+      "loss": 0.0006,
+      "step": 56144
+    },
+    {
+      "epoch": 52.38,
+      "learning_rate": 4.345289179104478e-05,
+      "loss": 0.0001,
+      "step": 56148
+    },
+    {
+      "epoch": 52.38,
+      "learning_rate": 4.345242537313433e-05,
+      "loss": 0.0,
+      "step": 56152
+    },
+    {
+      "epoch": 52.38,
+      "learning_rate": 4.3451958955223885e-05,
+      "loss": 0.0,
+      "step": 56156
+    },
+    {
+      "epoch": 52.39,
+      "learning_rate": 4.345149253731344e-05,
+      "loss": 0.0,
+      "step": 56160
+    },
+    {
+      "epoch": 52.39,
+      "learning_rate": 4.345102611940299e-05,
+      "loss": 0.0001,
+      "step": 56164
+    },
+    {
+      "epoch": 52.4,
+      "learning_rate": 4.3450559701492536e-05,
+      "loss": 0.0,
+      "step": 56168
+    },
+    {
+      "epoch": 52.4,
+      "learning_rate": 4.345009328358209e-05,
+      "loss": 0.0,
+      "step": 56172
+    },
+    {
+      "epoch": 52.4,
+      "learning_rate": 4.3449626865671646e-05,
+      "loss": 0.0,
+      "step": 56176
+    },
+    {
+      "epoch": 52.41,
+      "learning_rate": 4.3449160447761194e-05,
+      "loss": 0.0,
+      "step": 56180
+    },
+    {
+      "epoch": 52.41,
+      "learning_rate": 4.344869402985075e-05,
+      "loss": 0.0,
+      "step": 56184
+    },
+    {
+      "epoch": 52.41,
+      "learning_rate": 4.3448227611940304e-05,
+      "loss": 0.0,
+      "step": 56188
+    },
+    {
+      "epoch": 52.42,
+      "learning_rate": 4.344776119402985e-05,
+      "loss": 0.0,
+      "step": 56192
+    },
+    {
+      "epoch": 52.42,
+      "learning_rate": 4.344729477611941e-05,
+      "loss": 0.0,
+      "step": 56196
+    },
+    {
+      "epoch": 52.43,
+      "learning_rate": 4.3446828358208955e-05,
+      "loss": 0.0001,
+      "step": 56200
+    },
+    {
+      "epoch": 52.43,
+      "learning_rate": 4.344636194029851e-05,
+      "loss": 0.0002,
+      "step": 56204
+    },
+    {
+      "epoch": 52.43,
+      "learning_rate": 4.3445895522388065e-05,
+      "loss": 0.0,
+      "step": 56208
+    },
+    {
+      "epoch": 52.44,
+      "learning_rate": 4.344542910447761e-05,
+      "loss": 0.0,
+      "step": 56212
+    },
+    {
+      "epoch": 52.44,
+      "learning_rate": 4.344496268656717e-05,
+      "loss": 0.0,
+      "step": 56216
+    },
+    {
+      "epoch": 52.44,
+      "learning_rate": 4.344449626865672e-05,
+      "loss": 0.0,
+      "step": 56220
+    },
+    {
+      "epoch": 52.45,
+      "learning_rate": 4.344402985074627e-05,
+      "loss": 0.0009,
+      "step": 56224
+    },
+    {
+      "epoch": 52.45,
+      "learning_rate": 4.344356343283582e-05,
+      "loss": 0.0008,
+      "step": 56228
+    },
+    {
+      "epoch": 52.46,
+      "learning_rate": 4.3443097014925374e-05,
+      "loss": 0.0,
+      "step": 56232
+    },
+    {
+      "epoch": 52.46,
+      "learning_rate": 4.344263059701493e-05,
+      "loss": 0.0014,
+      "step": 56236
+    },
+    {
+      "epoch": 52.46,
+      "learning_rate": 4.344216417910448e-05,
+      "loss": 0.0,
+      "step": 56240
+    },
+    {
+      "epoch": 52.47,
+      "learning_rate": 4.344169776119403e-05,
+      "loss": 0.0,
+      "step": 56244
+    },
+    {
+      "epoch": 52.47,
+      "learning_rate": 4.3441231343283587e-05,
+      "loss": 0.0002,
+      "step": 56248
+    },
+    {
+      "epoch": 52.47,
+      "learning_rate": 4.3440764925373135e-05,
+      "loss": 0.0,
+      "step": 56252
+    },
+    {
+      "epoch": 52.48,
+      "learning_rate": 4.344029850746269e-05,
+      "loss": 0.0,
+      "step": 56256
+    },
+    {
+      "epoch": 52.48,
+      "learning_rate": 4.343983208955224e-05,
+      "loss": 0.0002,
+      "step": 56260
+    },
+    {
+      "epoch": 52.49,
+      "learning_rate": 4.343936567164179e-05,
+      "loss": 0.0,
+      "step": 56264
+    },
+    {
+      "epoch": 52.49,
+      "learning_rate": 4.343889925373135e-05,
+      "loss": 0.0004,
+      "step": 56268
+    },
+    {
+      "epoch": 52.49,
+      "learning_rate": 4.3438432835820896e-05,
+      "loss": 0.0,
+      "step": 56272
+    },
+    {
+      "epoch": 52.5,
+      "learning_rate": 4.343796641791045e-05,
+      "loss": 0.0,
+      "step": 56276
+    },
+    {
+      "epoch": 52.5,
+      "learning_rate": 4.3437500000000006e-05,
+      "loss": 0.0,
+      "step": 56280
+    },
+    {
+      "epoch": 52.5,
+      "learning_rate": 4.3437033582089554e-05,
+      "loss": 0.0,
+      "step": 56284
+    },
+    {
+      "epoch": 52.51,
+      "learning_rate": 4.34365671641791e-05,
+      "loss": 0.0,
+      "step": 56288
+    },
+    {
+      "epoch": 52.51,
+      "learning_rate": 4.343610074626866e-05,
+      "loss": 0.0,
+      "step": 56292
+    },
+    {
+      "epoch": 52.51,
+      "learning_rate": 4.343563432835821e-05,
+      "loss": 0.0,
+      "step": 56296
+    },
+    {
+      "epoch": 52.52,
+      "learning_rate": 4.3435167910447766e-05,
+      "loss": 0.0,
+      "step": 56300
+    },
+    {
+      "epoch": 52.52,
+      "learning_rate": 4.3434701492537315e-05,
+      "loss": 0.0,
+      "step": 56304
+    },
+    {
+      "epoch": 52.53,
+      "learning_rate": 4.343423507462687e-05,
+      "loss": 0.0,
+      "step": 56308
+    },
+    {
+      "epoch": 52.53,
+      "learning_rate": 4.3433768656716424e-05,
+      "loss": 0.0,
+      "step": 56312
+    },
+    {
+      "epoch": 52.53,
+      "learning_rate": 4.343330223880597e-05,
+      "loss": 0.0,
+      "step": 56316
+    },
+    {
+      "epoch": 52.54,
+      "learning_rate": 4.343283582089552e-05,
+      "loss": 0.0002,
+      "step": 56320
+    },
+    {
+      "epoch": 52.54,
+      "learning_rate": 4.343236940298508e-05,
+      "loss": 0.0,
+      "step": 56324
+    },
+    {
+      "epoch": 52.54,
+      "learning_rate": 4.343190298507463e-05,
+      "loss": 0.0004,
+      "step": 56328
+    },
+    {
+      "epoch": 52.55,
+      "learning_rate": 4.343143656716418e-05,
+      "loss": 0.0,
+      "step": 56332
+    },
+    {
+      "epoch": 52.55,
+      "learning_rate": 4.3430970149253734e-05,
+      "loss": 0.0,
+      "step": 56336
+    },
+    {
+      "epoch": 52.56,
+      "learning_rate": 4.343050373134329e-05,
+      "loss": 0.0,
+      "step": 56340
+    },
+    {
+      "epoch": 52.56,
+      "learning_rate": 4.3430037313432837e-05,
+      "loss": 0.0,
+      "step": 56344
+    },
+    {
+      "epoch": 52.56,
+      "learning_rate": 4.342957089552239e-05,
+      "loss": 0.0,
+      "step": 56348
+    },
+    {
+      "epoch": 52.57,
+      "learning_rate": 4.342910447761194e-05,
+      "loss": 0.0,
+      "step": 56352
+    },
+    {
+      "epoch": 52.57,
+      "learning_rate": 4.3428638059701495e-05,
+      "loss": 0.0,
+      "step": 56356
+    },
+    {
+      "epoch": 52.57,
+      "learning_rate": 4.342817164179105e-05,
+      "loss": 0.0,
+      "step": 56360
+    },
+    {
+      "epoch": 52.58,
+      "learning_rate": 4.34277052238806e-05,
+      "loss": 0.0,
+      "step": 56364
+    },
+    {
+      "epoch": 52.58,
+      "learning_rate": 4.342723880597015e-05,
+      "loss": 0.0,
+      "step": 56368
+    },
+    {
+      "epoch": 52.59,
+      "learning_rate": 4.342677238805971e-05,
+      "loss": 0.0,
+      "step": 56372
+    },
+    {
+      "epoch": 52.59,
+      "learning_rate": 4.3426305970149256e-05,
+      "loss": 0.0,
+      "step": 56376
+    },
+    {
+      "epoch": 52.59,
+      "learning_rate": 4.3425839552238804e-05,
+      "loss": 0.0,
+      "step": 56380
+    },
+    {
+      "epoch": 52.6,
+      "learning_rate": 4.3425373134328365e-05,
+      "loss": 0.0004,
+      "step": 56384
+    },
+    {
+      "epoch": 52.6,
+      "learning_rate": 4.3424906716417913e-05,
+      "loss": 0.0,
+      "step": 56388
+    },
+    {
+      "epoch": 52.6,
+      "learning_rate": 4.342444029850746e-05,
+      "loss": 0.0001,
+      "step": 56392
+    },
+    {
+      "epoch": 52.61,
+      "learning_rate": 4.3423973880597017e-05,
+      "loss": 0.0,
+      "step": 56396
+    },
+    {
+      "epoch": 52.61,
+      "learning_rate": 4.342350746268657e-05,
+      "loss": 0.0002,
+      "step": 56400
+    },
+    {
+      "epoch": 52.62,
+      "learning_rate": 4.342304104477612e-05,
+      "loss": 0.0004,
+      "step": 56404
+    },
+    {
+      "epoch": 52.62,
+      "learning_rate": 4.3422574626865674e-05,
+      "loss": 0.0007,
+      "step": 56408
+    },
+    {
+      "epoch": 52.62,
+      "learning_rate": 4.342210820895522e-05,
+      "loss": 0.0,
+      "step": 56412
+    },
+    {
+      "epoch": 52.63,
+      "learning_rate": 4.342164179104478e-05,
+      "loss": 0.0,
+      "step": 56416
+    },
+    {
+      "epoch": 52.63,
+      "learning_rate": 4.342117537313433e-05,
+      "loss": 0.0,
+      "step": 56420
+    },
+    {
+      "epoch": 52.63,
+      "learning_rate": 4.342070895522388e-05,
+      "loss": 0.0,
+      "step": 56424
+    },
+    {
+      "epoch": 52.64,
+      "learning_rate": 4.3420242537313435e-05,
+      "loss": 0.0016,
+      "step": 56428
+    },
+    {
+      "epoch": 52.64,
+      "learning_rate": 4.341977611940299e-05,
+      "loss": 0.0,
+      "step": 56432
+    },
+    {
+      "epoch": 52.65,
+      "learning_rate": 4.341930970149254e-05,
+      "loss": 0.0,
+      "step": 56436
+    },
+    {
+      "epoch": 52.65,
+      "learning_rate": 4.341884328358209e-05,
+      "loss": 0.0006,
+      "step": 56440
+    },
+    {
+      "epoch": 52.65,
+      "learning_rate": 4.341837686567165e-05,
+      "loss": 0.0004,
+      "step": 56444
+    },
+    {
+      "epoch": 52.66,
+      "learning_rate": 4.3417910447761196e-05,
+      "loss": 0.0,
+      "step": 56448
+    },
+    {
+      "epoch": 52.66,
+      "learning_rate": 4.3417444029850745e-05,
+      "loss": 0.0003,
+      "step": 56452
+    },
+    {
+      "epoch": 52.66,
+      "learning_rate": 4.34169776119403e-05,
+      "loss": 0.0,
+      "step": 56456
+    },
+    {
+      "epoch": 52.67,
+      "learning_rate": 4.3416511194029854e-05,
+      "loss": 0.0,
+      "step": 56460
+    },
+    {
+      "epoch": 52.67,
+      "learning_rate": 4.341604477611941e-05,
+      "loss": 0.0,
+      "step": 56464
+    },
+    {
+      "epoch": 52.68,
+      "learning_rate": 4.341557835820896e-05,
+      "loss": 0.0001,
+      "step": 56468
+    },
+    {
+      "epoch": 52.68,
+      "learning_rate": 4.3415111940298506e-05,
+      "loss": 0.0,
+      "step": 56472
+    },
+    {
+      "epoch": 52.68,
+      "learning_rate": 4.341464552238807e-05,
+      "loss": 0.0,
+      "step": 56476
+    },
+    {
+      "epoch": 52.69,
+      "learning_rate": 4.3414179104477615e-05,
+      "loss": 0.0,
+      "step": 56480
+    },
+    {
+      "epoch": 52.69,
+      "learning_rate": 4.3413712686567163e-05,
+      "loss": 0.0,
+      "step": 56484
+    },
+    {
+      "epoch": 52.69,
+      "learning_rate": 4.341324626865672e-05,
+      "loss": 0.0,
+      "step": 56488
+    },
+    {
+      "epoch": 52.7,
+      "learning_rate": 4.341277985074627e-05,
+      "loss": 0.0007,
+      "step": 56492
+    },
+    {
+      "epoch": 52.7,
+      "learning_rate": 4.341231343283582e-05,
+      "loss": 0.0,
+      "step": 56496
+    },
+    {
+      "epoch": 52.71,
+      "learning_rate": 4.3411847014925376e-05,
+      "loss": 0.0,
+      "step": 56500
+    },
+    {
+      "epoch": 52.71,
+      "eval_exact_match": 0.7398452611218569,
+      "eval_exec": 0.7688588007736944,
+      "eval_loss": 0.48737451434135437,
+      "eval_runtime": 1152.9321,
+      "eval_samples_per_second": 0.897,
+      "step": 56500
+    },
+    {
+      "epoch": 52.71,
+      "learning_rate": 4.341138059701493e-05,
+      "loss": 0.0008,
+      "step": 56504
+    },
+    {
+      "epoch": 52.71,
+      "learning_rate": 4.341091417910448e-05,
+      "loss": 0.0,
+      "step": 56508
+    },
+    {
+      "epoch": 52.72,
+      "learning_rate": 4.3410447761194034e-05,
+      "loss": 0.0,
+      "step": 56512
+    },
+    {
+      "epoch": 52.72,
+      "learning_rate": 4.340998134328358e-05,
+      "loss": 0.0,
+      "step": 56516
+    },
+    {
+      "epoch": 52.72,
+      "learning_rate": 4.340951492537314e-05,
+      "loss": 0.0002,
+      "step": 56520
+    },
+    {
+      "epoch": 52.73,
+      "learning_rate": 4.340904850746269e-05,
+      "loss": 0.0,
+      "step": 56524
+    },
+    {
+      "epoch": 52.73,
+      "learning_rate": 4.340858208955224e-05,
+      "loss": 0.0,
+      "step": 56528
+    },
+    {
+      "epoch": 52.73,
+      "learning_rate": 4.340811567164179e-05,
+      "loss": 0.0,
+      "step": 56532
+    },
+    {
+      "epoch": 52.74,
+      "learning_rate": 4.340764925373135e-05,
+      "loss": 0.0022,
+      "step": 56536
+    },
+    {
+      "epoch": 52.74,
+      "learning_rate": 4.34071828358209e-05,
+      "loss": 0.0025,
+      "step": 56540
+    },
+    {
+      "epoch": 52.75,
+      "learning_rate": 4.3406716417910446e-05,
+      "loss": 0.0,
+      "step": 56544
+    },
+    {
+      "epoch": 52.75,
+      "learning_rate": 4.340625e-05,
+      "loss": 0.0001,
+      "step": 56548
+    },
+    {
+      "epoch": 52.75,
+      "learning_rate": 4.3405783582089556e-05,
+      "loss": 0.0001,
+      "step": 56552
+    },
+    {
+      "epoch": 52.76,
+      "learning_rate": 4.3405317164179104e-05,
+      "loss": 0.0001,
+      "step": 56556
+    },
+    {
+      "epoch": 52.76,
+      "learning_rate": 4.340485074626866e-05,
+      "loss": 0.0003,
+      "step": 56560
+    },
+    {
+      "epoch": 52.76,
+      "learning_rate": 4.340438432835821e-05,
+      "loss": 0.0004,
+      "step": 56564
+    },
+    {
+      "epoch": 52.77,
+      "learning_rate": 4.340391791044776e-05,
+      "loss": 0.0,
+      "step": 56568
+    },
+    {
+      "epoch": 52.77,
+      "learning_rate": 4.340345149253732e-05,
+      "loss": 0.0,
+      "step": 56572
+    },
+    {
+      "epoch": 52.78,
+      "learning_rate": 4.3402985074626865e-05,
+      "loss": 0.0004,
+      "step": 56576
+    },
+    {
+      "epoch": 52.78,
+      "learning_rate": 4.340251865671642e-05,
+      "loss": 0.0,
+      "step": 56580
+    },
+    {
+      "epoch": 52.78,
+      "learning_rate": 4.3402052238805975e-05,
+      "loss": 0.0,
+      "step": 56584
+    },
+    {
+      "epoch": 52.79,
+      "learning_rate": 4.340158582089552e-05,
+      "loss": 0.0001,
+      "step": 56588
+    },
+    {
+      "epoch": 52.79,
+      "learning_rate": 4.340111940298507e-05,
+      "loss": 0.0,
+      "step": 56592
+    },
+    {
+      "epoch": 52.79,
+      "learning_rate": 4.340065298507463e-05,
+      "loss": 0.0,
+      "step": 56596
+    },
+    {
+      "epoch": 52.8,
+      "learning_rate": 4.340018656716418e-05,
+      "loss": 0.0,
+      "step": 56600
+    },
+    {
+      "epoch": 52.8,
+      "learning_rate": 4.339972014925373e-05,
+      "loss": 0.0003,
+      "step": 56604
+    },
+    {
+      "epoch": 52.81,
+      "learning_rate": 4.3399253731343284e-05,
+      "loss": 0.0002,
+      "step": 56608
+    },
+    {
+      "epoch": 52.81,
+      "learning_rate": 4.339878731343284e-05,
+      "loss": 0.0009,
+      "step": 56612
+    },
+    {
+      "epoch": 52.81,
+      "learning_rate": 4.339832089552239e-05,
+      "loss": 0.0001,
+      "step": 56616
+    },
+    {
+      "epoch": 52.82,
+      "learning_rate": 4.339785447761194e-05,
+      "loss": 0.0009,
+      "step": 56620
+    },
+    {
+      "epoch": 52.82,
+      "learning_rate": 4.339738805970149e-05,
+      "loss": 0.0001,
+      "step": 56624
+    },
+    {
+      "epoch": 52.82,
+      "learning_rate": 4.339692164179105e-05,
+      "loss": 0.0,
+      "step": 56628
+    },
+    {
+      "epoch": 52.83,
+      "learning_rate": 4.33964552238806e-05,
+      "loss": 0.0001,
+      "step": 56632
+    },
+    {
+      "epoch": 52.83,
+      "learning_rate": 4.339598880597015e-05,
+      "loss": 0.0005,
+      "step": 56636
+    },
+    {
+      "epoch": 52.84,
+      "learning_rate": 4.33955223880597e-05,
+      "loss": 0.0,
+      "step": 56640
+    },
+    {
+      "epoch": 52.84,
+      "learning_rate": 4.339505597014926e-05,
+      "loss": 0.0,
+      "step": 56644
+    },
+    {
+      "epoch": 52.84,
+      "learning_rate": 4.3394589552238806e-05,
+      "loss": 0.0001,
+      "step": 56648
+    },
+    {
+      "epoch": 52.85,
+      "learning_rate": 4.339412313432836e-05,
+      "loss": 0.0001,
+      "step": 56652
+    },
+    {
+      "epoch": 52.85,
+      "learning_rate": 4.3393656716417916e-05,
+      "loss": 0.0003,
+      "step": 56656
+    },
+    {
+      "epoch": 52.85,
+      "learning_rate": 4.3393190298507464e-05,
+      "loss": 0.0084,
+      "step": 56660
+    },
+    {
+      "epoch": 52.86,
+      "learning_rate": 4.339272388059702e-05,
+      "loss": 0.0001,
+      "step": 56664
+    },
+    {
+      "epoch": 52.86,
+      "learning_rate": 4.339225746268657e-05,
+      "loss": 0.0,
+      "step": 56668
+    },
+    {
+      "epoch": 52.87,
+      "learning_rate": 4.339179104477612e-05,
+      "loss": 0.0,
+      "step": 56672
+    },
+    {
+      "epoch": 52.87,
+      "learning_rate": 4.339132462686568e-05,
+      "loss": 0.0,
+      "step": 56676
+    },
+    {
+      "epoch": 52.87,
+      "learning_rate": 4.3390858208955225e-05,
+      "loss": 0.0,
+      "step": 56680
+    },
+    {
+      "epoch": 52.88,
+      "learning_rate": 4.339039179104477e-05,
+      "loss": 0.0,
+      "step": 56684
+    },
+    {
+      "epoch": 52.88,
+      "learning_rate": 4.3389925373134335e-05,
+      "loss": 0.0,
+      "step": 56688
+    },
+    {
+      "epoch": 52.88,
+      "learning_rate": 4.338945895522388e-05,
+      "loss": 0.0,
+      "step": 56692
+    },
+    {
+      "epoch": 52.89,
+      "learning_rate": 4.338899253731343e-05,
+      "loss": 0.0035,
+      "step": 56696
+    },
+    {
+      "epoch": 52.89,
+      "learning_rate": 4.3388526119402986e-05,
+      "loss": 0.0,
+      "step": 56700
+    },
+    {
+      "epoch": 52.9,
+      "learning_rate": 4.338805970149254e-05,
+      "loss": 0.0001,
+      "step": 56704
+    },
+    {
+      "epoch": 52.9,
+      "learning_rate": 4.338759328358209e-05,
+      "loss": 0.0,
+      "step": 56708
+    },
+    {
+      "epoch": 52.9,
+      "learning_rate": 4.3387126865671644e-05,
+      "loss": 0.0,
+      "step": 56712
+    },
+    {
+      "epoch": 52.91,
+      "learning_rate": 4.33866604477612e-05,
+      "loss": 0.0005,
+      "step": 56716
+    },
+    {
+      "epoch": 52.91,
+      "learning_rate": 4.338619402985075e-05,
+      "loss": 0.0,
+      "step": 56720
+    },
+    {
+      "epoch": 52.91,
+      "learning_rate": 4.33857276119403e-05,
+      "loss": 0.0,
+      "step": 56724
+    },
+    {
+      "epoch": 52.92,
+      "learning_rate": 4.338526119402985e-05,
+      "loss": 0.0,
+      "step": 56728
+    },
+    {
+      "epoch": 52.92,
+      "learning_rate": 4.3384794776119405e-05,
+      "loss": 0.0001,
+      "step": 56732
+    },
+    {
+      "epoch": 52.93,
+      "learning_rate": 4.338432835820896e-05,
+      "loss": 0.0007,
+      "step": 56736
+    },
+    {
+      "epoch": 52.93,
+      "learning_rate": 4.338386194029851e-05,
+      "loss": 0.0,
+      "step": 56740
+    },
+    {
+      "epoch": 52.93,
+      "learning_rate": 4.3383395522388056e-05,
+      "loss": 0.0088,
+      "step": 56744
+    },
+    {
+      "epoch": 52.94,
+      "learning_rate": 4.338292910447762e-05,
+      "loss": 0.0,
+      "step": 56748
+    },
+    {
+      "epoch": 52.94,
+      "learning_rate": 4.3382462686567166e-05,
+      "loss": 0.0001,
+      "step": 56752
+    },
+    {
+      "epoch": 52.94,
+      "learning_rate": 4.3381996268656714e-05,
+      "loss": 0.0002,
+      "step": 56756
+    },
+    {
+      "epoch": 52.95,
+      "learning_rate": 4.338152985074627e-05,
+      "loss": 0.0001,
+      "step": 56760
+    },
+    {
+      "epoch": 52.95,
+      "learning_rate": 4.3381063432835824e-05,
+      "loss": 0.0003,
+      "step": 56764
+    },
+    {
+      "epoch": 52.96,
+      "learning_rate": 4.338059701492537e-05,
+      "loss": 0.0,
+      "step": 56768
+    },
+    {
+      "epoch": 52.96,
+      "learning_rate": 4.338013059701493e-05,
+      "loss": 0.0,
+      "step": 56772
+    },
+    {
+      "epoch": 52.96,
+      "learning_rate": 4.337966417910448e-05,
+      "loss": 0.0,
+      "step": 56776
+    },
+    {
+      "epoch": 52.97,
+      "learning_rate": 4.337919776119403e-05,
+      "loss": 0.0056,
+      "step": 56780
+    },
+    {
+      "epoch": 52.97,
+      "learning_rate": 4.3378731343283585e-05,
+      "loss": 0.0001,
+      "step": 56784
+    },
+    {
+      "epoch": 52.97,
+      "learning_rate": 4.337826492537313e-05,
+      "loss": 0.0,
+      "step": 56788
+    },
+    {
+      "epoch": 52.98,
+      "learning_rate": 4.337779850746269e-05,
+      "loss": 0.0008,
+      "step": 56792
+    },
+    {
+      "epoch": 52.98,
+      "learning_rate": 4.337733208955224e-05,
+      "loss": 0.0007,
+      "step": 56796
+    },
+    {
+      "epoch": 52.98,
+      "learning_rate": 4.337686567164179e-05,
+      "loss": 0.0002,
+      "step": 56800
+    },
+    {
+      "epoch": 52.99,
+      "learning_rate": 4.3376399253731346e-05,
+      "loss": 0.0001,
+      "step": 56804
+    },
+    {
+      "epoch": 52.99,
+      "learning_rate": 4.33759328358209e-05,
+      "loss": 0.0,
+      "step": 56808
+    },
+    {
+      "epoch": 53.0,
+      "learning_rate": 4.337546641791045e-05,
+      "loss": 0.0,
+      "step": 56812
+    },
+    {
+      "epoch": 53.0,
+      "learning_rate": 4.3375000000000004e-05,
+      "loss": 0.0005,
+      "step": 56816
+    },
+    {
+      "epoch": 53.0,
+      "learning_rate": 4.337453358208955e-05,
+      "loss": 0.0018,
+      "step": 56820
+    },
+    {
+      "epoch": 53.01,
+      "learning_rate": 4.337406716417911e-05,
+      "loss": 0.0017,
+      "step": 56824
+    },
+    {
+      "epoch": 53.01,
+      "learning_rate": 4.337360074626866e-05,
+      "loss": 0.0,
+      "step": 56828
+    },
+    {
+      "epoch": 53.01,
+      "learning_rate": 4.337313432835821e-05,
+      "loss": 0.0003,
+      "step": 56832
+    },
+    {
+      "epoch": 53.02,
+      "learning_rate": 4.3372667910447765e-05,
+      "loss": 0.0,
+      "step": 56836
+    },
+    {
+      "epoch": 53.02,
+      "learning_rate": 4.337220149253732e-05,
+      "loss": 0.0007,
+      "step": 56840
+    },
+    {
+      "epoch": 53.03,
+      "learning_rate": 4.337173507462687e-05,
+      "loss": 0.0,
+      "step": 56844
+    },
+    {
+      "epoch": 53.03,
+      "learning_rate": 4.3371268656716416e-05,
+      "loss": 0.0001,
+      "step": 56848
+    },
+    {
+      "epoch": 53.03,
+      "learning_rate": 4.337080223880597e-05,
+      "loss": 0.0,
+      "step": 56852
+    },
+    {
+      "epoch": 53.04,
+      "learning_rate": 4.3370335820895526e-05,
+      "loss": 0.0,
+      "step": 56856
+    },
+    {
+      "epoch": 53.04,
+      "learning_rate": 4.3369869402985074e-05,
+      "loss": 0.0003,
+      "step": 56860
+    },
+    {
+      "epoch": 53.04,
+      "learning_rate": 4.336940298507463e-05,
+      "loss": 0.0,
+      "step": 56864
+    },
+    {
+      "epoch": 53.05,
+      "learning_rate": 4.3368936567164184e-05,
+      "loss": 0.0,
+      "step": 56868
+    },
+    {
+      "epoch": 53.05,
+      "learning_rate": 4.336847014925373e-05,
+      "loss": 0.0,
+      "step": 56872
+    },
+    {
+      "epoch": 53.06,
+      "learning_rate": 4.336800373134329e-05,
+      "loss": 0.0,
+      "step": 56876
+    },
+    {
+      "epoch": 53.06,
+      "learning_rate": 4.3367537313432835e-05,
+      "loss": 0.0,
+      "step": 56880
+    },
+    {
+      "epoch": 53.06,
+      "learning_rate": 4.336707089552239e-05,
+      "loss": 0.0,
+      "step": 56884
+    },
+    {
+      "epoch": 53.07,
+      "learning_rate": 4.3366604477611945e-05,
+      "loss": 0.0,
+      "step": 56888
+    },
+    {
+      "epoch": 53.07,
+      "learning_rate": 4.336613805970149e-05,
+      "loss": 0.0,
+      "step": 56892
+    },
+    {
+      "epoch": 53.07,
+      "learning_rate": 4.336567164179105e-05,
+      "loss": 0.0,
+      "step": 56896
+    },
+    {
+      "epoch": 53.08,
+      "learning_rate": 4.33652052238806e-05,
+      "loss": 0.0,
+      "step": 56900
+    },
+    {
+      "epoch": 53.08,
+      "learning_rate": 4.336473880597015e-05,
+      "loss": 0.0008,
+      "step": 56904
+    },
+    {
+      "epoch": 53.09,
+      "learning_rate": 4.33642723880597e-05,
+      "loss": 0.0002,
+      "step": 56908
+    },
+    {
+      "epoch": 53.09,
+      "learning_rate": 4.3363805970149254e-05,
+      "loss": 0.0,
+      "step": 56912
+    },
+    {
+      "epoch": 53.09,
+      "learning_rate": 4.336333955223881e-05,
+      "loss": 0.0,
+      "step": 56916
+    },
+    {
+      "epoch": 53.1,
+      "learning_rate": 4.336287313432836e-05,
+      "loss": 0.0034,
+      "step": 56920
+    },
+    {
+      "epoch": 53.1,
+      "learning_rate": 4.336240671641791e-05,
+      "loss": 0.0001,
+      "step": 56924
+    },
+    {
+      "epoch": 53.1,
+      "learning_rate": 4.336194029850747e-05,
+      "loss": 0.0,
+      "step": 56928
+    },
+    {
+      "epoch": 53.11,
+      "learning_rate": 4.3361473880597015e-05,
+      "loss": 0.0,
+      "step": 56932
+    },
+    {
+      "epoch": 53.11,
+      "learning_rate": 4.336100746268657e-05,
+      "loss": 0.0001,
+      "step": 56936
+    },
+    {
+      "epoch": 53.12,
+      "learning_rate": 4.336054104477612e-05,
+      "loss": 0.0013,
+      "step": 56940
+    },
+    {
+      "epoch": 53.12,
+      "learning_rate": 4.336007462686567e-05,
+      "loss": 0.0,
+      "step": 56944
+    },
+    {
+      "epoch": 53.12,
+      "learning_rate": 4.335960820895523e-05,
+      "loss": 0.0,
+      "step": 56948
+    },
+    {
+      "epoch": 53.13,
+      "learning_rate": 4.3359141791044776e-05,
+      "loss": 0.0006,
+      "step": 56952
+    },
+    {
+      "epoch": 53.13,
+      "learning_rate": 4.335867537313433e-05,
+      "loss": 0.0,
+      "step": 56956
+    },
+    {
+      "epoch": 53.13,
+      "learning_rate": 4.3358208955223886e-05,
+      "loss": 0.0,
+      "step": 56960
+    },
+    {
+      "epoch": 53.14,
+      "learning_rate": 4.3357742537313434e-05,
+      "loss": 0.0,
+      "step": 56964
+    },
+    {
+      "epoch": 53.14,
+      "learning_rate": 4.335727611940299e-05,
+      "loss": 0.0,
+      "step": 56968
+    },
+    {
+      "epoch": 53.15,
+      "learning_rate": 4.335680970149254e-05,
+      "loss": 0.0,
+      "step": 56972
+    },
+    {
+      "epoch": 53.15,
+      "learning_rate": 4.335634328358209e-05,
+      "loss": 0.0,
+      "step": 56976
+    },
+    {
+      "epoch": 53.15,
+      "learning_rate": 4.335587686567165e-05,
+      "loss": 0.0003,
+      "step": 56980
+    },
+    {
+      "epoch": 53.16,
+      "learning_rate": 4.3355410447761195e-05,
+      "loss": 0.0,
+      "step": 56984
+    },
+    {
+      "epoch": 53.16,
+      "learning_rate": 4.335494402985075e-05,
+      "loss": 0.0003,
+      "step": 56988
+    },
+    {
+      "epoch": 53.16,
+      "learning_rate": 4.3354477611940305e-05,
+      "loss": 0.0,
+      "step": 56992
+    },
+    {
+      "epoch": 53.17,
+      "learning_rate": 4.335401119402985e-05,
+      "loss": 0.0,
+      "step": 56996
+    },
+    {
+      "epoch": 53.17,
+      "learning_rate": 4.33535447761194e-05,
+      "loss": 0.0,
+      "step": 57000
+    },
+    {
+      "epoch": 53.17,
+      "eval_exact_match": 0.7311411992263056,
+      "eval_exec": 0.7756286266924565,
+      "eval_loss": 0.4817250370979309,
+      "eval_runtime": 1118.231,
+      "eval_samples_per_second": 0.925,
+      "step": 57000
+    },
+    {
+      "epoch": 53.18,
+      "learning_rate": 4.335307835820896e-05,
+      "loss": 0.0001,
+      "step": 57004
+    },
+    {
+      "epoch": 53.18,
+      "learning_rate": 4.335261194029851e-05,
+      "loss": 0.0,
+      "step": 57008
+    },
+    {
+      "epoch": 53.18,
+      "learning_rate": 4.335214552238806e-05,
+      "loss": 0.0,
+      "step": 57012
+    },
+    {
+      "epoch": 53.19,
+      "learning_rate": 4.3351679104477614e-05,
+      "loss": 0.0,
+      "step": 57016
+    },
+    {
+      "epoch": 53.19,
+      "learning_rate": 4.335121268656717e-05,
+      "loss": 0.0,
+      "step": 57020
+    },
+    {
+      "epoch": 53.19,
+      "learning_rate": 4.335074626865672e-05,
+      "loss": 0.0001,
+      "step": 57024
+    },
+    {
+      "epoch": 53.2,
+      "learning_rate": 4.335027985074627e-05,
+      "loss": 0.0001,
+      "step": 57028
+    },
+    {
+      "epoch": 53.2,
+      "learning_rate": 4.334981343283582e-05,
+      "loss": 0.0,
+      "step": 57032
+    },
+    {
+      "epoch": 53.21,
+      "learning_rate": 4.3349347014925375e-05,
+      "loss": 0.0,
+      "step": 57036
+    },
+    {
+      "epoch": 53.21,
+      "learning_rate": 4.334888059701493e-05,
+      "loss": 0.0,
+      "step": 57040
+    },
+    {
+      "epoch": 53.21,
+      "learning_rate": 4.334841417910448e-05,
+      "loss": 0.0,
+      "step": 57044
+    },
+    {
+      "epoch": 53.22,
+      "learning_rate": 4.334794776119403e-05,
+      "loss": 0.0,
+      "step": 57048
+    },
+    {
+      "epoch": 53.22,
+      "learning_rate": 4.334748134328359e-05,
+      "loss": 0.0031,
+      "step": 57052
+    },
+    {
+      "epoch": 53.22,
+      "learning_rate": 4.3347014925373136e-05,
+      "loss": 0.0001,
+      "step": 57056
+    },
+    {
+      "epoch": 53.23,
+      "learning_rate": 4.3346548507462684e-05,
+      "loss": 0.0,
+      "step": 57060
+    },
+    {
+      "epoch": 53.23,
+      "learning_rate": 4.3346082089552246e-05,
+      "loss": 0.0117,
+      "step": 57064
+    },
+    {
+      "epoch": 53.24,
+      "learning_rate": 4.3345615671641794e-05,
+      "loss": 0.0,
+      "step": 57068
+    },
+    {
+      "epoch": 53.24,
+      "learning_rate": 4.334514925373134e-05,
+      "loss": 0.0,
+      "step": 57072
+    },
+    {
+      "epoch": 53.24,
+      "learning_rate": 4.33446828358209e-05,
+      "loss": 0.0,
+      "step": 57076
+    },
+    {
+      "epoch": 53.25,
+      "learning_rate": 4.334421641791045e-05,
+      "loss": 0.0,
+      "step": 57080
+    },
+    {
+      "epoch": 53.25,
+      "learning_rate": 4.334375e-05,
+      "loss": 0.0,
+      "step": 57084
+    },
+    {
+      "epoch": 53.25,
+      "learning_rate": 4.3343283582089555e-05,
+      "loss": 0.0006,
+      "step": 57088
+    },
+    {
+      "epoch": 53.26,
+      "learning_rate": 4.33428171641791e-05,
+      "loss": 0.0,
+      "step": 57092
+    },
+    {
+      "epoch": 53.26,
+      "learning_rate": 4.334235074626866e-05,
+      "loss": 0.0,
+      "step": 57096
+    },
+    {
+      "epoch": 53.26,
+      "learning_rate": 4.334188432835821e-05,
+      "loss": 0.0007,
+      "step": 57100
+    },
+    {
+      "epoch": 53.27,
+      "learning_rate": 4.334141791044776e-05,
+      "loss": 0.0014,
+      "step": 57104
+    },
+    {
+      "epoch": 53.27,
+      "learning_rate": 4.3340951492537316e-05,
+      "loss": 0.0,
+      "step": 57108
+    },
+    {
+      "epoch": 53.28,
+      "learning_rate": 4.334048507462687e-05,
+      "loss": 0.0001,
+      "step": 57112
+    },
+    {
+      "epoch": 53.28,
+      "learning_rate": 4.334001865671642e-05,
+      "loss": 0.0,
+      "step": 57116
+    },
+    {
+      "epoch": 53.28,
+      "learning_rate": 4.3339552238805974e-05,
+      "loss": 0.0065,
+      "step": 57120
+    },
+    {
+      "epoch": 53.29,
+      "learning_rate": 4.333908582089553e-05,
+      "loss": 0.0,
+      "step": 57124
+    },
+    {
+      "epoch": 53.29,
+      "learning_rate": 4.333861940298508e-05,
+      "loss": 0.0,
+      "step": 57128
+    },
+    {
+      "epoch": 53.29,
+      "learning_rate": 4.333815298507463e-05,
+      "loss": 0.0,
+      "step": 57132
+    },
+    {
+      "epoch": 53.3,
+      "learning_rate": 4.333768656716418e-05,
+      "loss": 0.0,
+      "step": 57136
+    },
+    {
+      "epoch": 53.3,
+      "learning_rate": 4.3337220149253735e-05,
+      "loss": 0.0,
+      "step": 57140
+    },
+    {
+      "epoch": 53.31,
+      "learning_rate": 4.333675373134329e-05,
+      "loss": 0.0,
+      "step": 57144
+    },
+    {
+      "epoch": 53.31,
+      "learning_rate": 4.333628731343284e-05,
+      "loss": 0.0001,
+      "step": 57148
+    },
+    {
+      "epoch": 53.31,
+      "learning_rate": 4.3335820895522386e-05,
+      "loss": 0.0,
+      "step": 57152
+    },
+    {
+      "epoch": 53.32,
+      "learning_rate": 4.333535447761195e-05,
+      "loss": 0.0005,
+      "step": 57156
+    },
+    {
+      "epoch": 53.32,
+      "learning_rate": 4.3334888059701496e-05,
+      "loss": 0.0,
+      "step": 57160
+    },
+    {
+      "epoch": 53.32,
+      "learning_rate": 4.3334421641791044e-05,
+      "loss": 0.0,
+      "step": 57164
+    },
+    {
+      "epoch": 53.33,
+      "learning_rate": 4.33339552238806e-05,
+      "loss": 0.0234,
+      "step": 57168
+    },
+    {
+      "epoch": 53.33,
+      "learning_rate": 4.3333488805970154e-05,
+      "loss": 0.0,
+      "step": 57172
+    },
+    {
+      "epoch": 53.34,
+      "learning_rate": 4.33330223880597e-05,
+      "loss": 0.0,
+      "step": 57176
+    },
+    {
+      "epoch": 53.34,
+      "learning_rate": 4.333255597014926e-05,
+      "loss": 0.0001,
+      "step": 57180
+    },
+    {
+      "epoch": 53.34,
+      "learning_rate": 4.333208955223881e-05,
+      "loss": 0.0041,
+      "step": 57184
+    },
+    {
+      "epoch": 53.35,
+      "learning_rate": 4.333162313432836e-05,
+      "loss": 0.0,
+      "step": 57188
+    },
+    {
+      "epoch": 53.35,
+      "learning_rate": 4.3331156716417915e-05,
+      "loss": 0.0,
+      "step": 57192
+    },
+    {
+      "epoch": 53.35,
+      "learning_rate": 4.333069029850746e-05,
+      "loss": 0.0,
+      "step": 57196
+    },
+    {
+      "epoch": 53.36,
+      "learning_rate": 4.333022388059702e-05,
+      "loss": 0.0,
+      "step": 57200
+    },
+    {
+      "epoch": 53.36,
+      "learning_rate": 4.332975746268657e-05,
+      "loss": 0.0001,
+      "step": 57204
+    },
+    {
+      "epoch": 53.37,
+      "learning_rate": 4.332929104477612e-05,
+      "loss": 0.0,
+      "step": 57208
+    },
+    {
+      "epoch": 53.37,
+      "learning_rate": 4.332882462686567e-05,
+      "loss": 0.0001,
+      "step": 57212
+    },
+    {
+      "epoch": 53.37,
+      "learning_rate": 4.332835820895523e-05,
+      "loss": 0.0,
+      "step": 57216
+    },
+    {
+      "epoch": 53.38,
+      "learning_rate": 4.332789179104478e-05,
+      "loss": 0.0005,
+      "step": 57220
+    },
+    {
+      "epoch": 53.38,
+      "learning_rate": 4.332742537313433e-05,
+      "loss": 0.0,
+      "step": 57224
+    },
+    {
+      "epoch": 53.38,
+      "learning_rate": 4.332695895522388e-05,
+      "loss": 0.0,
+      "step": 57228
+    },
+    {
+      "epoch": 53.39,
+      "learning_rate": 4.3326492537313437e-05,
+      "loss": 0.0003,
+      "step": 57232
+    },
+    {
+      "epoch": 53.39,
+      "learning_rate": 4.3326026119402985e-05,
+      "loss": 0.0,
+      "step": 57236
+    },
+    {
+      "epoch": 53.4,
+      "learning_rate": 4.332555970149254e-05,
+      "loss": 0.0,
+      "step": 57240
+    },
+    {
+      "epoch": 53.4,
+      "learning_rate": 4.332509328358209e-05,
+      "loss": 0.0,
+      "step": 57244
+    },
+    {
+      "epoch": 53.4,
+      "learning_rate": 4.332462686567164e-05,
+      "loss": 0.0,
+      "step": 57248
+    },
+    {
+      "epoch": 53.41,
+      "learning_rate": 4.33241604477612e-05,
+      "loss": 0.0,
+      "step": 57252
+    },
+    {
+      "epoch": 53.41,
+      "learning_rate": 4.3323694029850746e-05,
+      "loss": 0.0,
+      "step": 57256
+    },
+    {
+      "epoch": 53.41,
+      "learning_rate": 4.33232276119403e-05,
+      "loss": 0.0,
+      "step": 57260
+    },
+    {
+      "epoch": 53.42,
+      "learning_rate": 4.3322761194029855e-05,
+      "loss": 0.0,
+      "step": 57264
+    },
+    {
+      "epoch": 53.42,
+      "learning_rate": 4.3322294776119404e-05,
+      "loss": 0.0,
+      "step": 57268
+    },
+    {
+      "epoch": 53.43,
+      "learning_rate": 4.332182835820895e-05,
+      "loss": 0.0,
+      "step": 57272
+    },
+    {
+      "epoch": 53.43,
+      "learning_rate": 4.3321361940298513e-05,
+      "loss": 0.0,
+      "step": 57276
+    },
+    {
+      "epoch": 53.43,
+      "learning_rate": 4.332089552238806e-05,
+      "loss": 0.0,
+      "step": 57280
+    },
+    {
+      "epoch": 53.44,
+      "learning_rate": 4.3320429104477616e-05,
+      "loss": 0.0,
+      "step": 57284
+    },
+    {
+      "epoch": 53.44,
+      "learning_rate": 4.3319962686567165e-05,
+      "loss": 0.0,
+      "step": 57288
+    },
+    {
+      "epoch": 53.44,
+      "learning_rate": 4.331949626865672e-05,
+      "loss": 0.0,
+      "step": 57292
+    },
+    {
+      "epoch": 53.45,
+      "learning_rate": 4.3319029850746274e-05,
+      "loss": 0.0,
+      "step": 57296
+    },
+    {
+      "epoch": 53.45,
+      "learning_rate": 4.331856343283582e-05,
+      "loss": 0.0,
+      "step": 57300
+    },
+    {
+      "epoch": 53.46,
+      "learning_rate": 4.331809701492537e-05,
+      "loss": 0.0,
+      "step": 57304
+    },
+    {
+      "epoch": 53.46,
+      "learning_rate": 4.331763059701493e-05,
+      "loss": 0.0023,
+      "step": 57308
+    },
+    {
+      "epoch": 53.46,
+      "learning_rate": 4.331716417910448e-05,
+      "loss": 0.0001,
+      "step": 57312
+    },
+    {
+      "epoch": 53.47,
+      "learning_rate": 4.331669776119403e-05,
+      "loss": 0.0,
+      "step": 57316
+    },
+    {
+      "epoch": 53.47,
+      "learning_rate": 4.3316231343283584e-05,
+      "loss": 0.0006,
+      "step": 57320
+    },
+    {
+      "epoch": 53.47,
+      "learning_rate": 4.331576492537314e-05,
+      "loss": 0.0091,
+      "step": 57324
+    },
+    {
+      "epoch": 53.48,
+      "learning_rate": 4.3315298507462687e-05,
+      "loss": 0.0,
+      "step": 57328
+    },
+    {
+      "epoch": 53.48,
+      "learning_rate": 4.331483208955224e-05,
+      "loss": 0.0,
+      "step": 57332
+    },
+    {
+      "epoch": 53.49,
+      "learning_rate": 4.3314365671641796e-05,
+      "loss": 0.0,
+      "step": 57336
+    },
+    {
+      "epoch": 53.49,
+      "learning_rate": 4.3313899253731345e-05,
+      "loss": 0.0,
+      "step": 57340
+    },
+    {
+      "epoch": 53.49,
+      "learning_rate": 4.33134328358209e-05,
+      "loss": 0.0002,
+      "step": 57344
+    },
+    {
+      "epoch": 53.5,
+      "learning_rate": 4.331296641791045e-05,
+      "loss": 0.0,
+      "step": 57348
+    },
+    {
+      "epoch": 53.5,
+      "learning_rate": 4.33125e-05,
+      "loss": 0.0,
+      "step": 57352
+    },
+    {
+      "epoch": 53.5,
+      "learning_rate": 4.331203358208956e-05,
+      "loss": 0.0022,
+      "step": 57356
+    },
+    {
+      "epoch": 53.51,
+      "learning_rate": 4.3311567164179106e-05,
+      "loss": 0.0,
+      "step": 57360
+    },
+    {
+      "epoch": 53.51,
+      "learning_rate": 4.3311100746268654e-05,
+      "loss": 0.0,
+      "step": 57364
+    },
+    {
+      "epoch": 53.51,
+      "learning_rate": 4.3310634328358215e-05,
+      "loss": 0.0,
+      "step": 57368
+    },
+    {
+      "epoch": 53.52,
+      "learning_rate": 4.3310167910447763e-05,
+      "loss": 0.0001,
+      "step": 57372
+    },
+    {
+      "epoch": 53.52,
+      "learning_rate": 4.330970149253731e-05,
+      "loss": 0.0,
+      "step": 57376
+    },
+    {
+      "epoch": 53.53,
+      "learning_rate": 4.3309235074626866e-05,
+      "loss": 0.0,
+      "step": 57380
+    },
+    {
+      "epoch": 53.53,
+      "learning_rate": 4.330876865671642e-05,
+      "loss": 0.0,
+      "step": 57384
+    },
+    {
+      "epoch": 53.53,
+      "learning_rate": 4.330830223880597e-05,
+      "loss": 0.002,
+      "step": 57388
+    },
+    {
+      "epoch": 53.54,
+      "learning_rate": 4.3307835820895524e-05,
+      "loss": 0.0001,
+      "step": 57392
+    },
+    {
+      "epoch": 53.54,
+      "learning_rate": 4.330736940298508e-05,
+      "loss": 0.0001,
+      "step": 57396
+    },
+    {
+      "epoch": 53.54,
+      "learning_rate": 4.330690298507463e-05,
+      "loss": 0.0,
+      "step": 57400
+    },
+    {
+      "epoch": 53.55,
+      "learning_rate": 4.330643656716418e-05,
+      "loss": 0.0,
+      "step": 57404
+    },
+    {
+      "epoch": 53.55,
+      "learning_rate": 4.330597014925373e-05,
+      "loss": 0.0,
+      "step": 57408
+    },
+    {
+      "epoch": 53.56,
+      "learning_rate": 4.3305503731343285e-05,
+      "loss": 0.0021,
+      "step": 57412
+    },
+    {
+      "epoch": 53.56,
+      "learning_rate": 4.330503731343284e-05,
+      "loss": 0.0,
+      "step": 57416
+    },
+    {
+      "epoch": 53.56,
+      "learning_rate": 4.330457089552239e-05,
+      "loss": 0.0008,
+      "step": 57420
+    },
+    {
+      "epoch": 53.57,
+      "learning_rate": 4.3304104477611937e-05,
+      "loss": 0.0001,
+      "step": 57424
+    },
+    {
+      "epoch": 53.57,
+      "learning_rate": 4.33036380597015e-05,
+      "loss": 0.0,
+      "step": 57428
+    },
+    {
+      "epoch": 53.57,
+      "learning_rate": 4.3303171641791046e-05,
+      "loss": 0.0,
+      "step": 57432
+    },
+    {
+      "epoch": 53.58,
+      "learning_rate": 4.3302705223880595e-05,
+      "loss": 0.0005,
+      "step": 57436
+    },
+    {
+      "epoch": 53.58,
+      "learning_rate": 4.330223880597015e-05,
+      "loss": 0.0,
+      "step": 57440
+    },
+    {
+      "epoch": 53.59,
+      "learning_rate": 4.3301772388059704e-05,
+      "loss": 0.0,
+      "step": 57444
+    },
+    {
+      "epoch": 53.59,
+      "learning_rate": 4.330130597014926e-05,
+      "loss": 0.0,
+      "step": 57448
+    },
+    {
+      "epoch": 53.59,
+      "learning_rate": 4.330083955223881e-05,
+      "loss": 0.0,
+      "step": 57452
+    },
+    {
+      "epoch": 53.6,
+      "learning_rate": 4.330037313432836e-05,
+      "loss": 0.0,
+      "step": 57456
+    },
+    {
+      "epoch": 53.6,
+      "learning_rate": 4.329990671641792e-05,
+      "loss": 0.0001,
+      "step": 57460
+    },
+    {
+      "epoch": 53.6,
+      "learning_rate": 4.3299440298507465e-05,
+      "loss": 0.0,
+      "step": 57464
+    },
+    {
+      "epoch": 53.61,
+      "learning_rate": 4.3298973880597013e-05,
+      "loss": 0.0014,
+      "step": 57468
+    },
+    {
+      "epoch": 53.61,
+      "learning_rate": 4.329850746268657e-05,
+      "loss": 0.0,
+      "step": 57472
+    },
+    {
+      "epoch": 53.62,
+      "learning_rate": 4.329804104477612e-05,
+      "loss": 0.0,
+      "step": 57476
+    },
+    {
+      "epoch": 53.62,
+      "learning_rate": 4.329757462686567e-05,
+      "loss": 0.0076,
+      "step": 57480
+    },
+    {
+      "epoch": 53.62,
+      "learning_rate": 4.3297108208955226e-05,
+      "loss": 0.0,
+      "step": 57484
+    },
+    {
+      "epoch": 53.63,
+      "learning_rate": 4.329664179104478e-05,
+      "loss": 0.0,
+      "step": 57488
+    },
+    {
+      "epoch": 53.63,
+      "learning_rate": 4.329617537313433e-05,
+      "loss": 0.0,
+      "step": 57492
+    },
+    {
+      "epoch": 53.63,
+      "learning_rate": 4.3295708955223884e-05,
+      "loss": 0.0,
+      "step": 57496
+    },
+    {
+      "epoch": 53.64,
+      "learning_rate": 4.329524253731343e-05,
+      "loss": 0.001,
+      "step": 57500
+    },
+    {
+      "epoch": 53.64,
+      "eval_exact_match": 0.7446808510638298,
+      "eval_exec": 0.7823984526112185,
+      "eval_loss": 0.4641968011856079,
+      "eval_runtime": 1155.653,
+      "eval_samples_per_second": 0.895,
+      "step": 57500
+    },
+    {
+      "epoch": 53.64,
+      "learning_rate": 4.329477611940299e-05,
+      "loss": 0.0,
+      "step": 57504
+    },
+    {
+      "epoch": 53.65,
+      "learning_rate": 4.329430970149254e-05,
+      "loss": 0.0,
+      "step": 57508
+    },
+    {
+      "epoch": 53.65,
+      "learning_rate": 4.329384328358209e-05,
+      "loss": 0.0005,
+      "step": 57512
+    },
+    {
+      "epoch": 53.65,
+      "learning_rate": 4.3293376865671645e-05,
+      "loss": 0.0,
+      "step": 57516
+    },
+    {
+      "epoch": 53.66,
+      "learning_rate": 4.32929104477612e-05,
+      "loss": 0.0051,
+      "step": 57520
+    },
+    {
+      "epoch": 53.66,
+      "learning_rate": 4.329244402985075e-05,
+      "loss": 0.0001,
+      "step": 57524
+    },
+    {
+      "epoch": 53.66,
+      "learning_rate": 4.3291977611940296e-05,
+      "loss": 0.0,
+      "step": 57528
+    },
+    {
+      "epoch": 53.67,
+      "learning_rate": 4.329151119402985e-05,
+      "loss": 0.0,
+      "step": 57532
+    },
+    {
+      "epoch": 53.67,
+      "learning_rate": 4.3291044776119406e-05,
+      "loss": 0.0,
+      "step": 57536
+    },
+    {
+      "epoch": 53.68,
+      "learning_rate": 4.3290578358208954e-05,
+      "loss": 0.0,
+      "step": 57540
+    },
+    {
+      "epoch": 53.68,
+      "learning_rate": 4.329011194029851e-05,
+      "loss": 0.0,
+      "step": 57544
+    },
+    {
+      "epoch": 53.68,
+      "learning_rate": 4.3289645522388064e-05,
+      "loss": 0.0,
+      "step": 57548
+    },
+    {
+      "epoch": 53.69,
+      "learning_rate": 4.328917910447761e-05,
+      "loss": 0.0008,
+      "step": 57552
+    },
+    {
+      "epoch": 53.69,
+      "learning_rate": 4.328871268656717e-05,
+      "loss": 0.0002,
+      "step": 57556
+    },
+    {
+      "epoch": 53.69,
+      "learning_rate": 4.3288246268656715e-05,
+      "loss": 0.0001,
+      "step": 57560
+    },
+    {
+      "epoch": 53.7,
+      "learning_rate": 4.328777985074627e-05,
+      "loss": 0.0,
+      "step": 57564
+    },
+    {
+      "epoch": 53.7,
+      "learning_rate": 4.3287313432835825e-05,
+      "loss": 0.0,
+      "step": 57568
+    },
+    {
+      "epoch": 53.71,
+      "learning_rate": 4.328684701492537e-05,
+      "loss": 0.0,
+      "step": 57572
+    },
+    {
+      "epoch": 53.71,
+      "learning_rate": 4.328638059701493e-05,
+      "loss": 0.0,
+      "step": 57576
+    },
+    {
+      "epoch": 53.71,
+      "learning_rate": 4.328591417910448e-05,
+      "loss": 0.0015,
+      "step": 57580
+    },
+    {
+      "epoch": 53.72,
+      "learning_rate": 4.328544776119403e-05,
+      "loss": 0.0,
+      "step": 57584
+    },
+    {
+      "epoch": 53.72,
+      "learning_rate": 4.328498134328358e-05,
+      "loss": 0.0013,
+      "step": 57588
+    },
+    {
+      "epoch": 53.72,
+      "learning_rate": 4.3284514925373134e-05,
+      "loss": 0.0,
+      "step": 57592
+    },
+    {
+      "epoch": 53.73,
+      "learning_rate": 4.328404850746269e-05,
+      "loss": 0.0049,
+      "step": 57596
+    },
+    {
+      "epoch": 53.73,
+      "learning_rate": 4.328358208955224e-05,
+      "loss": 0.0,
+      "step": 57600
+    },
+    {
+      "epoch": 53.73,
+      "learning_rate": 4.328311567164179e-05,
+      "loss": 0.0,
+      "step": 57604
+    },
+    {
+      "epoch": 53.74,
+      "learning_rate": 4.328264925373135e-05,
+      "loss": 0.0,
+      "step": 57608
+    },
+    {
+      "epoch": 53.74,
+      "learning_rate": 4.32821828358209e-05,
+      "loss": 0.0,
+      "step": 57612
+    },
+    {
+      "epoch": 53.75,
+      "learning_rate": 4.328171641791045e-05,
+      "loss": 0.0001,
+      "step": 57616
+    },
+    {
+      "epoch": 53.75,
+      "learning_rate": 4.328125e-05,
+      "loss": 0.0006,
+      "step": 57620
+    },
+    {
+      "epoch": 53.75,
+      "learning_rate": 4.328078358208956e-05,
+      "loss": 0.0,
+      "step": 57624
+    },
+    {
+      "epoch": 53.76,
+      "learning_rate": 4.328031716417911e-05,
+      "loss": 0.0,
+      "step": 57628
+    },
+    {
+      "epoch": 53.76,
+      "learning_rate": 4.3279850746268656e-05,
+      "loss": 0.0004,
+      "step": 57632
+    },
+    {
+      "epoch": 53.76,
+      "learning_rate": 4.327938432835821e-05,
+      "loss": 0.0,
+      "step": 57636
+    },
+    {
+      "epoch": 53.77,
+      "learning_rate": 4.3278917910447766e-05,
+      "loss": 0.0,
+      "step": 57640
+    },
+    {
+      "epoch": 53.77,
+      "learning_rate": 4.3278451492537314e-05,
+      "loss": 0.0,
+      "step": 57644
+    },
+    {
+      "epoch": 53.78,
+      "learning_rate": 4.327798507462687e-05,
+      "loss": 0.0,
+      "step": 57648
+    },
+    {
+      "epoch": 53.78,
+      "learning_rate": 4.327751865671642e-05,
+      "loss": 0.0,
+      "step": 57652
+    },
+    {
+      "epoch": 53.78,
+      "learning_rate": 4.327705223880597e-05,
+      "loss": 0.0,
+      "step": 57656
+    },
+    {
+      "epoch": 53.79,
+      "learning_rate": 4.327658582089553e-05,
+      "loss": 0.0,
+      "step": 57660
+    },
+    {
+      "epoch": 53.79,
+      "learning_rate": 4.3276119402985075e-05,
+      "loss": 0.0001,
+      "step": 57664
+    },
+    {
+      "epoch": 53.79,
+      "learning_rate": 4.327565298507463e-05,
+      "loss": 0.0001,
+      "step": 57668
+    },
+    {
+      "epoch": 53.8,
+      "learning_rate": 4.3275186567164185e-05,
+      "loss": 0.0001,
+      "step": 57672
+    },
+    {
+      "epoch": 53.8,
+      "learning_rate": 4.327472014925373e-05,
+      "loss": 0.0,
+      "step": 57676
+    },
+    {
+      "epoch": 53.81,
+      "learning_rate": 4.327425373134328e-05,
+      "loss": 0.0,
+      "step": 57680
+    },
+    {
+      "epoch": 53.81,
+      "learning_rate": 4.327378731343284e-05,
+      "loss": 0.0,
+      "step": 57684
+    },
+    {
+      "epoch": 53.81,
+      "learning_rate": 4.327332089552239e-05,
+      "loss": 0.0,
+      "step": 57688
+    },
+    {
+      "epoch": 53.82,
+      "learning_rate": 4.327285447761194e-05,
+      "loss": 0.0,
+      "step": 57692
+    },
+    {
+      "epoch": 53.82,
+      "learning_rate": 4.3272388059701494e-05,
+      "loss": 0.0002,
+      "step": 57696
+    },
+    {
+      "epoch": 53.82,
+      "learning_rate": 4.327192164179105e-05,
+      "loss": 0.0,
+      "step": 57700
+    },
+    {
+      "epoch": 53.83,
+      "learning_rate": 4.32714552238806e-05,
+      "loss": 0.0,
+      "step": 57704
+    },
+    {
+      "epoch": 53.83,
+      "learning_rate": 4.327098880597015e-05,
+      "loss": 0.0,
+      "step": 57708
+    },
+    {
+      "epoch": 53.84,
+      "learning_rate": 4.32705223880597e-05,
+      "loss": 0.0,
+      "step": 57712
+    },
+    {
+      "epoch": 53.84,
+      "learning_rate": 4.3270055970149255e-05,
+      "loss": 0.0,
+      "step": 57716
+    },
+    {
+      "epoch": 53.84,
+      "learning_rate": 4.326958955223881e-05,
+      "loss": 0.0,
+      "step": 57720
+    },
+    {
+      "epoch": 53.85,
+      "learning_rate": 4.326912313432836e-05,
+      "loss": 0.0,
+      "step": 57724
+    },
+    {
+      "epoch": 53.85,
+      "learning_rate": 4.326865671641791e-05,
+      "loss": 0.0001,
+      "step": 57728
+    },
+    {
+      "epoch": 53.85,
+      "learning_rate": 4.326819029850747e-05,
+      "loss": 0.003,
+      "step": 57732
+    },
+    {
+      "epoch": 53.86,
+      "learning_rate": 4.3267723880597016e-05,
+      "loss": 0.0,
+      "step": 57736
+    },
+    {
+      "epoch": 53.86,
+      "learning_rate": 4.3267257462686564e-05,
+      "loss": 0.0003,
+      "step": 57740
+    },
+    {
+      "epoch": 53.87,
+      "learning_rate": 4.3266791044776126e-05,
+      "loss": 0.0,
+      "step": 57744
+    },
+    {
+      "epoch": 53.87,
+      "learning_rate": 4.3266324626865674e-05,
+      "loss": 0.0,
+      "step": 57748
+    },
+    {
+      "epoch": 53.87,
+      "learning_rate": 4.326585820895522e-05,
+      "loss": 0.0001,
+      "step": 57752
+    },
+    {
+      "epoch": 53.88,
+      "learning_rate": 4.326539179104478e-05,
+      "loss": 0.0,
+      "step": 57756
+    },
+    {
+      "epoch": 53.88,
+      "learning_rate": 4.326492537313433e-05,
+      "loss": 0.0001,
+      "step": 57760
+    },
+    {
+      "epoch": 53.88,
+      "learning_rate": 4.326445895522388e-05,
+      "loss": 0.0002,
+      "step": 57764
+    },
+    {
+      "epoch": 53.89,
+      "learning_rate": 4.3263992537313435e-05,
+      "loss": 0.0003,
+      "step": 57768
+    },
+    {
+      "epoch": 53.89,
+      "learning_rate": 4.326352611940298e-05,
+      "loss": 0.0,
+      "step": 57772
+    },
+    {
+      "epoch": 53.9,
+      "learning_rate": 4.3263059701492545e-05,
+      "loss": 0.0,
+      "step": 57776
+    },
+    {
+      "epoch": 53.9,
+      "learning_rate": 4.326259328358209e-05,
+      "loss": 0.0,
+      "step": 57780
+    },
+    {
+      "epoch": 53.9,
+      "learning_rate": 4.326212686567164e-05,
+      "loss": 0.0001,
+      "step": 57784
+    },
+    {
+      "epoch": 53.91,
+      "learning_rate": 4.3261660447761196e-05,
+      "loss": 0.0,
+      "step": 57788
+    },
+    {
+      "epoch": 53.91,
+      "learning_rate": 4.326119402985075e-05,
+      "loss": 0.0,
+      "step": 57792
+    },
+    {
+      "epoch": 53.91,
+      "learning_rate": 4.32607276119403e-05,
+      "loss": 0.0,
+      "step": 57796
+    },
+    {
+      "epoch": 53.92,
+      "learning_rate": 4.3260261194029854e-05,
+      "loss": 0.0,
+      "step": 57800
+    },
+    {
+      "epoch": 53.92,
+      "learning_rate": 4.325979477611941e-05,
+      "loss": 0.0,
+      "step": 57804
+    },
+    {
+      "epoch": 53.93,
+      "learning_rate": 4.325932835820896e-05,
+      "loss": 0.0,
+      "step": 57808
+    },
+    {
+      "epoch": 53.93,
+      "learning_rate": 4.325886194029851e-05,
+      "loss": 0.0,
+      "step": 57812
+    },
+    {
+      "epoch": 53.93,
+      "learning_rate": 4.325839552238806e-05,
+      "loss": 0.0,
+      "step": 57816
+    },
+    {
+      "epoch": 53.94,
+      "learning_rate": 4.3257929104477615e-05,
+      "loss": 0.0,
+      "step": 57820
+    },
+    {
+      "epoch": 53.94,
+      "learning_rate": 4.325746268656717e-05,
+      "loss": 0.0,
+      "step": 57824
+    },
+    {
+      "epoch": 53.94,
+      "learning_rate": 4.325699626865672e-05,
+      "loss": 0.0001,
+      "step": 57828
+    },
+    {
+      "epoch": 53.95,
+      "learning_rate": 4.3256529850746266e-05,
+      "loss": 0.0,
+      "step": 57832
+    },
+    {
+      "epoch": 53.95,
+      "learning_rate": 4.325606343283583e-05,
+      "loss": 0.0,
+      "step": 57836
+    },
+    {
+      "epoch": 53.96,
+      "learning_rate": 4.3255597014925376e-05,
+      "loss": 0.0,
+      "step": 57840
+    },
+    {
+      "epoch": 53.96,
+      "learning_rate": 4.3255130597014924e-05,
+      "loss": 0.0,
+      "step": 57844
+    },
+    {
+      "epoch": 53.96,
+      "learning_rate": 4.325466417910448e-05,
+      "loss": 0.0,
+      "step": 57848
+    },
+    {
+      "epoch": 53.97,
+      "learning_rate": 4.3254197761194034e-05,
+      "loss": 0.002,
+      "step": 57852
+    },
+    {
+      "epoch": 53.97,
+      "learning_rate": 4.325373134328358e-05,
+      "loss": 0.0,
+      "step": 57856
+    },
+    {
+      "epoch": 53.97,
+      "learning_rate": 4.325326492537314e-05,
+      "loss": 0.0,
+      "step": 57860
+    },
+    {
+      "epoch": 53.98,
+      "learning_rate": 4.325279850746269e-05,
+      "loss": 0.0,
+      "step": 57864
+    },
+    {
+      "epoch": 53.98,
+      "learning_rate": 4.325233208955224e-05,
+      "loss": 0.0,
+      "step": 57868
+    },
+    {
+      "epoch": 53.98,
+      "learning_rate": 4.3251865671641795e-05,
+      "loss": 0.0003,
+      "step": 57872
+    },
+    {
+      "epoch": 53.99,
+      "learning_rate": 4.325139925373134e-05,
+      "loss": 0.0004,
+      "step": 57876
+    },
+    {
+      "epoch": 53.99,
+      "learning_rate": 4.32509328358209e-05,
+      "loss": 0.0002,
+      "step": 57880
+    },
+    {
+      "epoch": 54.0,
+      "learning_rate": 4.325046641791045e-05,
+      "loss": 0.0,
+      "step": 57884
+    },
+    {
+      "epoch": 54.0,
+      "learning_rate": 4.325e-05,
+      "loss": 0.0001,
+      "step": 57888
+    },
+    {
+      "epoch": 54.0,
+      "learning_rate": 4.324953358208955e-05,
+      "loss": 0.0,
+      "step": 57892
+    },
+    {
+      "epoch": 54.01,
+      "learning_rate": 4.324906716417911e-05,
+      "loss": 0.0,
+      "step": 57896
+    },
+    {
+      "epoch": 54.01,
+      "learning_rate": 4.324860074626866e-05,
+      "loss": 0.0,
+      "step": 57900
+    },
+    {
+      "epoch": 54.01,
+      "learning_rate": 4.324813432835821e-05,
+      "loss": 0.0,
+      "step": 57904
+    },
+    {
+      "epoch": 54.02,
+      "learning_rate": 4.324766791044776e-05,
+      "loss": 0.0,
+      "step": 57908
+    },
+    {
+      "epoch": 54.02,
+      "learning_rate": 4.324720149253732e-05,
+      "loss": 0.0,
+      "step": 57912
+    },
+    {
+      "epoch": 54.03,
+      "learning_rate": 4.3246735074626865e-05,
+      "loss": 0.0001,
+      "step": 57916
+    },
+    {
+      "epoch": 54.03,
+      "learning_rate": 4.324626865671642e-05,
+      "loss": 0.0,
+      "step": 57920
+    },
+    {
+      "epoch": 54.03,
+      "learning_rate": 4.324580223880597e-05,
+      "loss": 0.0004,
+      "step": 57924
+    },
+    {
+      "epoch": 54.04,
+      "learning_rate": 4.324533582089552e-05,
+      "loss": 0.0,
+      "step": 57928
+    },
+    {
+      "epoch": 54.04,
+      "learning_rate": 4.324486940298508e-05,
+      "loss": 0.0,
+      "step": 57932
+    },
+    {
+      "epoch": 54.04,
+      "learning_rate": 4.3244402985074626e-05,
+      "loss": 0.0,
+      "step": 57936
+    },
+    {
+      "epoch": 54.05,
+      "learning_rate": 4.324393656716418e-05,
+      "loss": 0.0,
+      "step": 57940
+    },
+    {
+      "epoch": 54.05,
+      "learning_rate": 4.3243470149253736e-05,
+      "loss": 0.0,
+      "step": 57944
+    },
+    {
+      "epoch": 54.06,
+      "learning_rate": 4.3243003731343284e-05,
+      "loss": 0.0001,
+      "step": 57948
+    },
+    {
+      "epoch": 54.06,
+      "learning_rate": 4.324253731343284e-05,
+      "loss": 0.0013,
+      "step": 57952
+    },
+    {
+      "epoch": 54.06,
+      "learning_rate": 4.3242070895522394e-05,
+      "loss": 0.0,
+      "step": 57956
+    },
+    {
+      "epoch": 54.07,
+      "learning_rate": 4.324160447761194e-05,
+      "loss": 0.0001,
+      "step": 57960
+    },
+    {
+      "epoch": 54.07,
+      "learning_rate": 4.32411380597015e-05,
+      "loss": 0.0001,
+      "step": 57964
+    },
+    {
+      "epoch": 54.07,
+      "learning_rate": 4.3240671641791045e-05,
+      "loss": 0.0,
+      "step": 57968
+    },
+    {
+      "epoch": 54.08,
+      "learning_rate": 4.32402052238806e-05,
+      "loss": 0.0,
+      "step": 57972
+    },
+    {
+      "epoch": 54.08,
+      "learning_rate": 4.3239738805970155e-05,
+      "loss": 0.0,
+      "step": 57976
+    },
+    {
+      "epoch": 54.09,
+      "learning_rate": 4.32392723880597e-05,
+      "loss": 0.0001,
+      "step": 57980
+    },
+    {
+      "epoch": 54.09,
+      "learning_rate": 4.323880597014925e-05,
+      "loss": 0.0001,
+      "step": 57984
+    },
+    {
+      "epoch": 54.09,
+      "learning_rate": 4.323833955223881e-05,
+      "loss": 0.0,
+      "step": 57988
+    },
+    {
+      "epoch": 54.1,
+      "learning_rate": 4.323787313432836e-05,
+      "loss": 0.0,
+      "step": 57992
+    },
+    {
+      "epoch": 54.1,
+      "learning_rate": 4.323740671641791e-05,
+      "loss": 0.0002,
+      "step": 57996
+    },
+    {
+      "epoch": 54.1,
+      "learning_rate": 4.3236940298507464e-05,
+      "loss": 0.0001,
+      "step": 58000
+    },
+    {
+      "epoch": 54.1,
+      "eval_exact_match": 0.7427466150870407,
+      "eval_exec": 0.7785299806576402,
+      "eval_loss": 0.49178797006607056,
+      "eval_runtime": 1102.7711,
+      "eval_samples_per_second": 0.938,
+      "step": 58000
+    },
+    {
+      "epoch": 54.11,
+      "learning_rate": 4.323647388059702e-05,
+      "loss": 0.0,
+      "step": 58004
+    },
+    {
+      "epoch": 54.11,
+      "learning_rate": 4.323600746268657e-05,
+      "loss": 0.0001,
+      "step": 58008
+    },
+    {
+      "epoch": 54.12,
+      "learning_rate": 4.323554104477612e-05,
+      "loss": 0.0,
+      "step": 58012
+    },
+    {
+      "epoch": 54.12,
+      "learning_rate": 4.323507462686568e-05,
+      "loss": 0.0008,
+      "step": 58016
+    },
+    {
+      "epoch": 54.12,
+      "learning_rate": 4.3234608208955225e-05,
+      "loss": 0.0,
+      "step": 58020
+    },
+    {
+      "epoch": 54.13,
+      "learning_rate": 4.323414179104478e-05,
+      "loss": 0.0,
+      "step": 58024
+    },
+    {
+      "epoch": 54.13,
+      "learning_rate": 4.323367537313433e-05,
+      "loss": 0.0,
+      "step": 58028
+    },
+    {
+      "epoch": 54.13,
+      "learning_rate": 4.323320895522388e-05,
+      "loss": 0.0,
+      "step": 58032
+    },
+    {
+      "epoch": 54.14,
+      "learning_rate": 4.323274253731344e-05,
+      "loss": 0.0,
+      "step": 58036
+    },
+    {
+      "epoch": 54.14,
+      "learning_rate": 4.3232276119402986e-05,
+      "loss": 0.0,
+      "step": 58040
+    },
+    {
+      "epoch": 54.15,
+      "learning_rate": 4.3231809701492534e-05,
+      "loss": 0.0,
+      "step": 58044
+    },
+    {
+      "epoch": 54.15,
+      "learning_rate": 4.3231343283582096e-05,
+      "loss": 0.0005,
+      "step": 58048
+    },
+    {
+      "epoch": 54.15,
+      "learning_rate": 4.3230876865671644e-05,
+      "loss": 0.0,
+      "step": 58052
+    },
+    {
+      "epoch": 54.16,
+      "learning_rate": 4.323041044776119e-05,
+      "loss": 0.0002,
+      "step": 58056
+    },
+    {
+      "epoch": 54.16,
+      "learning_rate": 4.322994402985075e-05,
+      "loss": 0.0,
+      "step": 58060
+    },
+    {
+      "epoch": 54.16,
+      "learning_rate": 4.32294776119403e-05,
+      "loss": 0.0003,
+      "step": 58064
+    },
+    {
+      "epoch": 54.17,
+      "learning_rate": 4.322901119402985e-05,
+      "loss": 0.0,
+      "step": 58068
+    },
+    {
+      "epoch": 54.17,
+      "learning_rate": 4.3228544776119405e-05,
+      "loss": 0.0,
+      "step": 58072
+    },
+    {
+      "epoch": 54.18,
+      "learning_rate": 4.322807835820896e-05,
+      "loss": 0.0,
+      "step": 58076
+    },
+    {
+      "epoch": 54.18,
+      "learning_rate": 4.322761194029851e-05,
+      "loss": 0.0043,
+      "step": 58080
+    },
+    {
+      "epoch": 54.18,
+      "learning_rate": 4.322714552238806e-05,
+      "loss": 0.0,
+      "step": 58084
+    },
+    {
+      "epoch": 54.19,
+      "learning_rate": 4.322667910447761e-05,
+      "loss": 0.0,
+      "step": 58088
+    },
+    {
+      "epoch": 54.19,
+      "learning_rate": 4.3226212686567166e-05,
+      "loss": 0.0,
+      "step": 58092
+    },
+    {
+      "epoch": 54.19,
+      "learning_rate": 4.322574626865672e-05,
+      "loss": 0.0,
+      "step": 58096
+    },
+    {
+      "epoch": 54.2,
+      "learning_rate": 4.322527985074627e-05,
+      "loss": 0.0,
+      "step": 58100
+    },
+    {
+      "epoch": 54.2,
+      "learning_rate": 4.3224813432835824e-05,
+      "loss": 0.0,
+      "step": 58104
+    },
+    {
+      "epoch": 54.21,
+      "learning_rate": 4.322434701492538e-05,
+      "loss": 0.0,
+      "step": 58108
+    },
+    {
+      "epoch": 54.21,
+      "learning_rate": 4.322388059701493e-05,
+      "loss": 0.0,
+      "step": 58112
+    },
+    {
+      "epoch": 54.21,
+      "learning_rate": 4.322341417910448e-05,
+      "loss": 0.0003,
+      "step": 58116
+    },
+    {
+      "epoch": 54.22,
+      "learning_rate": 4.322294776119403e-05,
+      "loss": 0.0082,
+      "step": 58120
+    },
+    {
+      "epoch": 54.22,
+      "learning_rate": 4.3222481343283585e-05,
+      "loss": 0.0,
+      "step": 58124
+    },
+    {
+      "epoch": 54.22,
+      "learning_rate": 4.322201492537314e-05,
+      "loss": 0.0003,
+      "step": 58128
+    },
+    {
+      "epoch": 54.23,
+      "learning_rate": 4.322154850746269e-05,
+      "loss": 0.0063,
+      "step": 58132
+    },
+    {
+      "epoch": 54.23,
+      "learning_rate": 4.322108208955224e-05,
+      "loss": 0.0,
+      "step": 58136
+    },
+    {
+      "epoch": 54.24,
+      "learning_rate": 4.32206156716418e-05,
+      "loss": 0.0001,
+      "step": 58140
+    },
+    {
+      "epoch": 54.24,
+      "learning_rate": 4.3220149253731346e-05,
+      "loss": 0.0002,
+      "step": 58144
+    },
+    {
+      "epoch": 54.24,
+      "learning_rate": 4.3219682835820894e-05,
+      "loss": 0.0,
+      "step": 58148
+    },
+    {
+      "epoch": 54.25,
+      "learning_rate": 4.321921641791045e-05,
+      "loss": 0.0004,
+      "step": 58152
+    },
+    {
+      "epoch": 54.25,
+      "learning_rate": 4.3218750000000004e-05,
+      "loss": 0.0001,
+      "step": 58156
+    },
+    {
+      "epoch": 54.25,
+      "learning_rate": 4.321828358208955e-05,
+      "loss": 0.0,
+      "step": 58160
+    },
+    {
+      "epoch": 54.26,
+      "learning_rate": 4.321781716417911e-05,
+      "loss": 0.0,
+      "step": 58164
+    },
+    {
+      "epoch": 54.26,
+      "learning_rate": 4.321735074626866e-05,
+      "loss": 0.0,
+      "step": 58168
+    },
+    {
+      "epoch": 54.26,
+      "learning_rate": 4.321688432835821e-05,
+      "loss": 0.0,
+      "step": 58172
+    },
+    {
+      "epoch": 54.27,
+      "learning_rate": 4.3216417910447765e-05,
+      "loss": 0.0,
+      "step": 58176
+    },
+    {
+      "epoch": 54.27,
+      "learning_rate": 4.321595149253731e-05,
+      "loss": 0.0,
+      "step": 58180
+    },
+    {
+      "epoch": 54.28,
+      "learning_rate": 4.321548507462687e-05,
+      "loss": 0.0,
+      "step": 58184
+    },
+    {
+      "epoch": 54.28,
+      "learning_rate": 4.321501865671642e-05,
+      "loss": 0.0001,
+      "step": 58188
+    },
+    {
+      "epoch": 54.28,
+      "learning_rate": 4.321455223880597e-05,
+      "loss": 0.0,
+      "step": 58192
+    },
+    {
+      "epoch": 54.29,
+      "learning_rate": 4.3214085820895526e-05,
+      "loss": 0.001,
+      "step": 58196
+    },
+    {
+      "epoch": 54.29,
+      "learning_rate": 4.321361940298508e-05,
+      "loss": 0.0019,
+      "step": 58200
+    },
+    {
+      "epoch": 54.29,
+      "learning_rate": 4.321315298507463e-05,
+      "loss": 0.0,
+      "step": 58204
+    },
+    {
+      "epoch": 54.3,
+      "learning_rate": 4.321268656716418e-05,
+      "loss": 0.0,
+      "step": 58208
+    },
+    {
+      "epoch": 54.3,
+      "learning_rate": 4.321222014925373e-05,
+      "loss": 0.0,
+      "step": 58212
+    },
+    {
+      "epoch": 54.31,
+      "learning_rate": 4.3211753731343287e-05,
+      "loss": 0.0026,
+      "step": 58216
+    },
+    {
+      "epoch": 54.31,
+      "learning_rate": 4.3211287313432835e-05,
+      "loss": 0.0001,
+      "step": 58220
+    },
+    {
+      "epoch": 54.31,
+      "learning_rate": 4.321082089552239e-05,
+      "loss": 0.0003,
+      "step": 58224
+    },
+    {
+      "epoch": 54.32,
+      "learning_rate": 4.3210354477611944e-05,
+      "loss": 0.0,
+      "step": 58228
+    },
+    {
+      "epoch": 54.32,
+      "learning_rate": 4.320988805970149e-05,
+      "loss": 0.0001,
+      "step": 58232
+    },
+    {
+      "epoch": 54.32,
+      "learning_rate": 4.320942164179105e-05,
+      "loss": 0.0054,
+      "step": 58236
+    },
+    {
+      "epoch": 54.33,
+      "learning_rate": 4.3208955223880596e-05,
+      "loss": 0.0,
+      "step": 58240
+    },
+    {
+      "epoch": 54.33,
+      "learning_rate": 4.320848880597015e-05,
+      "loss": 0.0,
+      "step": 58244
+    },
+    {
+      "epoch": 54.34,
+      "learning_rate": 4.3208022388059705e-05,
+      "loss": 0.0,
+      "step": 58248
+    },
+    {
+      "epoch": 54.34,
+      "learning_rate": 4.3207555970149254e-05,
+      "loss": 0.0005,
+      "step": 58252
+    },
+    {
+      "epoch": 54.34,
+      "learning_rate": 4.320708955223881e-05,
+      "loss": 0.0003,
+      "step": 58256
+    },
+    {
+      "epoch": 54.35,
+      "learning_rate": 4.3206623134328363e-05,
+      "loss": 0.0,
+      "step": 58260
+    },
+    {
+      "epoch": 54.35,
+      "learning_rate": 4.320615671641791e-05,
+      "loss": 0.0,
+      "step": 58264
+    },
+    {
+      "epoch": 54.35,
+      "learning_rate": 4.3205690298507466e-05,
+      "loss": 0.0,
+      "step": 58268
+    },
+    {
+      "epoch": 54.36,
+      "learning_rate": 4.3205223880597015e-05,
+      "loss": 0.0,
+      "step": 58272
+    },
+    {
+      "epoch": 54.36,
+      "learning_rate": 4.320475746268657e-05,
+      "loss": 0.0,
+      "step": 58276
+    },
+    {
+      "epoch": 54.37,
+      "learning_rate": 4.3204291044776124e-05,
+      "loss": 0.0,
+      "step": 58280
+    },
+    {
+      "epoch": 54.37,
+      "learning_rate": 4.320382462686567e-05,
+      "loss": 0.0,
+      "step": 58284
+    },
+    {
+      "epoch": 54.37,
+      "learning_rate": 4.320335820895523e-05,
+      "loss": 0.0,
+      "step": 58288
+    },
+    {
+      "epoch": 54.38,
+      "learning_rate": 4.320289179104478e-05,
+      "loss": 0.0004,
+      "step": 58292
+    },
+    {
+      "epoch": 54.38,
+      "learning_rate": 4.320242537313433e-05,
+      "loss": 0.0,
+      "step": 58296
+    },
+    {
+      "epoch": 54.38,
+      "learning_rate": 4.320195895522388e-05,
+      "loss": 0.0,
+      "step": 58300
+    },
+    {
+      "epoch": 54.39,
+      "learning_rate": 4.320149253731344e-05,
+      "loss": 0.011,
+      "step": 58304
+    },
+    {
+      "epoch": 54.39,
+      "learning_rate": 4.320102611940299e-05,
+      "loss": 0.0052,
+      "step": 58308
+    },
+    {
+      "epoch": 54.4,
+      "learning_rate": 4.3200559701492537e-05,
+      "loss": 0.0001,
+      "step": 58312
+    },
+    {
+      "epoch": 54.4,
+      "learning_rate": 4.320009328358209e-05,
+      "loss": 0.0,
+      "step": 58316
+    },
+    {
+      "epoch": 54.4,
+      "learning_rate": 4.3199626865671646e-05,
+      "loss": 0.0,
+      "step": 58320
+    },
+    {
+      "epoch": 54.41,
+      "learning_rate": 4.3199160447761195e-05,
+      "loss": 0.0001,
+      "step": 58324
+    },
+    {
+      "epoch": 54.41,
+      "learning_rate": 4.319869402985075e-05,
+      "loss": 0.0,
+      "step": 58328
+    },
+    {
+      "epoch": 54.41,
+      "learning_rate": 4.31982276119403e-05,
+      "loss": 0.0003,
+      "step": 58332
+    },
+    {
+      "epoch": 54.42,
+      "learning_rate": 4.319776119402985e-05,
+      "loss": 0.0001,
+      "step": 58336
+    },
+    {
+      "epoch": 54.42,
+      "learning_rate": 4.319729477611941e-05,
+      "loss": 0.001,
+      "step": 58340
+    },
+    {
+      "epoch": 54.43,
+      "learning_rate": 4.3196828358208955e-05,
+      "loss": 0.0021,
+      "step": 58344
+    },
+    {
+      "epoch": 54.43,
+      "learning_rate": 4.319636194029851e-05,
+      "loss": 0.0003,
+      "step": 58348
+    },
+    {
+      "epoch": 54.43,
+      "learning_rate": 4.3195895522388065e-05,
+      "loss": 0.0,
+      "step": 58352
+    },
+    {
+      "epoch": 54.44,
+      "learning_rate": 4.3195429104477613e-05,
+      "loss": 0.0001,
+      "step": 58356
+    },
+    {
+      "epoch": 54.44,
+      "learning_rate": 4.319496268656716e-05,
+      "loss": 0.0001,
+      "step": 58360
+    },
+    {
+      "epoch": 54.44,
+      "learning_rate": 4.319449626865672e-05,
+      "loss": 0.0,
+      "step": 58364
+    },
+    {
+      "epoch": 54.45,
+      "learning_rate": 4.319402985074627e-05,
+      "loss": 0.0,
+      "step": 58368
+    },
+    {
+      "epoch": 54.45,
+      "learning_rate": 4.319356343283582e-05,
+      "loss": 0.0,
+      "step": 58372
+    },
+    {
+      "epoch": 54.46,
+      "learning_rate": 4.3193097014925374e-05,
+      "loss": 0.0016,
+      "step": 58376
+    },
+    {
+      "epoch": 54.46,
+      "learning_rate": 4.319263059701493e-05,
+      "loss": 0.0,
+      "step": 58380
+    },
+    {
+      "epoch": 54.46,
+      "learning_rate": 4.319216417910448e-05,
+      "loss": 0.0001,
+      "step": 58384
+    },
+    {
+      "epoch": 54.47,
+      "learning_rate": 4.319169776119403e-05,
+      "loss": 0.0,
+      "step": 58388
+    },
+    {
+      "epoch": 54.47,
+      "learning_rate": 4.319123134328358e-05,
+      "loss": 0.0,
+      "step": 58392
+    },
+    {
+      "epoch": 54.47,
+      "learning_rate": 4.3190764925373135e-05,
+      "loss": 0.0,
+      "step": 58396
+    },
+    {
+      "epoch": 54.48,
+      "learning_rate": 4.319029850746269e-05,
+      "loss": 0.0001,
+      "step": 58400
+    },
+    {
+      "epoch": 54.48,
+      "learning_rate": 4.318983208955224e-05,
+      "loss": 0.0,
+      "step": 58404
+    },
+    {
+      "epoch": 54.49,
+      "learning_rate": 4.318936567164179e-05,
+      "loss": 0.0001,
+      "step": 58408
+    },
+    {
+      "epoch": 54.49,
+      "learning_rate": 4.318889925373135e-05,
+      "loss": 0.0004,
+      "step": 58412
+    },
+    {
+      "epoch": 54.49,
+      "learning_rate": 4.3188432835820896e-05,
+      "loss": 0.0004,
+      "step": 58416
+    },
+    {
+      "epoch": 54.5,
+      "learning_rate": 4.3187966417910445e-05,
+      "loss": 0.0,
+      "step": 58420
+    },
+    {
+      "epoch": 54.5,
+      "learning_rate": 4.3187500000000006e-05,
+      "loss": 0.0001,
+      "step": 58424
+    },
+    {
+      "epoch": 54.5,
+      "learning_rate": 4.3187033582089554e-05,
+      "loss": 0.0,
+      "step": 58428
+    },
+    {
+      "epoch": 54.51,
+      "learning_rate": 4.318656716417911e-05,
+      "loss": 0.0004,
+      "step": 58432
+    },
+    {
+      "epoch": 54.51,
+      "learning_rate": 4.318610074626866e-05,
+      "loss": 0.0001,
+      "step": 58436
+    },
+    {
+      "epoch": 54.51,
+      "learning_rate": 4.318563432835821e-05,
+      "loss": 0.0,
+      "step": 58440
+    },
+    {
+      "epoch": 54.52,
+      "learning_rate": 4.318516791044777e-05,
+      "loss": 0.0,
+      "step": 58444
+    },
+    {
+      "epoch": 54.52,
+      "learning_rate": 4.3184701492537315e-05,
+      "loss": 0.0008,
+      "step": 58448
+    },
+    {
+      "epoch": 54.53,
+      "learning_rate": 4.3184235074626863e-05,
+      "loss": 0.0001,
+      "step": 58452
+    },
+    {
+      "epoch": 54.53,
+      "learning_rate": 4.3183768656716425e-05,
+      "loss": 0.0,
+      "step": 58456
+    },
+    {
+      "epoch": 54.53,
+      "learning_rate": 4.318330223880597e-05,
+      "loss": 0.0,
+      "step": 58460
+    },
+    {
+      "epoch": 54.54,
+      "learning_rate": 4.318283582089552e-05,
+      "loss": 0.0017,
+      "step": 58464
+    },
+    {
+      "epoch": 54.54,
+      "learning_rate": 4.3182369402985076e-05,
+      "loss": 0.0006,
+      "step": 58468
+    },
+    {
+      "epoch": 54.54,
+      "learning_rate": 4.318190298507463e-05,
+      "loss": 0.0001,
+      "step": 58472
+    },
+    {
+      "epoch": 54.55,
+      "learning_rate": 4.318143656716418e-05,
+      "loss": 0.0,
+      "step": 58476
+    },
+    {
+      "epoch": 54.55,
+      "learning_rate": 4.3180970149253734e-05,
+      "loss": 0.0,
+      "step": 58480
+    },
+    {
+      "epoch": 54.56,
+      "learning_rate": 4.318050373134329e-05,
+      "loss": 0.0,
+      "step": 58484
+    },
+    {
+      "epoch": 54.56,
+      "learning_rate": 4.318003731343284e-05,
+      "loss": 0.0029,
+      "step": 58488
+    },
+    {
+      "epoch": 54.56,
+      "learning_rate": 4.317957089552239e-05,
+      "loss": 0.0,
+      "step": 58492
+    },
+    {
+      "epoch": 54.57,
+      "learning_rate": 4.317910447761194e-05,
+      "loss": 0.0005,
+      "step": 58496
+    },
+    {
+      "epoch": 54.57,
+      "learning_rate": 4.3178638059701495e-05,
+      "loss": 0.0001,
+      "step": 58500
+    },
+    {
+      "epoch": 54.57,
+      "eval_exact_match": 0.7350096711798839,
+      "eval_exec": 0.7717601547388782,
+      "eval_loss": 0.4618161916732788,
+      "eval_runtime": 1164.5608,
+      "eval_samples_per_second": 0.888,
+      "step": 58500
+    },
+    {
+      "epoch": 54.57,
+      "learning_rate": 4.317817164179105e-05,
+      "loss": 0.0,
+      "step": 58504
+    },
+    {
+      "epoch": 54.58,
+      "learning_rate": 4.31777052238806e-05,
+      "loss": 0.0,
+      "step": 58508
+    },
+    {
+      "epoch": 54.58,
+      "learning_rate": 4.3177238805970146e-05,
+      "loss": 0.0,
+      "step": 58512
+    },
+    {
+      "epoch": 54.59,
+      "learning_rate": 4.317677238805971e-05,
+      "loss": 0.0,
+      "step": 58516
+    },
+    {
+      "epoch": 54.59,
+      "learning_rate": 4.3176305970149256e-05,
+      "loss": 0.0,
+      "step": 58520
+    },
+    {
+      "epoch": 54.59,
+      "learning_rate": 4.3175839552238804e-05,
+      "loss": 0.0,
+      "step": 58524
+    },
+    {
+      "epoch": 54.6,
+      "learning_rate": 4.317537313432836e-05,
+      "loss": 0.0,
+      "step": 58528
+    },
+    {
+      "epoch": 54.6,
+      "learning_rate": 4.3174906716417914e-05,
+      "loss": 0.0001,
+      "step": 58532
+    },
+    {
+      "epoch": 54.6,
+      "learning_rate": 4.317444029850746e-05,
+      "loss": 0.0,
+      "step": 58536
+    },
+    {
+      "epoch": 54.61,
+      "learning_rate": 4.317397388059702e-05,
+      "loss": 0.0,
+      "step": 58540
+    },
+    {
+      "epoch": 54.61,
+      "learning_rate": 4.317350746268657e-05,
+      "loss": 0.0,
+      "step": 58544
+    },
+    {
+      "epoch": 54.62,
+      "learning_rate": 4.317304104477612e-05,
+      "loss": 0.0,
+      "step": 58548
+    },
+    {
+      "epoch": 54.62,
+      "learning_rate": 4.3172574626865675e-05,
+      "loss": 0.0,
+      "step": 58552
+    },
+    {
+      "epoch": 54.62,
+      "learning_rate": 4.317210820895522e-05,
+      "loss": 0.0,
+      "step": 58556
+    },
+    {
+      "epoch": 54.63,
+      "learning_rate": 4.317164179104478e-05,
+      "loss": 0.0,
+      "step": 58560
+    },
+    {
+      "epoch": 54.63,
+      "learning_rate": 4.317117537313433e-05,
+      "loss": 0.0,
+      "step": 58564
+    },
+    {
+      "epoch": 54.63,
+      "learning_rate": 4.317070895522388e-05,
+      "loss": 0.0,
+      "step": 58568
+    },
+    {
+      "epoch": 54.64,
+      "learning_rate": 4.317024253731343e-05,
+      "loss": 0.004,
+      "step": 58572
+    },
+    {
+      "epoch": 54.64,
+      "learning_rate": 4.316977611940299e-05,
+      "loss": 0.0,
+      "step": 58576
+    },
+    {
+      "epoch": 54.65,
+      "learning_rate": 4.316930970149254e-05,
+      "loss": 0.0,
+      "step": 58580
+    },
+    {
+      "epoch": 54.65,
+      "learning_rate": 4.316884328358209e-05,
+      "loss": 0.0,
+      "step": 58584
+    },
+    {
+      "epoch": 54.65,
+      "learning_rate": 4.316837686567164e-05,
+      "loss": 0.0,
+      "step": 58588
+    },
+    {
+      "epoch": 54.66,
+      "learning_rate": 4.31679104477612e-05,
+      "loss": 0.0,
+      "step": 58592
+    },
+    {
+      "epoch": 54.66,
+      "learning_rate": 4.316744402985075e-05,
+      "loss": 0.0,
+      "step": 58596
+    },
+    {
+      "epoch": 54.66,
+      "learning_rate": 4.31669776119403e-05,
+      "loss": 0.0,
+      "step": 58600
+    },
+    {
+      "epoch": 54.67,
+      "learning_rate": 4.316651119402985e-05,
+      "loss": 0.0,
+      "step": 58604
+    },
+    {
+      "epoch": 54.67,
+      "learning_rate": 4.316604477611941e-05,
+      "loss": 0.0,
+      "step": 58608
+    },
+    {
+      "epoch": 54.68,
+      "learning_rate": 4.316557835820896e-05,
+      "loss": 0.0,
+      "step": 58612
+    },
+    {
+      "epoch": 54.68,
+      "learning_rate": 4.3165111940298506e-05,
+      "loss": 0.0,
+      "step": 58616
+    },
+    {
+      "epoch": 54.68,
+      "learning_rate": 4.316464552238806e-05,
+      "loss": 0.0,
+      "step": 58620
+    },
+    {
+      "epoch": 54.69,
+      "learning_rate": 4.3164179104477616e-05,
+      "loss": 0.0,
+      "step": 58624
+    },
+    {
+      "epoch": 54.69,
+      "learning_rate": 4.3163712686567164e-05,
+      "loss": 0.0,
+      "step": 58628
+    },
+    {
+      "epoch": 54.69,
+      "learning_rate": 4.316324626865672e-05,
+      "loss": 0.0001,
+      "step": 58632
+    },
+    {
+      "epoch": 54.7,
+      "learning_rate": 4.3162779850746274e-05,
+      "loss": 0.0,
+      "step": 58636
+    },
+    {
+      "epoch": 54.7,
+      "learning_rate": 4.316231343283582e-05,
+      "loss": 0.0,
+      "step": 58640
+    },
+    {
+      "epoch": 54.71,
+      "learning_rate": 4.316184701492538e-05,
+      "loss": 0.0,
+      "step": 58644
+    },
+    {
+      "epoch": 54.71,
+      "learning_rate": 4.3161380597014925e-05,
+      "loss": 0.0,
+      "step": 58648
+    },
+    {
+      "epoch": 54.71,
+      "learning_rate": 4.316091417910448e-05,
+      "loss": 0.0,
+      "step": 58652
+    },
+    {
+      "epoch": 54.72,
+      "learning_rate": 4.3160447761194035e-05,
+      "loss": 0.0,
+      "step": 58656
+    },
+    {
+      "epoch": 54.72,
+      "learning_rate": 4.315998134328358e-05,
+      "loss": 0.0,
+      "step": 58660
+    },
+    {
+      "epoch": 54.72,
+      "learning_rate": 4.315951492537313e-05,
+      "loss": 0.0001,
+      "step": 58664
+    },
+    {
+      "epoch": 54.73,
+      "learning_rate": 4.315904850746269e-05,
+      "loss": 0.0,
+      "step": 58668
+    },
+    {
+      "epoch": 54.73,
+      "learning_rate": 4.315858208955224e-05,
+      "loss": 0.0,
+      "step": 58672
+    },
+    {
+      "epoch": 54.73,
+      "learning_rate": 4.315811567164179e-05,
+      "loss": 0.0,
+      "step": 58676
+    },
+    {
+      "epoch": 54.74,
+      "learning_rate": 4.3157649253731344e-05,
+      "loss": 0.0004,
+      "step": 58680
+    },
+    {
+      "epoch": 54.74,
+      "learning_rate": 4.31571828358209e-05,
+      "loss": 0.0,
+      "step": 58684
+    },
+    {
+      "epoch": 54.75,
+      "learning_rate": 4.315671641791045e-05,
+      "loss": 0.0,
+      "step": 58688
+    },
+    {
+      "epoch": 54.75,
+      "learning_rate": 4.315625e-05,
+      "loss": 0.0,
+      "step": 58692
+    },
+    {
+      "epoch": 54.75,
+      "learning_rate": 4.315578358208956e-05,
+      "loss": 0.0,
+      "step": 58696
+    },
+    {
+      "epoch": 54.76,
+      "learning_rate": 4.3155317164179105e-05,
+      "loss": 0.0023,
+      "step": 58700
+    },
+    {
+      "epoch": 54.76,
+      "learning_rate": 4.315485074626866e-05,
+      "loss": 0.0039,
+      "step": 58704
+    },
+    {
+      "epoch": 54.76,
+      "learning_rate": 4.315438432835821e-05,
+      "loss": 0.0,
+      "step": 58708
+    },
+    {
+      "epoch": 54.77,
+      "learning_rate": 4.315391791044776e-05,
+      "loss": 0.0001,
+      "step": 58712
+    },
+    {
+      "epoch": 54.77,
+      "learning_rate": 4.315345149253732e-05,
+      "loss": 0.0,
+      "step": 58716
+    },
+    {
+      "epoch": 54.78,
+      "learning_rate": 4.3152985074626866e-05,
+      "loss": 0.0002,
+      "step": 58720
+    },
+    {
+      "epoch": 54.78,
+      "learning_rate": 4.3152518656716414e-05,
+      "loss": 0.0,
+      "step": 58724
+    },
+    {
+      "epoch": 54.78,
+      "learning_rate": 4.3152052238805976e-05,
+      "loss": 0.0,
+      "step": 58728
+    },
+    {
+      "epoch": 54.79,
+      "learning_rate": 4.3151585820895524e-05,
+      "loss": 0.0127,
+      "step": 58732
+    },
+    {
+      "epoch": 54.79,
+      "learning_rate": 4.315111940298507e-05,
+      "loss": 0.0,
+      "step": 58736
+    },
+    {
+      "epoch": 54.79,
+      "learning_rate": 4.315065298507463e-05,
+      "loss": 0.0006,
+      "step": 58740
+    },
+    {
+      "epoch": 54.8,
+      "learning_rate": 4.315018656716418e-05,
+      "loss": 0.0,
+      "step": 58744
+    },
+    {
+      "epoch": 54.8,
+      "learning_rate": 4.314972014925374e-05,
+      "loss": 0.0,
+      "step": 58748
+    },
+    {
+      "epoch": 54.81,
+      "learning_rate": 4.3149253731343285e-05,
+      "loss": 0.0009,
+      "step": 58752
+    },
+    {
+      "epoch": 54.81,
+      "learning_rate": 4.314878731343284e-05,
+      "loss": 0.0003,
+      "step": 58756
+    },
+    {
+      "epoch": 54.81,
+      "learning_rate": 4.3148320895522395e-05,
+      "loss": 0.0002,
+      "step": 58760
+    },
+    {
+      "epoch": 54.82,
+      "learning_rate": 4.314785447761194e-05,
+      "loss": 0.0001,
+      "step": 58764
+    },
+    {
+      "epoch": 54.82,
+      "learning_rate": 4.314738805970149e-05,
+      "loss": 0.0011,
+      "step": 58768
+    },
+    {
+      "epoch": 54.82,
+      "learning_rate": 4.3146921641791046e-05,
+      "loss": 0.0,
+      "step": 58772
+    },
+    {
+      "epoch": 54.83,
+      "learning_rate": 4.31464552238806e-05,
+      "loss": 0.0,
+      "step": 58776
+    },
+    {
+      "epoch": 54.83,
+      "learning_rate": 4.314598880597015e-05,
+      "loss": 0.0,
+      "step": 58780
+    },
+    {
+      "epoch": 54.84,
+      "learning_rate": 4.3145522388059704e-05,
+      "loss": 0.0,
+      "step": 58784
+    },
+    {
+      "epoch": 54.84,
+      "learning_rate": 4.314505597014926e-05,
+      "loss": 0.0003,
+      "step": 58788
+    },
+    {
+      "epoch": 54.84,
+      "learning_rate": 4.314458955223881e-05,
+      "loss": 0.0003,
+      "step": 58792
+    },
+    {
+      "epoch": 54.85,
+      "learning_rate": 4.314412313432836e-05,
+      "loss": 0.0005,
+      "step": 58796
+    },
+    {
+      "epoch": 54.85,
+      "learning_rate": 4.314365671641791e-05,
+      "loss": 0.0005,
+      "step": 58800
+    },
+    {
+      "epoch": 54.85,
+      "learning_rate": 4.3143190298507465e-05,
+      "loss": 0.0,
+      "step": 58804
+    },
+    {
+      "epoch": 54.86,
+      "learning_rate": 4.314272388059702e-05,
+      "loss": 0.0001,
+      "step": 58808
+    },
+    {
+      "epoch": 54.86,
+      "learning_rate": 4.314225746268657e-05,
+      "loss": 0.0,
+      "step": 58812
+    },
+    {
+      "epoch": 54.87,
+      "learning_rate": 4.314179104477612e-05,
+      "loss": 0.0,
+      "step": 58816
+    },
+    {
+      "epoch": 54.87,
+      "learning_rate": 4.314132462686568e-05,
+      "loss": 0.0002,
+      "step": 58820
+    },
+    {
+      "epoch": 54.87,
+      "learning_rate": 4.3140858208955226e-05,
+      "loss": 0.0,
+      "step": 58824
+    },
+    {
+      "epoch": 54.88,
+      "learning_rate": 4.3140391791044774e-05,
+      "loss": 0.0011,
+      "step": 58828
+    },
+    {
+      "epoch": 54.88,
+      "learning_rate": 4.313992537313433e-05,
+      "loss": 0.0,
+      "step": 58832
+    },
+    {
+      "epoch": 54.88,
+      "learning_rate": 4.3139458955223884e-05,
+      "loss": 0.0,
+      "step": 58836
+    },
+    {
+      "epoch": 54.89,
+      "learning_rate": 4.313899253731343e-05,
+      "loss": 0.0,
+      "step": 58840
+    },
+    {
+      "epoch": 54.89,
+      "learning_rate": 4.313852611940299e-05,
+      "loss": 0.0,
+      "step": 58844
+    },
+    {
+      "epoch": 54.9,
+      "learning_rate": 4.313805970149254e-05,
+      "loss": 0.0001,
+      "step": 58848
+    },
+    {
+      "epoch": 54.9,
+      "learning_rate": 4.313759328358209e-05,
+      "loss": 0.0,
+      "step": 58852
+    },
+    {
+      "epoch": 54.9,
+      "learning_rate": 4.3137126865671645e-05,
+      "loss": 0.0,
+      "step": 58856
+    },
+    {
+      "epoch": 54.91,
+      "learning_rate": 4.313666044776119e-05,
+      "loss": 0.0001,
+      "step": 58860
+    },
+    {
+      "epoch": 54.91,
+      "learning_rate": 4.313619402985075e-05,
+      "loss": 0.0001,
+      "step": 58864
+    },
+    {
+      "epoch": 54.91,
+      "learning_rate": 4.31357276119403e-05,
+      "loss": 0.0005,
+      "step": 58868
+    },
+    {
+      "epoch": 54.92,
+      "learning_rate": 4.313526119402985e-05,
+      "loss": 0.0,
+      "step": 58872
+    },
+    {
+      "epoch": 54.92,
+      "learning_rate": 4.3134794776119406e-05,
+      "loss": 0.0002,
+      "step": 58876
+    },
+    {
+      "epoch": 54.93,
+      "learning_rate": 4.313432835820896e-05,
+      "loss": 0.0,
+      "step": 58880
+    },
+    {
+      "epoch": 54.93,
+      "learning_rate": 4.313386194029851e-05,
+      "loss": 0.0006,
+      "step": 58884
+    },
+    {
+      "epoch": 54.93,
+      "learning_rate": 4.313339552238806e-05,
+      "loss": 0.0,
+      "step": 58888
+    },
+    {
+      "epoch": 54.94,
+      "learning_rate": 4.313292910447761e-05,
+      "loss": 0.0,
+      "step": 58892
+    },
+    {
+      "epoch": 54.94,
+      "learning_rate": 4.313246268656717e-05,
+      "loss": 0.0003,
+      "step": 58896
+    },
+    {
+      "epoch": 54.94,
+      "learning_rate": 4.3131996268656715e-05,
+      "loss": 0.0002,
+      "step": 58900
+    },
+    {
+      "epoch": 54.95,
+      "learning_rate": 4.313152985074627e-05,
+      "loss": 0.0,
+      "step": 58904
+    },
+    {
+      "epoch": 54.95,
+      "learning_rate": 4.3131063432835825e-05,
+      "loss": 0.0,
+      "step": 58908
+    },
+    {
+      "epoch": 54.96,
+      "learning_rate": 4.313059701492538e-05,
+      "loss": 0.0,
+      "step": 58912
+    },
+    {
+      "epoch": 54.96,
+      "learning_rate": 4.313013059701493e-05,
+      "loss": 0.0,
+      "step": 58916
+    },
+    {
+      "epoch": 54.96,
+      "learning_rate": 4.3129664179104476e-05,
+      "loss": 0.0001,
+      "step": 58920
+    },
+    {
+      "epoch": 54.97,
+      "learning_rate": 4.312919776119404e-05,
+      "loss": 0.0,
+      "step": 58924
+    },
+    {
+      "epoch": 54.97,
+      "learning_rate": 4.3128731343283586e-05,
+      "loss": 0.0,
+      "step": 58928
+    },
+    {
+      "epoch": 54.97,
+      "learning_rate": 4.3128264925373134e-05,
+      "loss": 0.0001,
+      "step": 58932
+    },
+    {
+      "epoch": 54.98,
+      "learning_rate": 4.312779850746269e-05,
+      "loss": 0.0,
+      "step": 58936
+    },
+    {
+      "epoch": 54.98,
+      "learning_rate": 4.3127332089552244e-05,
+      "loss": 0.0001,
+      "step": 58940
+    },
+    {
+      "epoch": 54.98,
+      "learning_rate": 4.312686567164179e-05,
+      "loss": 0.0003,
+      "step": 58944
+    },
+    {
+      "epoch": 54.99,
+      "learning_rate": 4.312639925373135e-05,
+      "loss": 0.0,
+      "step": 58948
+    },
+    {
+      "epoch": 54.99,
+      "learning_rate": 4.3125932835820895e-05,
+      "loss": 0.0,
+      "step": 58952
+    },
+    {
+      "epoch": 55.0,
+      "learning_rate": 4.312546641791045e-05,
+      "loss": 0.0008,
+      "step": 58956
+    },
+    {
+      "epoch": 55.0,
+      "learning_rate": 4.3125000000000005e-05,
+      "loss": 0.0004,
+      "step": 58960
+    },
+    {
+      "epoch": 55.0,
+      "learning_rate": 4.312453358208955e-05,
+      "loss": 0.0001,
+      "step": 58964
+    },
+    {
+      "epoch": 55.01,
+      "learning_rate": 4.312406716417911e-05,
+      "loss": 0.0,
+      "step": 58968
+    },
+    {
+      "epoch": 55.01,
+      "learning_rate": 4.312360074626866e-05,
+      "loss": 0.0,
+      "step": 58972
+    },
+    {
+      "epoch": 55.01,
+      "learning_rate": 4.312313432835821e-05,
+      "loss": 0.0,
+      "step": 58976
+    },
+    {
+      "epoch": 55.02,
+      "learning_rate": 4.312266791044776e-05,
+      "loss": 0.0,
+      "step": 58980
+    },
+    {
+      "epoch": 55.02,
+      "learning_rate": 4.312220149253732e-05,
+      "loss": 0.0,
+      "step": 58984
+    },
+    {
+      "epoch": 55.03,
+      "learning_rate": 4.312173507462687e-05,
+      "loss": 0.0002,
+      "step": 58988
+    },
+    {
+      "epoch": 55.03,
+      "learning_rate": 4.312126865671642e-05,
+      "loss": 0.0001,
+      "step": 58992
+    },
+    {
+      "epoch": 55.03,
+      "learning_rate": 4.312080223880597e-05,
+      "loss": 0.0,
+      "step": 58996
+    },
+    {
+      "epoch": 55.04,
+      "learning_rate": 4.312033582089553e-05,
+      "loss": 0.0,
+      "step": 59000
+    },
+    {
+      "epoch": 55.04,
+      "eval_exact_match": 0.7446808510638298,
+      "eval_exec": 0.781431334622824,
+      "eval_loss": 0.4688223898410797,
+      "eval_runtime": 1134.1914,
+      "eval_samples_per_second": 0.912,
+      "step": 59000
+    },
+    {
+      "epoch": 55.04,
+      "learning_rate": 4.3119869402985075e-05,
+      "loss": 0.0026,
+      "step": 59004
+    },
+    {
+      "epoch": 55.04,
+      "learning_rate": 4.311940298507463e-05,
+      "loss": 0.0,
+      "step": 59008
+    },
+    {
+      "epoch": 55.05,
+      "learning_rate": 4.311893656716418e-05,
+      "loss": 0.0,
+      "step": 59012
+    },
+    {
+      "epoch": 55.05,
+      "learning_rate": 4.311847014925373e-05,
+      "loss": 0.0,
+      "step": 59016
+    },
+    {
+      "epoch": 55.06,
+      "learning_rate": 4.311800373134329e-05,
+      "loss": 0.0095,
+      "step": 59020
+    },
+    {
+      "epoch": 55.06,
+      "learning_rate": 4.3117537313432836e-05,
+      "loss": 0.0001,
+      "step": 59024
+    },
+    {
+      "epoch": 55.06,
+      "learning_rate": 4.311707089552239e-05,
+      "loss": 0.0,
+      "step": 59028
+    },
+    {
+      "epoch": 55.07,
+      "learning_rate": 4.3116604477611946e-05,
+      "loss": 0.0001,
+      "step": 59032
+    },
+    {
+      "epoch": 55.07,
+      "learning_rate": 4.3116138059701494e-05,
+      "loss": 0.0,
+      "step": 59036
+    },
+    {
+      "epoch": 55.07,
+      "learning_rate": 4.311567164179104e-05,
+      "loss": 0.0,
+      "step": 59040
+    },
+    {
+      "epoch": 55.08,
+      "learning_rate": 4.3115205223880604e-05,
+      "loss": 0.0,
+      "step": 59044
+    },
+    {
+      "epoch": 55.08,
+      "learning_rate": 4.311473880597015e-05,
+      "loss": 0.0002,
+      "step": 59048
+    },
+    {
+      "epoch": 55.09,
+      "learning_rate": 4.31142723880597e-05,
+      "loss": 0.0,
+      "step": 59052
+    },
+    {
+      "epoch": 55.09,
+      "learning_rate": 4.3113805970149255e-05,
+      "loss": 0.0,
+      "step": 59056
+    },
+    {
+      "epoch": 55.09,
+      "learning_rate": 4.311333955223881e-05,
+      "loss": 0.0,
+      "step": 59060
+    },
+    {
+      "epoch": 55.1,
+      "learning_rate": 4.311287313432836e-05,
+      "loss": 0.0004,
+      "step": 59064
+    },
+    {
+      "epoch": 55.1,
+      "learning_rate": 4.311240671641791e-05,
+      "loss": 0.0002,
+      "step": 59068
+    },
+    {
+      "epoch": 55.1,
+      "learning_rate": 4.311194029850746e-05,
+      "loss": 0.0,
+      "step": 59072
+    },
+    {
+      "epoch": 55.11,
+      "learning_rate": 4.311147388059702e-05,
+      "loss": 0.0008,
+      "step": 59076
+    },
+    {
+      "epoch": 55.11,
+      "learning_rate": 4.311100746268657e-05,
+      "loss": 0.0,
+      "step": 59080
+    },
+    {
+      "epoch": 55.12,
+      "learning_rate": 4.311054104477612e-05,
+      "loss": 0.0003,
+      "step": 59084
+    },
+    {
+      "epoch": 55.12,
+      "learning_rate": 4.3110074626865674e-05,
+      "loss": 0.0,
+      "step": 59088
+    },
+    {
+      "epoch": 55.12,
+      "learning_rate": 4.310960820895523e-05,
+      "loss": 0.0023,
+      "step": 59092
+    },
+    {
+      "epoch": 55.13,
+      "learning_rate": 4.310914179104478e-05,
+      "loss": 0.001,
+      "step": 59096
+    },
+    {
+      "epoch": 55.13,
+      "learning_rate": 4.310867537313433e-05,
+      "loss": 0.0,
+      "step": 59100
+    },
+    {
+      "epoch": 55.13,
+      "learning_rate": 4.3108208955223886e-05,
+      "loss": 0.0,
+      "step": 59104
+    },
+    {
+      "epoch": 55.14,
+      "learning_rate": 4.3107742537313435e-05,
+      "loss": 0.0007,
+      "step": 59108
+    },
+    {
+      "epoch": 55.14,
+      "learning_rate": 4.310727611940299e-05,
+      "loss": 0.0,
+      "step": 59112
+    },
+    {
+      "epoch": 55.15,
+      "learning_rate": 4.310680970149254e-05,
+      "loss": 0.0,
+      "step": 59116
+    },
+    {
+      "epoch": 55.15,
+      "learning_rate": 4.310634328358209e-05,
+      "loss": 0.0045,
+      "step": 59120
+    },
+    {
+      "epoch": 55.15,
+      "learning_rate": 4.310587686567165e-05,
+      "loss": 0.0012,
+      "step": 59124
+    },
+    {
+      "epoch": 55.16,
+      "learning_rate": 4.3105410447761196e-05,
+      "loss": 0.0028,
+      "step": 59128
+    },
+    {
+      "epoch": 55.16,
+      "learning_rate": 4.3104944029850744e-05,
+      "loss": 0.0,
+      "step": 59132
+    },
+    {
+      "epoch": 55.16,
+      "learning_rate": 4.3104477611940305e-05,
+      "loss": 0.0,
+      "step": 59136
+    },
+    {
+      "epoch": 55.17,
+      "learning_rate": 4.3104011194029854e-05,
+      "loss": 0.0,
+      "step": 59140
+    },
+    {
+      "epoch": 55.17,
+      "learning_rate": 4.31035447761194e-05,
+      "loss": 0.0,
+      "step": 59144
+    },
+    {
+      "epoch": 55.18,
+      "learning_rate": 4.3103078358208957e-05,
+      "loss": 0.0,
+      "step": 59148
+    },
+    {
+      "epoch": 55.18,
+      "learning_rate": 4.310261194029851e-05,
+      "loss": 0.0011,
+      "step": 59152
+    },
+    {
+      "epoch": 55.18,
+      "learning_rate": 4.310214552238806e-05,
+      "loss": 0.0,
+      "step": 59156
+    },
+    {
+      "epoch": 55.19,
+      "learning_rate": 4.3101679104477615e-05,
+      "loss": 0.0,
+      "step": 59160
+    },
+    {
+      "epoch": 55.19,
+      "learning_rate": 4.310121268656717e-05,
+      "loss": 0.0,
+      "step": 59164
+    },
+    {
+      "epoch": 55.19,
+      "learning_rate": 4.310074626865672e-05,
+      "loss": 0.0,
+      "step": 59168
+    },
+    {
+      "epoch": 55.2,
+      "learning_rate": 4.310027985074627e-05,
+      "loss": 0.0,
+      "step": 59172
+    },
+    {
+      "epoch": 55.2,
+      "learning_rate": 4.309981343283582e-05,
+      "loss": 0.0,
+      "step": 59176
+    },
+    {
+      "epoch": 55.21,
+      "learning_rate": 4.3099347014925376e-05,
+      "loss": 0.0,
+      "step": 59180
+    },
+    {
+      "epoch": 55.21,
+      "learning_rate": 4.309888059701493e-05,
+      "loss": 0.0001,
+      "step": 59184
+    },
+    {
+      "epoch": 55.21,
+      "learning_rate": 4.309841417910448e-05,
+      "loss": 0.0,
+      "step": 59188
+    },
+    {
+      "epoch": 55.22,
+      "learning_rate": 4.309794776119403e-05,
+      "loss": 0.0001,
+      "step": 59192
+    },
+    {
+      "epoch": 55.22,
+      "learning_rate": 4.309748134328359e-05,
+      "loss": 0.0001,
+      "step": 59196
+    },
+    {
+      "epoch": 55.22,
+      "learning_rate": 4.3097014925373137e-05,
+      "loss": 0.0,
+      "step": 59200
+    },
+    {
+      "epoch": 55.23,
+      "learning_rate": 4.3096548507462685e-05,
+      "loss": 0.0001,
+      "step": 59204
+    },
+    {
+      "epoch": 55.23,
+      "learning_rate": 4.309608208955224e-05,
+      "loss": 0.0,
+      "step": 59208
+    },
+    {
+      "epoch": 55.24,
+      "learning_rate": 4.3095615671641794e-05,
+      "loss": 0.0,
+      "step": 59212
+    },
+    {
+      "epoch": 55.24,
+      "learning_rate": 4.309514925373134e-05,
+      "loss": 0.0044,
+      "step": 59216
+    },
+    {
+      "epoch": 55.24,
+      "learning_rate": 4.30946828358209e-05,
+      "loss": 0.0009,
+      "step": 59220
+    },
+    {
+      "epoch": 55.25,
+      "learning_rate": 4.309421641791045e-05,
+      "loss": 0.0001,
+      "step": 59224
+    },
+    {
+      "epoch": 55.25,
+      "learning_rate": 4.309375e-05,
+      "loss": 0.0,
+      "step": 59228
+    },
+    {
+      "epoch": 55.25,
+      "learning_rate": 4.3093283582089555e-05,
+      "loss": 0.0001,
+      "step": 59232
+    },
+    {
+      "epoch": 55.26,
+      "learning_rate": 4.3092817164179104e-05,
+      "loss": 0.0,
+      "step": 59236
+    },
+    {
+      "epoch": 55.26,
+      "learning_rate": 4.309235074626866e-05,
+      "loss": 0.0001,
+      "step": 59240
+    },
+    {
+      "epoch": 55.26,
+      "learning_rate": 4.3091884328358213e-05,
+      "loss": 0.0,
+      "step": 59244
+    },
+    {
+      "epoch": 55.27,
+      "learning_rate": 4.309141791044776e-05,
+      "loss": 0.0011,
+      "step": 59248
+    },
+    {
+      "epoch": 55.27,
+      "learning_rate": 4.3090951492537316e-05,
+      "loss": 0.0,
+      "step": 59252
+    },
+    {
+      "epoch": 55.28,
+      "learning_rate": 4.309048507462687e-05,
+      "loss": 0.0,
+      "step": 59256
+    },
+    {
+      "epoch": 55.28,
+      "learning_rate": 4.309001865671642e-05,
+      "loss": 0.0,
+      "step": 59260
+    },
+    {
+      "epoch": 55.28,
+      "learning_rate": 4.3089552238805974e-05,
+      "loss": 0.0,
+      "step": 59264
+    },
+    {
+      "epoch": 55.29,
+      "learning_rate": 4.308908582089552e-05,
+      "loss": 0.0,
+      "step": 59268
+    },
+    {
+      "epoch": 55.29,
+      "learning_rate": 4.308861940298508e-05,
+      "loss": 0.0001,
+      "step": 59272
+    },
+    {
+      "epoch": 55.29,
+      "learning_rate": 4.308815298507463e-05,
+      "loss": 0.0044,
+      "step": 59276
+    },
+    {
+      "epoch": 55.3,
+      "learning_rate": 4.308768656716418e-05,
+      "loss": 0.0002,
+      "step": 59280
+    },
+    {
+      "epoch": 55.3,
+      "learning_rate": 4.308722014925373e-05,
+      "loss": 0.0,
+      "step": 59284
+    },
+    {
+      "epoch": 55.31,
+      "learning_rate": 4.308675373134329e-05,
+      "loss": 0.0,
+      "step": 59288
+    },
+    {
+      "epoch": 55.31,
+      "learning_rate": 4.308628731343284e-05,
+      "loss": 0.0,
+      "step": 59292
+    },
+    {
+      "epoch": 55.31,
+      "learning_rate": 4.3085820895522387e-05,
+      "loss": 0.0,
+      "step": 59296
+    },
+    {
+      "epoch": 55.32,
+      "learning_rate": 4.308535447761194e-05,
+      "loss": 0.0,
+      "step": 59300
+    },
+    {
+      "epoch": 55.32,
+      "learning_rate": 4.3084888059701496e-05,
+      "loss": 0.0,
+      "step": 59304
+    },
+    {
+      "epoch": 55.32,
+      "learning_rate": 4.3084421641791044e-05,
+      "loss": 0.0,
+      "step": 59308
+    },
+    {
+      "epoch": 55.33,
+      "learning_rate": 4.30839552238806e-05,
+      "loss": 0.0001,
+      "step": 59312
+    },
+    {
+      "epoch": 55.33,
+      "learning_rate": 4.3083488805970154e-05,
+      "loss": 0.0,
+      "step": 59316
+    },
+    {
+      "epoch": 55.34,
+      "learning_rate": 4.30830223880597e-05,
+      "loss": 0.0,
+      "step": 59320
+    },
+    {
+      "epoch": 55.34,
+      "learning_rate": 4.308255597014926e-05,
+      "loss": 0.0,
+      "step": 59324
+    },
+    {
+      "epoch": 55.34,
+      "learning_rate": 4.3082089552238805e-05,
+      "loss": 0.0001,
+      "step": 59328
+    },
+    {
+      "epoch": 55.35,
+      "learning_rate": 4.308162313432836e-05,
+      "loss": 0.0,
+      "step": 59332
+    },
+    {
+      "epoch": 55.35,
+      "learning_rate": 4.3081156716417915e-05,
+      "loss": 0.0,
+      "step": 59336
+    },
+    {
+      "epoch": 55.35,
+      "learning_rate": 4.3080690298507463e-05,
+      "loss": 0.0,
+      "step": 59340
+    },
+    {
+      "epoch": 55.36,
+      "learning_rate": 4.308022388059701e-05,
+      "loss": 0.0,
+      "step": 59344
+    },
+    {
+      "epoch": 55.36,
+      "learning_rate": 4.307975746268657e-05,
+      "loss": 0.0,
+      "step": 59348
+    },
+    {
+      "epoch": 55.37,
+      "learning_rate": 4.307929104477612e-05,
+      "loss": 0.0,
+      "step": 59352
+    },
+    {
+      "epoch": 55.37,
+      "learning_rate": 4.307882462686567e-05,
+      "loss": 0.0,
+      "step": 59356
+    },
+    {
+      "epoch": 55.37,
+      "learning_rate": 4.3078358208955224e-05,
+      "loss": 0.0,
+      "step": 59360
+    },
+    {
+      "epoch": 55.38,
+      "learning_rate": 4.307789179104478e-05,
+      "loss": 0.0,
+      "step": 59364
+    },
+    {
+      "epoch": 55.38,
+      "learning_rate": 4.307742537313433e-05,
+      "loss": 0.0004,
+      "step": 59368
+    },
+    {
+      "epoch": 55.38,
+      "learning_rate": 4.307695895522388e-05,
+      "loss": 0.0,
+      "step": 59372
+    },
+    {
+      "epoch": 55.39,
+      "learning_rate": 4.307649253731344e-05,
+      "loss": 0.0,
+      "step": 59376
+    },
+    {
+      "epoch": 55.39,
+      "learning_rate": 4.3076026119402985e-05,
+      "loss": 0.0,
+      "step": 59380
+    },
+    {
+      "epoch": 55.4,
+      "learning_rate": 4.307555970149254e-05,
+      "loss": 0.0045,
+      "step": 59384
+    },
+    {
+      "epoch": 55.4,
+      "learning_rate": 4.307509328358209e-05,
+      "loss": 0.0,
+      "step": 59388
+    },
+    {
+      "epoch": 55.4,
+      "learning_rate": 4.307462686567164e-05,
+      "loss": 0.0,
+      "step": 59392
+    },
+    {
+      "epoch": 55.41,
+      "learning_rate": 4.30741604477612e-05,
+      "loss": 0.0,
+      "step": 59396
+    },
+    {
+      "epoch": 55.41,
+      "learning_rate": 4.3073694029850746e-05,
+      "loss": 0.0001,
+      "step": 59400
+    },
+    {
+      "epoch": 55.41,
+      "learning_rate": 4.30732276119403e-05,
+      "loss": 0.0001,
+      "step": 59404
+    },
+    {
+      "epoch": 55.42,
+      "learning_rate": 4.3072761194029856e-05,
+      "loss": 0.0,
+      "step": 59408
+    },
+    {
+      "epoch": 55.42,
+      "learning_rate": 4.3072294776119404e-05,
+      "loss": 0.0,
+      "step": 59412
+    },
+    {
+      "epoch": 55.43,
+      "learning_rate": 4.307182835820896e-05,
+      "loss": 0.0001,
+      "step": 59416
+    },
+    {
+      "epoch": 55.43,
+      "learning_rate": 4.307136194029851e-05,
+      "loss": 0.0,
+      "step": 59420
+    },
+    {
+      "epoch": 55.43,
+      "learning_rate": 4.307089552238806e-05,
+      "loss": 0.0002,
+      "step": 59424
+    },
+    {
+      "epoch": 55.44,
+      "learning_rate": 4.307042910447762e-05,
+      "loss": 0.0025,
+      "step": 59428
+    },
+    {
+      "epoch": 55.44,
+      "learning_rate": 4.3069962686567165e-05,
+      "loss": 0.0,
+      "step": 59432
+    },
+    {
+      "epoch": 55.44,
+      "learning_rate": 4.306949626865672e-05,
+      "loss": 0.0,
+      "step": 59436
+    },
+    {
+      "epoch": 55.45,
+      "learning_rate": 4.3069029850746275e-05,
+      "loss": 0.0,
+      "step": 59440
+    },
+    {
+      "epoch": 55.45,
+      "learning_rate": 4.306856343283582e-05,
+      "loss": 0.0001,
+      "step": 59444
+    },
+    {
+      "epoch": 55.46,
+      "learning_rate": 4.306809701492537e-05,
+      "loss": 0.0,
+      "step": 59448
+    },
+    {
+      "epoch": 55.46,
+      "learning_rate": 4.3067630597014926e-05,
+      "loss": 0.0,
+      "step": 59452
+    },
+    {
+      "epoch": 55.46,
+      "learning_rate": 4.306716417910448e-05,
+      "loss": 0.0,
+      "step": 59456
+    },
+    {
+      "epoch": 55.47,
+      "learning_rate": 4.306669776119403e-05,
+      "loss": 0.0,
+      "step": 59460
+    },
+    {
+      "epoch": 55.47,
+      "learning_rate": 4.3066231343283584e-05,
+      "loss": 0.0,
+      "step": 59464
+    },
+    {
+      "epoch": 55.47,
+      "learning_rate": 4.306576492537314e-05,
+      "loss": 0.0,
+      "step": 59468
+    },
+    {
+      "epoch": 55.48,
+      "learning_rate": 4.306529850746269e-05,
+      "loss": 0.0003,
+      "step": 59472
+    },
+    {
+      "epoch": 55.48,
+      "learning_rate": 4.306483208955224e-05,
+      "loss": 0.0003,
+      "step": 59476
+    },
+    {
+      "epoch": 55.49,
+      "learning_rate": 4.306436567164179e-05,
+      "loss": 0.0002,
+      "step": 59480
+    },
+    {
+      "epoch": 55.49,
+      "learning_rate": 4.3063899253731345e-05,
+      "loss": 0.0,
+      "step": 59484
+    },
+    {
+      "epoch": 55.49,
+      "learning_rate": 4.30634328358209e-05,
+      "loss": 0.0006,
+      "step": 59488
+    },
+    {
+      "epoch": 55.5,
+      "learning_rate": 4.306296641791045e-05,
+      "loss": 0.0007,
+      "step": 59492
+    },
+    {
+      "epoch": 55.5,
+      "learning_rate": 4.30625e-05,
+      "loss": 0.0002,
+      "step": 59496
+    },
+    {
+      "epoch": 55.5,
+      "learning_rate": 4.306203358208956e-05,
+      "loss": 0.0003,
+      "step": 59500
+    },
+    {
+      "epoch": 55.5,
+      "eval_exact_match": 0.7446808510638298,
+      "eval_exec": 0.7756286266924565,
+      "eval_loss": 0.48064661026000977,
+      "eval_runtime": 1173.6666,
+      "eval_samples_per_second": 0.881,
+      "step": 59500
+    },
+    {
+      "epoch": 55.51,
+      "learning_rate": 4.3061567164179106e-05,
+      "loss": 0.0001,
+      "step": 59504
+    },
+    {
+      "epoch": 55.51,
+      "learning_rate": 4.3061100746268654e-05,
+      "loss": 0.0,
+      "step": 59508
+    },
+    {
+      "epoch": 55.51,
+      "learning_rate": 4.306063432835821e-05,
+      "loss": 0.0001,
+      "step": 59512
+    },
+    {
+      "epoch": 55.52,
+      "learning_rate": 4.3060167910447764e-05,
+      "loss": 0.0001,
+      "step": 59516
+    },
+    {
+      "epoch": 55.52,
+      "learning_rate": 4.305970149253731e-05,
+      "loss": 0.0,
+      "step": 59520
+    },
+    {
+      "epoch": 55.53,
+      "learning_rate": 4.305923507462687e-05,
+      "loss": 0.001,
+      "step": 59524
+    },
+    {
+      "epoch": 55.53,
+      "learning_rate": 4.305876865671642e-05,
+      "loss": 0.0004,
+      "step": 59528
+    },
+    {
+      "epoch": 55.53,
+      "learning_rate": 4.305830223880597e-05,
+      "loss": 0.0002,
+      "step": 59532
+    },
+    {
+      "epoch": 55.54,
+      "learning_rate": 4.3057835820895525e-05,
+      "loss": 0.0,
+      "step": 59536
+    },
+    {
+      "epoch": 55.54,
+      "learning_rate": 4.305736940298507e-05,
+      "loss": 0.0,
+      "step": 59540
+    },
+    {
+      "epoch": 55.54,
+      "learning_rate": 4.305690298507463e-05,
+      "loss": 0.0,
+      "step": 59544
+    },
+    {
+      "epoch": 55.55,
+      "learning_rate": 4.305643656716418e-05,
+      "loss": 0.0001,
+      "step": 59548
+    },
+    {
+      "epoch": 55.55,
+      "learning_rate": 4.305597014925373e-05,
+      "loss": 0.0,
+      "step": 59552
+    },
+    {
+      "epoch": 55.56,
+      "learning_rate": 4.3055503731343286e-05,
+      "loss": 0.0003,
+      "step": 59556
+    },
+    {
+      "epoch": 55.56,
+      "learning_rate": 4.305503731343284e-05,
+      "loss": 0.0019,
+      "step": 59560
+    },
+    {
+      "epoch": 55.56,
+      "learning_rate": 4.305457089552239e-05,
+      "loss": 0.0,
+      "step": 59564
+    },
+    {
+      "epoch": 55.57,
+      "learning_rate": 4.3054104477611944e-05,
+      "loss": 0.0,
+      "step": 59568
+    },
+    {
+      "epoch": 55.57,
+      "learning_rate": 4.305363805970149e-05,
+      "loss": 0.0012,
+      "step": 59572
+    },
+    {
+      "epoch": 55.57,
+      "learning_rate": 4.305317164179105e-05,
+      "loss": 0.0,
+      "step": 59576
+    },
+    {
+      "epoch": 55.58,
+      "learning_rate": 4.30527052238806e-05,
+      "loss": 0.0,
+      "step": 59580
+    },
+    {
+      "epoch": 55.58,
+      "learning_rate": 4.305223880597015e-05,
+      "loss": 0.0,
+      "step": 59584
+    },
+    {
+      "epoch": 55.59,
+      "learning_rate": 4.3051772388059705e-05,
+      "loss": 0.0,
+      "step": 59588
+    },
+    {
+      "epoch": 55.59,
+      "learning_rate": 4.305130597014926e-05,
+      "loss": 0.0,
+      "step": 59592
+    },
+    {
+      "epoch": 55.59,
+      "learning_rate": 4.305083955223881e-05,
+      "loss": 0.0,
+      "step": 59596
+    },
+    {
+      "epoch": 55.6,
+      "learning_rate": 4.3050373134328356e-05,
+      "loss": 0.0003,
+      "step": 59600
+    },
+    {
+      "epoch": 55.6,
+      "learning_rate": 4.304990671641792e-05,
+      "loss": 0.0084,
+      "step": 59604
+    },
+    {
+      "epoch": 55.6,
+      "learning_rate": 4.3049440298507466e-05,
+      "loss": 0.0001,
+      "step": 59608
+    },
+    {
+      "epoch": 55.61,
+      "learning_rate": 4.3048973880597014e-05,
+      "loss": 0.0,
+      "step": 59612
+    },
+    {
+      "epoch": 55.61,
+      "learning_rate": 4.304850746268657e-05,
+      "loss": 0.0,
+      "step": 59616
+    },
+    {
+      "epoch": 55.62,
+      "learning_rate": 4.3048041044776124e-05,
+      "loss": 0.0,
+      "step": 59620
+    },
+    {
+      "epoch": 55.62,
+      "learning_rate": 4.304757462686567e-05,
+      "loss": 0.0,
+      "step": 59624
+    },
+    {
+      "epoch": 55.62,
+      "learning_rate": 4.304710820895523e-05,
+      "loss": 0.0,
+      "step": 59628
+    },
+    {
+      "epoch": 55.63,
+      "learning_rate": 4.3046641791044775e-05,
+      "loss": 0.0,
+      "step": 59632
+    },
+    {
+      "epoch": 55.63,
+      "learning_rate": 4.304617537313433e-05,
+      "loss": 0.0,
+      "step": 59636
+    },
+    {
+      "epoch": 55.63,
+      "learning_rate": 4.3045708955223885e-05,
+      "loss": 0.0,
+      "step": 59640
+    },
+    {
+      "epoch": 55.64,
+      "learning_rate": 4.304524253731343e-05,
+      "loss": 0.0,
+      "step": 59644
+    },
+    {
+      "epoch": 55.64,
+      "learning_rate": 4.304477611940299e-05,
+      "loss": 0.0,
+      "step": 59648
+    },
+    {
+      "epoch": 55.65,
+      "learning_rate": 4.304430970149254e-05,
+      "loss": 0.0,
+      "step": 59652
+    },
+    {
+      "epoch": 55.65,
+      "learning_rate": 4.304384328358209e-05,
+      "loss": 0.0001,
+      "step": 59656
+    },
+    {
+      "epoch": 55.65,
+      "learning_rate": 4.304337686567164e-05,
+      "loss": 0.0,
+      "step": 59660
+    },
+    {
+      "epoch": 55.66,
+      "learning_rate": 4.30429104477612e-05,
+      "loss": 0.0,
+      "step": 59664
+    },
+    {
+      "epoch": 55.66,
+      "learning_rate": 4.304244402985075e-05,
+      "loss": 0.0,
+      "step": 59668
+    },
+    {
+      "epoch": 55.66,
+      "learning_rate": 4.30419776119403e-05,
+      "loss": 0.0,
+      "step": 59672
+    },
+    {
+      "epoch": 55.67,
+      "learning_rate": 4.304151119402985e-05,
+      "loss": 0.0,
+      "step": 59676
+    },
+    {
+      "epoch": 55.67,
+      "learning_rate": 4.304104477611941e-05,
+      "loss": 0.0,
+      "step": 59680
+    },
+    {
+      "epoch": 55.68,
+      "learning_rate": 4.3040578358208955e-05,
+      "loss": 0.0,
+      "step": 59684
+    },
+    {
+      "epoch": 55.68,
+      "learning_rate": 4.304011194029851e-05,
+      "loss": 0.0,
+      "step": 59688
+    },
+    {
+      "epoch": 55.68,
+      "learning_rate": 4.303964552238806e-05,
+      "loss": 0.0001,
+      "step": 59692
+    },
+    {
+      "epoch": 55.69,
+      "learning_rate": 4.303917910447761e-05,
+      "loss": 0.0,
+      "step": 59696
+    },
+    {
+      "epoch": 55.69,
+      "learning_rate": 4.303871268656717e-05,
+      "loss": 0.0,
+      "step": 59700
+    },
+    {
+      "epoch": 55.69,
+      "learning_rate": 4.3038246268656716e-05,
+      "loss": 0.0,
+      "step": 59704
+    },
+    {
+      "epoch": 55.7,
+      "learning_rate": 4.303777985074627e-05,
+      "loss": 0.0,
+      "step": 59708
+    },
+    {
+      "epoch": 55.7,
+      "learning_rate": 4.3037313432835826e-05,
+      "loss": 0.0001,
+      "step": 59712
+    },
+    {
+      "epoch": 55.71,
+      "learning_rate": 4.3036847014925374e-05,
+      "loss": 0.0,
+      "step": 59716
+    },
+    {
+      "epoch": 55.71,
+      "learning_rate": 4.303638059701492e-05,
+      "loss": 0.0,
+      "step": 59720
+    },
+    {
+      "epoch": 55.71,
+      "learning_rate": 4.3035914179104484e-05,
+      "loss": 0.0,
+      "step": 59724
+    },
+    {
+      "epoch": 55.72,
+      "learning_rate": 4.303544776119403e-05,
+      "loss": 0.0037,
+      "step": 59728
+    },
+    {
+      "epoch": 55.72,
+      "learning_rate": 4.303498134328359e-05,
+      "loss": 0.0,
+      "step": 59732
+    },
+    {
+      "epoch": 55.72,
+      "learning_rate": 4.3034514925373135e-05,
+      "loss": 0.0,
+      "step": 59736
+    },
+    {
+      "epoch": 55.73,
+      "learning_rate": 4.303404850746269e-05,
+      "loss": 0.0,
+      "step": 59740
+    },
+    {
+      "epoch": 55.73,
+      "learning_rate": 4.3033582089552245e-05,
+      "loss": 0.0,
+      "step": 59744
+    },
+    {
+      "epoch": 55.73,
+      "learning_rate": 4.303311567164179e-05,
+      "loss": 0.0002,
+      "step": 59748
+    },
+    {
+      "epoch": 55.74,
+      "learning_rate": 4.303264925373134e-05,
+      "loss": 0.0,
+      "step": 59752
+    },
+    {
+      "epoch": 55.74,
+      "learning_rate": 4.30321828358209e-05,
+      "loss": 0.0,
+      "step": 59756
+    },
+    {
+      "epoch": 55.75,
+      "learning_rate": 4.303171641791045e-05,
+      "loss": 0.0,
+      "step": 59760
+    },
+    {
+      "epoch": 55.75,
+      "learning_rate": 4.303125e-05,
+      "loss": 0.0,
+      "step": 59764
+    },
+    {
+      "epoch": 55.75,
+      "learning_rate": 4.3030783582089554e-05,
+      "loss": 0.0,
+      "step": 59768
+    },
+    {
+      "epoch": 55.76,
+      "learning_rate": 4.303031716417911e-05,
+      "loss": 0.0001,
+      "step": 59772
+    },
+    {
+      "epoch": 55.76,
+      "learning_rate": 4.302985074626866e-05,
+      "loss": 0.0,
+      "step": 59776
+    },
+    {
+      "epoch": 55.76,
+      "learning_rate": 4.302938432835821e-05,
+      "loss": 0.0,
+      "step": 59780
+    },
+    {
+      "epoch": 55.77,
+      "learning_rate": 4.302891791044777e-05,
+      "loss": 0.0,
+      "step": 59784
+    },
+    {
+      "epoch": 55.77,
+      "learning_rate": 4.3028451492537315e-05,
+      "loss": 0.0,
+      "step": 59788
+    },
+    {
+      "epoch": 55.78,
+      "learning_rate": 4.302798507462687e-05,
+      "loss": 0.0,
+      "step": 59792
+    },
+    {
+      "epoch": 55.78,
+      "learning_rate": 4.302751865671642e-05,
+      "loss": 0.0001,
+      "step": 59796
+    },
+    {
+      "epoch": 55.78,
+      "learning_rate": 4.302705223880597e-05,
+      "loss": 0.0,
+      "step": 59800
+    },
+    {
+      "epoch": 55.79,
+      "learning_rate": 4.302658582089553e-05,
+      "loss": 0.0,
+      "step": 59804
+    },
+    {
+      "epoch": 55.79,
+      "learning_rate": 4.3026119402985076e-05,
+      "loss": 0.0001,
+      "step": 59808
+    },
+    {
+      "epoch": 55.79,
+      "learning_rate": 4.3025652985074624e-05,
+      "loss": 0.0,
+      "step": 59812
+    },
+    {
+      "epoch": 55.8,
+      "learning_rate": 4.3025186567164186e-05,
+      "loss": 0.0007,
+      "step": 59816
+    },
+    {
+      "epoch": 55.8,
+      "learning_rate": 4.3024720149253734e-05,
+      "loss": 0.0,
+      "step": 59820
+    },
+    {
+      "epoch": 55.81,
+      "learning_rate": 4.302425373134328e-05,
+      "loss": 0.0,
+      "step": 59824
+    },
+    {
+      "epoch": 55.81,
+      "learning_rate": 4.302378731343284e-05,
+      "loss": 0.0003,
+      "step": 59828
+    },
+    {
+      "epoch": 55.81,
+      "learning_rate": 4.302332089552239e-05,
+      "loss": 0.0,
+      "step": 59832
+    },
+    {
+      "epoch": 55.82,
+      "learning_rate": 4.302285447761194e-05,
+      "loss": 0.0,
+      "step": 59836
+    },
+    {
+      "epoch": 55.82,
+      "learning_rate": 4.3022388059701495e-05,
+      "loss": 0.0004,
+      "step": 59840
+    },
+    {
+      "epoch": 55.82,
+      "learning_rate": 4.302192164179105e-05,
+      "loss": 0.0001,
+      "step": 59844
+    },
+    {
+      "epoch": 55.83,
+      "learning_rate": 4.30214552238806e-05,
+      "loss": 0.0001,
+      "step": 59848
+    },
+    {
+      "epoch": 55.83,
+      "learning_rate": 4.302098880597015e-05,
+      "loss": 0.0,
+      "step": 59852
+    },
+    {
+      "epoch": 55.84,
+      "learning_rate": 4.30205223880597e-05,
+      "loss": 0.0,
+      "step": 59856
+    },
+    {
+      "epoch": 55.84,
+      "learning_rate": 4.3020055970149256e-05,
+      "loss": 0.0,
+      "step": 59860
+    },
+    {
+      "epoch": 55.84,
+      "learning_rate": 4.301958955223881e-05,
+      "loss": 0.0,
+      "step": 59864
+    },
+    {
+      "epoch": 55.85,
+      "learning_rate": 4.301912313432836e-05,
+      "loss": 0.0015,
+      "step": 59868
+    },
+    {
+      "epoch": 55.85,
+      "learning_rate": 4.301865671641791e-05,
+      "loss": 0.0,
+      "step": 59872
+    },
+    {
+      "epoch": 55.85,
+      "learning_rate": 4.301819029850747e-05,
+      "loss": 0.0,
+      "step": 59876
+    },
+    {
+      "epoch": 55.86,
+      "learning_rate": 4.301772388059702e-05,
+      "loss": 0.0,
+      "step": 59880
+    },
+    {
+      "epoch": 55.86,
+      "learning_rate": 4.3017257462686565e-05,
+      "loss": 0.0001,
+      "step": 59884
+    },
+    {
+      "epoch": 55.87,
+      "learning_rate": 4.301679104477612e-05,
+      "loss": 0.0001,
+      "step": 59888
+    },
+    {
+      "epoch": 55.87,
+      "learning_rate": 4.3016324626865675e-05,
+      "loss": 0.0,
+      "step": 59892
+    },
+    {
+      "epoch": 55.87,
+      "learning_rate": 4.301585820895523e-05,
+      "loss": 0.0,
+      "step": 59896
+    },
+    {
+      "epoch": 55.88,
+      "learning_rate": 4.301539179104478e-05,
+      "loss": 0.0001,
+      "step": 59900
+    },
+    {
+      "epoch": 55.88,
+      "learning_rate": 4.301492537313433e-05,
+      "loss": 0.0,
+      "step": 59904
+    },
+    {
+      "epoch": 55.88,
+      "learning_rate": 4.301445895522389e-05,
+      "loss": 0.0,
+      "step": 59908
+    },
+    {
+      "epoch": 55.89,
+      "learning_rate": 4.3013992537313436e-05,
+      "loss": 0.0009,
+      "step": 59912
+    },
+    {
+      "epoch": 55.89,
+      "learning_rate": 4.3013526119402984e-05,
+      "loss": 0.0,
+      "step": 59916
+    },
+    {
+      "epoch": 55.9,
+      "learning_rate": 4.301305970149254e-05,
+      "loss": 0.0,
+      "step": 59920
+    },
+    {
+      "epoch": 55.9,
+      "learning_rate": 4.3012593283582094e-05,
+      "loss": 0.0,
+      "step": 59924
+    },
+    {
+      "epoch": 55.9,
+      "learning_rate": 4.301212686567164e-05,
+      "loss": 0.0,
+      "step": 59928
+    },
+    {
+      "epoch": 55.91,
+      "learning_rate": 4.30116604477612e-05,
+      "loss": 0.0,
+      "step": 59932
+    },
+    {
+      "epoch": 55.91,
+      "learning_rate": 4.301119402985075e-05,
+      "loss": 0.0,
+      "step": 59936
+    },
+    {
+      "epoch": 55.91,
+      "learning_rate": 4.30107276119403e-05,
+      "loss": 0.0,
+      "step": 59940
+    },
+    {
+      "epoch": 55.92,
+      "learning_rate": 4.3010261194029855e-05,
+      "loss": 0.0,
+      "step": 59944
+    },
+    {
+      "epoch": 55.92,
+      "learning_rate": 4.30097947761194e-05,
+      "loss": 0.0,
+      "step": 59948
+    },
+    {
+      "epoch": 55.93,
+      "learning_rate": 4.300932835820896e-05,
+      "loss": 0.0,
+      "step": 59952
+    },
+    {
+      "epoch": 55.93,
+      "learning_rate": 4.300886194029851e-05,
+      "loss": 0.0,
+      "step": 59956
+    },
+    {
+      "epoch": 55.93,
+      "learning_rate": 4.300839552238806e-05,
+      "loss": 0.0,
+      "step": 59960
+    },
+    {
+      "epoch": 55.94,
+      "learning_rate": 4.300792910447761e-05,
+      "loss": 0.0,
+      "step": 59964
+    },
+    {
+      "epoch": 55.94,
+      "learning_rate": 4.300746268656717e-05,
+      "loss": 0.0008,
+      "step": 59968
+    },
+    {
+      "epoch": 55.94,
+      "learning_rate": 4.300699626865672e-05,
+      "loss": 0.0001,
+      "step": 59972
+    },
+    {
+      "epoch": 55.95,
+      "learning_rate": 4.300652985074627e-05,
+      "loss": 0.0,
+      "step": 59976
+    },
+    {
+      "epoch": 55.95,
+      "learning_rate": 4.300606343283582e-05,
+      "loss": 0.0,
+      "step": 59980
+    },
+    {
+      "epoch": 55.96,
+      "learning_rate": 4.300559701492538e-05,
+      "loss": 0.0,
+      "step": 59984
+    },
+    {
+      "epoch": 55.96,
+      "learning_rate": 4.3005130597014925e-05,
+      "loss": 0.0004,
+      "step": 59988
+    },
+    {
+      "epoch": 55.96,
+      "learning_rate": 4.300466417910448e-05,
+      "loss": 0.0,
+      "step": 59992
+    },
+    {
+      "epoch": 55.97,
+      "learning_rate": 4.3004197761194035e-05,
+      "loss": 0.0,
+      "step": 59996
+    },
+    {
+      "epoch": 55.97,
+      "learning_rate": 4.300373134328358e-05,
+      "loss": 0.0,
+      "step": 60000
+    },
+    {
+      "epoch": 55.97,
+      "eval_exact_match": 0.7514506769825918,
+      "eval_exec": 0.7843326885880078,
+      "eval_loss": 0.5125755667686462,
+      "eval_runtime": 1138.056,
+      "eval_samples_per_second": 0.909,
+      "step": 60000
+    },
+    {
+      "epoch": 55.97,
+      "learning_rate": 4.300326492537314e-05,
+      "loss": 0.0,
+      "step": 60004
+    },
+    {
+      "epoch": 55.98,
+      "learning_rate": 4.3002798507462686e-05,
+      "loss": 0.0,
+      "step": 60008
+    },
+    {
+      "epoch": 55.98,
+      "learning_rate": 4.300233208955224e-05,
+      "loss": 0.0001,
+      "step": 60012
+    },
+    {
+      "epoch": 55.98,
+      "learning_rate": 4.3001865671641796e-05,
+      "loss": 0.0,
+      "step": 60016
+    },
+    {
+      "epoch": 55.99,
+      "learning_rate": 4.3001399253731344e-05,
+      "loss": 0.0,
+      "step": 60020
+    },
+    {
+      "epoch": 55.99,
+      "learning_rate": 4.300093283582089e-05,
+      "loss": 0.0,
+      "step": 60024
+    },
+    {
+      "epoch": 56.0,
+      "learning_rate": 4.3000466417910454e-05,
+      "loss": 0.0,
+      "step": 60028
+    },
+    {
+      "epoch": 56.0,
+      "learning_rate": 4.3e-05,
+      "loss": 0.0,
+      "step": 60032
+    },
+    {
+      "epoch": 56.0,
+      "learning_rate": 4.299953358208955e-05,
+      "loss": 0.0,
+      "step": 60036
+    },
+    {
+      "epoch": 56.01,
+      "learning_rate": 4.2999067164179105e-05,
+      "loss": 0.0001,
+      "step": 60040
+    },
+    {
+      "epoch": 56.01,
+      "learning_rate": 4.299860074626866e-05,
+      "loss": 0.0001,
+      "step": 60044
+    },
+    {
+      "epoch": 56.01,
+      "learning_rate": 4.299813432835821e-05,
+      "loss": 0.0014,
+      "step": 60048
+    },
+    {
+      "epoch": 56.02,
+      "learning_rate": 4.299766791044776e-05,
+      "loss": 0.0,
+      "step": 60052
+    },
+    {
+      "epoch": 56.02,
+      "learning_rate": 4.299720149253732e-05,
+      "loss": 0.0,
+      "step": 60056
+    },
+    {
+      "epoch": 56.03,
+      "learning_rate": 4.299673507462687e-05,
+      "loss": 0.0,
+      "step": 60060
+    },
+    {
+      "epoch": 56.03,
+      "learning_rate": 4.299626865671642e-05,
+      "loss": 0.0011,
+      "step": 60064
+    },
+    {
+      "epoch": 56.03,
+      "learning_rate": 4.299580223880597e-05,
+      "loss": 0.0005,
+      "step": 60068
+    },
+    {
+      "epoch": 56.04,
+      "learning_rate": 4.299533582089553e-05,
+      "loss": 0.0,
+      "step": 60072
+    },
+    {
+      "epoch": 56.04,
+      "learning_rate": 4.299486940298508e-05,
+      "loss": 0.0,
+      "step": 60076
+    },
+    {
+      "epoch": 56.04,
+      "learning_rate": 4.299440298507463e-05,
+      "loss": 0.0,
+      "step": 60080
+    },
+    {
+      "epoch": 56.05,
+      "learning_rate": 4.299393656716418e-05,
+      "loss": 0.0,
+      "step": 60084
+    },
+    {
+      "epoch": 56.05,
+      "learning_rate": 4.2993470149253736e-05,
+      "loss": 0.0,
+      "step": 60088
+    },
+    {
+      "epoch": 56.06,
+      "learning_rate": 4.2993003731343285e-05,
+      "loss": 0.0,
+      "step": 60092
+    },
+    {
+      "epoch": 56.06,
+      "learning_rate": 4.299253731343284e-05,
+      "loss": 0.0,
+      "step": 60096
+    },
+    {
+      "epoch": 56.06,
+      "learning_rate": 4.299207089552239e-05,
+      "loss": 0.0,
+      "step": 60100
+    },
+    {
+      "epoch": 56.07,
+      "learning_rate": 4.299160447761194e-05,
+      "loss": 0.0,
+      "step": 60104
+    },
+    {
+      "epoch": 56.07,
+      "learning_rate": 4.29911380597015e-05,
+      "loss": 0.0,
+      "step": 60108
+    },
+    {
+      "epoch": 56.07,
+      "learning_rate": 4.2990671641791046e-05,
+      "loss": 0.0005,
+      "step": 60112
+    },
+    {
+      "epoch": 56.08,
+      "learning_rate": 4.29902052238806e-05,
+      "loss": 0.0001,
+      "step": 60116
+    },
+    {
+      "epoch": 56.08,
+      "learning_rate": 4.2989738805970155e-05,
+      "loss": 0.0,
+      "step": 60120
+    },
+    {
+      "epoch": 56.09,
+      "learning_rate": 4.2989272388059704e-05,
+      "loss": 0.0,
+      "step": 60124
+    },
+    {
+      "epoch": 56.09,
+      "learning_rate": 4.298880597014925e-05,
+      "loss": 0.0001,
+      "step": 60128
+    },
+    {
+      "epoch": 56.09,
+      "learning_rate": 4.2988339552238807e-05,
+      "loss": 0.0,
+      "step": 60132
+    },
+    {
+      "epoch": 56.1,
+      "learning_rate": 4.298787313432836e-05,
+      "loss": 0.0008,
+      "step": 60136
+    },
+    {
+      "epoch": 56.1,
+      "learning_rate": 4.298740671641791e-05,
+      "loss": 0.0001,
+      "step": 60140
+    },
+    {
+      "epoch": 56.1,
+      "learning_rate": 4.2986940298507465e-05,
+      "loss": 0.0009,
+      "step": 60144
+    },
+    {
+      "epoch": 56.11,
+      "learning_rate": 4.298647388059702e-05,
+      "loss": 0.0,
+      "step": 60148
+    },
+    {
+      "epoch": 56.11,
+      "learning_rate": 4.298600746268657e-05,
+      "loss": 0.0021,
+      "step": 60152
+    },
+    {
+      "epoch": 56.12,
+      "learning_rate": 4.298554104477612e-05,
+      "loss": 0.0009,
+      "step": 60156
+    },
+    {
+      "epoch": 56.12,
+      "learning_rate": 4.298507462686567e-05,
+      "loss": 0.0,
+      "step": 60160
+    },
+    {
+      "epoch": 56.12,
+      "learning_rate": 4.2984608208955226e-05,
+      "loss": 0.0001,
+      "step": 60164
+    },
+    {
+      "epoch": 56.13,
+      "learning_rate": 4.298414179104478e-05,
+      "loss": 0.0008,
+      "step": 60168
+    },
+    {
+      "epoch": 56.13,
+      "learning_rate": 4.298367537313433e-05,
+      "loss": 0.0,
+      "step": 60172
+    },
+    {
+      "epoch": 56.13,
+      "learning_rate": 4.2983208955223883e-05,
+      "loss": 0.0,
+      "step": 60176
+    },
+    {
+      "epoch": 56.14,
+      "learning_rate": 4.298274253731344e-05,
+      "loss": 0.0,
+      "step": 60180
+    },
+    {
+      "epoch": 56.14,
+      "learning_rate": 4.2982276119402987e-05,
+      "loss": 0.0,
+      "step": 60184
+    },
+    {
+      "epoch": 56.15,
+      "learning_rate": 4.2981809701492535e-05,
+      "loss": 0.0,
+      "step": 60188
+    },
+    {
+      "epoch": 56.15,
+      "learning_rate": 4.298134328358209e-05,
+      "loss": 0.0,
+      "step": 60192
+    },
+    {
+      "epoch": 56.15,
+      "learning_rate": 4.2980876865671644e-05,
+      "loss": 0.0,
+      "step": 60196
+    },
+    {
+      "epoch": 56.16,
+      "learning_rate": 4.298041044776119e-05,
+      "loss": 0.0001,
+      "step": 60200
+    },
+    {
+      "epoch": 56.16,
+      "learning_rate": 4.297994402985075e-05,
+      "loss": 0.001,
+      "step": 60204
+    },
+    {
+      "epoch": 56.16,
+      "learning_rate": 4.29794776119403e-05,
+      "loss": 0.0,
+      "step": 60208
+    },
+    {
+      "epoch": 56.17,
+      "learning_rate": 4.297901119402985e-05,
+      "loss": 0.0,
+      "step": 60212
+    },
+    {
+      "epoch": 56.17,
+      "learning_rate": 4.2978544776119405e-05,
+      "loss": 0.0,
+      "step": 60216
+    },
+    {
+      "epoch": 56.18,
+      "learning_rate": 4.2978078358208954e-05,
+      "loss": 0.0,
+      "step": 60220
+    },
+    {
+      "epoch": 56.18,
+      "learning_rate": 4.2977611940298515e-05,
+      "loss": 0.0,
+      "step": 60224
+    },
+    {
+      "epoch": 56.18,
+      "learning_rate": 4.297714552238806e-05,
+      "loss": 0.0019,
+      "step": 60228
+    },
+    {
+      "epoch": 56.19,
+      "learning_rate": 4.297667910447761e-05,
+      "loss": 0.0,
+      "step": 60232
+    },
+    {
+      "epoch": 56.19,
+      "learning_rate": 4.2976212686567166e-05,
+      "loss": 0.0,
+      "step": 60236
+    },
+    {
+      "epoch": 56.19,
+      "learning_rate": 4.297574626865672e-05,
+      "loss": 0.0001,
+      "step": 60240
+    },
+    {
+      "epoch": 56.2,
+      "learning_rate": 4.297527985074627e-05,
+      "loss": 0.0001,
+      "step": 60244
+    },
+    {
+      "epoch": 56.2,
+      "learning_rate": 4.2974813432835824e-05,
+      "loss": 0.0,
+      "step": 60248
+    },
+    {
+      "epoch": 56.21,
+      "learning_rate": 4.297434701492537e-05,
+      "loss": 0.0,
+      "step": 60252
+    },
+    {
+      "epoch": 56.21,
+      "learning_rate": 4.297388059701493e-05,
+      "loss": 0.0056,
+      "step": 60256
+    },
+    {
+      "epoch": 56.21,
+      "learning_rate": 4.297341417910448e-05,
+      "loss": 0.0,
+      "step": 60260
+    },
+    {
+      "epoch": 56.22,
+      "learning_rate": 4.297294776119403e-05,
+      "loss": 0.0,
+      "step": 60264
+    },
+    {
+      "epoch": 56.22,
+      "learning_rate": 4.2972481343283585e-05,
+      "loss": 0.0,
+      "step": 60268
+    },
+    {
+      "epoch": 56.22,
+      "learning_rate": 4.297201492537314e-05,
+      "loss": 0.0,
+      "step": 60272
+    },
+    {
+      "epoch": 56.23,
+      "learning_rate": 4.297154850746269e-05,
+      "loss": 0.0,
+      "step": 60276
+    },
+    {
+      "epoch": 56.23,
+      "learning_rate": 4.2971082089552237e-05,
+      "loss": 0.0,
+      "step": 60280
+    },
+    {
+      "epoch": 56.24,
+      "learning_rate": 4.29706156716418e-05,
+      "loss": 0.0002,
+      "step": 60284
+    },
+    {
+      "epoch": 56.24,
+      "learning_rate": 4.2970149253731346e-05,
+      "loss": 0.0,
+      "step": 60288
+    },
+    {
+      "epoch": 56.24,
+      "learning_rate": 4.2969682835820894e-05,
+      "loss": 0.0,
+      "step": 60292
+    },
+    {
+      "epoch": 56.25,
+      "learning_rate": 4.296921641791045e-05,
+      "loss": 0.0,
+      "step": 60296
+    },
+    {
+      "epoch": 56.25,
+      "learning_rate": 4.2968750000000004e-05,
+      "loss": 0.0,
+      "step": 60300
+    },
+    {
+      "epoch": 56.25,
+      "learning_rate": 4.296828358208955e-05,
+      "loss": 0.0,
+      "step": 60304
+    },
+    {
+      "epoch": 56.26,
+      "learning_rate": 4.296781716417911e-05,
+      "loss": 0.0,
+      "step": 60308
+    },
+    {
+      "epoch": 56.26,
+      "learning_rate": 4.2967350746268655e-05,
+      "loss": 0.0,
+      "step": 60312
+    },
+    {
+      "epoch": 56.26,
+      "learning_rate": 4.296688432835821e-05,
+      "loss": 0.0,
+      "step": 60316
+    },
+    {
+      "epoch": 56.27,
+      "learning_rate": 4.2966417910447765e-05,
+      "loss": 0.0,
+      "step": 60320
+    },
+    {
+      "epoch": 56.27,
+      "learning_rate": 4.2965951492537313e-05,
+      "loss": 0.0,
+      "step": 60324
+    },
+    {
+      "epoch": 56.28,
+      "learning_rate": 4.296548507462687e-05,
+      "loss": 0.0,
+      "step": 60328
+    },
+    {
+      "epoch": 56.28,
+      "learning_rate": 4.296501865671642e-05,
+      "loss": 0.0,
+      "step": 60332
+    },
+    {
+      "epoch": 56.28,
+      "learning_rate": 4.296455223880597e-05,
+      "loss": 0.0005,
+      "step": 60336
+    },
+    {
+      "epoch": 56.29,
+      "learning_rate": 4.296408582089552e-05,
+      "loss": 0.0,
+      "step": 60340
+    },
+    {
+      "epoch": 56.29,
+      "learning_rate": 4.296361940298508e-05,
+      "loss": 0.0,
+      "step": 60344
+    },
+    {
+      "epoch": 56.29,
+      "learning_rate": 4.296315298507463e-05,
+      "loss": 0.0,
+      "step": 60348
+    },
+    {
+      "epoch": 56.3,
+      "learning_rate": 4.296268656716418e-05,
+      "loss": 0.0,
+      "step": 60352
+    },
+    {
+      "epoch": 56.3,
+      "learning_rate": 4.296222014925373e-05,
+      "loss": 0.0,
+      "step": 60356
+    },
+    {
+      "epoch": 56.31,
+      "learning_rate": 4.296175373134329e-05,
+      "loss": 0.0004,
+      "step": 60360
+    },
+    {
+      "epoch": 56.31,
+      "learning_rate": 4.2961287313432835e-05,
+      "loss": 0.0,
+      "step": 60364
+    },
+    {
+      "epoch": 56.31,
+      "learning_rate": 4.296082089552239e-05,
+      "loss": 0.0,
+      "step": 60368
+    },
+    {
+      "epoch": 56.32,
+      "learning_rate": 4.296035447761194e-05,
+      "loss": 0.0,
+      "step": 60372
+    },
+    {
+      "epoch": 56.32,
+      "learning_rate": 4.295988805970149e-05,
+      "loss": 0.0,
+      "step": 60376
+    },
+    {
+      "epoch": 56.32,
+      "learning_rate": 4.295942164179105e-05,
+      "loss": 0.0,
+      "step": 60380
+    },
+    {
+      "epoch": 56.33,
+      "learning_rate": 4.2958955223880596e-05,
+      "loss": 0.0,
+      "step": 60384
+    },
+    {
+      "epoch": 56.33,
+      "learning_rate": 4.295848880597015e-05,
+      "loss": 0.001,
+      "step": 60388
+    },
+    {
+      "epoch": 56.34,
+      "learning_rate": 4.2958022388059706e-05,
+      "loss": 0.0,
+      "step": 60392
+    },
+    {
+      "epoch": 56.34,
+      "learning_rate": 4.2957555970149254e-05,
+      "loss": 0.0,
+      "step": 60396
+    },
+    {
+      "epoch": 56.34,
+      "learning_rate": 4.295708955223881e-05,
+      "loss": 0.0032,
+      "step": 60400
+    },
+    {
+      "epoch": 56.35,
+      "learning_rate": 4.2956623134328364e-05,
+      "loss": 0.0,
+      "step": 60404
+    },
+    {
+      "epoch": 56.35,
+      "learning_rate": 4.295615671641791e-05,
+      "loss": 0.0,
+      "step": 60408
+    },
+    {
+      "epoch": 56.35,
+      "learning_rate": 4.295569029850747e-05,
+      "loss": 0.0,
+      "step": 60412
+    },
+    {
+      "epoch": 56.36,
+      "learning_rate": 4.2955223880597015e-05,
+      "loss": 0.0001,
+      "step": 60416
+    },
+    {
+      "epoch": 56.36,
+      "learning_rate": 4.295475746268657e-05,
+      "loss": 0.0008,
+      "step": 60420
+    },
+    {
+      "epoch": 56.37,
+      "learning_rate": 4.2954291044776125e-05,
+      "loss": 0.0,
+      "step": 60424
+    },
+    {
+      "epoch": 56.37,
+      "learning_rate": 4.295382462686567e-05,
+      "loss": 0.0,
+      "step": 60428
+    },
+    {
+      "epoch": 56.37,
+      "learning_rate": 4.295335820895522e-05,
+      "loss": 0.0,
+      "step": 60432
+    },
+    {
+      "epoch": 56.38,
+      "learning_rate": 4.295289179104478e-05,
+      "loss": 0.001,
+      "step": 60436
+    },
+    {
+      "epoch": 56.38,
+      "learning_rate": 4.295242537313433e-05,
+      "loss": 0.0001,
+      "step": 60440
+    },
+    {
+      "epoch": 56.38,
+      "learning_rate": 4.295195895522388e-05,
+      "loss": 0.0003,
+      "step": 60444
+    },
+    {
+      "epoch": 56.39,
+      "learning_rate": 4.2951492537313434e-05,
+      "loss": 0.0,
+      "step": 60448
+    },
+    {
+      "epoch": 56.39,
+      "learning_rate": 4.295102611940299e-05,
+      "loss": 0.0,
+      "step": 60452
+    },
+    {
+      "epoch": 56.4,
+      "learning_rate": 4.295055970149254e-05,
+      "loss": 0.0,
+      "step": 60456
+    },
+    {
+      "epoch": 56.4,
+      "learning_rate": 4.295009328358209e-05,
+      "loss": 0.0,
+      "step": 60460
+    },
+    {
+      "epoch": 56.4,
+      "learning_rate": 4.294962686567165e-05,
+      "loss": 0.0,
+      "step": 60464
+    },
+    {
+      "epoch": 56.41,
+      "learning_rate": 4.2949160447761195e-05,
+      "loss": 0.0,
+      "step": 60468
+    },
+    {
+      "epoch": 56.41,
+      "learning_rate": 4.294869402985075e-05,
+      "loss": 0.0,
+      "step": 60472
+    },
+    {
+      "epoch": 56.41,
+      "learning_rate": 4.29482276119403e-05,
+      "loss": 0.0,
+      "step": 60476
+    },
+    {
+      "epoch": 56.42,
+      "learning_rate": 4.294776119402985e-05,
+      "loss": 0.0004,
+      "step": 60480
+    },
+    {
+      "epoch": 56.42,
+      "learning_rate": 4.294729477611941e-05,
+      "loss": 0.0088,
+      "step": 60484
+    },
+    {
+      "epoch": 56.43,
+      "learning_rate": 4.2946828358208956e-05,
+      "loss": 0.0,
+      "step": 60488
+    },
+    {
+      "epoch": 56.43,
+      "learning_rate": 4.2946361940298504e-05,
+      "loss": 0.0,
+      "step": 60492
+    },
+    {
+      "epoch": 56.43,
+      "learning_rate": 4.2945895522388066e-05,
+      "loss": 0.0009,
+      "step": 60496
+    },
+    {
+      "epoch": 56.44,
+      "learning_rate": 4.2945429104477614e-05,
+      "loss": 0.0002,
+      "step": 60500
+    },
+    {
+      "epoch": 56.44,
+      "eval_exact_match": 0.7524177949709865,
+      "eval_exec": 0.7833655705996132,
+      "eval_loss": 0.480731725692749,
+      "eval_runtime": 1132.1694,
+      "eval_samples_per_second": 0.913,
+      "step": 60500
+    },
+    {
+      "epoch": 56.44,
+      "learning_rate": 4.294496268656716e-05,
+      "loss": 0.0,
+      "step": 60504
+    },
+    {
+      "epoch": 56.44,
+      "learning_rate": 4.294449626865672e-05,
+      "loss": 0.0,
+      "step": 60508
+    },
+    {
+      "epoch": 56.45,
+      "learning_rate": 4.294402985074627e-05,
+      "loss": 0.0,
+      "step": 60512
+    },
+    {
+      "epoch": 56.45,
+      "learning_rate": 4.294356343283582e-05,
+      "loss": 0.0,
+      "step": 60516
+    },
+    {
+      "epoch": 56.46,
+      "learning_rate": 4.2943097014925375e-05,
+      "loss": 0.0,
+      "step": 60520
+    },
+    {
+      "epoch": 56.46,
+      "learning_rate": 4.294263059701493e-05,
+      "loss": 0.0,
+      "step": 60524
+    },
+    {
+      "epoch": 56.46,
+      "learning_rate": 4.294216417910448e-05,
+      "loss": 0.0,
+      "step": 60528
+    },
+    {
+      "epoch": 56.47,
+      "learning_rate": 4.294169776119403e-05,
+      "loss": 0.0,
+      "step": 60532
+    },
+    {
+      "epoch": 56.47,
+      "learning_rate": 4.294123134328358e-05,
+      "loss": 0.0,
+      "step": 60536
+    },
+    {
+      "epoch": 56.47,
+      "learning_rate": 4.2940764925373136e-05,
+      "loss": 0.0,
+      "step": 60540
+    },
+    {
+      "epoch": 56.48,
+      "learning_rate": 4.294029850746269e-05,
+      "loss": 0.0001,
+      "step": 60544
+    },
+    {
+      "epoch": 56.48,
+      "learning_rate": 4.293983208955224e-05,
+      "loss": 0.0,
+      "step": 60548
+    },
+    {
+      "epoch": 56.49,
+      "learning_rate": 4.2939365671641794e-05,
+      "loss": 0.0,
+      "step": 60552
+    },
+    {
+      "epoch": 56.49,
+      "learning_rate": 4.293889925373135e-05,
+      "loss": 0.0013,
+      "step": 60556
+    },
+    {
+      "epoch": 56.49,
+      "learning_rate": 4.29384328358209e-05,
+      "loss": 0.0,
+      "step": 60560
+    },
+    {
+      "epoch": 56.5,
+      "learning_rate": 4.293796641791045e-05,
+      "loss": 0.0,
+      "step": 60564
+    },
+    {
+      "epoch": 56.5,
+      "learning_rate": 4.29375e-05,
+      "loss": 0.0,
+      "step": 60568
+    },
+    {
+      "epoch": 56.5,
+      "learning_rate": 4.2937033582089555e-05,
+      "loss": 0.0,
+      "step": 60572
+    },
+    {
+      "epoch": 56.51,
+      "learning_rate": 4.293656716417911e-05,
+      "loss": 0.0,
+      "step": 60576
+    },
+    {
+      "epoch": 56.51,
+      "learning_rate": 4.293610074626866e-05,
+      "loss": 0.0,
+      "step": 60580
+    },
+    {
+      "epoch": 56.51,
+      "learning_rate": 4.293563432835821e-05,
+      "loss": 0.0,
+      "step": 60584
+    },
+    {
+      "epoch": 56.52,
+      "learning_rate": 4.293516791044777e-05,
+      "loss": 0.0,
+      "step": 60588
+    },
+    {
+      "epoch": 56.52,
+      "learning_rate": 4.2934701492537316e-05,
+      "loss": 0.0103,
+      "step": 60592
+    },
+    {
+      "epoch": 56.53,
+      "learning_rate": 4.2934235074626864e-05,
+      "loss": 0.0007,
+      "step": 60596
+    },
+    {
+      "epoch": 56.53,
+      "learning_rate": 4.293376865671642e-05,
+      "loss": 0.0001,
+      "step": 60600
+    },
+    {
+      "epoch": 56.53,
+      "learning_rate": 4.2933302238805974e-05,
+      "loss": 0.0,
+      "step": 60604
+    },
+    {
+      "epoch": 56.54,
+      "learning_rate": 4.293283582089552e-05,
+      "loss": 0.0,
+      "step": 60608
+    },
+    {
+      "epoch": 56.54,
+      "learning_rate": 4.293236940298508e-05,
+      "loss": 0.0007,
+      "step": 60612
+    },
+    {
+      "epoch": 56.54,
+      "learning_rate": 4.293190298507463e-05,
+      "loss": 0.0,
+      "step": 60616
+    },
+    {
+      "epoch": 56.55,
+      "learning_rate": 4.293143656716418e-05,
+      "loss": 0.0002,
+      "step": 60620
+    },
+    {
+      "epoch": 56.55,
+      "learning_rate": 4.2930970149253735e-05,
+      "loss": 0.0011,
+      "step": 60624
+    },
+    {
+      "epoch": 56.56,
+      "learning_rate": 4.293050373134328e-05,
+      "loss": 0.0001,
+      "step": 60628
+    },
+    {
+      "epoch": 56.56,
+      "learning_rate": 4.293003731343284e-05,
+      "loss": 0.0,
+      "step": 60632
+    },
+    {
+      "epoch": 56.56,
+      "learning_rate": 4.292957089552239e-05,
+      "loss": 0.0,
+      "step": 60636
+    },
+    {
+      "epoch": 56.57,
+      "learning_rate": 4.292910447761194e-05,
+      "loss": 0.0005,
+      "step": 60640
+    },
+    {
+      "epoch": 56.57,
+      "learning_rate": 4.292863805970149e-05,
+      "loss": 0.0,
+      "step": 60644
+    },
+    {
+      "epoch": 56.57,
+      "learning_rate": 4.292817164179105e-05,
+      "loss": 0.0,
+      "step": 60648
+    },
+    {
+      "epoch": 56.58,
+      "learning_rate": 4.29277052238806e-05,
+      "loss": 0.0,
+      "step": 60652
+    },
+    {
+      "epoch": 56.58,
+      "learning_rate": 4.292723880597015e-05,
+      "loss": 0.0,
+      "step": 60656
+    },
+    {
+      "epoch": 56.59,
+      "learning_rate": 4.29267723880597e-05,
+      "loss": 0.0002,
+      "step": 60660
+    },
+    {
+      "epoch": 56.59,
+      "learning_rate": 4.292630597014926e-05,
+      "loss": 0.0,
+      "step": 60664
+    },
+    {
+      "epoch": 56.59,
+      "learning_rate": 4.2925839552238805e-05,
+      "loss": 0.0001,
+      "step": 60668
+    },
+    {
+      "epoch": 56.6,
+      "learning_rate": 4.292537313432836e-05,
+      "loss": 0.0,
+      "step": 60672
+    },
+    {
+      "epoch": 56.6,
+      "learning_rate": 4.2924906716417915e-05,
+      "loss": 0.0001,
+      "step": 60676
+    },
+    {
+      "epoch": 56.6,
+      "learning_rate": 4.292444029850746e-05,
+      "loss": 0.0,
+      "step": 60680
+    },
+    {
+      "epoch": 56.61,
+      "learning_rate": 4.292397388059702e-05,
+      "loss": 0.0,
+      "step": 60684
+    },
+    {
+      "epoch": 56.61,
+      "learning_rate": 4.2923507462686566e-05,
+      "loss": 0.0,
+      "step": 60688
+    },
+    {
+      "epoch": 56.62,
+      "learning_rate": 4.292304104477612e-05,
+      "loss": 0.0,
+      "step": 60692
+    },
+    {
+      "epoch": 56.62,
+      "learning_rate": 4.2922574626865676e-05,
+      "loss": 0.0,
+      "step": 60696
+    },
+    {
+      "epoch": 56.62,
+      "learning_rate": 4.2922108208955224e-05,
+      "loss": 0.0001,
+      "step": 60700
+    },
+    {
+      "epoch": 56.63,
+      "learning_rate": 4.292164179104477e-05,
+      "loss": 0.0,
+      "step": 60704
+    },
+    {
+      "epoch": 56.63,
+      "learning_rate": 4.2921175373134334e-05,
+      "loss": 0.0,
+      "step": 60708
+    },
+    {
+      "epoch": 56.63,
+      "learning_rate": 4.292070895522388e-05,
+      "loss": 0.0,
+      "step": 60712
+    },
+    {
+      "epoch": 56.64,
+      "learning_rate": 4.292024253731344e-05,
+      "loss": 0.0,
+      "step": 60716
+    },
+    {
+      "epoch": 56.64,
+      "learning_rate": 4.2919776119402985e-05,
+      "loss": 0.0,
+      "step": 60720
+    },
+    {
+      "epoch": 56.65,
+      "learning_rate": 4.291930970149254e-05,
+      "loss": 0.0003,
+      "step": 60724
+    },
+    {
+      "epoch": 56.65,
+      "learning_rate": 4.2918843283582095e-05,
+      "loss": 0.0016,
+      "step": 60728
+    },
+    {
+      "epoch": 56.65,
+      "learning_rate": 4.291837686567164e-05,
+      "loss": 0.0197,
+      "step": 60732
+    },
+    {
+      "epoch": 56.66,
+      "learning_rate": 4.29179104477612e-05,
+      "loss": 0.0,
+      "step": 60736
+    },
+    {
+      "epoch": 56.66,
+      "learning_rate": 4.291744402985075e-05,
+      "loss": 0.0001,
+      "step": 60740
+    },
+    {
+      "epoch": 56.66,
+      "learning_rate": 4.29169776119403e-05,
+      "loss": 0.0,
+      "step": 60744
+    },
+    {
+      "epoch": 56.67,
+      "learning_rate": 4.291651119402985e-05,
+      "loss": 0.0002,
+      "step": 60748
+    },
+    {
+      "epoch": 56.67,
+      "learning_rate": 4.291604477611941e-05,
+      "loss": 0.0,
+      "step": 60752
+    },
+    {
+      "epoch": 56.68,
+      "learning_rate": 4.291557835820896e-05,
+      "loss": 0.0,
+      "step": 60756
+    },
+    {
+      "epoch": 56.68,
+      "learning_rate": 4.291511194029851e-05,
+      "loss": 0.0,
+      "step": 60760
+    },
+    {
+      "epoch": 56.68,
+      "learning_rate": 4.291464552238806e-05,
+      "loss": 0.0,
+      "step": 60764
+    },
+    {
+      "epoch": 56.69,
+      "learning_rate": 4.291417910447762e-05,
+      "loss": 0.0,
+      "step": 60768
+    },
+    {
+      "epoch": 56.69,
+      "learning_rate": 4.2913712686567165e-05,
+      "loss": 0.0,
+      "step": 60772
+    },
+    {
+      "epoch": 56.69,
+      "learning_rate": 4.291324626865672e-05,
+      "loss": 0.0,
+      "step": 60776
+    },
+    {
+      "epoch": 56.7,
+      "learning_rate": 4.291277985074627e-05,
+      "loss": 0.0001,
+      "step": 60780
+    },
+    {
+      "epoch": 56.7,
+      "learning_rate": 4.291231343283582e-05,
+      "loss": 0.0,
+      "step": 60784
+    },
+    {
+      "epoch": 56.71,
+      "learning_rate": 4.291184701492538e-05,
+      "loss": 0.0002,
+      "step": 60788
+    },
+    {
+      "epoch": 56.71,
+      "learning_rate": 4.2911380597014926e-05,
+      "loss": 0.0,
+      "step": 60792
+    },
+    {
+      "epoch": 56.71,
+      "learning_rate": 4.291091417910448e-05,
+      "loss": 0.0,
+      "step": 60796
+    },
+    {
+      "epoch": 56.72,
+      "learning_rate": 4.2910447761194036e-05,
+      "loss": 0.0001,
+      "step": 60800
+    },
+    {
+      "epoch": 56.72,
+      "learning_rate": 4.2909981343283584e-05,
+      "loss": 0.0001,
+      "step": 60804
+    },
+    {
+      "epoch": 56.72,
+      "learning_rate": 4.290951492537313e-05,
+      "loss": 0.0006,
+      "step": 60808
+    },
+    {
+      "epoch": 56.73,
+      "learning_rate": 4.290904850746269e-05,
+      "loss": 0.0,
+      "step": 60812
+    },
+    {
+      "epoch": 56.73,
+      "learning_rate": 4.290858208955224e-05,
+      "loss": 0.0006,
+      "step": 60816
+    },
+    {
+      "epoch": 56.73,
+      "learning_rate": 4.290811567164179e-05,
+      "loss": 0.0,
+      "step": 60820
+    },
+    {
+      "epoch": 56.74,
+      "learning_rate": 4.2907649253731345e-05,
+      "loss": 0.0,
+      "step": 60824
+    },
+    {
+      "epoch": 56.74,
+      "learning_rate": 4.29071828358209e-05,
+      "loss": 0.0,
+      "step": 60828
+    },
+    {
+      "epoch": 56.75,
+      "learning_rate": 4.290671641791045e-05,
+      "loss": 0.0011,
+      "step": 60832
+    },
+    {
+      "epoch": 56.75,
+      "learning_rate": 4.290625e-05,
+      "loss": 0.0,
+      "step": 60836
+    },
+    {
+      "epoch": 56.75,
+      "learning_rate": 4.290578358208955e-05,
+      "loss": 0.0003,
+      "step": 60840
+    },
+    {
+      "epoch": 56.76,
+      "learning_rate": 4.2905317164179106e-05,
+      "loss": 0.0002,
+      "step": 60844
+    },
+    {
+      "epoch": 56.76,
+      "learning_rate": 4.290485074626866e-05,
+      "loss": 0.0001,
+      "step": 60848
+    },
+    {
+      "epoch": 56.76,
+      "learning_rate": 4.290438432835821e-05,
+      "loss": 0.0,
+      "step": 60852
+    },
+    {
+      "epoch": 56.77,
+      "learning_rate": 4.2903917910447764e-05,
+      "loss": 0.0001,
+      "step": 60856
+    },
+    {
+      "epoch": 56.77,
+      "learning_rate": 4.290345149253732e-05,
+      "loss": 0.0,
+      "step": 60860
+    },
+    {
+      "epoch": 56.78,
+      "learning_rate": 4.290298507462687e-05,
+      "loss": 0.0,
+      "step": 60864
+    },
+    {
+      "epoch": 56.78,
+      "learning_rate": 4.2902518656716415e-05,
+      "loss": 0.0,
+      "step": 60868
+    },
+    {
+      "epoch": 56.78,
+      "learning_rate": 4.290205223880597e-05,
+      "loss": 0.001,
+      "step": 60872
+    },
+    {
+      "epoch": 56.79,
+      "learning_rate": 4.2901585820895525e-05,
+      "loss": 0.0,
+      "step": 60876
+    },
+    {
+      "epoch": 56.79,
+      "learning_rate": 4.290111940298508e-05,
+      "loss": 0.0,
+      "step": 60880
+    },
+    {
+      "epoch": 56.79,
+      "learning_rate": 4.290065298507463e-05,
+      "loss": 0.0,
+      "step": 60884
+    },
+    {
+      "epoch": 56.8,
+      "learning_rate": 4.290018656716418e-05,
+      "loss": 0.0,
+      "step": 60888
+    },
+    {
+      "epoch": 56.8,
+      "learning_rate": 4.289972014925374e-05,
+      "loss": 0.0,
+      "step": 60892
+    },
+    {
+      "epoch": 56.81,
+      "learning_rate": 4.2899253731343286e-05,
+      "loss": 0.0004,
+      "step": 60896
+    },
+    {
+      "epoch": 56.81,
+      "learning_rate": 4.2898787313432834e-05,
+      "loss": 0.0001,
+      "step": 60900
+    },
+    {
+      "epoch": 56.81,
+      "learning_rate": 4.2898320895522396e-05,
+      "loss": 0.0,
+      "step": 60904
+    },
+    {
+      "epoch": 56.82,
+      "learning_rate": 4.2897854477611944e-05,
+      "loss": 0.0,
+      "step": 60908
+    },
+    {
+      "epoch": 56.82,
+      "learning_rate": 4.289738805970149e-05,
+      "loss": 0.0002,
+      "step": 60912
+    },
+    {
+      "epoch": 56.82,
+      "learning_rate": 4.289692164179105e-05,
+      "loss": 0.0,
+      "step": 60916
+    },
+    {
+      "epoch": 56.83,
+      "learning_rate": 4.28964552238806e-05,
+      "loss": 0.0001,
+      "step": 60920
+    },
+    {
+      "epoch": 56.83,
+      "learning_rate": 4.289598880597015e-05,
+      "loss": 0.0,
+      "step": 60924
+    },
+    {
+      "epoch": 56.84,
+      "learning_rate": 4.2895522388059705e-05,
+      "loss": 0.0,
+      "step": 60928
+    },
+    {
+      "epoch": 56.84,
+      "learning_rate": 4.289505597014925e-05,
+      "loss": 0.0,
+      "step": 60932
+    },
+    {
+      "epoch": 56.84,
+      "learning_rate": 4.289458955223881e-05,
+      "loss": 0.0001,
+      "step": 60936
+    },
+    {
+      "epoch": 56.85,
+      "learning_rate": 4.289412313432836e-05,
+      "loss": 0.0,
+      "step": 60940
+    },
+    {
+      "epoch": 56.85,
+      "learning_rate": 4.289365671641791e-05,
+      "loss": 0.0001,
+      "step": 60944
+    },
+    {
+      "epoch": 56.85,
+      "learning_rate": 4.2893190298507466e-05,
+      "loss": 0.0,
+      "step": 60948
+    },
+    {
+      "epoch": 56.86,
+      "learning_rate": 4.289272388059702e-05,
+      "loss": 0.0,
+      "step": 60952
+    },
+    {
+      "epoch": 56.86,
+      "learning_rate": 4.289225746268657e-05,
+      "loss": 0.0,
+      "step": 60956
+    },
+    {
+      "epoch": 56.87,
+      "learning_rate": 4.289179104477612e-05,
+      "loss": 0.0,
+      "step": 60960
+    },
+    {
+      "epoch": 56.87,
+      "learning_rate": 4.289132462686568e-05,
+      "loss": 0.0,
+      "step": 60964
+    },
+    {
+      "epoch": 56.87,
+      "learning_rate": 4.289085820895523e-05,
+      "loss": 0.0,
+      "step": 60968
+    },
+    {
+      "epoch": 56.88,
+      "learning_rate": 4.2890391791044775e-05,
+      "loss": 0.0,
+      "step": 60972
+    },
+    {
+      "epoch": 56.88,
+      "learning_rate": 4.288992537313433e-05,
+      "loss": 0.0,
+      "step": 60976
+    },
+    {
+      "epoch": 56.88,
+      "learning_rate": 4.2889458955223885e-05,
+      "loss": 0.0001,
+      "step": 60980
+    },
+    {
+      "epoch": 56.89,
+      "learning_rate": 4.288899253731343e-05,
+      "loss": 0.0001,
+      "step": 60984
+    },
+    {
+      "epoch": 56.89,
+      "learning_rate": 4.288852611940299e-05,
+      "loss": 0.0,
+      "step": 60988
+    },
+    {
+      "epoch": 56.9,
+      "learning_rate": 4.2888059701492536e-05,
+      "loss": 0.0,
+      "step": 60992
+    },
+    {
+      "epoch": 56.9,
+      "learning_rate": 4.288759328358209e-05,
+      "loss": 0.0024,
+      "step": 60996
+    },
+    {
+      "epoch": 56.9,
+      "learning_rate": 4.2887126865671646e-05,
+      "loss": 0.0,
+      "step": 61000
+    },
+    {
+      "epoch": 56.9,
+      "eval_exact_match": 0.7495164410058027,
+      "eval_exec": 0.7775628626692457,
+      "eval_loss": 0.4850376546382904,
+      "eval_runtime": 1127.2294,
+      "eval_samples_per_second": 0.917,
+      "step": 61000
+    },
+    {
+      "epoch": 56.91,
+      "learning_rate": 4.2886660447761194e-05,
+      "loss": 0.0001,
+      "step": 61004
+    },
+    {
+      "epoch": 56.91,
+      "learning_rate": 4.288619402985075e-05,
+      "loss": 0.0,
+      "step": 61008
+    },
+    {
+      "epoch": 56.91,
+      "learning_rate": 4.2885727611940304e-05,
+      "loss": 0.0,
+      "step": 61012
+    },
+    {
+      "epoch": 56.92,
+      "learning_rate": 4.288526119402985e-05,
+      "loss": 0.0,
+      "step": 61016
+    },
+    {
+      "epoch": 56.92,
+      "learning_rate": 4.28847947761194e-05,
+      "loss": 0.0,
+      "step": 61020
+    },
+    {
+      "epoch": 56.93,
+      "learning_rate": 4.288432835820896e-05,
+      "loss": 0.0,
+      "step": 61024
+    },
+    {
+      "epoch": 56.93,
+      "learning_rate": 4.288386194029851e-05,
+      "loss": 0.0,
+      "step": 61028
+    },
+    {
+      "epoch": 56.93,
+      "learning_rate": 4.288339552238806e-05,
+      "loss": 0.0023,
+      "step": 61032
+    },
+    {
+      "epoch": 56.94,
+      "learning_rate": 4.288292910447761e-05,
+      "loss": 0.0001,
+      "step": 61036
+    },
+    {
+      "epoch": 56.94,
+      "learning_rate": 4.288246268656717e-05,
+      "loss": 0.0,
+      "step": 61040
+    },
+    {
+      "epoch": 56.94,
+      "learning_rate": 4.288199626865672e-05,
+      "loss": 0.0,
+      "step": 61044
+    },
+    {
+      "epoch": 56.95,
+      "learning_rate": 4.288152985074627e-05,
+      "loss": 0.0,
+      "step": 61048
+    },
+    {
+      "epoch": 56.95,
+      "learning_rate": 4.288106343283582e-05,
+      "loss": 0.0,
+      "step": 61052
+    },
+    {
+      "epoch": 56.96,
+      "learning_rate": 4.288059701492538e-05,
+      "loss": 0.0,
+      "step": 61056
+    },
+    {
+      "epoch": 56.96,
+      "learning_rate": 4.288013059701493e-05,
+      "loss": 0.0,
+      "step": 61060
+    },
+    {
+      "epoch": 56.96,
+      "learning_rate": 4.287966417910448e-05,
+      "loss": 0.0,
+      "step": 61064
+    },
+    {
+      "epoch": 56.97,
+      "learning_rate": 4.287919776119403e-05,
+      "loss": 0.0008,
+      "step": 61068
+    },
+    {
+      "epoch": 56.97,
+      "learning_rate": 4.2878731343283586e-05,
+      "loss": 0.0,
+      "step": 61072
+    },
+    {
+      "epoch": 56.97,
+      "learning_rate": 4.2878264925373135e-05,
+      "loss": 0.0,
+      "step": 61076
+    },
+    {
+      "epoch": 56.98,
+      "learning_rate": 4.287779850746269e-05,
+      "loss": 0.0,
+      "step": 61080
+    },
+    {
+      "epoch": 56.98,
+      "learning_rate": 4.2877332089552244e-05,
+      "loss": 0.0,
+      "step": 61084
+    },
+    {
+      "epoch": 56.98,
+      "learning_rate": 4.287686567164179e-05,
+      "loss": 0.0,
+      "step": 61088
+    },
+    {
+      "epoch": 56.99,
+      "learning_rate": 4.287639925373135e-05,
+      "loss": 0.0,
+      "step": 61092
+    },
+    {
+      "epoch": 56.99,
+      "learning_rate": 4.2875932835820896e-05,
+      "loss": 0.0001,
+      "step": 61096
+    },
+    {
+      "epoch": 57.0,
+      "learning_rate": 4.287546641791045e-05,
+      "loss": 0.0,
+      "step": 61100
+    },
+    {
+      "epoch": 57.0,
+      "learning_rate": 4.2875000000000005e-05,
+      "loss": 0.0,
+      "step": 61104
+    },
+    {
+      "epoch": 57.0,
+      "learning_rate": 4.2874533582089554e-05,
+      "loss": 0.0003,
+      "step": 61108
+    },
+    {
+      "epoch": 57.01,
+      "learning_rate": 4.28740671641791e-05,
+      "loss": 0.0,
+      "step": 61112
+    },
+    {
+      "epoch": 57.01,
+      "learning_rate": 4.287360074626866e-05,
+      "loss": 0.0178,
+      "step": 61116
+    },
+    {
+      "epoch": 57.01,
+      "learning_rate": 4.287313432835821e-05,
+      "loss": 0.0,
+      "step": 61120
+    },
+    {
+      "epoch": 57.02,
+      "learning_rate": 4.287266791044776e-05,
+      "loss": 0.0,
+      "step": 61124
+    },
+    {
+      "epoch": 57.02,
+      "learning_rate": 4.2872201492537315e-05,
+      "loss": 0.0,
+      "step": 61128
+    },
+    {
+      "epoch": 57.03,
+      "learning_rate": 4.287173507462687e-05,
+      "loss": 0.0001,
+      "step": 61132
+    },
+    {
+      "epoch": 57.03,
+      "learning_rate": 4.287126865671642e-05,
+      "loss": 0.0103,
+      "step": 61136
+    },
+    {
+      "epoch": 57.03,
+      "learning_rate": 4.287080223880597e-05,
+      "loss": 0.0001,
+      "step": 61140
+    },
+    {
+      "epoch": 57.04,
+      "learning_rate": 4.287033582089553e-05,
+      "loss": 0.0,
+      "step": 61144
+    },
+    {
+      "epoch": 57.04,
+      "learning_rate": 4.2869869402985075e-05,
+      "loss": 0.0,
+      "step": 61148
+    },
+    {
+      "epoch": 57.04,
+      "learning_rate": 4.286940298507463e-05,
+      "loss": 0.0,
+      "step": 61152
+    },
+    {
+      "epoch": 57.05,
+      "learning_rate": 4.286893656716418e-05,
+      "loss": 0.0,
+      "step": 61156
+    },
+    {
+      "epoch": 57.05,
+      "learning_rate": 4.2868470149253733e-05,
+      "loss": 0.0006,
+      "step": 61160
+    },
+    {
+      "epoch": 57.06,
+      "learning_rate": 4.286800373134329e-05,
+      "loss": 0.0,
+      "step": 61164
+    },
+    {
+      "epoch": 57.06,
+      "learning_rate": 4.2867537313432836e-05,
+      "loss": 0.0,
+      "step": 61168
+    },
+    {
+      "epoch": 57.06,
+      "learning_rate": 4.2867070895522385e-05,
+      "loss": 0.0,
+      "step": 61172
+    },
+    {
+      "epoch": 57.07,
+      "learning_rate": 4.2866604477611946e-05,
+      "loss": 0.0,
+      "step": 61176
+    },
+    {
+      "epoch": 57.07,
+      "learning_rate": 4.2866138059701494e-05,
+      "loss": 0.0,
+      "step": 61180
+    },
+    {
+      "epoch": 57.07,
+      "learning_rate": 4.286567164179104e-05,
+      "loss": 0.0001,
+      "step": 61184
+    },
+    {
+      "epoch": 57.08,
+      "learning_rate": 4.28652052238806e-05,
+      "loss": 0.0,
+      "step": 61188
+    },
+    {
+      "epoch": 57.08,
+      "learning_rate": 4.286473880597015e-05,
+      "loss": 0.0,
+      "step": 61192
+    },
+    {
+      "epoch": 57.09,
+      "learning_rate": 4.28642723880597e-05,
+      "loss": 0.0,
+      "step": 61196
+    },
+    {
+      "epoch": 57.09,
+      "learning_rate": 4.2863805970149255e-05,
+      "loss": 0.0,
+      "step": 61200
+    },
+    {
+      "epoch": 57.09,
+      "learning_rate": 4.286333955223881e-05,
+      "loss": 0.0011,
+      "step": 61204
+    },
+    {
+      "epoch": 57.1,
+      "learning_rate": 4.2862873134328365e-05,
+      "loss": 0.0,
+      "step": 61208
+    },
+    {
+      "epoch": 57.1,
+      "learning_rate": 4.286240671641791e-05,
+      "loss": 0.0,
+      "step": 61212
+    },
+    {
+      "epoch": 57.1,
+      "learning_rate": 4.286194029850746e-05,
+      "loss": 0.0009,
+      "step": 61216
+    },
+    {
+      "epoch": 57.11,
+      "learning_rate": 4.2861473880597016e-05,
+      "loss": 0.0,
+      "step": 61220
+    },
+    {
+      "epoch": 57.11,
+      "learning_rate": 4.286100746268657e-05,
+      "loss": 0.0,
+      "step": 61224
+    },
+    {
+      "epoch": 57.12,
+      "learning_rate": 4.286054104477612e-05,
+      "loss": 0.0044,
+      "step": 61228
+    },
+    {
+      "epoch": 57.12,
+      "learning_rate": 4.2860074626865674e-05,
+      "loss": 0.0,
+      "step": 61232
+    },
+    {
+      "epoch": 57.12,
+      "learning_rate": 4.285960820895523e-05,
+      "loss": 0.0,
+      "step": 61236
+    },
+    {
+      "epoch": 57.13,
+      "learning_rate": 4.285914179104478e-05,
+      "loss": 0.0,
+      "step": 61240
+    },
+    {
+      "epoch": 57.13,
+      "learning_rate": 4.285867537313433e-05,
+      "loss": 0.0001,
+      "step": 61244
+    },
+    {
+      "epoch": 57.13,
+      "learning_rate": 4.285820895522388e-05,
+      "loss": 0.0,
+      "step": 61248
+    },
+    {
+      "epoch": 57.14,
+      "learning_rate": 4.2857742537313435e-05,
+      "loss": 0.0,
+      "step": 61252
+    },
+    {
+      "epoch": 57.14,
+      "learning_rate": 4.285727611940299e-05,
+      "loss": 0.0001,
+      "step": 61256
+    },
+    {
+      "epoch": 57.15,
+      "learning_rate": 4.285680970149254e-05,
+      "loss": 0.0,
+      "step": 61260
+    },
+    {
+      "epoch": 57.15,
+      "learning_rate": 4.285634328358209e-05,
+      "loss": 0.0,
+      "step": 61264
+    },
+    {
+      "epoch": 57.15,
+      "learning_rate": 4.285587686567165e-05,
+      "loss": 0.0,
+      "step": 61268
+    },
+    {
+      "epoch": 57.16,
+      "learning_rate": 4.2855410447761196e-05,
+      "loss": 0.0,
+      "step": 61272
+    },
+    {
+      "epoch": 57.16,
+      "learning_rate": 4.2854944029850744e-05,
+      "loss": 0.0,
+      "step": 61276
+    },
+    {
+      "epoch": 57.16,
+      "learning_rate": 4.28544776119403e-05,
+      "loss": 0.0,
+      "step": 61280
+    },
+    {
+      "epoch": 57.17,
+      "learning_rate": 4.2854011194029854e-05,
+      "loss": 0.0,
+      "step": 61284
+    },
+    {
+      "epoch": 57.17,
+      "learning_rate": 4.28535447761194e-05,
+      "loss": 0.0,
+      "step": 61288
+    },
+    {
+      "epoch": 57.18,
+      "learning_rate": 4.285307835820896e-05,
+      "loss": 0.0,
+      "step": 61292
+    },
+    {
+      "epoch": 57.18,
+      "learning_rate": 4.285261194029851e-05,
+      "loss": 0.0,
+      "step": 61296
+    },
+    {
+      "epoch": 57.18,
+      "learning_rate": 4.285214552238806e-05,
+      "loss": 0.0,
+      "step": 61300
+    },
+    {
+      "epoch": 57.19,
+      "learning_rate": 4.2851679104477615e-05,
+      "loss": 0.0,
+      "step": 61304
+    },
+    {
+      "epoch": 57.19,
+      "learning_rate": 4.285121268656716e-05,
+      "loss": 0.0,
+      "step": 61308
+    },
+    {
+      "epoch": 57.19,
+      "learning_rate": 4.285074626865672e-05,
+      "loss": 0.0007,
+      "step": 61312
+    },
+    {
+      "epoch": 57.2,
+      "learning_rate": 4.285027985074627e-05,
+      "loss": 0.0,
+      "step": 61316
+    },
+    {
+      "epoch": 57.2,
+      "learning_rate": 4.284981343283582e-05,
+      "loss": 0.0,
+      "step": 61320
+    },
+    {
+      "epoch": 57.21,
+      "learning_rate": 4.284934701492537e-05,
+      "loss": 0.0001,
+      "step": 61324
+    },
+    {
+      "epoch": 57.21,
+      "learning_rate": 4.284888059701493e-05,
+      "loss": 0.0,
+      "step": 61328
+    },
+    {
+      "epoch": 57.21,
+      "learning_rate": 4.284841417910448e-05,
+      "loss": 0.0008,
+      "step": 61332
+    },
+    {
+      "epoch": 57.22,
+      "learning_rate": 4.284794776119403e-05,
+      "loss": 0.0001,
+      "step": 61336
+    },
+    {
+      "epoch": 57.22,
+      "learning_rate": 4.284748134328358e-05,
+      "loss": 0.0001,
+      "step": 61340
+    },
+    {
+      "epoch": 57.22,
+      "learning_rate": 4.284701492537314e-05,
+      "loss": 0.0003,
+      "step": 61344
+    },
+    {
+      "epoch": 57.23,
+      "learning_rate": 4.2846548507462685e-05,
+      "loss": 0.0,
+      "step": 61348
+    },
+    {
+      "epoch": 57.23,
+      "learning_rate": 4.284608208955224e-05,
+      "loss": 0.0,
+      "step": 61352
+    },
+    {
+      "epoch": 57.24,
+      "learning_rate": 4.2845615671641795e-05,
+      "loss": 0.0,
+      "step": 61356
+    },
+    {
+      "epoch": 57.24,
+      "learning_rate": 4.284514925373134e-05,
+      "loss": 0.0001,
+      "step": 61360
+    },
+    {
+      "epoch": 57.24,
+      "learning_rate": 4.28446828358209e-05,
+      "loss": 0.0001,
+      "step": 61364
+    },
+    {
+      "epoch": 57.25,
+      "learning_rate": 4.2844216417910446e-05,
+      "loss": 0.0,
+      "step": 61368
+    },
+    {
+      "epoch": 57.25,
+      "learning_rate": 4.284375000000001e-05,
+      "loss": 0.0,
+      "step": 61372
+    },
+    {
+      "epoch": 57.25,
+      "learning_rate": 4.2843283582089556e-05,
+      "loss": 0.0,
+      "step": 61376
+    },
+    {
+      "epoch": 57.26,
+      "learning_rate": 4.2842817164179104e-05,
+      "loss": 0.0,
+      "step": 61380
+    },
+    {
+      "epoch": 57.26,
+      "learning_rate": 4.284235074626866e-05,
+      "loss": 0.0,
+      "step": 61384
+    },
+    {
+      "epoch": 57.26,
+      "learning_rate": 4.2841884328358214e-05,
+      "loss": 0.0,
+      "step": 61388
+    },
+    {
+      "epoch": 57.27,
+      "learning_rate": 4.284141791044776e-05,
+      "loss": 0.0,
+      "step": 61392
+    },
+    {
+      "epoch": 57.27,
+      "learning_rate": 4.284095149253732e-05,
+      "loss": 0.0001,
+      "step": 61396
+    },
+    {
+      "epoch": 57.28,
+      "learning_rate": 4.2840485074626865e-05,
+      "loss": 0.0001,
+      "step": 61400
+    },
+    {
+      "epoch": 57.28,
+      "learning_rate": 4.284001865671642e-05,
+      "loss": 0.0,
+      "step": 61404
+    },
+    {
+      "epoch": 57.28,
+      "learning_rate": 4.2839552238805975e-05,
+      "loss": 0.0,
+      "step": 61408
+    },
+    {
+      "epoch": 57.29,
+      "learning_rate": 4.283908582089552e-05,
+      "loss": 0.0,
+      "step": 61412
+    },
+    {
+      "epoch": 57.29,
+      "learning_rate": 4.283861940298508e-05,
+      "loss": 0.0,
+      "step": 61416
+    },
+    {
+      "epoch": 57.29,
+      "learning_rate": 4.283815298507463e-05,
+      "loss": 0.0,
+      "step": 61420
+    },
+    {
+      "epoch": 57.3,
+      "learning_rate": 4.283768656716418e-05,
+      "loss": 0.0,
+      "step": 61424
+    },
+    {
+      "epoch": 57.3,
+      "learning_rate": 4.283722014925373e-05,
+      "loss": 0.0,
+      "step": 61428
+    },
+    {
+      "epoch": 57.31,
+      "learning_rate": 4.283675373134329e-05,
+      "loss": 0.0004,
+      "step": 61432
+    },
+    {
+      "epoch": 57.31,
+      "learning_rate": 4.283628731343284e-05,
+      "loss": 0.0001,
+      "step": 61436
+    },
+    {
+      "epoch": 57.31,
+      "learning_rate": 4.283582089552239e-05,
+      "loss": 0.0,
+      "step": 61440
+    },
+    {
+      "epoch": 57.32,
+      "learning_rate": 4.283535447761194e-05,
+      "loss": 0.0004,
+      "step": 61444
+    },
+    {
+      "epoch": 57.32,
+      "learning_rate": 4.28348880597015e-05,
+      "loss": 0.0,
+      "step": 61448
+    },
+    {
+      "epoch": 57.32,
+      "learning_rate": 4.2834421641791045e-05,
+      "loss": 0.0,
+      "step": 61452
+    },
+    {
+      "epoch": 57.33,
+      "learning_rate": 4.28339552238806e-05,
+      "loss": 0.0025,
+      "step": 61456
+    },
+    {
+      "epoch": 57.33,
+      "learning_rate": 4.283348880597015e-05,
+      "loss": 0.0002,
+      "step": 61460
+    },
+    {
+      "epoch": 57.34,
+      "learning_rate": 4.28330223880597e-05,
+      "loss": 0.0,
+      "step": 61464
+    },
+    {
+      "epoch": 57.34,
+      "learning_rate": 4.283255597014926e-05,
+      "loss": 0.0,
+      "step": 61468
+    },
+    {
+      "epoch": 57.34,
+      "learning_rate": 4.2832089552238806e-05,
+      "loss": 0.0,
+      "step": 61472
+    },
+    {
+      "epoch": 57.35,
+      "learning_rate": 4.283162313432836e-05,
+      "loss": 0.0,
+      "step": 61476
+    },
+    {
+      "epoch": 57.35,
+      "learning_rate": 4.2831156716417916e-05,
+      "loss": 0.0,
+      "step": 61480
+    },
+    {
+      "epoch": 57.35,
+      "learning_rate": 4.2830690298507464e-05,
+      "loss": 0.0,
+      "step": 61484
+    },
+    {
+      "epoch": 57.36,
+      "learning_rate": 4.283022388059701e-05,
+      "loss": 0.0,
+      "step": 61488
+    },
+    {
+      "epoch": 57.36,
+      "learning_rate": 4.282975746268657e-05,
+      "loss": 0.0009,
+      "step": 61492
+    },
+    {
+      "epoch": 57.37,
+      "learning_rate": 4.282929104477612e-05,
+      "loss": 0.0001,
+      "step": 61496
+    },
+    {
+      "epoch": 57.37,
+      "learning_rate": 4.282882462686567e-05,
+      "loss": 0.0,
+      "step": 61500
+    },
+    {
+      "epoch": 57.37,
+      "eval_exact_match": 0.7514506769825918,
+      "eval_exec": 0.7843326885880078,
+      "eval_loss": 0.4877242147922516,
+      "eval_runtime": 1153.3063,
+      "eval_samples_per_second": 0.897,
+      "step": 61500
+    },
+    {
+      "epoch": 57.37,
+      "learning_rate": 4.2828358208955225e-05,
+      "loss": 0.0,
+      "step": 61504
+    },
+    {
+      "epoch": 57.38,
+      "learning_rate": 4.282789179104478e-05,
+      "loss": 0.0,
+      "step": 61508
+    },
+    {
+      "epoch": 57.38,
+      "learning_rate": 4.282742537313433e-05,
+      "loss": 0.0001,
+      "step": 61512
+    },
+    {
+      "epoch": 57.38,
+      "learning_rate": 4.282695895522388e-05,
+      "loss": 0.0,
+      "step": 61516
+    },
+    {
+      "epoch": 57.39,
+      "learning_rate": 4.282649253731343e-05,
+      "loss": 0.0,
+      "step": 61520
+    },
+    {
+      "epoch": 57.39,
+      "learning_rate": 4.2826026119402986e-05,
+      "loss": 0.0038,
+      "step": 61524
+    },
+    {
+      "epoch": 57.4,
+      "learning_rate": 4.282555970149254e-05,
+      "loss": 0.0001,
+      "step": 61528
+    },
+    {
+      "epoch": 57.4,
+      "learning_rate": 4.282509328358209e-05,
+      "loss": 0.0,
+      "step": 61532
+    },
+    {
+      "epoch": 57.4,
+      "learning_rate": 4.2824626865671644e-05,
+      "loss": 0.0,
+      "step": 61536
+    },
+    {
+      "epoch": 57.41,
+      "learning_rate": 4.28241604477612e-05,
+      "loss": 0.0,
+      "step": 61540
+    },
+    {
+      "epoch": 57.41,
+      "learning_rate": 4.282369402985075e-05,
+      "loss": 0.0,
+      "step": 61544
+    },
+    {
+      "epoch": 57.41,
+      "learning_rate": 4.28232276119403e-05,
+      "loss": 0.0,
+      "step": 61548
+    },
+    {
+      "epoch": 57.42,
+      "learning_rate": 4.282276119402985e-05,
+      "loss": 0.0,
+      "step": 61552
+    },
+    {
+      "epoch": 57.42,
+      "learning_rate": 4.2822294776119405e-05,
+      "loss": 0.0,
+      "step": 61556
+    },
+    {
+      "epoch": 57.43,
+      "learning_rate": 4.282182835820896e-05,
+      "loss": 0.0,
+      "step": 61560
+    },
+    {
+      "epoch": 57.43,
+      "learning_rate": 4.282136194029851e-05,
+      "loss": 0.0,
+      "step": 61564
+    },
+    {
+      "epoch": 57.43,
+      "learning_rate": 4.282089552238806e-05,
+      "loss": 0.0,
+      "step": 61568
+    },
+    {
+      "epoch": 57.44,
+      "learning_rate": 4.282042910447762e-05,
+      "loss": 0.0,
+      "step": 61572
+    },
+    {
+      "epoch": 57.44,
+      "learning_rate": 4.2819962686567166e-05,
+      "loss": 0.0,
+      "step": 61576
+    },
+    {
+      "epoch": 57.44,
+      "learning_rate": 4.2819496268656714e-05,
+      "loss": 0.0002,
+      "step": 61580
+    },
+    {
+      "epoch": 57.45,
+      "learning_rate": 4.2819029850746276e-05,
+      "loss": 0.0007,
+      "step": 61584
+    },
+    {
+      "epoch": 57.45,
+      "learning_rate": 4.2818563432835824e-05,
+      "loss": 0.0,
+      "step": 61588
+    },
+    {
+      "epoch": 57.46,
+      "learning_rate": 4.281809701492537e-05,
+      "loss": 0.0,
+      "step": 61592
+    },
+    {
+      "epoch": 57.46,
+      "learning_rate": 4.281763059701493e-05,
+      "loss": 0.0,
+      "step": 61596
+    },
+    {
+      "epoch": 57.46,
+      "learning_rate": 4.281716417910448e-05,
+      "loss": 0.0,
+      "step": 61600
+    },
+    {
+      "epoch": 57.47,
+      "learning_rate": 4.281669776119403e-05,
+      "loss": 0.0001,
+      "step": 61604
+    },
+    {
+      "epoch": 57.47,
+      "learning_rate": 4.2816231343283585e-05,
+      "loss": 0.0,
+      "step": 61608
+    },
+    {
+      "epoch": 57.47,
+      "learning_rate": 4.281576492537313e-05,
+      "loss": 0.0,
+      "step": 61612
+    },
+    {
+      "epoch": 57.48,
+      "learning_rate": 4.281529850746269e-05,
+      "loss": 0.0,
+      "step": 61616
+    },
+    {
+      "epoch": 57.48,
+      "learning_rate": 4.281483208955224e-05,
+      "loss": 0.0,
+      "step": 61620
+    },
+    {
+      "epoch": 57.49,
+      "learning_rate": 4.281436567164179e-05,
+      "loss": 0.0003,
+      "step": 61624
+    },
+    {
+      "epoch": 57.49,
+      "learning_rate": 4.2813899253731346e-05,
+      "loss": 0.0,
+      "step": 61628
+    },
+    {
+      "epoch": 57.49,
+      "learning_rate": 4.28134328358209e-05,
+      "loss": 0.0,
+      "step": 61632
+    },
+    {
+      "epoch": 57.5,
+      "learning_rate": 4.281296641791045e-05,
+      "loss": 0.0,
+      "step": 61636
+    },
+    {
+      "epoch": 57.5,
+      "learning_rate": 4.28125e-05,
+      "loss": 0.0,
+      "step": 61640
+    },
+    {
+      "epoch": 57.5,
+      "learning_rate": 4.281203358208956e-05,
+      "loss": 0.0,
+      "step": 61644
+    },
+    {
+      "epoch": 57.51,
+      "learning_rate": 4.281156716417911e-05,
+      "loss": 0.0035,
+      "step": 61648
+    },
+    {
+      "epoch": 57.51,
+      "learning_rate": 4.2811100746268655e-05,
+      "loss": 0.0,
+      "step": 61652
+    },
+    {
+      "epoch": 57.51,
+      "learning_rate": 4.281063432835821e-05,
+      "loss": 0.0002,
+      "step": 61656
+    },
+    {
+      "epoch": 57.52,
+      "learning_rate": 4.2810167910447765e-05,
+      "loss": 0.0,
+      "step": 61660
+    },
+    {
+      "epoch": 57.52,
+      "learning_rate": 4.280970149253731e-05,
+      "loss": 0.0,
+      "step": 61664
+    },
+    {
+      "epoch": 57.53,
+      "learning_rate": 4.280923507462687e-05,
+      "loss": 0.0,
+      "step": 61668
+    },
+    {
+      "epoch": 57.53,
+      "learning_rate": 4.2808768656716416e-05,
+      "loss": 0.0,
+      "step": 61672
+    },
+    {
+      "epoch": 57.53,
+      "learning_rate": 4.280830223880597e-05,
+      "loss": 0.0001,
+      "step": 61676
+    },
+    {
+      "epoch": 57.54,
+      "learning_rate": 4.2807835820895526e-05,
+      "loss": 0.0,
+      "step": 61680
+    },
+    {
+      "epoch": 57.54,
+      "learning_rate": 4.2807369402985074e-05,
+      "loss": 0.0,
+      "step": 61684
+    },
+    {
+      "epoch": 57.54,
+      "learning_rate": 4.280690298507463e-05,
+      "loss": 0.0,
+      "step": 61688
+    },
+    {
+      "epoch": 57.55,
+      "learning_rate": 4.2806436567164184e-05,
+      "loss": 0.0,
+      "step": 61692
+    },
+    {
+      "epoch": 57.55,
+      "learning_rate": 4.280597014925373e-05,
+      "loss": 0.0,
+      "step": 61696
+    },
+    {
+      "epoch": 57.56,
+      "learning_rate": 4.280550373134329e-05,
+      "loss": 0.0,
+      "step": 61700
+    },
+    {
+      "epoch": 57.56,
+      "learning_rate": 4.280503731343284e-05,
+      "loss": 0.0,
+      "step": 61704
+    },
+    {
+      "epoch": 57.56,
+      "learning_rate": 4.280457089552239e-05,
+      "loss": 0.0,
+      "step": 61708
+    },
+    {
+      "epoch": 57.57,
+      "learning_rate": 4.2804104477611945e-05,
+      "loss": 0.0,
+      "step": 61712
+    },
+    {
+      "epoch": 57.57,
+      "learning_rate": 4.280363805970149e-05,
+      "loss": 0.0,
+      "step": 61716
+    },
+    {
+      "epoch": 57.57,
+      "learning_rate": 4.280317164179105e-05,
+      "loss": 0.0,
+      "step": 61720
+    },
+    {
+      "epoch": 57.58,
+      "learning_rate": 4.28027052238806e-05,
+      "loss": 0.0001,
+      "step": 61724
+    },
+    {
+      "epoch": 57.58,
+      "learning_rate": 4.280223880597015e-05,
+      "loss": 0.0,
+      "step": 61728
+    },
+    {
+      "epoch": 57.59,
+      "learning_rate": 4.28017723880597e-05,
+      "loss": 0.0,
+      "step": 61732
+    },
+    {
+      "epoch": 57.59,
+      "learning_rate": 4.280130597014926e-05,
+      "loss": 0.0,
+      "step": 61736
+    },
+    {
+      "epoch": 57.59,
+      "learning_rate": 4.280083955223881e-05,
+      "loss": 0.0,
+      "step": 61740
+    },
+    {
+      "epoch": 57.6,
+      "learning_rate": 4.280037313432836e-05,
+      "loss": 0.0142,
+      "step": 61744
+    },
+    {
+      "epoch": 57.6,
+      "learning_rate": 4.279990671641791e-05,
+      "loss": 0.0,
+      "step": 61748
+    },
+    {
+      "epoch": 57.6,
+      "learning_rate": 4.279944029850747e-05,
+      "loss": 0.0003,
+      "step": 61752
+    },
+    {
+      "epoch": 57.61,
+      "learning_rate": 4.2798973880597015e-05,
+      "loss": 0.0,
+      "step": 61756
+    },
+    {
+      "epoch": 57.61,
+      "learning_rate": 4.279850746268657e-05,
+      "loss": 0.0003,
+      "step": 61760
+    },
+    {
+      "epoch": 57.62,
+      "learning_rate": 4.2798041044776125e-05,
+      "loss": 0.0,
+      "step": 61764
+    },
+    {
+      "epoch": 57.62,
+      "learning_rate": 4.279757462686567e-05,
+      "loss": 0.0,
+      "step": 61768
+    },
+    {
+      "epoch": 57.62,
+      "learning_rate": 4.279710820895523e-05,
+      "loss": 0.0,
+      "step": 61772
+    },
+    {
+      "epoch": 57.63,
+      "learning_rate": 4.2796641791044776e-05,
+      "loss": 0.0002,
+      "step": 61776
+    },
+    {
+      "epoch": 57.63,
+      "learning_rate": 4.279617537313433e-05,
+      "loss": 0.0009,
+      "step": 61780
+    },
+    {
+      "epoch": 57.63,
+      "learning_rate": 4.2795708955223886e-05,
+      "loss": 0.0,
+      "step": 61784
+    },
+    {
+      "epoch": 57.64,
+      "learning_rate": 4.2795242537313434e-05,
+      "loss": 0.0,
+      "step": 61788
+    },
+    {
+      "epoch": 57.64,
+      "learning_rate": 4.279477611940298e-05,
+      "loss": 0.0003,
+      "step": 61792
+    },
+    {
+      "epoch": 57.65,
+      "learning_rate": 4.2794309701492544e-05,
+      "loss": 0.0002,
+      "step": 61796
+    },
+    {
+      "epoch": 57.65,
+      "learning_rate": 4.279384328358209e-05,
+      "loss": 0.0,
+      "step": 61800
+    },
+    {
+      "epoch": 57.65,
+      "learning_rate": 4.279337686567164e-05,
+      "loss": 0.0,
+      "step": 61804
+    },
+    {
+      "epoch": 57.66,
+      "learning_rate": 4.2792910447761195e-05,
+      "loss": 0.0,
+      "step": 61808
+    },
+    {
+      "epoch": 57.66,
+      "learning_rate": 4.279244402985075e-05,
+      "loss": 0.0,
+      "step": 61812
+    },
+    {
+      "epoch": 57.66,
+      "learning_rate": 4.27919776119403e-05,
+      "loss": 0.0,
+      "step": 61816
+    },
+    {
+      "epoch": 57.67,
+      "learning_rate": 4.279151119402985e-05,
+      "loss": 0.0,
+      "step": 61820
+    },
+    {
+      "epoch": 57.67,
+      "learning_rate": 4.279104477611941e-05,
+      "loss": 0.0003,
+      "step": 61824
+    },
+    {
+      "epoch": 57.68,
+      "learning_rate": 4.2790578358208956e-05,
+      "loss": 0.0,
+      "step": 61828
+    },
+    {
+      "epoch": 57.68,
+      "learning_rate": 4.279011194029851e-05,
+      "loss": 0.0001,
+      "step": 61832
+    },
+    {
+      "epoch": 57.68,
+      "learning_rate": 4.278964552238806e-05,
+      "loss": 0.0,
+      "step": 61836
+    },
+    {
+      "epoch": 57.69,
+      "learning_rate": 4.2789179104477614e-05,
+      "loss": 0.0,
+      "step": 61840
+    },
+    {
+      "epoch": 57.69,
+      "learning_rate": 4.278871268656717e-05,
+      "loss": 0.0011,
+      "step": 61844
+    },
+    {
+      "epoch": 57.69,
+      "learning_rate": 4.278824626865672e-05,
+      "loss": 0.0,
+      "step": 61848
+    },
+    {
+      "epoch": 57.7,
+      "learning_rate": 4.2787779850746265e-05,
+      "loss": 0.0,
+      "step": 61852
+    },
+    {
+      "epoch": 57.7,
+      "learning_rate": 4.2787313432835827e-05,
+      "loss": 0.0005,
+      "step": 61856
+    },
+    {
+      "epoch": 57.71,
+      "learning_rate": 4.2786847014925375e-05,
+      "loss": 0.0,
+      "step": 61860
+    },
+    {
+      "epoch": 57.71,
+      "learning_rate": 4.278638059701493e-05,
+      "loss": 0.0001,
+      "step": 61864
+    },
+    {
+      "epoch": 57.71,
+      "learning_rate": 4.278591417910448e-05,
+      "loss": 0.0,
+      "step": 61868
+    },
+    {
+      "epoch": 57.72,
+      "learning_rate": 4.278544776119403e-05,
+      "loss": 0.0003,
+      "step": 61872
+    },
+    {
+      "epoch": 57.72,
+      "learning_rate": 4.278498134328359e-05,
+      "loss": 0.0014,
+      "step": 61876
+    },
+    {
+      "epoch": 57.72,
+      "learning_rate": 4.2784514925373136e-05,
+      "loss": 0.0,
+      "step": 61880
+    },
+    {
+      "epoch": 57.73,
+      "learning_rate": 4.278404850746269e-05,
+      "loss": 0.0,
+      "step": 61884
+    },
+    {
+      "epoch": 57.73,
+      "learning_rate": 4.2783582089552246e-05,
+      "loss": 0.0,
+      "step": 61888
+    },
+    {
+      "epoch": 57.73,
+      "learning_rate": 4.2783115671641794e-05,
+      "loss": 0.0,
+      "step": 61892
+    },
+    {
+      "epoch": 57.74,
+      "learning_rate": 4.278264925373134e-05,
+      "loss": 0.0,
+      "step": 61896
+    },
+    {
+      "epoch": 57.74,
+      "learning_rate": 4.27821828358209e-05,
+      "loss": 0.0,
+      "step": 61900
+    },
+    {
+      "epoch": 57.75,
+      "learning_rate": 4.278171641791045e-05,
+      "loss": 0.0,
+      "step": 61904
+    },
+    {
+      "epoch": 57.75,
+      "learning_rate": 4.278125e-05,
+      "loss": 0.0,
+      "step": 61908
+    },
+    {
+      "epoch": 57.75,
+      "learning_rate": 4.2780783582089555e-05,
+      "loss": 0.0,
+      "step": 61912
+    },
+    {
+      "epoch": 57.76,
+      "learning_rate": 4.278031716417911e-05,
+      "loss": 0.0,
+      "step": 61916
+    },
+    {
+      "epoch": 57.76,
+      "learning_rate": 4.277985074626866e-05,
+      "loss": 0.0,
+      "step": 61920
+    },
+    {
+      "epoch": 57.76,
+      "learning_rate": 4.277938432835821e-05,
+      "loss": 0.0,
+      "step": 61924
+    },
+    {
+      "epoch": 57.77,
+      "learning_rate": 4.277891791044776e-05,
+      "loss": 0.0024,
+      "step": 61928
+    },
+    {
+      "epoch": 57.77,
+      "learning_rate": 4.2778451492537316e-05,
+      "loss": 0.0,
+      "step": 61932
+    },
+    {
+      "epoch": 57.78,
+      "learning_rate": 4.277798507462687e-05,
+      "loss": 0.0019,
+      "step": 61936
+    },
+    {
+      "epoch": 57.78,
+      "learning_rate": 4.277751865671642e-05,
+      "loss": 0.0,
+      "step": 61940
+    },
+    {
+      "epoch": 57.78,
+      "learning_rate": 4.2777052238805974e-05,
+      "loss": 0.0001,
+      "step": 61944
+    },
+    {
+      "epoch": 57.79,
+      "learning_rate": 4.277658582089553e-05,
+      "loss": 0.0,
+      "step": 61948
+    },
+    {
+      "epoch": 57.79,
+      "learning_rate": 4.2776119402985077e-05,
+      "loss": 0.0,
+      "step": 61952
+    },
+    {
+      "epoch": 57.79,
+      "learning_rate": 4.2775652985074625e-05,
+      "loss": 0.0,
+      "step": 61956
+    },
+    {
+      "epoch": 57.8,
+      "learning_rate": 4.277518656716418e-05,
+      "loss": 0.0,
+      "step": 61960
+    },
+    {
+      "epoch": 57.8,
+      "learning_rate": 4.2774720149253735e-05,
+      "loss": 0.0,
+      "step": 61964
+    },
+    {
+      "epoch": 57.81,
+      "learning_rate": 4.277425373134328e-05,
+      "loss": 0.0,
+      "step": 61968
+    },
+    {
+      "epoch": 57.81,
+      "learning_rate": 4.277378731343284e-05,
+      "loss": 0.0,
+      "step": 61972
+    },
+    {
+      "epoch": 57.81,
+      "learning_rate": 4.277332089552239e-05,
+      "loss": 0.0,
+      "step": 61976
+    },
+    {
+      "epoch": 57.82,
+      "learning_rate": 4.277285447761194e-05,
+      "loss": 0.0,
+      "step": 61980
+    },
+    {
+      "epoch": 57.82,
+      "learning_rate": 4.2772388059701496e-05,
+      "loss": 0.0,
+      "step": 61984
+    },
+    {
+      "epoch": 57.82,
+      "learning_rate": 4.2771921641791044e-05,
+      "loss": 0.0,
+      "step": 61988
+    },
+    {
+      "epoch": 57.83,
+      "learning_rate": 4.27714552238806e-05,
+      "loss": 0.0,
+      "step": 61992
+    },
+    {
+      "epoch": 57.83,
+      "learning_rate": 4.2770988805970153e-05,
+      "loss": 0.0,
+      "step": 61996
+    },
+    {
+      "epoch": 57.84,
+      "learning_rate": 4.27705223880597e-05,
+      "loss": 0.0,
+      "step": 62000
+    },
+    {
+      "epoch": 57.84,
+      "eval_exact_match": 0.7514506769825918,
+      "eval_exec": 0.7872340425531915,
+      "eval_loss": 0.4987480640411377,
+      "eval_runtime": 1158.3298,
+      "eval_samples_per_second": 0.893,
+      "step": 62000
+    },
+    {
+      "epoch": 57.84,
+      "learning_rate": 4.2770055970149257e-05,
+      "loss": 0.0,
+      "step": 62004
+    },
+    {
+      "epoch": 57.84,
+      "learning_rate": 4.276958955223881e-05,
+      "loss": 0.0,
+      "step": 62008
+    },
+    {
+      "epoch": 57.85,
+      "learning_rate": 4.276912313432836e-05,
+      "loss": 0.0,
+      "step": 62012
+    },
+    {
+      "epoch": 57.85,
+      "learning_rate": 4.2768656716417914e-05,
+      "loss": 0.0003,
+      "step": 62016
+    },
+    {
+      "epoch": 57.85,
+      "learning_rate": 4.276819029850746e-05,
+      "loss": 0.0,
+      "step": 62020
+    },
+    {
+      "epoch": 57.86,
+      "learning_rate": 4.276772388059702e-05,
+      "loss": 0.0035,
+      "step": 62024
+    },
+    {
+      "epoch": 57.86,
+      "learning_rate": 4.276725746268657e-05,
+      "loss": 0.0,
+      "step": 62028
+    },
+    {
+      "epoch": 57.87,
+      "learning_rate": 4.276679104477612e-05,
+      "loss": 0.0,
+      "step": 62032
+    },
+    {
+      "epoch": 57.87,
+      "learning_rate": 4.2766324626865675e-05,
+      "loss": 0.0,
+      "step": 62036
+    },
+    {
+      "epoch": 57.87,
+      "learning_rate": 4.276585820895523e-05,
+      "loss": 0.0,
+      "step": 62040
+    },
+    {
+      "epoch": 57.88,
+      "learning_rate": 4.276539179104478e-05,
+      "loss": 0.0,
+      "step": 62044
+    },
+    {
+      "epoch": 57.88,
+      "learning_rate": 4.276492537313433e-05,
+      "loss": 0.0,
+      "step": 62048
+    },
+    {
+      "epoch": 57.88,
+      "learning_rate": 4.276445895522389e-05,
+      "loss": 0.0,
+      "step": 62052
+    },
+    {
+      "epoch": 57.89,
+      "learning_rate": 4.2763992537313436e-05,
+      "loss": 0.0,
+      "step": 62056
+    },
+    {
+      "epoch": 57.89,
+      "learning_rate": 4.2763526119402985e-05,
+      "loss": 0.0,
+      "step": 62060
+    },
+    {
+      "epoch": 57.9,
+      "learning_rate": 4.276305970149254e-05,
+      "loss": 0.0,
+      "step": 62064
+    },
+    {
+      "epoch": 57.9,
+      "learning_rate": 4.2762593283582094e-05,
+      "loss": 0.0,
+      "step": 62068
+    },
+    {
+      "epoch": 57.9,
+      "learning_rate": 4.276212686567164e-05,
+      "loss": 0.0,
+      "step": 62072
+    },
+    {
+      "epoch": 57.91,
+      "learning_rate": 4.27616604477612e-05,
+      "loss": 0.0,
+      "step": 62076
+    },
+    {
+      "epoch": 57.91,
+      "learning_rate": 4.2761194029850746e-05,
+      "loss": 0.0,
+      "step": 62080
+    },
+    {
+      "epoch": 57.91,
+      "learning_rate": 4.27607276119403e-05,
+      "loss": 0.0001,
+      "step": 62084
+    },
+    {
+      "epoch": 57.92,
+      "learning_rate": 4.2760261194029855e-05,
+      "loss": 0.0001,
+      "step": 62088
+    },
+    {
+      "epoch": 57.92,
+      "learning_rate": 4.2759794776119404e-05,
+      "loss": 0.0,
+      "step": 62092
+    },
+    {
+      "epoch": 57.93,
+      "learning_rate": 4.275932835820896e-05,
+      "loss": 0.0,
+      "step": 62096
+    },
+    {
+      "epoch": 57.93,
+      "learning_rate": 4.275886194029851e-05,
+      "loss": 0.0,
+      "step": 62100
+    },
+    {
+      "epoch": 57.93,
+      "learning_rate": 4.275839552238806e-05,
+      "loss": 0.0027,
+      "step": 62104
+    },
+    {
+      "epoch": 57.94,
+      "learning_rate": 4.275792910447761e-05,
+      "loss": 0.0,
+      "step": 62108
+    },
+    {
+      "epoch": 57.94,
+      "learning_rate": 4.275746268656717e-05,
+      "loss": 0.0,
+      "step": 62112
+    },
+    {
+      "epoch": 57.94,
+      "learning_rate": 4.275699626865672e-05,
+      "loss": 0.0001,
+      "step": 62116
+    },
+    {
+      "epoch": 57.95,
+      "learning_rate": 4.275652985074627e-05,
+      "loss": 0.0,
+      "step": 62120
+    },
+    {
+      "epoch": 57.95,
+      "learning_rate": 4.275606343283582e-05,
+      "loss": 0.0,
+      "step": 62124
+    },
+    {
+      "epoch": 57.96,
+      "learning_rate": 4.275559701492538e-05,
+      "loss": 0.0,
+      "step": 62128
+    },
+    {
+      "epoch": 57.96,
+      "learning_rate": 4.2755130597014925e-05,
+      "loss": 0.0001,
+      "step": 62132
+    },
+    {
+      "epoch": 57.96,
+      "learning_rate": 4.275466417910448e-05,
+      "loss": 0.0001,
+      "step": 62136
+    },
+    {
+      "epoch": 57.97,
+      "learning_rate": 4.275419776119403e-05,
+      "loss": 0.0,
+      "step": 62140
+    },
+    {
+      "epoch": 57.97,
+      "learning_rate": 4.2753731343283583e-05,
+      "loss": 0.0,
+      "step": 62144
+    },
+    {
+      "epoch": 57.97,
+      "learning_rate": 4.275326492537314e-05,
+      "loss": 0.0014,
+      "step": 62148
+    },
+    {
+      "epoch": 57.98,
+      "learning_rate": 4.2752798507462686e-05,
+      "loss": 0.0,
+      "step": 62152
+    },
+    {
+      "epoch": 57.98,
+      "learning_rate": 4.275233208955224e-05,
+      "loss": 0.0,
+      "step": 62156
+    },
+    {
+      "epoch": 57.98,
+      "learning_rate": 4.2751865671641796e-05,
+      "loss": 0.0001,
+      "step": 62160
+    },
+    {
+      "epoch": 57.99,
+      "learning_rate": 4.2751399253731344e-05,
+      "loss": 0.0025,
+      "step": 62164
+    },
+    {
+      "epoch": 57.99,
+      "learning_rate": 4.275093283582089e-05,
+      "loss": 0.0,
+      "step": 62168
+    },
+    {
+      "epoch": 58.0,
+      "learning_rate": 4.275046641791045e-05,
+      "loss": 0.003,
+      "step": 62172
+    },
+    {
+      "epoch": 58.0,
+      "learning_rate": 4.275e-05,
+      "loss": 0.0001,
+      "step": 62176
+    },
+    {
+      "epoch": 58.0,
+      "learning_rate": 4.274953358208956e-05,
+      "loss": 0.0,
+      "step": 62180
+    },
+    {
+      "epoch": 58.01,
+      "learning_rate": 4.2749067164179105e-05,
+      "loss": 0.0007,
+      "step": 62184
+    },
+    {
+      "epoch": 58.01,
+      "learning_rate": 4.274860074626866e-05,
+      "loss": 0.0,
+      "step": 62188
+    },
+    {
+      "epoch": 58.01,
+      "learning_rate": 4.2748134328358215e-05,
+      "loss": 0.0001,
+      "step": 62192
+    },
+    {
+      "epoch": 58.02,
+      "learning_rate": 4.274766791044776e-05,
+      "loss": 0.0,
+      "step": 62196
+    },
+    {
+      "epoch": 58.02,
+      "learning_rate": 4.274720149253731e-05,
+      "loss": 0.0002,
+      "step": 62200
+    },
+    {
+      "epoch": 58.03,
+      "learning_rate": 4.274673507462687e-05,
+      "loss": 0.0,
+      "step": 62204
+    },
+    {
+      "epoch": 58.03,
+      "learning_rate": 4.274626865671642e-05,
+      "loss": 0.0,
+      "step": 62208
+    },
+    {
+      "epoch": 58.03,
+      "learning_rate": 4.274580223880597e-05,
+      "loss": 0.0,
+      "step": 62212
+    },
+    {
+      "epoch": 58.04,
+      "learning_rate": 4.2745335820895524e-05,
+      "loss": 0.0,
+      "step": 62216
+    },
+    {
+      "epoch": 58.04,
+      "learning_rate": 4.274486940298508e-05,
+      "loss": 0.0,
+      "step": 62220
+    },
+    {
+      "epoch": 58.04,
+      "learning_rate": 4.274440298507463e-05,
+      "loss": 0.0,
+      "step": 62224
+    },
+    {
+      "epoch": 58.05,
+      "learning_rate": 4.274393656716418e-05,
+      "loss": 0.0,
+      "step": 62228
+    },
+    {
+      "epoch": 58.05,
+      "learning_rate": 4.274347014925373e-05,
+      "loss": 0.0,
+      "step": 62232
+    },
+    {
+      "epoch": 58.06,
+      "learning_rate": 4.2743003731343285e-05,
+      "loss": 0.0,
+      "step": 62236
+    },
+    {
+      "epoch": 58.06,
+      "learning_rate": 4.274253731343284e-05,
+      "loss": 0.0,
+      "step": 62240
+    },
+    {
+      "epoch": 58.06,
+      "learning_rate": 4.274207089552239e-05,
+      "loss": 0.0,
+      "step": 62244
+    },
+    {
+      "epoch": 58.07,
+      "learning_rate": 4.274160447761194e-05,
+      "loss": 0.0,
+      "step": 62248
+    },
+    {
+      "epoch": 58.07,
+      "learning_rate": 4.27411380597015e-05,
+      "loss": 0.0,
+      "step": 62252
+    },
+    {
+      "epoch": 58.07,
+      "learning_rate": 4.2740671641791046e-05,
+      "loss": 0.0,
+      "step": 62256
+    },
+    {
+      "epoch": 58.08,
+      "learning_rate": 4.2740205223880594e-05,
+      "loss": 0.0,
+      "step": 62260
+    },
+    {
+      "epoch": 58.08,
+      "learning_rate": 4.2739738805970156e-05,
+      "loss": 0.0066,
+      "step": 62264
+    },
+    {
+      "epoch": 58.09,
+      "learning_rate": 4.2739272388059704e-05,
+      "loss": 0.0,
+      "step": 62268
+    },
+    {
+      "epoch": 58.09,
+      "learning_rate": 4.273880597014925e-05,
+      "loss": 0.0,
+      "step": 62272
+    },
+    {
+      "epoch": 58.09,
+      "learning_rate": 4.273833955223881e-05,
+      "loss": 0.0009,
+      "step": 62276
+    },
+    {
+      "epoch": 58.1,
+      "learning_rate": 4.273787313432836e-05,
+      "loss": 0.0,
+      "step": 62280
+    },
+    {
+      "epoch": 58.1,
+      "learning_rate": 4.273740671641791e-05,
+      "loss": 0.0,
+      "step": 62284
+    },
+    {
+      "epoch": 58.1,
+      "learning_rate": 4.2736940298507465e-05,
+      "loss": 0.0015,
+      "step": 62288
+    },
+    {
+      "epoch": 58.11,
+      "learning_rate": 4.273647388059701e-05,
+      "loss": 0.0,
+      "step": 62292
+    },
+    {
+      "epoch": 58.11,
+      "learning_rate": 4.273600746268657e-05,
+      "loss": 0.0005,
+      "step": 62296
+    },
+    {
+      "epoch": 58.12,
+      "learning_rate": 4.273554104477612e-05,
+      "loss": 0.002,
+      "step": 62300
+    },
+    {
+      "epoch": 58.12,
+      "learning_rate": 4.273507462686567e-05,
+      "loss": 0.0,
+      "step": 62304
+    },
+    {
+      "epoch": 58.12,
+      "learning_rate": 4.2734608208955226e-05,
+      "loss": 0.0006,
+      "step": 62308
+    },
+    {
+      "epoch": 58.13,
+      "learning_rate": 4.273414179104478e-05,
+      "loss": 0.0007,
+      "step": 62312
+    },
+    {
+      "epoch": 58.13,
+      "learning_rate": 4.273367537313433e-05,
+      "loss": 0.0,
+      "step": 62316
+    },
+    {
+      "epoch": 58.13,
+      "learning_rate": 4.273320895522388e-05,
+      "loss": 0.0,
+      "step": 62320
+    },
+    {
+      "epoch": 58.14,
+      "learning_rate": 4.273274253731344e-05,
+      "loss": 0.0,
+      "step": 62324
+    },
+    {
+      "epoch": 58.14,
+      "learning_rate": 4.273227611940299e-05,
+      "loss": 0.0,
+      "step": 62328
+    },
+    {
+      "epoch": 58.15,
+      "learning_rate": 4.2731809701492535e-05,
+      "loss": 0.0,
+      "step": 62332
+    },
+    {
+      "epoch": 58.15,
+      "learning_rate": 4.273134328358209e-05,
+      "loss": 0.0,
+      "step": 62336
+    },
+    {
+      "epoch": 58.15,
+      "learning_rate": 4.2730876865671645e-05,
+      "loss": 0.0009,
+      "step": 62340
+    },
+    {
+      "epoch": 58.16,
+      "learning_rate": 4.27304104477612e-05,
+      "loss": 0.0021,
+      "step": 62344
+    },
+    {
+      "epoch": 58.16,
+      "learning_rate": 4.272994402985075e-05,
+      "loss": 0.0,
+      "step": 62348
+    },
+    {
+      "epoch": 58.16,
+      "learning_rate": 4.2729477611940296e-05,
+      "loss": 0.0,
+      "step": 62352
+    },
+    {
+      "epoch": 58.17,
+      "learning_rate": 4.272901119402986e-05,
+      "loss": 0.0,
+      "step": 62356
+    },
+    {
+      "epoch": 58.17,
+      "learning_rate": 4.2728544776119406e-05,
+      "loss": 0.0006,
+      "step": 62360
+    },
+    {
+      "epoch": 58.18,
+      "learning_rate": 4.2728078358208954e-05,
+      "loss": 0.0,
+      "step": 62364
+    },
+    {
+      "epoch": 58.18,
+      "learning_rate": 4.272761194029851e-05,
+      "loss": 0.0,
+      "step": 62368
+    },
+    {
+      "epoch": 58.18,
+      "learning_rate": 4.2727145522388064e-05,
+      "loss": 0.0,
+      "step": 62372
+    },
+    {
+      "epoch": 58.19,
+      "learning_rate": 4.272667910447761e-05,
+      "loss": 0.0,
+      "step": 62376
+    },
+    {
+      "epoch": 58.19,
+      "learning_rate": 4.272621268656717e-05,
+      "loss": 0.0109,
+      "step": 62380
+    },
+    {
+      "epoch": 58.19,
+      "learning_rate": 4.272574626865672e-05,
+      "loss": 0.0007,
+      "step": 62384
+    },
+    {
+      "epoch": 58.2,
+      "learning_rate": 4.272527985074627e-05,
+      "loss": 0.0,
+      "step": 62388
+    },
+    {
+      "epoch": 58.2,
+      "learning_rate": 4.2724813432835825e-05,
+      "loss": 0.0,
+      "step": 62392
+    },
+    {
+      "epoch": 58.21,
+      "learning_rate": 4.272434701492537e-05,
+      "loss": 0.0,
+      "step": 62396
+    },
+    {
+      "epoch": 58.21,
+      "learning_rate": 4.272388059701493e-05,
+      "loss": 0.0001,
+      "step": 62400
+    },
+    {
+      "epoch": 58.21,
+      "learning_rate": 4.272341417910448e-05,
+      "loss": 0.0,
+      "step": 62404
+    },
+    {
+      "epoch": 58.22,
+      "learning_rate": 4.272294776119403e-05,
+      "loss": 0.002,
+      "step": 62408
+    },
+    {
+      "epoch": 58.22,
+      "learning_rate": 4.272248134328358e-05,
+      "loss": 0.0001,
+      "step": 62412
+    },
+    {
+      "epoch": 58.22,
+      "learning_rate": 4.272201492537314e-05,
+      "loss": 0.0003,
+      "step": 62416
+    },
+    {
+      "epoch": 58.23,
+      "learning_rate": 4.272154850746269e-05,
+      "loss": 0.0,
+      "step": 62420
+    },
+    {
+      "epoch": 58.23,
+      "learning_rate": 4.272108208955224e-05,
+      "loss": 0.0,
+      "step": 62424
+    },
+    {
+      "epoch": 58.24,
+      "learning_rate": 4.272061567164179e-05,
+      "loss": 0.0,
+      "step": 62428
+    },
+    {
+      "epoch": 58.24,
+      "learning_rate": 4.272014925373135e-05,
+      "loss": 0.0,
+      "step": 62432
+    },
+    {
+      "epoch": 58.24,
+      "learning_rate": 4.2719682835820895e-05,
+      "loss": 0.0,
+      "step": 62436
+    },
+    {
+      "epoch": 58.25,
+      "learning_rate": 4.271921641791045e-05,
+      "loss": 0.0,
+      "step": 62440
+    },
+    {
+      "epoch": 58.25,
+      "learning_rate": 4.2718750000000005e-05,
+      "loss": 0.0,
+      "step": 62444
+    },
+    {
+      "epoch": 58.25,
+      "learning_rate": 4.271828358208955e-05,
+      "loss": 0.0,
+      "step": 62448
+    },
+    {
+      "epoch": 58.26,
+      "learning_rate": 4.271781716417911e-05,
+      "loss": 0.0002,
+      "step": 62452
+    },
+    {
+      "epoch": 58.26,
+      "learning_rate": 4.2717350746268656e-05,
+      "loss": 0.0002,
+      "step": 62456
+    },
+    {
+      "epoch": 58.26,
+      "learning_rate": 4.271688432835821e-05,
+      "loss": 0.0,
+      "step": 62460
+    },
+    {
+      "epoch": 58.27,
+      "learning_rate": 4.2716417910447766e-05,
+      "loss": 0.0002,
+      "step": 62464
+    },
+    {
+      "epoch": 58.27,
+      "learning_rate": 4.2715951492537314e-05,
+      "loss": 0.0,
+      "step": 62468
+    },
+    {
+      "epoch": 58.28,
+      "learning_rate": 4.271548507462686e-05,
+      "loss": 0.0,
+      "step": 62472
+    },
+    {
+      "epoch": 58.28,
+      "learning_rate": 4.2715018656716424e-05,
+      "loss": 0.0,
+      "step": 62476
+    },
+    {
+      "epoch": 58.28,
+      "learning_rate": 4.271455223880597e-05,
+      "loss": 0.0,
+      "step": 62480
+    },
+    {
+      "epoch": 58.29,
+      "learning_rate": 4.271408582089552e-05,
+      "loss": 0.0,
+      "step": 62484
+    },
+    {
+      "epoch": 58.29,
+      "learning_rate": 4.2713619402985075e-05,
+      "loss": 0.0,
+      "step": 62488
+    },
+    {
+      "epoch": 58.29,
+      "learning_rate": 4.271315298507463e-05,
+      "loss": 0.0026,
+      "step": 62492
+    },
+    {
+      "epoch": 58.3,
+      "learning_rate": 4.271268656716418e-05,
+      "loss": 0.0002,
+      "step": 62496
+    },
+    {
+      "epoch": 58.3,
+      "learning_rate": 4.271222014925373e-05,
+      "loss": 0.0005,
+      "step": 62500
+    },
+    {
+      "epoch": 58.3,
+      "eval_exact_match": 0.7446808510638298,
+      "eval_exec": 0.7823984526112185,
+      "eval_loss": 0.46804332733154297,
+      "eval_runtime": 1167.1187,
+      "eval_samples_per_second": 0.886,
+      "step": 62500
+    },
+    {
+      "epoch": 58.31,
+      "learning_rate": 4.271175373134329e-05,
+      "loss": 0.0018,
+      "step": 62504
+    },
+    {
+      "epoch": 58.31,
+      "learning_rate": 4.271128731343284e-05,
+      "loss": 0.0,
+      "step": 62508
+    },
+    {
+      "epoch": 58.31,
+      "learning_rate": 4.271082089552239e-05,
+      "loss": 0.0,
+      "step": 62512
+    },
+    {
+      "epoch": 58.32,
+      "learning_rate": 4.271035447761194e-05,
+      "loss": 0.0,
+      "step": 62516
+    },
+    {
+      "epoch": 58.32,
+      "learning_rate": 4.2709888059701494e-05,
+      "loss": 0.0,
+      "step": 62520
+    },
+    {
+      "epoch": 58.32,
+      "learning_rate": 4.270942164179105e-05,
+      "loss": 0.0,
+      "step": 62524
+    },
+    {
+      "epoch": 58.33,
+      "learning_rate": 4.27089552238806e-05,
+      "loss": 0.0027,
+      "step": 62528
+    },
+    {
+      "epoch": 58.33,
+      "learning_rate": 4.270848880597015e-05,
+      "loss": 0.0,
+      "step": 62532
+    },
+    {
+      "epoch": 58.34,
+      "learning_rate": 4.270802238805971e-05,
+      "loss": 0.0003,
+      "step": 62536
+    },
+    {
+      "epoch": 58.34,
+      "learning_rate": 4.2707555970149255e-05,
+      "loss": 0.0,
+      "step": 62540
+    },
+    {
+      "epoch": 58.34,
+      "learning_rate": 4.270708955223881e-05,
+      "loss": 0.0,
+      "step": 62544
+    },
+    {
+      "epoch": 58.35,
+      "learning_rate": 4.270662313432836e-05,
+      "loss": 0.0,
+      "step": 62548
+    },
+    {
+      "epoch": 58.35,
+      "learning_rate": 4.270615671641791e-05,
+      "loss": 0.0,
+      "step": 62552
+    },
+    {
+      "epoch": 58.35,
+      "learning_rate": 4.270569029850747e-05,
+      "loss": 0.0003,
+      "step": 62556
+    },
+    {
+      "epoch": 58.36,
+      "learning_rate": 4.2705223880597016e-05,
+      "loss": 0.0,
+      "step": 62560
+    },
+    {
+      "epoch": 58.36,
+      "learning_rate": 4.270475746268657e-05,
+      "loss": 0.0003,
+      "step": 62564
+    },
+    {
+      "epoch": 58.37,
+      "learning_rate": 4.2704291044776126e-05,
+      "loss": 0.0,
+      "step": 62568
+    },
+    {
+      "epoch": 58.37,
+      "learning_rate": 4.2703824626865674e-05,
+      "loss": 0.0,
+      "step": 62572
+    },
+    {
+      "epoch": 58.37,
+      "learning_rate": 4.270335820895522e-05,
+      "loss": 0.0,
+      "step": 62576
+    },
+    {
+      "epoch": 58.38,
+      "learning_rate": 4.270289179104478e-05,
+      "loss": 0.0002,
+      "step": 62580
+    },
+    {
+      "epoch": 58.38,
+      "learning_rate": 4.270242537313433e-05,
+      "loss": 0.0,
+      "step": 62584
+    },
+    {
+      "epoch": 58.38,
+      "learning_rate": 4.270195895522388e-05,
+      "loss": 0.0,
+      "step": 62588
+    },
+    {
+      "epoch": 58.39,
+      "learning_rate": 4.2701492537313435e-05,
+      "loss": 0.0012,
+      "step": 62592
+    },
+    {
+      "epoch": 58.39,
+      "learning_rate": 4.270102611940299e-05,
+      "loss": 0.0004,
+      "step": 62596
+    },
+    {
+      "epoch": 58.4,
+      "learning_rate": 4.270055970149254e-05,
+      "loss": 0.0,
+      "step": 62600
+    },
+    {
+      "epoch": 58.4,
+      "learning_rate": 4.270009328358209e-05,
+      "loss": 0.0,
+      "step": 62604
+    },
+    {
+      "epoch": 58.4,
+      "learning_rate": 4.269962686567164e-05,
+      "loss": 0.0007,
+      "step": 62608
+    },
+    {
+      "epoch": 58.41,
+      "learning_rate": 4.2699160447761196e-05,
+      "loss": 0.0,
+      "step": 62612
+    },
+    {
+      "epoch": 58.41,
+      "learning_rate": 4.269869402985075e-05,
+      "loss": 0.0,
+      "step": 62616
+    },
+    {
+      "epoch": 58.41,
+      "learning_rate": 4.26982276119403e-05,
+      "loss": 0.0,
+      "step": 62620
+    },
+    {
+      "epoch": 58.42,
+      "learning_rate": 4.2697761194029854e-05,
+      "loss": 0.0,
+      "step": 62624
+    },
+    {
+      "epoch": 58.42,
+      "learning_rate": 4.269729477611941e-05,
+      "loss": 0.0,
+      "step": 62628
+    },
+    {
+      "epoch": 58.43,
+      "learning_rate": 4.269682835820896e-05,
+      "loss": 0.0059,
+      "step": 62632
+    },
+    {
+      "epoch": 58.43,
+      "learning_rate": 4.2696361940298505e-05,
+      "loss": 0.0003,
+      "step": 62636
+    },
+    {
+      "epoch": 58.43,
+      "learning_rate": 4.269589552238806e-05,
+      "loss": 0.0,
+      "step": 62640
+    },
+    {
+      "epoch": 58.44,
+      "learning_rate": 4.2695429104477615e-05,
+      "loss": 0.0,
+      "step": 62644
+    },
+    {
+      "epoch": 58.44,
+      "learning_rate": 4.269496268656716e-05,
+      "loss": 0.0,
+      "step": 62648
+    },
+    {
+      "epoch": 58.44,
+      "learning_rate": 4.269449626865672e-05,
+      "loss": 0.0009,
+      "step": 62652
+    },
+    {
+      "epoch": 58.45,
+      "learning_rate": 4.269402985074627e-05,
+      "loss": 0.0,
+      "step": 62656
+    },
+    {
+      "epoch": 58.45,
+      "learning_rate": 4.269356343283582e-05,
+      "loss": 0.0,
+      "step": 62660
+    },
+    {
+      "epoch": 58.46,
+      "learning_rate": 4.2693097014925376e-05,
+      "loss": 0.0,
+      "step": 62664
+    },
+    {
+      "epoch": 58.46,
+      "learning_rate": 4.2692630597014924e-05,
+      "loss": 0.0001,
+      "step": 62668
+    },
+    {
+      "epoch": 58.46,
+      "learning_rate": 4.2692164179104486e-05,
+      "loss": 0.0001,
+      "step": 62672
+    },
+    {
+      "epoch": 58.47,
+      "learning_rate": 4.2691697761194034e-05,
+      "loss": 0.0,
+      "step": 62676
+    },
+    {
+      "epoch": 58.47,
+      "learning_rate": 4.269123134328358e-05,
+      "loss": 0.0002,
+      "step": 62680
+    },
+    {
+      "epoch": 58.47,
+      "learning_rate": 4.269076492537314e-05,
+      "loss": 0.0,
+      "step": 62684
+    },
+    {
+      "epoch": 58.48,
+      "learning_rate": 4.269029850746269e-05,
+      "loss": 0.0002,
+      "step": 62688
+    },
+    {
+      "epoch": 58.48,
+      "learning_rate": 4.268983208955224e-05,
+      "loss": 0.0,
+      "step": 62692
+    },
+    {
+      "epoch": 58.49,
+      "learning_rate": 4.2689365671641795e-05,
+      "loss": 0.0001,
+      "step": 62696
+    },
+    {
+      "epoch": 58.49,
+      "learning_rate": 4.268889925373134e-05,
+      "loss": 0.0001,
+      "step": 62700
+    },
+    {
+      "epoch": 58.49,
+      "learning_rate": 4.26884328358209e-05,
+      "loss": 0.0,
+      "step": 62704
+    },
+    {
+      "epoch": 58.5,
+      "learning_rate": 4.268796641791045e-05,
+      "loss": 0.0,
+      "step": 62708
+    },
+    {
+      "epoch": 58.5,
+      "learning_rate": 4.26875e-05,
+      "loss": 0.0029,
+      "step": 62712
+    },
+    {
+      "epoch": 58.5,
+      "learning_rate": 4.2687033582089556e-05,
+      "loss": 0.0001,
+      "step": 62716
+    },
+    {
+      "epoch": 58.51,
+      "learning_rate": 4.268656716417911e-05,
+      "loss": 0.0004,
+      "step": 62720
+    },
+    {
+      "epoch": 58.51,
+      "learning_rate": 4.268610074626866e-05,
+      "loss": 0.0001,
+      "step": 62724
+    },
+    {
+      "epoch": 58.51,
+      "learning_rate": 4.268563432835821e-05,
+      "loss": 0.0,
+      "step": 62728
+    },
+    {
+      "epoch": 58.52,
+      "learning_rate": 4.268516791044777e-05,
+      "loss": 0.0,
+      "step": 62732
+    },
+    {
+      "epoch": 58.52,
+      "learning_rate": 4.268470149253732e-05,
+      "loss": 0.0,
+      "step": 62736
+    },
+    {
+      "epoch": 58.53,
+      "learning_rate": 4.2684235074626865e-05,
+      "loss": 0.0002,
+      "step": 62740
+    },
+    {
+      "epoch": 58.53,
+      "learning_rate": 4.268376865671642e-05,
+      "loss": 0.0,
+      "step": 62744
+    },
+    {
+      "epoch": 58.53,
+      "learning_rate": 4.2683302238805975e-05,
+      "loss": 0.002,
+      "step": 62748
+    },
+    {
+      "epoch": 58.54,
+      "learning_rate": 4.268283582089552e-05,
+      "loss": 0.0001,
+      "step": 62752
+    },
+    {
+      "epoch": 58.54,
+      "learning_rate": 4.268236940298508e-05,
+      "loss": 0.0,
+      "step": 62756
+    },
+    {
+      "epoch": 58.54,
+      "learning_rate": 4.2681902985074626e-05,
+      "loss": 0.0139,
+      "step": 62760
+    },
+    {
+      "epoch": 58.55,
+      "learning_rate": 4.268143656716418e-05,
+      "loss": 0.0,
+      "step": 62764
+    },
+    {
+      "epoch": 58.55,
+      "learning_rate": 4.2680970149253736e-05,
+      "loss": 0.0,
+      "step": 62768
+    },
+    {
+      "epoch": 58.56,
+      "learning_rate": 4.2680503731343284e-05,
+      "loss": 0.0,
+      "step": 62772
+    },
+    {
+      "epoch": 58.56,
+      "learning_rate": 4.268003731343284e-05,
+      "loss": 0.0001,
+      "step": 62776
+    },
+    {
+      "epoch": 58.56,
+      "learning_rate": 4.2679570895522394e-05,
+      "loss": 0.0,
+      "step": 62780
+    },
+    {
+      "epoch": 58.57,
+      "learning_rate": 4.267910447761194e-05,
+      "loss": 0.0,
+      "step": 62784
+    },
+    {
+      "epoch": 58.57,
+      "learning_rate": 4.267863805970149e-05,
+      "loss": 0.0,
+      "step": 62788
+    },
+    {
+      "epoch": 58.57,
+      "learning_rate": 4.267817164179105e-05,
+      "loss": 0.0,
+      "step": 62792
+    },
+    {
+      "epoch": 58.58,
+      "learning_rate": 4.26777052238806e-05,
+      "loss": 0.0,
+      "step": 62796
+    },
+    {
+      "epoch": 58.58,
+      "learning_rate": 4.267723880597015e-05,
+      "loss": 0.0002,
+      "step": 62800
+    },
+    {
+      "epoch": 58.59,
+      "learning_rate": 4.26767723880597e-05,
+      "loss": 0.0,
+      "step": 62804
+    },
+    {
+      "epoch": 58.59,
+      "learning_rate": 4.267630597014926e-05,
+      "loss": 0.0,
+      "step": 62808
+    },
+    {
+      "epoch": 58.59,
+      "learning_rate": 4.2675839552238806e-05,
+      "loss": 0.0,
+      "step": 62812
+    },
+    {
+      "epoch": 58.6,
+      "learning_rate": 4.267537313432836e-05,
+      "loss": 0.0002,
+      "step": 62816
+    },
+    {
+      "epoch": 58.6,
+      "learning_rate": 4.267490671641791e-05,
+      "loss": 0.0,
+      "step": 62820
+    },
+    {
+      "epoch": 58.6,
+      "learning_rate": 4.2674440298507464e-05,
+      "loss": 0.0,
+      "step": 62824
+    },
+    {
+      "epoch": 58.61,
+      "learning_rate": 4.267397388059702e-05,
+      "loss": 0.0001,
+      "step": 62828
+    },
+    {
+      "epoch": 58.61,
+      "learning_rate": 4.267350746268657e-05,
+      "loss": 0.0,
+      "step": 62832
+    },
+    {
+      "epoch": 58.62,
+      "learning_rate": 4.267304104477612e-05,
+      "loss": 0.0,
+      "step": 62836
+    },
+    {
+      "epoch": 58.62,
+      "learning_rate": 4.2672574626865677e-05,
+      "loss": 0.0,
+      "step": 62840
+    },
+    {
+      "epoch": 58.62,
+      "learning_rate": 4.2672108208955225e-05,
+      "loss": 0.0,
+      "step": 62844
+    },
+    {
+      "epoch": 58.63,
+      "learning_rate": 4.267164179104478e-05,
+      "loss": 0.0001,
+      "step": 62848
+    },
+    {
+      "epoch": 58.63,
+      "learning_rate": 4.267117537313433e-05,
+      "loss": 0.0003,
+      "step": 62852
+    },
+    {
+      "epoch": 58.63,
+      "learning_rate": 4.267070895522388e-05,
+      "loss": 0.0,
+      "step": 62856
+    },
+    {
+      "epoch": 58.64,
+      "learning_rate": 4.267024253731344e-05,
+      "loss": 0.0,
+      "step": 62860
+    },
+    {
+      "epoch": 58.64,
+      "learning_rate": 4.2669776119402986e-05,
+      "loss": 0.0,
+      "step": 62864
+    },
+    {
+      "epoch": 58.65,
+      "learning_rate": 4.266930970149254e-05,
+      "loss": 0.0,
+      "step": 62868
+    },
+    {
+      "epoch": 58.65,
+      "learning_rate": 4.2668843283582095e-05,
+      "loss": 0.0004,
+      "step": 62872
+    },
+    {
+      "epoch": 58.65,
+      "learning_rate": 4.2668376865671644e-05,
+      "loss": 0.0,
+      "step": 62876
+    },
+    {
+      "epoch": 58.66,
+      "learning_rate": 4.266791044776119e-05,
+      "loss": 0.0,
+      "step": 62880
+    },
+    {
+      "epoch": 58.66,
+      "learning_rate": 4.2667444029850753e-05,
+      "loss": 0.0001,
+      "step": 62884
+    },
+    {
+      "epoch": 58.66,
+      "learning_rate": 4.26669776119403e-05,
+      "loss": 0.0,
+      "step": 62888
+    },
+    {
+      "epoch": 58.67,
+      "learning_rate": 4.266651119402985e-05,
+      "loss": 0.0,
+      "step": 62892
+    },
+    {
+      "epoch": 58.67,
+      "learning_rate": 4.2666044776119405e-05,
+      "loss": 0.0,
+      "step": 62896
+    },
+    {
+      "epoch": 58.68,
+      "learning_rate": 4.266557835820896e-05,
+      "loss": 0.0002,
+      "step": 62900
+    },
+    {
+      "epoch": 58.68,
+      "learning_rate": 4.266511194029851e-05,
+      "loss": 0.0,
+      "step": 62904
+    },
+    {
+      "epoch": 58.68,
+      "learning_rate": 4.266464552238806e-05,
+      "loss": 0.0009,
+      "step": 62908
+    },
+    {
+      "epoch": 58.69,
+      "learning_rate": 4.266417910447761e-05,
+      "loss": 0.0,
+      "step": 62912
+    },
+    {
+      "epoch": 58.69,
+      "learning_rate": 4.2663712686567166e-05,
+      "loss": 0.0,
+      "step": 62916
+    },
+    {
+      "epoch": 58.69,
+      "learning_rate": 4.266324626865672e-05,
+      "loss": 0.0,
+      "step": 62920
+    },
+    {
+      "epoch": 58.7,
+      "learning_rate": 4.266277985074627e-05,
+      "loss": 0.0,
+      "step": 62924
+    },
+    {
+      "epoch": 58.7,
+      "learning_rate": 4.2662313432835824e-05,
+      "loss": 0.0006,
+      "step": 62928
+    },
+    {
+      "epoch": 58.71,
+      "learning_rate": 4.266184701492538e-05,
+      "loss": 0.0001,
+      "step": 62932
+    },
+    {
+      "epoch": 58.71,
+      "learning_rate": 4.2661380597014927e-05,
+      "loss": 0.0,
+      "step": 62936
+    },
+    {
+      "epoch": 58.71,
+      "learning_rate": 4.2660914179104475e-05,
+      "loss": 0.0,
+      "step": 62940
+    },
+    {
+      "epoch": 58.72,
+      "learning_rate": 4.2660447761194036e-05,
+      "loss": 0.0,
+      "step": 62944
+    },
+    {
+      "epoch": 58.72,
+      "learning_rate": 4.2659981343283585e-05,
+      "loss": 0.0003,
+      "step": 62948
+    },
+    {
+      "epoch": 58.72,
+      "learning_rate": 4.265951492537313e-05,
+      "loss": 0.0003,
+      "step": 62952
+    },
+    {
+      "epoch": 58.73,
+      "learning_rate": 4.265904850746269e-05,
+      "loss": 0.0,
+      "step": 62956
+    },
+    {
+      "epoch": 58.73,
+      "learning_rate": 4.265858208955224e-05,
+      "loss": 0.0031,
+      "step": 62960
+    },
+    {
+      "epoch": 58.73,
+      "learning_rate": 4.265811567164179e-05,
+      "loss": 0.0,
+      "step": 62964
+    },
+    {
+      "epoch": 58.74,
+      "learning_rate": 4.2657649253731346e-05,
+      "loss": 0.0,
+      "step": 62968
+    },
+    {
+      "epoch": 58.74,
+      "learning_rate": 4.2657182835820894e-05,
+      "loss": 0.0002,
+      "step": 62972
+    },
+    {
+      "epoch": 58.75,
+      "learning_rate": 4.265671641791045e-05,
+      "loss": 0.0,
+      "step": 62976
+    },
+    {
+      "epoch": 58.75,
+      "learning_rate": 4.2656250000000003e-05,
+      "loss": 0.0001,
+      "step": 62980
+    },
+    {
+      "epoch": 58.75,
+      "learning_rate": 4.265578358208955e-05,
+      "loss": 0.0,
+      "step": 62984
+    },
+    {
+      "epoch": 58.76,
+      "learning_rate": 4.2655317164179107e-05,
+      "loss": 0.0004,
+      "step": 62988
+    },
+    {
+      "epoch": 58.76,
+      "learning_rate": 4.265485074626866e-05,
+      "loss": 0.0001,
+      "step": 62992
+    },
+    {
+      "epoch": 58.76,
+      "learning_rate": 4.265438432835821e-05,
+      "loss": 0.0,
+      "step": 62996
+    },
+    {
+      "epoch": 58.77,
+      "learning_rate": 4.2653917910447764e-05,
+      "loss": 0.0,
+      "step": 63000
+    },
+    {
+      "epoch": 58.77,
+      "eval_exact_match": 0.741779497098646,
+      "eval_exec": 0.7872340425531915,
+      "eval_loss": 0.4804081916809082,
+      "eval_runtime": 1136.1578,
+      "eval_samples_per_second": 0.91,
+      "step": 63000
+    },
+    {
+      "epoch": 58.77,
+      "learning_rate": 4.265345149253732e-05,
+      "loss": 0.0,
+      "step": 63004
+    },
+    {
+      "epoch": 58.78,
+      "learning_rate": 4.265298507462687e-05,
+      "loss": 0.0,
+      "step": 63008
+    },
+    {
+      "epoch": 58.78,
+      "learning_rate": 4.265251865671642e-05,
+      "loss": 0.0003,
+      "step": 63012
+    },
+    {
+      "epoch": 58.78,
+      "learning_rate": 4.265205223880597e-05,
+      "loss": 0.0001,
+      "step": 63016
+    },
+    {
+      "epoch": 58.79,
+      "learning_rate": 4.2651585820895525e-05,
+      "loss": 0.0001,
+      "step": 63020
+    },
+    {
+      "epoch": 58.79,
+      "learning_rate": 4.265111940298508e-05,
+      "loss": 0.0,
+      "step": 63024
+    },
+    {
+      "epoch": 58.79,
+      "learning_rate": 4.265065298507463e-05,
+      "loss": 0.0,
+      "step": 63028
+    },
+    {
+      "epoch": 58.8,
+      "learning_rate": 4.2650186567164177e-05,
+      "loss": 0.0,
+      "step": 63032
+    },
+    {
+      "epoch": 58.8,
+      "learning_rate": 4.264972014925374e-05,
+      "loss": 0.0075,
+      "step": 63036
+    },
+    {
+      "epoch": 58.81,
+      "learning_rate": 4.2649253731343286e-05,
+      "loss": 0.0001,
+      "step": 63040
+    },
+    {
+      "epoch": 58.81,
+      "learning_rate": 4.2648787313432835e-05,
+      "loss": 0.0001,
+      "step": 63044
+    },
+    {
+      "epoch": 58.81,
+      "learning_rate": 4.264832089552239e-05,
+      "loss": 0.0,
+      "step": 63048
+    },
+    {
+      "epoch": 58.82,
+      "learning_rate": 4.2647854477611944e-05,
+      "loss": 0.0,
+      "step": 63052
+    },
+    {
+      "epoch": 58.82,
+      "learning_rate": 4.264738805970149e-05,
+      "loss": 0.0002,
+      "step": 63056
+    },
+    {
+      "epoch": 58.82,
+      "learning_rate": 4.264692164179105e-05,
+      "loss": 0.0,
+      "step": 63060
+    },
+    {
+      "epoch": 58.83,
+      "learning_rate": 4.26464552238806e-05,
+      "loss": 0.0,
+      "step": 63064
+    },
+    {
+      "epoch": 58.83,
+      "learning_rate": 4.264598880597015e-05,
+      "loss": 0.0001,
+      "step": 63068
+    },
+    {
+      "epoch": 58.84,
+      "learning_rate": 4.2645522388059705e-05,
+      "loss": 0.0001,
+      "step": 63072
+    },
+    {
+      "epoch": 58.84,
+      "learning_rate": 4.2645055970149253e-05,
+      "loss": 0.0,
+      "step": 63076
+    },
+    {
+      "epoch": 58.84,
+      "learning_rate": 4.264458955223881e-05,
+      "loss": 0.0,
+      "step": 63080
+    },
+    {
+      "epoch": 58.85,
+      "learning_rate": 4.264412313432836e-05,
+      "loss": 0.0,
+      "step": 63084
+    },
+    {
+      "epoch": 58.85,
+      "learning_rate": 4.264365671641791e-05,
+      "loss": 0.0,
+      "step": 63088
+    },
+    {
+      "epoch": 58.85,
+      "learning_rate": 4.264319029850746e-05,
+      "loss": 0.0,
+      "step": 63092
+    },
+    {
+      "epoch": 58.86,
+      "learning_rate": 4.264272388059702e-05,
+      "loss": 0.0,
+      "step": 63096
+    },
+    {
+      "epoch": 58.86,
+      "learning_rate": 4.264225746268657e-05,
+      "loss": 0.0,
+      "step": 63100
+    },
+    {
+      "epoch": 58.87,
+      "learning_rate": 4.264179104477612e-05,
+      "loss": 0.0,
+      "step": 63104
+    },
+    {
+      "epoch": 58.87,
+      "learning_rate": 4.264132462686567e-05,
+      "loss": 0.0019,
+      "step": 63108
+    },
+    {
+      "epoch": 58.87,
+      "learning_rate": 4.264085820895523e-05,
+      "loss": 0.0,
+      "step": 63112
+    },
+    {
+      "epoch": 58.88,
+      "learning_rate": 4.2640391791044775e-05,
+      "loss": 0.0,
+      "step": 63116
+    },
+    {
+      "epoch": 58.88,
+      "learning_rate": 4.263992537313433e-05,
+      "loss": 0.0,
+      "step": 63120
+    },
+    {
+      "epoch": 58.88,
+      "learning_rate": 4.2639458955223885e-05,
+      "loss": 0.0,
+      "step": 63124
+    },
+    {
+      "epoch": 58.89,
+      "learning_rate": 4.2638992537313433e-05,
+      "loss": 0.0001,
+      "step": 63128
+    },
+    {
+      "epoch": 58.89,
+      "learning_rate": 4.263852611940299e-05,
+      "loss": 0.0,
+      "step": 63132
+    },
+    {
+      "epoch": 58.9,
+      "learning_rate": 4.2638059701492536e-05,
+      "loss": 0.0001,
+      "step": 63136
+    },
+    {
+      "epoch": 58.9,
+      "learning_rate": 4.263759328358209e-05,
+      "loss": 0.0012,
+      "step": 63140
+    },
+    {
+      "epoch": 58.9,
+      "learning_rate": 4.2637126865671646e-05,
+      "loss": 0.0004,
+      "step": 63144
+    },
+    {
+      "epoch": 58.91,
+      "learning_rate": 4.2636660447761194e-05,
+      "loss": 0.0002,
+      "step": 63148
+    },
+    {
+      "epoch": 58.91,
+      "learning_rate": 4.263619402985074e-05,
+      "loss": 0.0,
+      "step": 63152
+    },
+    {
+      "epoch": 58.91,
+      "learning_rate": 4.2635727611940304e-05,
+      "loss": 0.0,
+      "step": 63156
+    },
+    {
+      "epoch": 58.92,
+      "learning_rate": 4.263526119402985e-05,
+      "loss": 0.0,
+      "step": 63160
+    },
+    {
+      "epoch": 58.92,
+      "learning_rate": 4.263479477611941e-05,
+      "loss": 0.0,
+      "step": 63164
+    },
+    {
+      "epoch": 58.93,
+      "learning_rate": 4.2634328358208955e-05,
+      "loss": 0.0002,
+      "step": 63168
+    },
+    {
+      "epoch": 58.93,
+      "learning_rate": 4.263386194029851e-05,
+      "loss": 0.0004,
+      "step": 63172
+    },
+    {
+      "epoch": 58.93,
+      "learning_rate": 4.2633395522388065e-05,
+      "loss": 0.0,
+      "step": 63176
+    },
+    {
+      "epoch": 58.94,
+      "learning_rate": 4.263292910447761e-05,
+      "loss": 0.0,
+      "step": 63180
+    },
+    {
+      "epoch": 58.94,
+      "learning_rate": 4.263246268656717e-05,
+      "loss": 0.0,
+      "step": 63184
+    },
+    {
+      "epoch": 58.94,
+      "learning_rate": 4.263199626865672e-05,
+      "loss": 0.0,
+      "step": 63188
+    },
+    {
+      "epoch": 58.95,
+      "learning_rate": 4.263152985074627e-05,
+      "loss": 0.0,
+      "step": 63192
+    },
+    {
+      "epoch": 58.95,
+      "learning_rate": 4.263106343283582e-05,
+      "loss": 0.0,
+      "step": 63196
+    },
+    {
+      "epoch": 58.96,
+      "learning_rate": 4.2630597014925374e-05,
+      "loss": 0.0,
+      "step": 63200
+    },
+    {
+      "epoch": 58.96,
+      "learning_rate": 4.263013059701493e-05,
+      "loss": 0.0038,
+      "step": 63204
+    },
+    {
+      "epoch": 58.96,
+      "learning_rate": 4.262966417910448e-05,
+      "loss": 0.0,
+      "step": 63208
+    },
+    {
+      "epoch": 58.97,
+      "learning_rate": 4.262919776119403e-05,
+      "loss": 0.0001,
+      "step": 63212
+    },
+    {
+      "epoch": 58.97,
+      "learning_rate": 4.262873134328359e-05,
+      "loss": 0.0156,
+      "step": 63216
+    },
+    {
+      "epoch": 58.97,
+      "learning_rate": 4.2628264925373135e-05,
+      "loss": 0.0,
+      "step": 63220
+    },
+    {
+      "epoch": 58.98,
+      "learning_rate": 4.262779850746269e-05,
+      "loss": 0.0001,
+      "step": 63224
+    },
+    {
+      "epoch": 58.98,
+      "learning_rate": 4.262733208955224e-05,
+      "loss": 0.0003,
+      "step": 63228
+    },
+    {
+      "epoch": 58.98,
+      "learning_rate": 4.262686567164179e-05,
+      "loss": 0.0,
+      "step": 63232
+    },
+    {
+      "epoch": 58.99,
+      "learning_rate": 4.262639925373135e-05,
+      "loss": 0.0,
+      "step": 63236
+    },
+    {
+      "epoch": 58.99,
+      "learning_rate": 4.2625932835820896e-05,
+      "loss": 0.0,
+      "step": 63240
+    },
+    {
+      "epoch": 59.0,
+      "learning_rate": 4.262546641791045e-05,
+      "loss": 0.0,
+      "step": 63244
+    },
+    {
+      "epoch": 59.0,
+      "learning_rate": 4.2625000000000006e-05,
+      "loss": 0.0,
+      "step": 63248
+    },
+    {
+      "epoch": 59.0,
+      "learning_rate": 4.2624533582089554e-05,
+      "loss": 0.0,
+      "step": 63252
+    },
+    {
+      "epoch": 59.01,
+      "learning_rate": 4.26240671641791e-05,
+      "loss": 0.0,
+      "step": 63256
+    },
+    {
+      "epoch": 59.01,
+      "learning_rate": 4.262360074626866e-05,
+      "loss": 0.0001,
+      "step": 63260
+    },
+    {
+      "epoch": 59.01,
+      "learning_rate": 4.262313432835821e-05,
+      "loss": 0.0,
+      "step": 63264
+    },
+    {
+      "epoch": 59.02,
+      "learning_rate": 4.262266791044776e-05,
+      "loss": 0.0,
+      "step": 63268
+    },
+    {
+      "epoch": 59.02,
+      "learning_rate": 4.2622201492537315e-05,
+      "loss": 0.0,
+      "step": 63272
+    },
+    {
+      "epoch": 59.03,
+      "learning_rate": 4.262173507462687e-05,
+      "loss": 0.0018,
+      "step": 63276
+    },
+    {
+      "epoch": 59.03,
+      "learning_rate": 4.262126865671642e-05,
+      "loss": 0.0002,
+      "step": 63280
+    },
+    {
+      "epoch": 59.03,
+      "learning_rate": 4.262080223880597e-05,
+      "loss": 0.0,
+      "step": 63284
+    },
+    {
+      "epoch": 59.04,
+      "learning_rate": 4.262033582089552e-05,
+      "loss": 0.0,
+      "step": 63288
+    },
+    {
+      "epoch": 59.04,
+      "learning_rate": 4.2619869402985076e-05,
+      "loss": 0.0002,
+      "step": 63292
+    },
+    {
+      "epoch": 59.04,
+      "learning_rate": 4.261940298507463e-05,
+      "loss": 0.0002,
+      "step": 63296
+    },
+    {
+      "epoch": 59.05,
+      "learning_rate": 4.261893656716418e-05,
+      "loss": 0.0,
+      "step": 63300
+    },
+    {
+      "epoch": 59.05,
+      "learning_rate": 4.2618470149253734e-05,
+      "loss": 0.0,
+      "step": 63304
+    },
+    {
+      "epoch": 59.06,
+      "learning_rate": 4.261800373134329e-05,
+      "loss": 0.0001,
+      "step": 63308
+    },
+    {
+      "epoch": 59.06,
+      "learning_rate": 4.261753731343284e-05,
+      "loss": 0.0,
+      "step": 63312
+    },
+    {
+      "epoch": 59.06,
+      "learning_rate": 4.2617070895522385e-05,
+      "loss": 0.0054,
+      "step": 63316
+    },
+    {
+      "epoch": 59.07,
+      "learning_rate": 4.261660447761194e-05,
+      "loss": 0.0,
+      "step": 63320
+    },
+    {
+      "epoch": 59.07,
+      "learning_rate": 4.2616138059701495e-05,
+      "loss": 0.0,
+      "step": 63324
+    },
+    {
+      "epoch": 59.07,
+      "learning_rate": 4.261567164179105e-05,
+      "loss": 0.0,
+      "step": 63328
+    },
+    {
+      "epoch": 59.08,
+      "learning_rate": 4.26152052238806e-05,
+      "loss": 0.0,
+      "step": 63332
+    },
+    {
+      "epoch": 59.08,
+      "learning_rate": 4.261473880597015e-05,
+      "loss": 0.0,
+      "step": 63336
+    },
+    {
+      "epoch": 59.09,
+      "learning_rate": 4.261427238805971e-05,
+      "loss": 0.0,
+      "step": 63340
+    },
+    {
+      "epoch": 59.09,
+      "learning_rate": 4.2613805970149256e-05,
+      "loss": 0.0007,
+      "step": 63344
+    },
+    {
+      "epoch": 59.09,
+      "learning_rate": 4.2613339552238804e-05,
+      "loss": 0.0,
+      "step": 63348
+    },
+    {
+      "epoch": 59.1,
+      "learning_rate": 4.2612873134328366e-05,
+      "loss": 0.0,
+      "step": 63352
+    },
+    {
+      "epoch": 59.1,
+      "learning_rate": 4.2612406716417914e-05,
+      "loss": 0.0,
+      "step": 63356
+    },
+    {
+      "epoch": 59.1,
+      "learning_rate": 4.261194029850746e-05,
+      "loss": 0.0,
+      "step": 63360
+    },
+    {
+      "epoch": 59.11,
+      "learning_rate": 4.261147388059702e-05,
+      "loss": 0.0,
+      "step": 63364
+    },
+    {
+      "epoch": 59.11,
+      "learning_rate": 4.261100746268657e-05,
+      "loss": 0.0,
+      "step": 63368
+    },
+    {
+      "epoch": 59.12,
+      "learning_rate": 4.261054104477612e-05,
+      "loss": 0.0001,
+      "step": 63372
+    },
+    {
+      "epoch": 59.12,
+      "learning_rate": 4.2610074626865675e-05,
+      "loss": 0.0,
+      "step": 63376
+    },
+    {
+      "epoch": 59.12,
+      "learning_rate": 4.260960820895522e-05,
+      "loss": 0.0,
+      "step": 63380
+    },
+    {
+      "epoch": 59.13,
+      "learning_rate": 4.260914179104478e-05,
+      "loss": 0.0,
+      "step": 63384
+    },
+    {
+      "epoch": 59.13,
+      "learning_rate": 4.260867537313433e-05,
+      "loss": 0.0007,
+      "step": 63388
+    },
+    {
+      "epoch": 59.13,
+      "learning_rate": 4.260820895522388e-05,
+      "loss": 0.0001,
+      "step": 63392
+    },
+    {
+      "epoch": 59.14,
+      "learning_rate": 4.2607742537313436e-05,
+      "loss": 0.0,
+      "step": 63396
+    },
+    {
+      "epoch": 59.14,
+      "learning_rate": 4.260727611940299e-05,
+      "loss": 0.0002,
+      "step": 63400
+    },
+    {
+      "epoch": 59.15,
+      "learning_rate": 4.260680970149254e-05,
+      "loss": 0.0,
+      "step": 63404
+    },
+    {
+      "epoch": 59.15,
+      "learning_rate": 4.260634328358209e-05,
+      "loss": 0.0,
+      "step": 63408
+    },
+    {
+      "epoch": 59.15,
+      "learning_rate": 4.260587686567165e-05,
+      "loss": 0.0019,
+      "step": 63412
+    },
+    {
+      "epoch": 59.16,
+      "learning_rate": 4.26054104477612e-05,
+      "loss": 0.0,
+      "step": 63416
+    },
+    {
+      "epoch": 59.16,
+      "learning_rate": 4.2604944029850745e-05,
+      "loss": 0.006,
+      "step": 63420
+    },
+    {
+      "epoch": 59.16,
+      "learning_rate": 4.26044776119403e-05,
+      "loss": 0.0,
+      "step": 63424
+    },
+    {
+      "epoch": 59.17,
+      "learning_rate": 4.2604011194029855e-05,
+      "loss": 0.0001,
+      "step": 63428
+    },
+    {
+      "epoch": 59.17,
+      "learning_rate": 4.26035447761194e-05,
+      "loss": 0.0008,
+      "step": 63432
+    },
+    {
+      "epoch": 59.18,
+      "learning_rate": 4.260307835820896e-05,
+      "loss": 0.0,
+      "step": 63436
+    },
+    {
+      "epoch": 59.18,
+      "learning_rate": 4.2602611940298506e-05,
+      "loss": 0.0032,
+      "step": 63440
+    },
+    {
+      "epoch": 59.18,
+      "learning_rate": 4.260214552238806e-05,
+      "loss": 0.0004,
+      "step": 63444
+    },
+    {
+      "epoch": 59.19,
+      "learning_rate": 4.2601679104477616e-05,
+      "loss": 0.0,
+      "step": 63448
+    },
+    {
+      "epoch": 59.19,
+      "learning_rate": 4.2601212686567164e-05,
+      "loss": 0.0,
+      "step": 63452
+    },
+    {
+      "epoch": 59.19,
+      "learning_rate": 4.260074626865672e-05,
+      "loss": 0.0,
+      "step": 63456
+    },
+    {
+      "epoch": 59.2,
+      "learning_rate": 4.2600279850746274e-05,
+      "loss": 0.0,
+      "step": 63460
+    },
+    {
+      "epoch": 59.2,
+      "learning_rate": 4.259981343283582e-05,
+      "loss": 0.0,
+      "step": 63464
+    },
+    {
+      "epoch": 59.21,
+      "learning_rate": 4.259934701492537e-05,
+      "loss": 0.0001,
+      "step": 63468
+    },
+    {
+      "epoch": 59.21,
+      "learning_rate": 4.259888059701493e-05,
+      "loss": 0.0001,
+      "step": 63472
+    },
+    {
+      "epoch": 59.21,
+      "learning_rate": 4.259841417910448e-05,
+      "loss": 0.0,
+      "step": 63476
+    },
+    {
+      "epoch": 59.22,
+      "learning_rate": 4.259794776119403e-05,
+      "loss": 0.0,
+      "step": 63480
+    },
+    {
+      "epoch": 59.22,
+      "learning_rate": 4.259748134328358e-05,
+      "loss": 0.0,
+      "step": 63484
+    },
+    {
+      "epoch": 59.22,
+      "learning_rate": 4.259701492537314e-05,
+      "loss": 0.0,
+      "step": 63488
+    },
+    {
+      "epoch": 59.23,
+      "learning_rate": 4.259654850746269e-05,
+      "loss": 0.0,
+      "step": 63492
+    },
+    {
+      "epoch": 59.23,
+      "learning_rate": 4.259608208955224e-05,
+      "loss": 0.0001,
+      "step": 63496
+    },
+    {
+      "epoch": 59.24,
+      "learning_rate": 4.259561567164179e-05,
+      "loss": 0.0002,
+      "step": 63500
+    },
+    {
+      "epoch": 59.24,
+      "eval_exact_match": 0.746615087040619,
+      "eval_exec": 0.7872340425531915,
+      "eval_loss": 0.48403027653694153,
+      "eval_runtime": 1101.428,
+      "eval_samples_per_second": 0.939,
+      "step": 63500
+    },
+    {
+      "epoch": 59.24,
+      "learning_rate": 4.259514925373135e-05,
+      "loss": 0.0,
+      "step": 63504
+    },
+    {
+      "epoch": 59.24,
+      "learning_rate": 4.25946828358209e-05,
+      "loss": 0.0001,
+      "step": 63508
+    },
+    {
+      "epoch": 59.25,
+      "learning_rate": 4.259421641791045e-05,
+      "loss": 0.0101,
+      "step": 63512
+    },
+    {
+      "epoch": 59.25,
+      "learning_rate": 4.259375e-05,
+      "loss": 0.0,
+      "step": 63516
+    },
+    {
+      "epoch": 59.25,
+      "learning_rate": 4.259328358208956e-05,
+      "loss": 0.0,
+      "step": 63520
+    },
+    {
+      "epoch": 59.26,
+      "learning_rate": 4.2592817164179105e-05,
+      "loss": 0.0,
+      "step": 63524
+    },
+    {
+      "epoch": 59.26,
+      "learning_rate": 4.259235074626866e-05,
+      "loss": 0.0,
+      "step": 63528
+    },
+    {
+      "epoch": 59.26,
+      "learning_rate": 4.259188432835821e-05,
+      "loss": 0.0,
+      "step": 63532
+    },
+    {
+      "epoch": 59.27,
+      "learning_rate": 4.259141791044776e-05,
+      "loss": 0.0,
+      "step": 63536
+    },
+    {
+      "epoch": 59.27,
+      "learning_rate": 4.259095149253732e-05,
+      "loss": 0.0,
+      "step": 63540
+    },
+    {
+      "epoch": 59.28,
+      "learning_rate": 4.2590485074626866e-05,
+      "loss": 0.0,
+      "step": 63544
+    },
+    {
+      "epoch": 59.28,
+      "learning_rate": 4.259001865671642e-05,
+      "loss": 0.0,
+      "step": 63548
+    },
+    {
+      "epoch": 59.28,
+      "learning_rate": 4.2589552238805976e-05,
+      "loss": 0.0,
+      "step": 63552
+    },
+    {
+      "epoch": 59.29,
+      "learning_rate": 4.2589085820895524e-05,
+      "loss": 0.0,
+      "step": 63556
+    },
+    {
+      "epoch": 59.29,
+      "learning_rate": 4.258861940298507e-05,
+      "loss": 0.0,
+      "step": 63560
+    },
+    {
+      "epoch": 59.29,
+      "learning_rate": 4.2588152985074634e-05,
+      "loss": 0.0,
+      "step": 63564
+    },
+    {
+      "epoch": 59.3,
+      "learning_rate": 4.258768656716418e-05,
+      "loss": 0.0011,
+      "step": 63568
+    },
+    {
+      "epoch": 59.3,
+      "learning_rate": 4.258722014925373e-05,
+      "loss": 0.0001,
+      "step": 63572
+    },
+    {
+      "epoch": 59.31,
+      "learning_rate": 4.2586753731343285e-05,
+      "loss": 0.0,
+      "step": 63576
+    },
+    {
+      "epoch": 59.31,
+      "learning_rate": 4.258628731343284e-05,
+      "loss": 0.0002,
+      "step": 63580
+    },
+    {
+      "epoch": 59.31,
+      "learning_rate": 4.258582089552239e-05,
+      "loss": 0.0,
+      "step": 63584
+    },
+    {
+      "epoch": 59.32,
+      "learning_rate": 4.258535447761194e-05,
+      "loss": 0.0,
+      "step": 63588
+    },
+    {
+      "epoch": 59.32,
+      "learning_rate": 4.258488805970149e-05,
+      "loss": 0.0,
+      "step": 63592
+    },
+    {
+      "epoch": 59.32,
+      "learning_rate": 4.2584421641791046e-05,
+      "loss": 0.0,
+      "step": 63596
+    },
+    {
+      "epoch": 59.33,
+      "learning_rate": 4.25839552238806e-05,
+      "loss": 0.0,
+      "step": 63600
+    },
+    {
+      "epoch": 59.33,
+      "learning_rate": 4.258348880597015e-05,
+      "loss": 0.0001,
+      "step": 63604
+    },
+    {
+      "epoch": 59.34,
+      "learning_rate": 4.2583022388059704e-05,
+      "loss": 0.0,
+      "step": 63608
+    },
+    {
+      "epoch": 59.34,
+      "learning_rate": 4.258255597014926e-05,
+      "loss": 0.0,
+      "step": 63612
+    },
+    {
+      "epoch": 59.34,
+      "learning_rate": 4.258208955223881e-05,
+      "loss": 0.0006,
+      "step": 63616
+    },
+    {
+      "epoch": 59.35,
+      "learning_rate": 4.2581623134328355e-05,
+      "loss": 0.0001,
+      "step": 63620
+    },
+    {
+      "epoch": 59.35,
+      "learning_rate": 4.258115671641792e-05,
+      "loss": 0.0,
+      "step": 63624
+    },
+    {
+      "epoch": 59.35,
+      "learning_rate": 4.2580690298507465e-05,
+      "loss": 0.0001,
+      "step": 63628
+    },
+    {
+      "epoch": 59.36,
+      "learning_rate": 4.258022388059701e-05,
+      "loss": 0.0,
+      "step": 63632
+    },
+    {
+      "epoch": 59.36,
+      "learning_rate": 4.257975746268657e-05,
+      "loss": 0.0,
+      "step": 63636
+    },
+    {
+      "epoch": 59.37,
+      "learning_rate": 4.257929104477612e-05,
+      "loss": 0.0001,
+      "step": 63640
+    },
+    {
+      "epoch": 59.37,
+      "learning_rate": 4.257882462686567e-05,
+      "loss": 0.0,
+      "step": 63644
+    },
+    {
+      "epoch": 59.37,
+      "learning_rate": 4.2578358208955226e-05,
+      "loss": 0.0006,
+      "step": 63648
+    },
+    {
+      "epoch": 59.38,
+      "learning_rate": 4.2577891791044774e-05,
+      "loss": 0.0,
+      "step": 63652
+    },
+    {
+      "epoch": 59.38,
+      "learning_rate": 4.2577425373134336e-05,
+      "loss": 0.0011,
+      "step": 63656
+    },
+    {
+      "epoch": 59.38,
+      "learning_rate": 4.2576958955223884e-05,
+      "loss": 0.0,
+      "step": 63660
+    },
+    {
+      "epoch": 59.39,
+      "learning_rate": 4.257649253731343e-05,
+      "loss": 0.0001,
+      "step": 63664
+    },
+    {
+      "epoch": 59.39,
+      "learning_rate": 4.257602611940299e-05,
+      "loss": 0.0,
+      "step": 63668
+    },
+    {
+      "epoch": 59.4,
+      "learning_rate": 4.257555970149254e-05,
+      "loss": 0.0,
+      "step": 63672
+    },
+    {
+      "epoch": 59.4,
+      "learning_rate": 4.257509328358209e-05,
+      "loss": 0.0,
+      "step": 63676
+    },
+    {
+      "epoch": 59.4,
+      "learning_rate": 4.2574626865671645e-05,
+      "loss": 0.0,
+      "step": 63680
+    },
+    {
+      "epoch": 59.41,
+      "learning_rate": 4.25741604477612e-05,
+      "loss": 0.0,
+      "step": 63684
+    },
+    {
+      "epoch": 59.41,
+      "learning_rate": 4.257369402985075e-05,
+      "loss": 0.0,
+      "step": 63688
+    },
+    {
+      "epoch": 59.41,
+      "learning_rate": 4.25732276119403e-05,
+      "loss": 0.0004,
+      "step": 63692
+    },
+    {
+      "epoch": 59.42,
+      "learning_rate": 4.257276119402985e-05,
+      "loss": 0.0,
+      "step": 63696
+    },
+    {
+      "epoch": 59.42,
+      "learning_rate": 4.2572294776119406e-05,
+      "loss": 0.0,
+      "step": 63700
+    },
+    {
+      "epoch": 59.43,
+      "learning_rate": 4.257182835820896e-05,
+      "loss": 0.0001,
+      "step": 63704
+    },
+    {
+      "epoch": 59.43,
+      "learning_rate": 4.257136194029851e-05,
+      "loss": 0.0001,
+      "step": 63708
+    },
+    {
+      "epoch": 59.43,
+      "learning_rate": 4.257089552238806e-05,
+      "loss": 0.0001,
+      "step": 63712
+    },
+    {
+      "epoch": 59.44,
+      "learning_rate": 4.257042910447762e-05,
+      "loss": 0.0,
+      "step": 63716
+    },
+    {
+      "epoch": 59.44,
+      "learning_rate": 4.256996268656717e-05,
+      "loss": 0.0,
+      "step": 63720
+    },
+    {
+      "epoch": 59.44,
+      "learning_rate": 4.2569496268656715e-05,
+      "loss": 0.0,
+      "step": 63724
+    },
+    {
+      "epoch": 59.45,
+      "learning_rate": 4.256902985074627e-05,
+      "loss": 0.0,
+      "step": 63728
+    },
+    {
+      "epoch": 59.45,
+      "learning_rate": 4.2568563432835825e-05,
+      "loss": 0.0,
+      "step": 63732
+    },
+    {
+      "epoch": 59.46,
+      "learning_rate": 4.256809701492537e-05,
+      "loss": 0.0,
+      "step": 63736
+    },
+    {
+      "epoch": 59.46,
+      "learning_rate": 4.256763059701493e-05,
+      "loss": 0.0001,
+      "step": 63740
+    },
+    {
+      "epoch": 59.46,
+      "learning_rate": 4.256716417910448e-05,
+      "loss": 0.0,
+      "step": 63744
+    },
+    {
+      "epoch": 59.47,
+      "learning_rate": 4.256669776119403e-05,
+      "loss": 0.0,
+      "step": 63748
+    },
+    {
+      "epoch": 59.47,
+      "learning_rate": 4.2566231343283586e-05,
+      "loss": 0.0001,
+      "step": 63752
+    },
+    {
+      "epoch": 59.47,
+      "learning_rate": 4.2565764925373134e-05,
+      "loss": 0.0,
+      "step": 63756
+    },
+    {
+      "epoch": 59.48,
+      "learning_rate": 4.256529850746269e-05,
+      "loss": 0.001,
+      "step": 63760
+    },
+    {
+      "epoch": 59.48,
+      "learning_rate": 4.2564832089552244e-05,
+      "loss": 0.0,
+      "step": 63764
+    },
+    {
+      "epoch": 59.49,
+      "learning_rate": 4.256436567164179e-05,
+      "loss": 0.0001,
+      "step": 63768
+    },
+    {
+      "epoch": 59.49,
+      "learning_rate": 4.256389925373134e-05,
+      "loss": 0.0001,
+      "step": 63772
+    },
+    {
+      "epoch": 59.49,
+      "learning_rate": 4.25634328358209e-05,
+      "loss": 0.0,
+      "step": 63776
+    },
+    {
+      "epoch": 59.5,
+      "learning_rate": 4.256296641791045e-05,
+      "loss": 0.0006,
+      "step": 63780
+    },
+    {
+      "epoch": 59.5,
+      "learning_rate": 4.25625e-05,
+      "loss": 0.0,
+      "step": 63784
+    },
+    {
+      "epoch": 59.5,
+      "learning_rate": 4.256203358208955e-05,
+      "loss": 0.0,
+      "step": 63788
+    },
+    {
+      "epoch": 59.51,
+      "learning_rate": 4.256156716417911e-05,
+      "loss": 0.0002,
+      "step": 63792
+    },
+    {
+      "epoch": 59.51,
+      "learning_rate": 4.2561100746268656e-05,
+      "loss": 0.0001,
+      "step": 63796
+    },
+    {
+      "epoch": 59.51,
+      "learning_rate": 4.256063432835821e-05,
+      "loss": 0.0,
+      "step": 63800
+    },
+    {
+      "epoch": 59.52,
+      "learning_rate": 4.2560167910447766e-05,
+      "loss": 0.0,
+      "step": 63804
+    },
+    {
+      "epoch": 59.52,
+      "learning_rate": 4.2559701492537314e-05,
+      "loss": 0.0,
+      "step": 63808
+    },
+    {
+      "epoch": 59.53,
+      "learning_rate": 4.255923507462687e-05,
+      "loss": 0.0,
+      "step": 63812
+    },
+    {
+      "epoch": 59.53,
+      "learning_rate": 4.255876865671642e-05,
+      "loss": 0.0,
+      "step": 63816
+    },
+    {
+      "epoch": 59.53,
+      "learning_rate": 4.255830223880597e-05,
+      "loss": 0.0,
+      "step": 63820
+    },
+    {
+      "epoch": 59.54,
+      "learning_rate": 4.2557835820895527e-05,
+      "loss": 0.0,
+      "step": 63824
+    },
+    {
+      "epoch": 59.54,
+      "learning_rate": 4.2557369402985075e-05,
+      "loss": 0.0,
+      "step": 63828
+    },
+    {
+      "epoch": 59.54,
+      "learning_rate": 4.255690298507463e-05,
+      "loss": 0.0,
+      "step": 63832
+    },
+    {
+      "epoch": 59.55,
+      "learning_rate": 4.2556436567164184e-05,
+      "loss": 0.0,
+      "step": 63836
+    },
+    {
+      "epoch": 59.55,
+      "learning_rate": 4.255597014925373e-05,
+      "loss": 0.0,
+      "step": 63840
+    },
+    {
+      "epoch": 59.56,
+      "learning_rate": 4.255550373134329e-05,
+      "loss": 0.0001,
+      "step": 63844
+    },
+    {
+      "epoch": 59.56,
+      "learning_rate": 4.2555037313432836e-05,
+      "loss": 0.0,
+      "step": 63848
+    },
+    {
+      "epoch": 59.56,
+      "learning_rate": 4.255457089552239e-05,
+      "loss": 0.0,
+      "step": 63852
+    },
+    {
+      "epoch": 59.57,
+      "learning_rate": 4.2554104477611945e-05,
+      "loss": 0.0,
+      "step": 63856
+    },
+    {
+      "epoch": 59.57,
+      "learning_rate": 4.2553638059701494e-05,
+      "loss": 0.0,
+      "step": 63860
+    },
+    {
+      "epoch": 59.57,
+      "learning_rate": 4.255317164179105e-05,
+      "loss": 0.0,
+      "step": 63864
+    },
+    {
+      "epoch": 59.58,
+      "learning_rate": 4.2552705223880603e-05,
+      "loss": 0.0,
+      "step": 63868
+    },
+    {
+      "epoch": 59.58,
+      "learning_rate": 4.255223880597015e-05,
+      "loss": 0.0,
+      "step": 63872
+    },
+    {
+      "epoch": 59.59,
+      "learning_rate": 4.25517723880597e-05,
+      "loss": 0.0,
+      "step": 63876
+    },
+    {
+      "epoch": 59.59,
+      "learning_rate": 4.2551305970149255e-05,
+      "loss": 0.0,
+      "step": 63880
+    },
+    {
+      "epoch": 59.59,
+      "learning_rate": 4.255083955223881e-05,
+      "loss": 0.0,
+      "step": 63884
+    },
+    {
+      "epoch": 59.6,
+      "learning_rate": 4.255037313432836e-05,
+      "loss": 0.0004,
+      "step": 63888
+    },
+    {
+      "epoch": 59.6,
+      "learning_rate": 4.254990671641791e-05,
+      "loss": 0.0004,
+      "step": 63892
+    },
+    {
+      "epoch": 59.6,
+      "learning_rate": 4.254944029850747e-05,
+      "loss": 0.0003,
+      "step": 63896
+    },
+    {
+      "epoch": 59.61,
+      "learning_rate": 4.2548973880597016e-05,
+      "loss": 0.0,
+      "step": 63900
+    },
+    {
+      "epoch": 59.61,
+      "learning_rate": 4.254850746268657e-05,
+      "loss": 0.0,
+      "step": 63904
+    },
+    {
+      "epoch": 59.62,
+      "learning_rate": 4.254804104477612e-05,
+      "loss": 0.0,
+      "step": 63908
+    },
+    {
+      "epoch": 59.62,
+      "learning_rate": 4.2547574626865674e-05,
+      "loss": 0.0,
+      "step": 63912
+    },
+    {
+      "epoch": 59.62,
+      "learning_rate": 4.254710820895523e-05,
+      "loss": 0.0006,
+      "step": 63916
+    },
+    {
+      "epoch": 59.63,
+      "learning_rate": 4.2546641791044777e-05,
+      "loss": 0.0,
+      "step": 63920
+    },
+    {
+      "epoch": 59.63,
+      "learning_rate": 4.254617537313433e-05,
+      "loss": 0.0,
+      "step": 63924
+    },
+    {
+      "epoch": 59.63,
+      "learning_rate": 4.2545708955223886e-05,
+      "loss": 0.0,
+      "step": 63928
+    },
+    {
+      "epoch": 59.64,
+      "learning_rate": 4.2545242537313435e-05,
+      "loss": 0.0,
+      "step": 63932
+    },
+    {
+      "epoch": 59.64,
+      "learning_rate": 4.254477611940298e-05,
+      "loss": 0.0002,
+      "step": 63936
+    },
+    {
+      "epoch": 59.65,
+      "learning_rate": 4.254430970149254e-05,
+      "loss": 0.0,
+      "step": 63940
+    },
+    {
+      "epoch": 59.65,
+      "learning_rate": 4.254384328358209e-05,
+      "loss": 0.0,
+      "step": 63944
+    },
+    {
+      "epoch": 59.65,
+      "learning_rate": 4.254337686567164e-05,
+      "loss": 0.0,
+      "step": 63948
+    },
+    {
+      "epoch": 59.66,
+      "learning_rate": 4.2542910447761195e-05,
+      "loss": 0.0,
+      "step": 63952
+    },
+    {
+      "epoch": 59.66,
+      "learning_rate": 4.254244402985075e-05,
+      "loss": 0.0,
+      "step": 63956
+    },
+    {
+      "epoch": 59.66,
+      "learning_rate": 4.25419776119403e-05,
+      "loss": 0.0,
+      "step": 63960
+    },
+    {
+      "epoch": 59.67,
+      "learning_rate": 4.2541511194029853e-05,
+      "loss": 0.0,
+      "step": 63964
+    },
+    {
+      "epoch": 59.67,
+      "learning_rate": 4.25410447761194e-05,
+      "loss": 0.0002,
+      "step": 63968
+    },
+    {
+      "epoch": 59.68,
+      "learning_rate": 4.2540578358208956e-05,
+      "loss": 0.0,
+      "step": 63972
+    },
+    {
+      "epoch": 59.68,
+      "learning_rate": 4.254011194029851e-05,
+      "loss": 0.0,
+      "step": 63976
+    },
+    {
+      "epoch": 59.68,
+      "learning_rate": 4.253964552238806e-05,
+      "loss": 0.0001,
+      "step": 63980
+    },
+    {
+      "epoch": 59.69,
+      "learning_rate": 4.2539179104477614e-05,
+      "loss": 0.0001,
+      "step": 63984
+    },
+    {
+      "epoch": 59.69,
+      "learning_rate": 4.253871268656717e-05,
+      "loss": 0.0,
+      "step": 63988
+    },
+    {
+      "epoch": 59.69,
+      "learning_rate": 4.253824626865672e-05,
+      "loss": 0.0,
+      "step": 63992
+    },
+    {
+      "epoch": 59.7,
+      "learning_rate": 4.253777985074627e-05,
+      "loss": 0.0,
+      "step": 63996
+    },
+    {
+      "epoch": 59.7,
+      "learning_rate": 4.253731343283582e-05,
+      "loss": 0.0014,
+      "step": 64000
+    },
+    {
+      "epoch": 59.7,
+      "eval_exact_match": 0.7437137330754352,
+      "eval_exec": 0.7891682785299806,
+      "eval_loss": 0.5204067826271057,
+      "eval_runtime": 1133.3553,
+      "eval_samples_per_second": 0.912,
+      "step": 64000
+    },
+    {
+      "epoch": 59.71,
+      "learning_rate": 4.2536847014925375e-05,
+      "loss": 0.0001,
+      "step": 64004
+    },
+    {
+      "epoch": 59.71,
+      "learning_rate": 4.253638059701493e-05,
+      "loss": 0.0,
+      "step": 64008
+    },
+    {
+      "epoch": 59.71,
+      "learning_rate": 4.253591417910448e-05,
+      "loss": 0.0,
+      "step": 64012
+    },
+    {
+      "epoch": 59.72,
+      "learning_rate": 4.253544776119403e-05,
+      "loss": 0.0,
+      "step": 64016
+    },
+    {
+      "epoch": 59.72,
+      "learning_rate": 4.253498134328359e-05,
+      "loss": 0.0,
+      "step": 64020
+    },
+    {
+      "epoch": 59.72,
+      "learning_rate": 4.2534514925373136e-05,
+      "loss": 0.003,
+      "step": 64024
+    },
+    {
+      "epoch": 59.73,
+      "learning_rate": 4.2534048507462685e-05,
+      "loss": 0.0,
+      "step": 64028
+    },
+    {
+      "epoch": 59.73,
+      "learning_rate": 4.2533582089552246e-05,
+      "loss": 0.0016,
+      "step": 64032
+    },
+    {
+      "epoch": 59.73,
+      "learning_rate": 4.2533115671641794e-05,
+      "loss": 0.0001,
+      "step": 64036
+    },
+    {
+      "epoch": 59.74,
+      "learning_rate": 4.253264925373134e-05,
+      "loss": 0.0,
+      "step": 64040
+    },
+    {
+      "epoch": 59.74,
+      "learning_rate": 4.25321828358209e-05,
+      "loss": 0.0,
+      "step": 64044
+    },
+    {
+      "epoch": 59.75,
+      "learning_rate": 4.253171641791045e-05,
+      "loss": 0.0014,
+      "step": 64048
+    },
+    {
+      "epoch": 59.75,
+      "learning_rate": 4.253125e-05,
+      "loss": 0.0,
+      "step": 64052
+    },
+    {
+      "epoch": 59.75,
+      "learning_rate": 4.2530783582089555e-05,
+      "loss": 0.0,
+      "step": 64056
+    },
+    {
+      "epoch": 59.76,
+      "learning_rate": 4.2530317164179103e-05,
+      "loss": 0.0,
+      "step": 64060
+    },
+    {
+      "epoch": 59.76,
+      "learning_rate": 4.252985074626866e-05,
+      "loss": 0.0,
+      "step": 64064
+    },
+    {
+      "epoch": 59.76,
+      "learning_rate": 4.252938432835821e-05,
+      "loss": 0.0,
+      "step": 64068
+    },
+    {
+      "epoch": 59.77,
+      "learning_rate": 4.252891791044776e-05,
+      "loss": 0.0,
+      "step": 64072
+    },
+    {
+      "epoch": 59.77,
+      "learning_rate": 4.2528451492537316e-05,
+      "loss": 0.0,
+      "step": 64076
+    },
+    {
+      "epoch": 59.78,
+      "learning_rate": 4.252798507462687e-05,
+      "loss": 0.0,
+      "step": 64080
+    },
+    {
+      "epoch": 59.78,
+      "learning_rate": 4.252751865671642e-05,
+      "loss": 0.0002,
+      "step": 64084
+    },
+    {
+      "epoch": 59.78,
+      "learning_rate": 4.252705223880597e-05,
+      "loss": 0.0,
+      "step": 64088
+    },
+    {
+      "epoch": 59.79,
+      "learning_rate": 4.252658582089553e-05,
+      "loss": 0.0,
+      "step": 64092
+    },
+    {
+      "epoch": 59.79,
+      "learning_rate": 4.252611940298508e-05,
+      "loss": 0.0,
+      "step": 64096
+    },
+    {
+      "epoch": 59.79,
+      "learning_rate": 4.2525652985074625e-05,
+      "loss": 0.0,
+      "step": 64100
+    },
+    {
+      "epoch": 59.8,
+      "learning_rate": 4.252518656716418e-05,
+      "loss": 0.0,
+      "step": 64104
+    },
+    {
+      "epoch": 59.8,
+      "learning_rate": 4.2524720149253735e-05,
+      "loss": 0.0,
+      "step": 64108
+    },
+    {
+      "epoch": 59.81,
+      "learning_rate": 4.252425373134328e-05,
+      "loss": 0.0,
+      "step": 64112
+    },
+    {
+      "epoch": 59.81,
+      "learning_rate": 4.252378731343284e-05,
+      "loss": 0.0,
+      "step": 64116
+    },
+    {
+      "epoch": 59.81,
+      "learning_rate": 4.2523320895522386e-05,
+      "loss": 0.0,
+      "step": 64120
+    },
+    {
+      "epoch": 59.82,
+      "learning_rate": 4.252285447761194e-05,
+      "loss": 0.0,
+      "step": 64124
+    },
+    {
+      "epoch": 59.82,
+      "learning_rate": 4.2522388059701496e-05,
+      "loss": 0.0,
+      "step": 64128
+    },
+    {
+      "epoch": 59.82,
+      "learning_rate": 4.2521921641791044e-05,
+      "loss": 0.0,
+      "step": 64132
+    },
+    {
+      "epoch": 59.83,
+      "learning_rate": 4.25214552238806e-05,
+      "loss": 0.0001,
+      "step": 64136
+    },
+    {
+      "epoch": 59.83,
+      "learning_rate": 4.2520988805970154e-05,
+      "loss": 0.0,
+      "step": 64140
+    },
+    {
+      "epoch": 59.84,
+      "learning_rate": 4.25205223880597e-05,
+      "loss": 0.0,
+      "step": 64144
+    },
+    {
+      "epoch": 59.84,
+      "learning_rate": 4.252005597014926e-05,
+      "loss": 0.0,
+      "step": 64148
+    },
+    {
+      "epoch": 59.84,
+      "learning_rate": 4.251958955223881e-05,
+      "loss": 0.0,
+      "step": 64152
+    },
+    {
+      "epoch": 59.85,
+      "learning_rate": 4.251912313432836e-05,
+      "loss": 0.0,
+      "step": 64156
+    },
+    {
+      "epoch": 59.85,
+      "learning_rate": 4.2518656716417915e-05,
+      "loss": 0.0,
+      "step": 64160
+    },
+    {
+      "epoch": 59.85,
+      "learning_rate": 4.251819029850746e-05,
+      "loss": 0.001,
+      "step": 64164
+    },
+    {
+      "epoch": 59.86,
+      "learning_rate": 4.251772388059702e-05,
+      "loss": 0.0001,
+      "step": 64168
+    },
+    {
+      "epoch": 59.86,
+      "learning_rate": 4.251725746268657e-05,
+      "loss": 0.0,
+      "step": 64172
+    },
+    {
+      "epoch": 59.87,
+      "learning_rate": 4.251679104477612e-05,
+      "loss": 0.0,
+      "step": 64176
+    },
+    {
+      "epoch": 59.87,
+      "learning_rate": 4.251632462686567e-05,
+      "loss": 0.0069,
+      "step": 64180
+    },
+    {
+      "epoch": 59.87,
+      "learning_rate": 4.251585820895523e-05,
+      "loss": 0.0,
+      "step": 64184
+    },
+    {
+      "epoch": 59.88,
+      "learning_rate": 4.251539179104478e-05,
+      "loss": 0.0,
+      "step": 64188
+    },
+    {
+      "epoch": 59.88,
+      "learning_rate": 4.251492537313433e-05,
+      "loss": 0.0004,
+      "step": 64192
+    },
+    {
+      "epoch": 59.88,
+      "learning_rate": 4.251445895522388e-05,
+      "loss": 0.0,
+      "step": 64196
+    },
+    {
+      "epoch": 59.89,
+      "learning_rate": 4.251399253731344e-05,
+      "loss": 0.0,
+      "step": 64200
+    },
+    {
+      "epoch": 59.89,
+      "learning_rate": 4.2513526119402985e-05,
+      "loss": 0.0,
+      "step": 64204
+    },
+    {
+      "epoch": 59.9,
+      "learning_rate": 4.251305970149254e-05,
+      "loss": 0.0,
+      "step": 64208
+    },
+    {
+      "epoch": 59.9,
+      "learning_rate": 4.2512593283582095e-05,
+      "loss": 0.0006,
+      "step": 64212
+    },
+    {
+      "epoch": 59.9,
+      "learning_rate": 4.251212686567164e-05,
+      "loss": 0.0,
+      "step": 64216
+    },
+    {
+      "epoch": 59.91,
+      "learning_rate": 4.25116604477612e-05,
+      "loss": 0.0,
+      "step": 64220
+    },
+    {
+      "epoch": 59.91,
+      "learning_rate": 4.2511194029850746e-05,
+      "loss": 0.0,
+      "step": 64224
+    },
+    {
+      "epoch": 59.91,
+      "learning_rate": 4.25107276119403e-05,
+      "loss": 0.0,
+      "step": 64228
+    },
+    {
+      "epoch": 59.92,
+      "learning_rate": 4.2510261194029856e-05,
+      "loss": 0.0001,
+      "step": 64232
+    },
+    {
+      "epoch": 59.92,
+      "learning_rate": 4.2509794776119404e-05,
+      "loss": 0.0,
+      "step": 64236
+    },
+    {
+      "epoch": 59.93,
+      "learning_rate": 4.250932835820895e-05,
+      "loss": 0.0,
+      "step": 64240
+    },
+    {
+      "epoch": 59.93,
+      "learning_rate": 4.2508861940298514e-05,
+      "loss": 0.0,
+      "step": 64244
+    },
+    {
+      "epoch": 59.93,
+      "learning_rate": 4.250839552238806e-05,
+      "loss": 0.0004,
+      "step": 64248
+    },
+    {
+      "epoch": 59.94,
+      "learning_rate": 4.250792910447761e-05,
+      "loss": 0.0001,
+      "step": 64252
+    },
+    {
+      "epoch": 59.94,
+      "learning_rate": 4.2507462686567165e-05,
+      "loss": 0.0,
+      "step": 64256
+    },
+    {
+      "epoch": 59.94,
+      "learning_rate": 4.250699626865672e-05,
+      "loss": 0.0,
+      "step": 64260
+    },
+    {
+      "epoch": 59.95,
+      "learning_rate": 4.250652985074627e-05,
+      "loss": 0.0,
+      "step": 64264
+    },
+    {
+      "epoch": 59.95,
+      "learning_rate": 4.250606343283582e-05,
+      "loss": 0.0,
+      "step": 64268
+    },
+    {
+      "epoch": 59.96,
+      "learning_rate": 4.250559701492537e-05,
+      "loss": 0.0003,
+      "step": 64272
+    },
+    {
+      "epoch": 59.96,
+      "learning_rate": 4.2505130597014926e-05,
+      "loss": 0.0,
+      "step": 64276
+    },
+    {
+      "epoch": 59.96,
+      "learning_rate": 4.250466417910448e-05,
+      "loss": 0.0,
+      "step": 64280
+    },
+    {
+      "epoch": 59.97,
+      "learning_rate": 4.250419776119403e-05,
+      "loss": 0.0,
+      "step": 64284
+    },
+    {
+      "epoch": 59.97,
+      "learning_rate": 4.2503731343283584e-05,
+      "loss": 0.0,
+      "step": 64288
+    },
+    {
+      "epoch": 59.97,
+      "learning_rate": 4.250326492537314e-05,
+      "loss": 0.0,
+      "step": 64292
+    },
+    {
+      "epoch": 59.98,
+      "learning_rate": 4.250279850746269e-05,
+      "loss": 0.0001,
+      "step": 64296
+    },
+    {
+      "epoch": 59.98,
+      "learning_rate": 4.2502332089552235e-05,
+      "loss": 0.0,
+      "step": 64300
+    },
+    {
+      "epoch": 59.98,
+      "learning_rate": 4.25018656716418e-05,
+      "loss": 0.0,
+      "step": 64304
+    },
+    {
+      "epoch": 59.99,
+      "learning_rate": 4.2501399253731345e-05,
+      "loss": 0.0,
+      "step": 64308
+    },
+    {
+      "epoch": 59.99,
+      "learning_rate": 4.25009328358209e-05,
+      "loss": 0.0,
+      "step": 64312
+    },
+    {
+      "epoch": 60.0,
+      "learning_rate": 4.250046641791045e-05,
+      "loss": 0.0,
+      "step": 64316
+    },
+    {
+      "epoch": 60.0,
+      "learning_rate": 4.25e-05,
+      "loss": 0.0,
+      "step": 64320
+    },
+    {
+      "epoch": 60.0,
+      "learning_rate": 4.249953358208956e-05,
+      "loss": 0.0,
+      "step": 64324
+    },
+    {
+      "epoch": 60.01,
+      "learning_rate": 4.2499067164179106e-05,
+      "loss": 0.0001,
+      "step": 64328
+    },
+    {
+      "epoch": 60.01,
+      "learning_rate": 4.2498600746268654e-05,
+      "loss": 0.0,
+      "step": 64332
+    },
+    {
+      "epoch": 60.01,
+      "learning_rate": 4.2498134328358216e-05,
+      "loss": 0.0,
+      "step": 64336
+    },
+    {
+      "epoch": 60.02,
+      "learning_rate": 4.2497667910447764e-05,
+      "loss": 0.0,
+      "step": 64340
+    },
+    {
+      "epoch": 60.02,
+      "learning_rate": 4.249720149253731e-05,
+      "loss": 0.0019,
+      "step": 64344
+    },
+    {
+      "epoch": 60.03,
+      "learning_rate": 4.249673507462687e-05,
+      "loss": 0.0013,
+      "step": 64348
+    },
+    {
+      "epoch": 60.03,
+      "learning_rate": 4.249626865671642e-05,
+      "loss": 0.0,
+      "step": 64352
+    },
+    {
+      "epoch": 60.03,
+      "learning_rate": 4.249580223880597e-05,
+      "loss": 0.0,
+      "step": 64356
+    },
+    {
+      "epoch": 60.04,
+      "learning_rate": 4.2495335820895525e-05,
+      "loss": 0.0,
+      "step": 64360
+    },
+    {
+      "epoch": 60.04,
+      "learning_rate": 4.249486940298508e-05,
+      "loss": 0.0,
+      "step": 64364
+    },
+    {
+      "epoch": 60.04,
+      "learning_rate": 4.249440298507463e-05,
+      "loss": 0.0,
+      "step": 64368
+    },
+    {
+      "epoch": 60.05,
+      "learning_rate": 4.249393656716418e-05,
+      "loss": 0.0,
+      "step": 64372
+    },
+    {
+      "epoch": 60.05,
+      "learning_rate": 4.249347014925373e-05,
+      "loss": 0.0,
+      "step": 64376
+    },
+    {
+      "epoch": 60.06,
+      "learning_rate": 4.2493003731343286e-05,
+      "loss": 0.0,
+      "step": 64380
+    },
+    {
+      "epoch": 60.06,
+      "learning_rate": 4.249253731343284e-05,
+      "loss": 0.0,
+      "step": 64384
+    },
+    {
+      "epoch": 60.06,
+      "learning_rate": 4.249207089552239e-05,
+      "loss": 0.0,
+      "step": 64388
+    },
+    {
+      "epoch": 60.07,
+      "learning_rate": 4.249160447761194e-05,
+      "loss": 0.0,
+      "step": 64392
+    },
+    {
+      "epoch": 60.07,
+      "learning_rate": 4.24911380597015e-05,
+      "loss": 0.0,
+      "step": 64396
+    },
+    {
+      "epoch": 60.07,
+      "learning_rate": 4.249067164179105e-05,
+      "loss": 0.0014,
+      "step": 64400
+    },
+    {
+      "epoch": 60.08,
+      "learning_rate": 4.2490205223880595e-05,
+      "loss": 0.0,
+      "step": 64404
+    },
+    {
+      "epoch": 60.08,
+      "learning_rate": 4.248973880597015e-05,
+      "loss": 0.0,
+      "step": 64408
+    },
+    {
+      "epoch": 60.09,
+      "learning_rate": 4.2489272388059705e-05,
+      "loss": 0.0,
+      "step": 64412
+    },
+    {
+      "epoch": 60.09,
+      "learning_rate": 4.248880597014925e-05,
+      "loss": 0.0,
+      "step": 64416
+    },
+    {
+      "epoch": 60.09,
+      "learning_rate": 4.248833955223881e-05,
+      "loss": 0.0,
+      "step": 64420
+    },
+    {
+      "epoch": 60.1,
+      "learning_rate": 4.248787313432836e-05,
+      "loss": 0.0001,
+      "step": 64424
+    },
+    {
+      "epoch": 60.1,
+      "learning_rate": 4.248740671641791e-05,
+      "loss": 0.0,
+      "step": 64428
+    },
+    {
+      "epoch": 60.1,
+      "learning_rate": 4.2486940298507466e-05,
+      "loss": 0.0001,
+      "step": 64432
+    },
+    {
+      "epoch": 60.11,
+      "learning_rate": 4.2486473880597014e-05,
+      "loss": 0.0,
+      "step": 64436
+    },
+    {
+      "epoch": 60.11,
+      "learning_rate": 4.248600746268657e-05,
+      "loss": 0.0,
+      "step": 64440
+    },
+    {
+      "epoch": 60.12,
+      "learning_rate": 4.2485541044776124e-05,
+      "loss": 0.0,
+      "step": 64444
+    },
+    {
+      "epoch": 60.12,
+      "learning_rate": 4.248507462686567e-05,
+      "loss": 0.0,
+      "step": 64448
+    },
+    {
+      "epoch": 60.12,
+      "learning_rate": 4.248460820895522e-05,
+      "loss": 0.0,
+      "step": 64452
+    },
+    {
+      "epoch": 60.13,
+      "learning_rate": 4.248414179104478e-05,
+      "loss": 0.0,
+      "step": 64456
+    },
+    {
+      "epoch": 60.13,
+      "learning_rate": 4.248367537313433e-05,
+      "loss": 0.0,
+      "step": 64460
+    },
+    {
+      "epoch": 60.13,
+      "learning_rate": 4.248320895522388e-05,
+      "loss": 0.0,
+      "step": 64464
+    },
+    {
+      "epoch": 60.14,
+      "learning_rate": 4.248274253731343e-05,
+      "loss": 0.0,
+      "step": 64468
+    },
+    {
+      "epoch": 60.14,
+      "learning_rate": 4.248227611940299e-05,
+      "loss": 0.0,
+      "step": 64472
+    },
+    {
+      "epoch": 60.15,
+      "learning_rate": 4.248180970149254e-05,
+      "loss": 0.0,
+      "step": 64476
+    },
+    {
+      "epoch": 60.15,
+      "learning_rate": 4.248134328358209e-05,
+      "loss": 0.0,
+      "step": 64480
+    },
+    {
+      "epoch": 60.15,
+      "learning_rate": 4.2480876865671646e-05,
+      "loss": 0.0,
+      "step": 64484
+    },
+    {
+      "epoch": 60.16,
+      "learning_rate": 4.24804104477612e-05,
+      "loss": 0.0,
+      "step": 64488
+    },
+    {
+      "epoch": 60.16,
+      "learning_rate": 4.247994402985075e-05,
+      "loss": 0.0,
+      "step": 64492
+    },
+    {
+      "epoch": 60.16,
+      "learning_rate": 4.24794776119403e-05,
+      "loss": 0.0002,
+      "step": 64496
+    },
+    {
+      "epoch": 60.17,
+      "learning_rate": 4.247901119402985e-05,
+      "loss": 0.0,
+      "step": 64500
+    },
+    {
+      "epoch": 60.17,
+      "eval_exact_match": 0.7485493230174082,
+      "eval_exec": 0.7911025145067698,
+      "eval_loss": 0.5219013690948486,
+      "eval_runtime": 1099.0283,
+      "eval_samples_per_second": 0.941,
+      "step": 64500
+    },
+    {
+      "epoch": 60.17,
+      "learning_rate": 4.247854477611941e-05,
+      "loss": 0.0,
+      "step": 64504
+    },
+    {
+      "epoch": 60.18,
+      "learning_rate": 4.2478078358208955e-05,
+      "loss": 0.0,
+      "step": 64508
+    },
+    {
+      "epoch": 60.18,
+      "learning_rate": 4.247761194029851e-05,
+      "loss": 0.0005,
+      "step": 64512
+    },
+    {
+      "epoch": 60.18,
+      "learning_rate": 4.2477145522388065e-05,
+      "loss": 0.0,
+      "step": 64516
+    },
+    {
+      "epoch": 60.19,
+      "learning_rate": 4.247667910447761e-05,
+      "loss": 0.0007,
+      "step": 64520
+    },
+    {
+      "epoch": 60.19,
+      "learning_rate": 4.247621268656717e-05,
+      "loss": 0.0,
+      "step": 64524
+    },
+    {
+      "epoch": 60.19,
+      "learning_rate": 4.2475746268656716e-05,
+      "loss": 0.0,
+      "step": 64528
+    },
+    {
+      "epoch": 60.2,
+      "learning_rate": 4.247527985074627e-05,
+      "loss": 0.0046,
+      "step": 64532
+    },
+    {
+      "epoch": 60.2,
+      "learning_rate": 4.2474813432835826e-05,
+      "loss": 0.0001,
+      "step": 64536
+    },
+    {
+      "epoch": 60.21,
+      "learning_rate": 4.2474347014925374e-05,
+      "loss": 0.0,
+      "step": 64540
+    },
+    {
+      "epoch": 60.21,
+      "learning_rate": 4.247388059701493e-05,
+      "loss": 0.0003,
+      "step": 64544
+    },
+    {
+      "epoch": 60.21,
+      "learning_rate": 4.2473414179104484e-05,
+      "loss": 0.0,
+      "step": 64548
+    },
+    {
+      "epoch": 60.22,
+      "learning_rate": 4.247294776119403e-05,
+      "loss": 0.0,
+      "step": 64552
+    },
+    {
+      "epoch": 60.22,
+      "learning_rate": 4.247248134328358e-05,
+      "loss": 0.0,
+      "step": 64556
+    },
+    {
+      "epoch": 60.22,
+      "learning_rate": 4.2472014925373135e-05,
+      "loss": 0.0001,
+      "step": 64560
+    },
+    {
+      "epoch": 60.23,
+      "learning_rate": 4.247154850746269e-05,
+      "loss": 0.0,
+      "step": 64564
+    },
+    {
+      "epoch": 60.23,
+      "learning_rate": 4.247108208955224e-05,
+      "loss": 0.0,
+      "step": 64568
+    },
+    {
+      "epoch": 60.24,
+      "learning_rate": 4.247061567164179e-05,
+      "loss": 0.0,
+      "step": 64572
+    },
+    {
+      "epoch": 60.24,
+      "learning_rate": 4.247014925373135e-05,
+      "loss": 0.0,
+      "step": 64576
+    },
+    {
+      "epoch": 60.24,
+      "learning_rate": 4.2469682835820896e-05,
+      "loss": 0.0,
+      "step": 64580
+    },
+    {
+      "epoch": 60.25,
+      "learning_rate": 4.246921641791045e-05,
+      "loss": 0.0004,
+      "step": 64584
+    },
+    {
+      "epoch": 60.25,
+      "learning_rate": 4.246875e-05,
+      "loss": 0.0009,
+      "step": 64588
+    },
+    {
+      "epoch": 60.25,
+      "learning_rate": 4.2468283582089554e-05,
+      "loss": 0.0002,
+      "step": 64592
+    },
+    {
+      "epoch": 60.26,
+      "learning_rate": 4.246781716417911e-05,
+      "loss": 0.0,
+      "step": 64596
+    },
+    {
+      "epoch": 60.26,
+      "learning_rate": 4.246735074626866e-05,
+      "loss": 0.0,
+      "step": 64600
+    },
+    {
+      "epoch": 60.26,
+      "learning_rate": 4.246688432835821e-05,
+      "loss": 0.0,
+      "step": 64604
+    },
+    {
+      "epoch": 60.27,
+      "learning_rate": 4.246641791044777e-05,
+      "loss": 0.0008,
+      "step": 64608
+    },
+    {
+      "epoch": 60.27,
+      "learning_rate": 4.2465951492537315e-05,
+      "loss": 0.0,
+      "step": 64612
+    },
+    {
+      "epoch": 60.28,
+      "learning_rate": 4.246548507462686e-05,
+      "loss": 0.0037,
+      "step": 64616
+    },
+    {
+      "epoch": 60.28,
+      "learning_rate": 4.246501865671642e-05,
+      "loss": 0.0,
+      "step": 64620
+    },
+    {
+      "epoch": 60.28,
+      "learning_rate": 4.246455223880597e-05,
+      "loss": 0.0,
+      "step": 64624
+    },
+    {
+      "epoch": 60.29,
+      "learning_rate": 4.246408582089552e-05,
+      "loss": 0.0,
+      "step": 64628
+    },
+    {
+      "epoch": 60.29,
+      "learning_rate": 4.2463619402985076e-05,
+      "loss": 0.0,
+      "step": 64632
+    },
+    {
+      "epoch": 60.29,
+      "learning_rate": 4.246315298507463e-05,
+      "loss": 0.0007,
+      "step": 64636
+    },
+    {
+      "epoch": 60.3,
+      "learning_rate": 4.2462686567164186e-05,
+      "loss": 0.0002,
+      "step": 64640
+    },
+    {
+      "epoch": 60.3,
+      "learning_rate": 4.2462220149253734e-05,
+      "loss": 0.0046,
+      "step": 64644
+    },
+    {
+      "epoch": 60.31,
+      "learning_rate": 4.246175373134328e-05,
+      "loss": 0.0,
+      "step": 64648
+    },
+    {
+      "epoch": 60.31,
+      "learning_rate": 4.2461287313432844e-05,
+      "loss": 0.0001,
+      "step": 64652
+    },
+    {
+      "epoch": 60.31,
+      "learning_rate": 4.246082089552239e-05,
+      "loss": 0.0,
+      "step": 64656
+    },
+    {
+      "epoch": 60.32,
+      "learning_rate": 4.246035447761194e-05,
+      "loss": 0.0,
+      "step": 64660
+    },
+    {
+      "epoch": 60.32,
+      "learning_rate": 4.2459888059701495e-05,
+      "loss": 0.0,
+      "step": 64664
+    },
+    {
+      "epoch": 60.32,
+      "learning_rate": 4.245942164179105e-05,
+      "loss": 0.0,
+      "step": 64668
+    },
+    {
+      "epoch": 60.33,
+      "learning_rate": 4.24589552238806e-05,
+      "loss": 0.0,
+      "step": 64672
+    },
+    {
+      "epoch": 60.33,
+      "learning_rate": 4.245848880597015e-05,
+      "loss": 0.0,
+      "step": 64676
+    },
+    {
+      "epoch": 60.34,
+      "learning_rate": 4.24580223880597e-05,
+      "loss": 0.0,
+      "step": 64680
+    },
+    {
+      "epoch": 60.34,
+      "learning_rate": 4.2457555970149256e-05,
+      "loss": 0.0,
+      "step": 64684
+    },
+    {
+      "epoch": 60.34,
+      "learning_rate": 4.245708955223881e-05,
+      "loss": 0.0,
+      "step": 64688
+    },
+    {
+      "epoch": 60.35,
+      "learning_rate": 4.245662313432836e-05,
+      "loss": 0.0,
+      "step": 64692
+    },
+    {
+      "epoch": 60.35,
+      "learning_rate": 4.2456156716417914e-05,
+      "loss": 0.0,
+      "step": 64696
+    },
+    {
+      "epoch": 60.35,
+      "learning_rate": 4.245569029850747e-05,
+      "loss": 0.0,
+      "step": 64700
+    },
+    {
+      "epoch": 60.36,
+      "learning_rate": 4.245522388059702e-05,
+      "loss": 0.0,
+      "step": 64704
+    },
+    {
+      "epoch": 60.36,
+      "learning_rate": 4.2454757462686565e-05,
+      "loss": 0.0001,
+      "step": 64708
+    },
+    {
+      "epoch": 60.37,
+      "learning_rate": 4.2454291044776127e-05,
+      "loss": 0.0021,
+      "step": 64712
+    },
+    {
+      "epoch": 60.37,
+      "learning_rate": 4.2453824626865675e-05,
+      "loss": 0.0004,
+      "step": 64716
+    },
+    {
+      "epoch": 60.37,
+      "learning_rate": 4.245335820895522e-05,
+      "loss": 0.0,
+      "step": 64720
+    },
+    {
+      "epoch": 60.38,
+      "learning_rate": 4.245289179104478e-05,
+      "loss": 0.0001,
+      "step": 64724
+    },
+    {
+      "epoch": 60.38,
+      "learning_rate": 4.245242537313433e-05,
+      "loss": 0.0,
+      "step": 64728
+    },
+    {
+      "epoch": 60.38,
+      "learning_rate": 4.245195895522388e-05,
+      "loss": 0.0,
+      "step": 64732
+    },
+    {
+      "epoch": 60.39,
+      "learning_rate": 4.2451492537313436e-05,
+      "loss": 0.0,
+      "step": 64736
+    },
+    {
+      "epoch": 60.39,
+      "learning_rate": 4.2451026119402984e-05,
+      "loss": 0.0,
+      "step": 64740
+    },
+    {
+      "epoch": 60.4,
+      "learning_rate": 4.245055970149254e-05,
+      "loss": 0.0011,
+      "step": 64744
+    },
+    {
+      "epoch": 60.4,
+      "learning_rate": 4.2450093283582094e-05,
+      "loss": 0.0,
+      "step": 64748
+    },
+    {
+      "epoch": 60.4,
+      "learning_rate": 4.244962686567164e-05,
+      "loss": 0.0001,
+      "step": 64752
+    },
+    {
+      "epoch": 60.41,
+      "learning_rate": 4.2449160447761197e-05,
+      "loss": 0.0,
+      "step": 64756
+    },
+    {
+      "epoch": 60.41,
+      "learning_rate": 4.244869402985075e-05,
+      "loss": 0.0001,
+      "step": 64760
+    },
+    {
+      "epoch": 60.41,
+      "learning_rate": 4.24482276119403e-05,
+      "loss": 0.0,
+      "step": 64764
+    },
+    {
+      "epoch": 60.42,
+      "learning_rate": 4.244776119402985e-05,
+      "loss": 0.0,
+      "step": 64768
+    },
+    {
+      "epoch": 60.42,
+      "learning_rate": 4.244729477611941e-05,
+      "loss": 0.0,
+      "step": 64772
+    },
+    {
+      "epoch": 60.43,
+      "learning_rate": 4.244682835820896e-05,
+      "loss": 0.0,
+      "step": 64776
+    },
+    {
+      "epoch": 60.43,
+      "learning_rate": 4.2446361940298506e-05,
+      "loss": 0.0,
+      "step": 64780
+    },
+    {
+      "epoch": 60.43,
+      "learning_rate": 4.244589552238806e-05,
+      "loss": 0.0001,
+      "step": 64784
+    },
+    {
+      "epoch": 60.44,
+      "learning_rate": 4.2445429104477616e-05,
+      "loss": 0.0,
+      "step": 64788
+    },
+    {
+      "epoch": 60.44,
+      "learning_rate": 4.2444962686567164e-05,
+      "loss": 0.0007,
+      "step": 64792
+    },
+    {
+      "epoch": 60.44,
+      "learning_rate": 4.244449626865672e-05,
+      "loss": 0.0086,
+      "step": 64796
+    },
+    {
+      "epoch": 60.45,
+      "learning_rate": 4.244402985074627e-05,
+      "loss": 0.0002,
+      "step": 64800
+    },
+    {
+      "epoch": 60.45,
+      "learning_rate": 4.244356343283583e-05,
+      "loss": 0.0,
+      "step": 64804
+    },
+    {
+      "epoch": 60.46,
+      "learning_rate": 4.2443097014925377e-05,
+      "loss": 0.0,
+      "step": 64808
+    },
+    {
+      "epoch": 60.46,
+      "learning_rate": 4.2442630597014925e-05,
+      "loss": 0.0,
+      "step": 64812
+    },
+    {
+      "epoch": 60.46,
+      "learning_rate": 4.244216417910448e-05,
+      "loss": 0.0,
+      "step": 64816
+    },
+    {
+      "epoch": 60.47,
+      "learning_rate": 4.2441697761194034e-05,
+      "loss": 0.0,
+      "step": 64820
+    },
+    {
+      "epoch": 60.47,
+      "learning_rate": 4.244123134328358e-05,
+      "loss": 0.0,
+      "step": 64824
+    },
+    {
+      "epoch": 60.47,
+      "learning_rate": 4.244076492537314e-05,
+      "loss": 0.0,
+      "step": 64828
+    },
+    {
+      "epoch": 60.48,
+      "learning_rate": 4.244029850746269e-05,
+      "loss": 0.0,
+      "step": 64832
+    },
+    {
+      "epoch": 60.48,
+      "learning_rate": 4.243983208955224e-05,
+      "loss": 0.0001,
+      "step": 64836
+    },
+    {
+      "epoch": 60.49,
+      "learning_rate": 4.2439365671641795e-05,
+      "loss": 0.0,
+      "step": 64840
+    },
+    {
+      "epoch": 60.49,
+      "learning_rate": 4.2438899253731344e-05,
+      "loss": 0.0,
+      "step": 64844
+    },
+    {
+      "epoch": 60.49,
+      "learning_rate": 4.24384328358209e-05,
+      "loss": 0.0,
+      "step": 64848
+    },
+    {
+      "epoch": 60.5,
+      "learning_rate": 4.2437966417910453e-05,
+      "loss": 0.0002,
+      "step": 64852
+    },
+    {
+      "epoch": 60.5,
+      "learning_rate": 4.24375e-05,
+      "loss": 0.0002,
+      "step": 64856
+    },
+    {
+      "epoch": 60.5,
+      "learning_rate": 4.243703358208955e-05,
+      "loss": 0.0,
+      "step": 64860
+    },
+    {
+      "epoch": 60.51,
+      "learning_rate": 4.243656716417911e-05,
+      "loss": 0.0001,
+      "step": 64864
+    },
+    {
+      "epoch": 60.51,
+      "learning_rate": 4.243610074626866e-05,
+      "loss": 0.0,
+      "step": 64868
+    },
+    {
+      "epoch": 60.51,
+      "learning_rate": 4.243563432835821e-05,
+      "loss": 0.0,
+      "step": 64872
+    },
+    {
+      "epoch": 60.52,
+      "learning_rate": 4.243516791044776e-05,
+      "loss": 0.0,
+      "step": 64876
+    },
+    {
+      "epoch": 60.52,
+      "learning_rate": 4.243470149253732e-05,
+      "loss": 0.0,
+      "step": 64880
+    },
+    {
+      "epoch": 60.53,
+      "learning_rate": 4.2434235074626866e-05,
+      "loss": 0.0001,
+      "step": 64884
+    },
+    {
+      "epoch": 60.53,
+      "learning_rate": 4.243376865671642e-05,
+      "loss": 0.0,
+      "step": 64888
+    },
+    {
+      "epoch": 60.53,
+      "learning_rate": 4.2433302238805975e-05,
+      "loss": 0.0001,
+      "step": 64892
+    },
+    {
+      "epoch": 60.54,
+      "learning_rate": 4.2432835820895524e-05,
+      "loss": 0.0,
+      "step": 64896
+    },
+    {
+      "epoch": 60.54,
+      "learning_rate": 4.243236940298508e-05,
+      "loss": 0.0003,
+      "step": 64900
+    },
+    {
+      "epoch": 60.54,
+      "learning_rate": 4.2431902985074627e-05,
+      "loss": 0.0009,
+      "step": 64904
+    },
+    {
+      "epoch": 60.55,
+      "learning_rate": 4.243143656716418e-05,
+      "loss": 0.0,
+      "step": 64908
+    },
+    {
+      "epoch": 60.55,
+      "learning_rate": 4.2430970149253736e-05,
+      "loss": 0.0,
+      "step": 64912
+    },
+    {
+      "epoch": 60.56,
+      "learning_rate": 4.2430503731343284e-05,
+      "loss": 0.0002,
+      "step": 64916
+    },
+    {
+      "epoch": 60.56,
+      "learning_rate": 4.243003731343283e-05,
+      "loss": 0.0004,
+      "step": 64920
+    },
+    {
+      "epoch": 60.56,
+      "learning_rate": 4.2429570895522394e-05,
+      "loss": 0.0,
+      "step": 64924
+    },
+    {
+      "epoch": 60.57,
+      "learning_rate": 4.242910447761194e-05,
+      "loss": 0.0,
+      "step": 64928
+    },
+    {
+      "epoch": 60.57,
+      "learning_rate": 4.242863805970149e-05,
+      "loss": 0.0,
+      "step": 64932
+    },
+    {
+      "epoch": 60.57,
+      "learning_rate": 4.2428171641791045e-05,
+      "loss": 0.0,
+      "step": 64936
+    },
+    {
+      "epoch": 60.58,
+      "learning_rate": 4.24277052238806e-05,
+      "loss": 0.0,
+      "step": 64940
+    },
+    {
+      "epoch": 60.58,
+      "learning_rate": 4.242723880597015e-05,
+      "loss": 0.0001,
+      "step": 64944
+    },
+    {
+      "epoch": 60.59,
+      "learning_rate": 4.2426772388059703e-05,
+      "loss": 0.0001,
+      "step": 64948
+    },
+    {
+      "epoch": 60.59,
+      "learning_rate": 4.242630597014925e-05,
+      "loss": 0.0001,
+      "step": 64952
+    },
+    {
+      "epoch": 60.59,
+      "learning_rate": 4.2425839552238806e-05,
+      "loss": 0.0001,
+      "step": 64956
+    },
+    {
+      "epoch": 60.6,
+      "learning_rate": 4.242537313432836e-05,
+      "loss": 0.0,
+      "step": 64960
+    },
+    {
+      "epoch": 60.6,
+      "learning_rate": 4.242490671641791e-05,
+      "loss": 0.0003,
+      "step": 64964
+    },
+    {
+      "epoch": 60.6,
+      "learning_rate": 4.2424440298507464e-05,
+      "loss": 0.0,
+      "step": 64968
+    },
+    {
+      "epoch": 60.61,
+      "learning_rate": 4.242397388059702e-05,
+      "loss": 0.0,
+      "step": 64972
+    },
+    {
+      "epoch": 60.61,
+      "learning_rate": 4.242350746268657e-05,
+      "loss": 0.0002,
+      "step": 64976
+    },
+    {
+      "epoch": 60.62,
+      "learning_rate": 4.242304104477612e-05,
+      "loss": 0.0,
+      "step": 64980
+    },
+    {
+      "epoch": 60.62,
+      "learning_rate": 4.242257462686568e-05,
+      "loss": 0.0,
+      "step": 64984
+    },
+    {
+      "epoch": 60.62,
+      "learning_rate": 4.2422108208955225e-05,
+      "loss": 0.0,
+      "step": 64988
+    },
+    {
+      "epoch": 60.63,
+      "learning_rate": 4.242164179104478e-05,
+      "loss": 0.0,
+      "step": 64992
+    },
+    {
+      "epoch": 60.63,
+      "learning_rate": 4.242117537313433e-05,
+      "loss": 0.0,
+      "step": 64996
+    },
+    {
+      "epoch": 60.63,
+      "learning_rate": 4.242070895522388e-05,
+      "loss": 0.0,
+      "step": 65000
+    },
+    {
+      "epoch": 60.63,
+      "eval_exact_match": 0.741779497098646,
+      "eval_exec": 0.7852998065764023,
+      "eval_loss": 0.5173532366752625,
+      "eval_runtime": 1115.6158,
+      "eval_samples_per_second": 0.927,
+      "step": 65000
+    },
+    {
+      "epoch": 60.64,
+      "learning_rate": 4.242024253731344e-05,
+      "loss": 0.0,
+      "step": 65004
+    },
+    {
+      "epoch": 60.64,
+      "learning_rate": 4.2419776119402986e-05,
+      "loss": 0.0,
+      "step": 65008
+    },
+    {
+      "epoch": 60.65,
+      "learning_rate": 4.2419309701492535e-05,
+      "loss": 0.0,
+      "step": 65012
+    },
+    {
+      "epoch": 60.65,
+      "learning_rate": 4.2418843283582096e-05,
+      "loss": 0.0,
+      "step": 65016
+    },
+    {
+      "epoch": 60.65,
+      "learning_rate": 4.2418376865671644e-05,
+      "loss": 0.0002,
+      "step": 65020
+    },
+    {
+      "epoch": 60.66,
+      "learning_rate": 4.241791044776119e-05,
+      "loss": 0.0001,
+      "step": 65024
+    },
+    {
+      "epoch": 60.66,
+      "learning_rate": 4.241744402985075e-05,
+      "loss": 0.0015,
+      "step": 65028
+    },
+    {
+      "epoch": 60.66,
+      "learning_rate": 4.24169776119403e-05,
+      "loss": 0.0004,
+      "step": 65032
+    },
+    {
+      "epoch": 60.67,
+      "learning_rate": 4.241651119402985e-05,
+      "loss": 0.0008,
+      "step": 65036
+    },
+    {
+      "epoch": 60.67,
+      "learning_rate": 4.2416044776119405e-05,
+      "loss": 0.0,
+      "step": 65040
+    },
+    {
+      "epoch": 60.68,
+      "learning_rate": 4.241557835820896e-05,
+      "loss": 0.0,
+      "step": 65044
+    },
+    {
+      "epoch": 60.68,
+      "learning_rate": 4.241511194029851e-05,
+      "loss": 0.0,
+      "step": 65048
+    },
+    {
+      "epoch": 60.68,
+      "learning_rate": 4.241464552238806e-05,
+      "loss": 0.0,
+      "step": 65052
+    },
+    {
+      "epoch": 60.69,
+      "learning_rate": 4.241417910447761e-05,
+      "loss": 0.0,
+      "step": 65056
+    },
+    {
+      "epoch": 60.69,
+      "learning_rate": 4.2413712686567166e-05,
+      "loss": 0.0001,
+      "step": 65060
+    },
+    {
+      "epoch": 60.69,
+      "learning_rate": 4.241324626865672e-05,
+      "loss": 0.0017,
+      "step": 65064
+    },
+    {
+      "epoch": 60.7,
+      "learning_rate": 4.241277985074627e-05,
+      "loss": 0.0,
+      "step": 65068
+    },
+    {
+      "epoch": 60.7,
+      "learning_rate": 4.241231343283582e-05,
+      "loss": 0.0,
+      "step": 65072
+    },
+    {
+      "epoch": 60.71,
+      "learning_rate": 4.241184701492538e-05,
+      "loss": 0.0,
+      "step": 65076
+    },
+    {
+      "epoch": 60.71,
+      "learning_rate": 4.241138059701493e-05,
+      "loss": 0.0002,
+      "step": 65080
+    },
+    {
+      "epoch": 60.71,
+      "learning_rate": 4.2410914179104475e-05,
+      "loss": 0.0,
+      "step": 65084
+    },
+    {
+      "epoch": 60.72,
+      "learning_rate": 4.241044776119403e-05,
+      "loss": 0.0,
+      "step": 65088
+    },
+    {
+      "epoch": 60.72,
+      "learning_rate": 4.2409981343283585e-05,
+      "loss": 0.0,
+      "step": 65092
+    },
+    {
+      "epoch": 60.72,
+      "learning_rate": 4.240951492537313e-05,
+      "loss": 0.0,
+      "step": 65096
+    },
+    {
+      "epoch": 60.73,
+      "learning_rate": 4.240904850746269e-05,
+      "loss": 0.0,
+      "step": 65100
+    },
+    {
+      "epoch": 60.73,
+      "learning_rate": 4.240858208955224e-05,
+      "loss": 0.0,
+      "step": 65104
+    },
+    {
+      "epoch": 60.73,
+      "learning_rate": 4.240811567164179e-05,
+      "loss": 0.0,
+      "step": 65108
+    },
+    {
+      "epoch": 60.74,
+      "learning_rate": 4.2407649253731346e-05,
+      "loss": 0.0042,
+      "step": 65112
+    },
+    {
+      "epoch": 60.74,
+      "learning_rate": 4.2407182835820894e-05,
+      "loss": 0.0,
+      "step": 65116
+    },
+    {
+      "epoch": 60.75,
+      "learning_rate": 4.240671641791045e-05,
+      "loss": 0.0,
+      "step": 65120
+    },
+    {
+      "epoch": 60.75,
+      "learning_rate": 4.2406250000000004e-05,
+      "loss": 0.0,
+      "step": 65124
+    },
+    {
+      "epoch": 60.75,
+      "learning_rate": 4.240578358208955e-05,
+      "loss": 0.0,
+      "step": 65128
+    },
+    {
+      "epoch": 60.76,
+      "learning_rate": 4.240531716417911e-05,
+      "loss": 0.0,
+      "step": 65132
+    },
+    {
+      "epoch": 60.76,
+      "learning_rate": 4.240485074626866e-05,
+      "loss": 0.0,
+      "step": 65136
+    },
+    {
+      "epoch": 60.76,
+      "learning_rate": 4.240438432835821e-05,
+      "loss": 0.0,
+      "step": 65140
+    },
+    {
+      "epoch": 60.77,
+      "learning_rate": 4.2403917910447765e-05,
+      "loss": 0.0001,
+      "step": 65144
+    },
+    {
+      "epoch": 60.77,
+      "learning_rate": 4.240345149253731e-05,
+      "loss": 0.0,
+      "step": 65148
+    },
+    {
+      "epoch": 60.78,
+      "learning_rate": 4.240298507462687e-05,
+      "loss": 0.0,
+      "step": 65152
+    },
+    {
+      "epoch": 60.78,
+      "learning_rate": 4.240251865671642e-05,
+      "loss": 0.0,
+      "step": 65156
+    },
+    {
+      "epoch": 60.78,
+      "learning_rate": 4.240205223880597e-05,
+      "loss": 0.0001,
+      "step": 65160
+    },
+    {
+      "epoch": 60.79,
+      "learning_rate": 4.2401585820895526e-05,
+      "loss": 0.0001,
+      "step": 65164
+    },
+    {
+      "epoch": 60.79,
+      "learning_rate": 4.240111940298508e-05,
+      "loss": 0.0,
+      "step": 65168
+    },
+    {
+      "epoch": 60.79,
+      "learning_rate": 4.240065298507463e-05,
+      "loss": 0.0001,
+      "step": 65172
+    },
+    {
+      "epoch": 60.8,
+      "learning_rate": 4.240018656716418e-05,
+      "loss": 0.0001,
+      "step": 65176
+    },
+    {
+      "epoch": 60.8,
+      "learning_rate": 4.239972014925373e-05,
+      "loss": 0.0,
+      "step": 65180
+    },
+    {
+      "epoch": 60.81,
+      "learning_rate": 4.239925373134329e-05,
+      "loss": 0.0001,
+      "step": 65184
+    },
+    {
+      "epoch": 60.81,
+      "learning_rate": 4.2398787313432835e-05,
+      "loss": 0.0,
+      "step": 65188
+    },
+    {
+      "epoch": 60.81,
+      "learning_rate": 4.239832089552239e-05,
+      "loss": 0.0,
+      "step": 65192
+    },
+    {
+      "epoch": 60.82,
+      "learning_rate": 4.2397854477611945e-05,
+      "loss": 0.0,
+      "step": 65196
+    },
+    {
+      "epoch": 60.82,
+      "learning_rate": 4.239738805970149e-05,
+      "loss": 0.0001,
+      "step": 65200
+    },
+    {
+      "epoch": 60.82,
+      "learning_rate": 4.239692164179105e-05,
+      "loss": 0.0,
+      "step": 65204
+    },
+    {
+      "epoch": 60.83,
+      "learning_rate": 4.2396455223880596e-05,
+      "loss": 0.0001,
+      "step": 65208
+    },
+    {
+      "epoch": 60.83,
+      "learning_rate": 4.239598880597015e-05,
+      "loss": 0.0,
+      "step": 65212
+    },
+    {
+      "epoch": 60.84,
+      "learning_rate": 4.2395522388059706e-05,
+      "loss": 0.0,
+      "step": 65216
+    },
+    {
+      "epoch": 60.84,
+      "learning_rate": 4.2395055970149254e-05,
+      "loss": 0.0,
+      "step": 65220
+    },
+    {
+      "epoch": 60.84,
+      "learning_rate": 4.239458955223881e-05,
+      "loss": 0.0098,
+      "step": 65224
+    },
+    {
+      "epoch": 60.85,
+      "learning_rate": 4.2394123134328364e-05,
+      "loss": 0.0,
+      "step": 65228
+    },
+    {
+      "epoch": 60.85,
+      "learning_rate": 4.239365671641791e-05,
+      "loss": 0.0,
+      "step": 65232
+    },
+    {
+      "epoch": 60.85,
+      "learning_rate": 4.239319029850746e-05,
+      "loss": 0.0,
+      "step": 65236
+    },
+    {
+      "epoch": 60.86,
+      "learning_rate": 4.2392723880597015e-05,
+      "loss": 0.0,
+      "step": 65240
+    },
+    {
+      "epoch": 60.86,
+      "learning_rate": 4.239225746268657e-05,
+      "loss": 0.0,
+      "step": 65244
+    },
+    {
+      "epoch": 60.87,
+      "learning_rate": 4.239179104477612e-05,
+      "loss": 0.0,
+      "step": 65248
+    },
+    {
+      "epoch": 60.87,
+      "learning_rate": 4.239132462686567e-05,
+      "loss": 0.0,
+      "step": 65252
+    },
+    {
+      "epoch": 60.87,
+      "learning_rate": 4.239085820895523e-05,
+      "loss": 0.0,
+      "step": 65256
+    },
+    {
+      "epoch": 60.88,
+      "learning_rate": 4.2390391791044776e-05,
+      "loss": 0.0,
+      "step": 65260
+    },
+    {
+      "epoch": 60.88,
+      "learning_rate": 4.238992537313433e-05,
+      "loss": 0.0,
+      "step": 65264
+    },
+    {
+      "epoch": 60.88,
+      "learning_rate": 4.238945895522388e-05,
+      "loss": 0.0028,
+      "step": 65268
+    },
+    {
+      "epoch": 60.89,
+      "learning_rate": 4.2388992537313434e-05,
+      "loss": 0.0,
+      "step": 65272
+    },
+    {
+      "epoch": 60.89,
+      "learning_rate": 4.238852611940299e-05,
+      "loss": 0.0002,
+      "step": 65276
+    },
+    {
+      "epoch": 60.9,
+      "learning_rate": 4.238805970149254e-05,
+      "loss": 0.0,
+      "step": 65280
+    },
+    {
+      "epoch": 60.9,
+      "learning_rate": 4.238759328358209e-05,
+      "loss": 0.0,
+      "step": 65284
+    },
+    {
+      "epoch": 60.9,
+      "learning_rate": 4.238712686567165e-05,
+      "loss": 0.0,
+      "step": 65288
+    },
+    {
+      "epoch": 60.91,
+      "learning_rate": 4.2386660447761195e-05,
+      "loss": 0.0,
+      "step": 65292
+    },
+    {
+      "epoch": 60.91,
+      "learning_rate": 4.238619402985075e-05,
+      "loss": 0.0,
+      "step": 65296
+    },
+    {
+      "epoch": 60.91,
+      "learning_rate": 4.23857276119403e-05,
+      "loss": 0.0,
+      "step": 65300
+    },
+    {
+      "epoch": 60.92,
+      "learning_rate": 4.238526119402985e-05,
+      "loss": 0.0011,
+      "step": 65304
+    },
+    {
+      "epoch": 60.92,
+      "learning_rate": 4.238479477611941e-05,
+      "loss": 0.0003,
+      "step": 65308
+    },
+    {
+      "epoch": 60.93,
+      "learning_rate": 4.2384328358208956e-05,
+      "loss": 0.001,
+      "step": 65312
+    },
+    {
+      "epoch": 60.93,
+      "learning_rate": 4.238386194029851e-05,
+      "loss": 0.0,
+      "step": 65316
+    },
+    {
+      "epoch": 60.93,
+      "learning_rate": 4.2383395522388066e-05,
+      "loss": 0.0,
+      "step": 65320
+    },
+    {
+      "epoch": 60.94,
+      "learning_rate": 4.2382929104477614e-05,
+      "loss": 0.0,
+      "step": 65324
+    },
+    {
+      "epoch": 60.94,
+      "learning_rate": 4.238246268656716e-05,
+      "loss": 0.0,
+      "step": 65328
+    },
+    {
+      "epoch": 60.94,
+      "learning_rate": 4.2381996268656724e-05,
+      "loss": 0.0,
+      "step": 65332
+    },
+    {
+      "epoch": 60.95,
+      "learning_rate": 4.238152985074627e-05,
+      "loss": 0.0,
+      "step": 65336
+    },
+    {
+      "epoch": 60.95,
+      "learning_rate": 4.238106343283582e-05,
+      "loss": 0.0,
+      "step": 65340
+    },
+    {
+      "epoch": 60.96,
+      "learning_rate": 4.2380597014925375e-05,
+      "loss": 0.0,
+      "step": 65344
+    },
+    {
+      "epoch": 60.96,
+      "learning_rate": 4.238013059701493e-05,
+      "loss": 0.0,
+      "step": 65348
+    },
+    {
+      "epoch": 60.96,
+      "learning_rate": 4.237966417910448e-05,
+      "loss": 0.0001,
+      "step": 65352
+    },
+    {
+      "epoch": 60.97,
+      "learning_rate": 4.237919776119403e-05,
+      "loss": 0.0019,
+      "step": 65356
+    },
+    {
+      "epoch": 60.97,
+      "learning_rate": 4.237873134328358e-05,
+      "loss": 0.0001,
+      "step": 65360
+    },
+    {
+      "epoch": 60.97,
+      "learning_rate": 4.2378264925373136e-05,
+      "loss": 0.0,
+      "step": 65364
+    },
+    {
+      "epoch": 60.98,
+      "learning_rate": 4.237779850746269e-05,
+      "loss": 0.0,
+      "step": 65368
+    },
+    {
+      "epoch": 60.98,
+      "learning_rate": 4.237733208955224e-05,
+      "loss": 0.0005,
+      "step": 65372
+    },
+    {
+      "epoch": 60.98,
+      "learning_rate": 4.2376865671641794e-05,
+      "loss": 0.0042,
+      "step": 65376
+    },
+    {
+      "epoch": 60.99,
+      "learning_rate": 4.237639925373135e-05,
+      "loss": 0.0,
+      "step": 65380
+    },
+    {
+      "epoch": 60.99,
+      "learning_rate": 4.23759328358209e-05,
+      "loss": 0.0,
+      "step": 65384
+    },
+    {
+      "epoch": 61.0,
+      "learning_rate": 4.2375466417910445e-05,
+      "loss": 0.0,
+      "step": 65388
+    },
+    {
+      "epoch": 61.0,
+      "learning_rate": 4.237500000000001e-05,
+      "loss": 0.0,
+      "step": 65392
+    },
+    {
+      "epoch": 61.0,
+      "learning_rate": 4.2374533582089555e-05,
+      "loss": 0.0,
+      "step": 65396
+    },
+    {
+      "epoch": 61.01,
+      "learning_rate": 4.23740671641791e-05,
+      "loss": 0.0,
+      "step": 65400
+    },
+    {
+      "epoch": 61.01,
+      "learning_rate": 4.237360074626866e-05,
+      "loss": 0.0,
+      "step": 65404
+    },
+    {
+      "epoch": 61.01,
+      "learning_rate": 4.237313432835821e-05,
+      "loss": 0.0,
+      "step": 65408
+    },
+    {
+      "epoch": 61.02,
+      "learning_rate": 4.237266791044776e-05,
+      "loss": 0.0,
+      "step": 65412
+    },
+    {
+      "epoch": 61.02,
+      "learning_rate": 4.2372201492537316e-05,
+      "loss": 0.0,
+      "step": 65416
+    },
+    {
+      "epoch": 61.03,
+      "learning_rate": 4.2371735074626864e-05,
+      "loss": 0.0,
+      "step": 65420
+    },
+    {
+      "epoch": 61.03,
+      "learning_rate": 4.237126865671642e-05,
+      "loss": 0.0126,
+      "step": 65424
+    },
+    {
+      "epoch": 61.03,
+      "learning_rate": 4.2370802238805974e-05,
+      "loss": 0.0,
+      "step": 65428
+    },
+    {
+      "epoch": 61.04,
+      "learning_rate": 4.237033582089552e-05,
+      "loss": 0.0,
+      "step": 65432
+    },
+    {
+      "epoch": 61.04,
+      "learning_rate": 4.236986940298508e-05,
+      "loss": 0.0,
+      "step": 65436
+    },
+    {
+      "epoch": 61.04,
+      "learning_rate": 4.236940298507463e-05,
+      "loss": 0.0,
+      "step": 65440
+    },
+    {
+      "epoch": 61.05,
+      "learning_rate": 4.236893656716418e-05,
+      "loss": 0.0,
+      "step": 65444
+    },
+    {
+      "epoch": 61.05,
+      "learning_rate": 4.2368470149253735e-05,
+      "loss": 0.0,
+      "step": 65448
+    },
+    {
+      "epoch": 61.06,
+      "learning_rate": 4.236800373134329e-05,
+      "loss": 0.0,
+      "step": 65452
+    },
+    {
+      "epoch": 61.06,
+      "learning_rate": 4.236753731343284e-05,
+      "loss": 0.0,
+      "step": 65456
+    },
+    {
+      "epoch": 61.06,
+      "learning_rate": 4.236707089552239e-05,
+      "loss": 0.0,
+      "step": 65460
+    },
+    {
+      "epoch": 61.07,
+      "learning_rate": 4.236660447761194e-05,
+      "loss": 0.0015,
+      "step": 65464
+    },
+    {
+      "epoch": 61.07,
+      "learning_rate": 4.2366138059701496e-05,
+      "loss": 0.0,
+      "step": 65468
+    },
+    {
+      "epoch": 61.07,
+      "learning_rate": 4.236567164179105e-05,
+      "loss": 0.0002,
+      "step": 65472
+    },
+    {
+      "epoch": 61.08,
+      "learning_rate": 4.23652052238806e-05,
+      "loss": 0.0,
+      "step": 65476
+    },
+    {
+      "epoch": 61.08,
+      "learning_rate": 4.236473880597015e-05,
+      "loss": 0.0,
+      "step": 65480
+    },
+    {
+      "epoch": 61.09,
+      "learning_rate": 4.236427238805971e-05,
+      "loss": 0.0001,
+      "step": 65484
+    },
+    {
+      "epoch": 61.09,
+      "learning_rate": 4.236380597014926e-05,
+      "loss": 0.0017,
+      "step": 65488
+    },
+    {
+      "epoch": 61.09,
+      "learning_rate": 4.2363339552238805e-05,
+      "loss": 0.0001,
+      "step": 65492
+    },
+    {
+      "epoch": 61.1,
+      "learning_rate": 4.236287313432836e-05,
+      "loss": 0.0,
+      "step": 65496
+    },
+    {
+      "epoch": 61.1,
+      "learning_rate": 4.2362406716417915e-05,
+      "loss": 0.0,
+      "step": 65500
+    },
+    {
+      "epoch": 61.1,
+      "eval_exact_match": 0.7475822050290135,
+      "eval_exec": 0.7833655705996132,
+      "eval_loss": 0.49484971165657043,
+      "eval_runtime": 1108.9884,
+      "eval_samples_per_second": 0.932,
+      "step": 65500
+    },
+    {
+      "epoch": 61.1,
+      "learning_rate": 4.236194029850746e-05,
+      "loss": 0.0,
+      "step": 65504
+    },
+    {
+      "epoch": 61.11,
+      "learning_rate": 4.236147388059702e-05,
+      "loss": 0.0014,
+      "step": 65508
+    },
+    {
+      "epoch": 61.11,
+      "learning_rate": 4.236100746268657e-05,
+      "loss": 0.0011,
+      "step": 65512
+    },
+    {
+      "epoch": 61.12,
+      "learning_rate": 4.236054104477612e-05,
+      "loss": 0.0,
+      "step": 65516
+    },
+    {
+      "epoch": 61.12,
+      "learning_rate": 4.2360074626865676e-05,
+      "loss": 0.0,
+      "step": 65520
+    },
+    {
+      "epoch": 61.12,
+      "learning_rate": 4.2359608208955224e-05,
+      "loss": 0.0001,
+      "step": 65524
+    },
+    {
+      "epoch": 61.13,
+      "learning_rate": 4.235914179104478e-05,
+      "loss": 0.0001,
+      "step": 65528
+    },
+    {
+      "epoch": 61.13,
+      "learning_rate": 4.2358675373134334e-05,
+      "loss": 0.0,
+      "step": 65532
+    },
+    {
+      "epoch": 61.13,
+      "learning_rate": 4.235820895522388e-05,
+      "loss": 0.0,
+      "step": 65536
+    },
+    {
+      "epoch": 61.14,
+      "learning_rate": 4.235774253731343e-05,
+      "loss": 0.0,
+      "step": 65540
+    },
+    {
+      "epoch": 61.14,
+      "learning_rate": 4.235727611940299e-05,
+      "loss": 0.0001,
+      "step": 65544
+    },
+    {
+      "epoch": 61.15,
+      "learning_rate": 4.235680970149254e-05,
+      "loss": 0.0,
+      "step": 65548
+    },
+    {
+      "epoch": 61.15,
+      "learning_rate": 4.235634328358209e-05,
+      "loss": 0.0,
+      "step": 65552
+    },
+    {
+      "epoch": 61.15,
+      "learning_rate": 4.235587686567164e-05,
+      "loss": 0.0,
+      "step": 65556
+    },
+    {
+      "epoch": 61.16,
+      "learning_rate": 4.23554104477612e-05,
+      "loss": 0.0,
+      "step": 65560
+    },
+    {
+      "epoch": 61.16,
+      "learning_rate": 4.2354944029850746e-05,
+      "loss": 0.0001,
+      "step": 65564
+    },
+    {
+      "epoch": 61.16,
+      "learning_rate": 4.23544776119403e-05,
+      "loss": 0.0026,
+      "step": 65568
+    },
+    {
+      "epoch": 61.17,
+      "learning_rate": 4.2354011194029856e-05,
+      "loss": 0.0001,
+      "step": 65572
+    },
+    {
+      "epoch": 61.17,
+      "learning_rate": 4.2353544776119404e-05,
+      "loss": 0.0001,
+      "step": 65576
+    },
+    {
+      "epoch": 61.18,
+      "learning_rate": 4.235307835820896e-05,
+      "loss": 0.0,
+      "step": 65580
+    },
+    {
+      "epoch": 61.18,
+      "learning_rate": 4.235261194029851e-05,
+      "loss": 0.0,
+      "step": 65584
+    },
+    {
+      "epoch": 61.18,
+      "learning_rate": 4.235214552238806e-05,
+      "loss": 0.0053,
+      "step": 65588
+    },
+    {
+      "epoch": 61.19,
+      "learning_rate": 4.235167910447762e-05,
+      "loss": 0.0003,
+      "step": 65592
+    },
+    {
+      "epoch": 61.19,
+      "learning_rate": 4.2351212686567165e-05,
+      "loss": 0.0006,
+      "step": 65596
+    },
+    {
+      "epoch": 61.19,
+      "learning_rate": 4.235074626865671e-05,
+      "loss": 0.0012,
+      "step": 65600
+    },
+    {
+      "epoch": 61.2,
+      "learning_rate": 4.2350279850746275e-05,
+      "loss": 0.0,
+      "step": 65604
+    },
+    {
+      "epoch": 61.2,
+      "learning_rate": 4.234981343283582e-05,
+      "loss": 0.0,
+      "step": 65608
+    },
+    {
+      "epoch": 61.21,
+      "learning_rate": 4.234934701492538e-05,
+      "loss": 0.0,
+      "step": 65612
+    },
+    {
+      "epoch": 61.21,
+      "learning_rate": 4.2348880597014926e-05,
+      "loss": 0.0,
+      "step": 65616
+    },
+    {
+      "epoch": 61.21,
+      "learning_rate": 4.234841417910448e-05,
+      "loss": 0.0004,
+      "step": 65620
+    },
+    {
+      "epoch": 61.22,
+      "learning_rate": 4.2347947761194036e-05,
+      "loss": 0.0,
+      "step": 65624
+    },
+    {
+      "epoch": 61.22,
+      "learning_rate": 4.2347481343283584e-05,
+      "loss": 0.0001,
+      "step": 65628
+    },
+    {
+      "epoch": 61.22,
+      "learning_rate": 4.234701492537313e-05,
+      "loss": 0.0,
+      "step": 65632
+    },
+    {
+      "epoch": 61.23,
+      "learning_rate": 4.2346548507462694e-05,
+      "loss": 0.0,
+      "step": 65636
+    },
+    {
+      "epoch": 61.23,
+      "learning_rate": 4.234608208955224e-05,
+      "loss": 0.0052,
+      "step": 65640
+    },
+    {
+      "epoch": 61.24,
+      "learning_rate": 4.234561567164179e-05,
+      "loss": 0.0,
+      "step": 65644
+    },
+    {
+      "epoch": 61.24,
+      "learning_rate": 4.2345149253731345e-05,
+      "loss": 0.0,
+      "step": 65648
+    },
+    {
+      "epoch": 61.24,
+      "learning_rate": 4.23446828358209e-05,
+      "loss": 0.0001,
+      "step": 65652
+    },
+    {
+      "epoch": 61.25,
+      "learning_rate": 4.234421641791045e-05,
+      "loss": 0.0001,
+      "step": 65656
+    },
+    {
+      "epoch": 61.25,
+      "learning_rate": 4.234375e-05,
+      "loss": 0.0,
+      "step": 65660
+    },
+    {
+      "epoch": 61.25,
+      "learning_rate": 4.234328358208956e-05,
+      "loss": 0.0,
+      "step": 65664
+    },
+    {
+      "epoch": 61.26,
+      "learning_rate": 4.2342817164179106e-05,
+      "loss": 0.0,
+      "step": 65668
+    },
+    {
+      "epoch": 61.26,
+      "learning_rate": 4.234235074626866e-05,
+      "loss": 0.0,
+      "step": 65672
+    },
+    {
+      "epoch": 61.26,
+      "learning_rate": 4.234188432835821e-05,
+      "loss": 0.0,
+      "step": 65676
+    },
+    {
+      "epoch": 61.27,
+      "learning_rate": 4.2341417910447764e-05,
+      "loss": 0.0,
+      "step": 65680
+    },
+    {
+      "epoch": 61.27,
+      "learning_rate": 4.234095149253732e-05,
+      "loss": 0.0004,
+      "step": 65684
+    },
+    {
+      "epoch": 61.28,
+      "learning_rate": 4.234048507462687e-05,
+      "loss": 0.0002,
+      "step": 65688
+    },
+    {
+      "epoch": 61.28,
+      "learning_rate": 4.2340018656716415e-05,
+      "loss": 0.0,
+      "step": 65692
+    },
+    {
+      "epoch": 61.28,
+      "learning_rate": 4.2339552238805976e-05,
+      "loss": 0.0003,
+      "step": 65696
+    },
+    {
+      "epoch": 61.29,
+      "learning_rate": 4.2339085820895525e-05,
+      "loss": 0.0021,
+      "step": 65700
+    },
+    {
+      "epoch": 61.29,
+      "learning_rate": 4.233861940298507e-05,
+      "loss": 0.0,
+      "step": 65704
+    },
+    {
+      "epoch": 61.29,
+      "learning_rate": 4.233815298507463e-05,
+      "loss": 0.0009,
+      "step": 65708
+    },
+    {
+      "epoch": 61.3,
+      "learning_rate": 4.233768656716418e-05,
+      "loss": 0.0008,
+      "step": 65712
+    },
+    {
+      "epoch": 61.3,
+      "learning_rate": 4.233722014925373e-05,
+      "loss": 0.0,
+      "step": 65716
+    },
+    {
+      "epoch": 61.31,
+      "learning_rate": 4.2336753731343286e-05,
+      "loss": 0.0,
+      "step": 65720
+    },
+    {
+      "epoch": 61.31,
+      "learning_rate": 4.233628731343284e-05,
+      "loss": 0.0,
+      "step": 65724
+    },
+    {
+      "epoch": 61.31,
+      "learning_rate": 4.233582089552239e-05,
+      "loss": 0.0,
+      "step": 65728
+    },
+    {
+      "epoch": 61.32,
+      "learning_rate": 4.2335354477611944e-05,
+      "loss": 0.0003,
+      "step": 65732
+    },
+    {
+      "epoch": 61.32,
+      "learning_rate": 4.233488805970149e-05,
+      "loss": 0.0015,
+      "step": 65736
+    },
+    {
+      "epoch": 61.32,
+      "learning_rate": 4.2334421641791047e-05,
+      "loss": 0.0,
+      "step": 65740
+    },
+    {
+      "epoch": 61.33,
+      "learning_rate": 4.23339552238806e-05,
+      "loss": 0.002,
+      "step": 65744
+    },
+    {
+      "epoch": 61.33,
+      "learning_rate": 4.233348880597015e-05,
+      "loss": 0.0,
+      "step": 65748
+    },
+    {
+      "epoch": 61.34,
+      "learning_rate": 4.23330223880597e-05,
+      "loss": 0.0051,
+      "step": 65752
+    },
+    {
+      "epoch": 61.34,
+      "learning_rate": 4.233255597014926e-05,
+      "loss": 0.0007,
+      "step": 65756
+    },
+    {
+      "epoch": 61.34,
+      "learning_rate": 4.233208955223881e-05,
+      "loss": 0.0,
+      "step": 65760
+    },
+    {
+      "epoch": 61.35,
+      "learning_rate": 4.2331623134328356e-05,
+      "loss": 0.0,
+      "step": 65764
+    },
+    {
+      "epoch": 61.35,
+      "learning_rate": 4.233115671641791e-05,
+      "loss": 0.0,
+      "step": 65768
+    },
+    {
+      "epoch": 61.35,
+      "learning_rate": 4.2330690298507466e-05,
+      "loss": 0.0,
+      "step": 65772
+    },
+    {
+      "epoch": 61.36,
+      "learning_rate": 4.233022388059702e-05,
+      "loss": 0.0,
+      "step": 65776
+    },
+    {
+      "epoch": 61.36,
+      "learning_rate": 4.232975746268657e-05,
+      "loss": 0.0,
+      "step": 65780
+    },
+    {
+      "epoch": 61.37,
+      "learning_rate": 4.2329291044776123e-05,
+      "loss": 0.0002,
+      "step": 65784
+    },
+    {
+      "epoch": 61.37,
+      "learning_rate": 4.232882462686568e-05,
+      "loss": 0.0001,
+      "step": 65788
+    },
+    {
+      "epoch": 61.37,
+      "learning_rate": 4.2328358208955227e-05,
+      "loss": 0.0089,
+      "step": 65792
+    },
+    {
+      "epoch": 61.38,
+      "learning_rate": 4.2327891791044775e-05,
+      "loss": 0.0,
+      "step": 65796
+    },
+    {
+      "epoch": 61.38,
+      "learning_rate": 4.232742537313433e-05,
+      "loss": 0.0,
+      "step": 65800
+    },
+    {
+      "epoch": 61.38,
+      "learning_rate": 4.2326958955223884e-05,
+      "loss": 0.0,
+      "step": 65804
+    },
+    {
+      "epoch": 61.39,
+      "learning_rate": 4.232649253731343e-05,
+      "loss": 0.0003,
+      "step": 65808
+    },
+    {
+      "epoch": 61.39,
+      "learning_rate": 4.232602611940299e-05,
+      "loss": 0.0,
+      "step": 65812
+    },
+    {
+      "epoch": 61.4,
+      "learning_rate": 4.232555970149254e-05,
+      "loss": 0.0,
+      "step": 65816
+    },
+    {
+      "epoch": 61.4,
+      "learning_rate": 4.232509328358209e-05,
+      "loss": 0.0,
+      "step": 65820
+    },
+    {
+      "epoch": 61.4,
+      "learning_rate": 4.2324626865671645e-05,
+      "loss": 0.0001,
+      "step": 65824
+    },
+    {
+      "epoch": 61.41,
+      "learning_rate": 4.2324160447761194e-05,
+      "loss": 0.0,
+      "step": 65828
+    },
+    {
+      "epoch": 61.41,
+      "learning_rate": 4.232369402985075e-05,
+      "loss": 0.0,
+      "step": 65832
+    },
+    {
+      "epoch": 61.41,
+      "learning_rate": 4.23232276119403e-05,
+      "loss": 0.0011,
+      "step": 65836
+    },
+    {
+      "epoch": 61.42,
+      "learning_rate": 4.232276119402985e-05,
+      "loss": 0.0,
+      "step": 65840
+    },
+    {
+      "epoch": 61.42,
+      "learning_rate": 4.2322294776119406e-05,
+      "loss": 0.0,
+      "step": 65844
+    },
+    {
+      "epoch": 61.43,
+      "learning_rate": 4.232182835820896e-05,
+      "loss": 0.0,
+      "step": 65848
+    },
+    {
+      "epoch": 61.43,
+      "learning_rate": 4.232136194029851e-05,
+      "loss": 0.0,
+      "step": 65852
+    },
+    {
+      "epoch": 61.43,
+      "learning_rate": 4.232089552238806e-05,
+      "loss": 0.0001,
+      "step": 65856
+    },
+    {
+      "epoch": 61.44,
+      "learning_rate": 4.232042910447761e-05,
+      "loss": 0.0,
+      "step": 65860
+    },
+    {
+      "epoch": 61.44,
+      "learning_rate": 4.231996268656717e-05,
+      "loss": 0.0,
+      "step": 65864
+    },
+    {
+      "epoch": 61.44,
+      "learning_rate": 4.2319496268656716e-05,
+      "loss": 0.0,
+      "step": 65868
+    },
+    {
+      "epoch": 61.45,
+      "learning_rate": 4.231902985074627e-05,
+      "loss": 0.0,
+      "step": 65872
+    },
+    {
+      "epoch": 61.45,
+      "learning_rate": 4.2318563432835825e-05,
+      "loss": 0.0,
+      "step": 65876
+    },
+    {
+      "epoch": 61.46,
+      "learning_rate": 4.2318097014925373e-05,
+      "loss": 0.0,
+      "step": 65880
+    },
+    {
+      "epoch": 61.46,
+      "learning_rate": 4.231763059701493e-05,
+      "loss": 0.0001,
+      "step": 65884
+    },
+    {
+      "epoch": 61.46,
+      "learning_rate": 4.2317164179104477e-05,
+      "loss": 0.0,
+      "step": 65888
+    },
+    {
+      "epoch": 61.47,
+      "learning_rate": 4.231669776119403e-05,
+      "loss": 0.0,
+      "step": 65892
+    },
+    {
+      "epoch": 61.47,
+      "learning_rate": 4.2316231343283586e-05,
+      "loss": 0.0,
+      "step": 65896
+    },
+    {
+      "epoch": 61.47,
+      "learning_rate": 4.2315764925373134e-05,
+      "loss": 0.0,
+      "step": 65900
+    },
+    {
+      "epoch": 61.48,
+      "learning_rate": 4.231529850746269e-05,
+      "loss": 0.0,
+      "step": 65904
+    },
+    {
+      "epoch": 61.48,
+      "learning_rate": 4.2314832089552244e-05,
+      "loss": 0.0,
+      "step": 65908
+    },
+    {
+      "epoch": 61.49,
+      "learning_rate": 4.231436567164179e-05,
+      "loss": 0.0,
+      "step": 65912
+    },
+    {
+      "epoch": 61.49,
+      "learning_rate": 4.231389925373134e-05,
+      "loss": 0.0,
+      "step": 65916
+    },
+    {
+      "epoch": 61.49,
+      "learning_rate": 4.2313432835820895e-05,
+      "loss": 0.0,
+      "step": 65920
+    },
+    {
+      "epoch": 61.5,
+      "learning_rate": 4.231296641791045e-05,
+      "loss": 0.0,
+      "step": 65924
+    },
+    {
+      "epoch": 61.5,
+      "learning_rate": 4.23125e-05,
+      "loss": 0.0,
+      "step": 65928
+    },
+    {
+      "epoch": 61.5,
+      "learning_rate": 4.2312033582089553e-05,
+      "loss": 0.0,
+      "step": 65932
+    },
+    {
+      "epoch": 61.51,
+      "learning_rate": 4.231156716417911e-05,
+      "loss": 0.0089,
+      "step": 65936
+    },
+    {
+      "epoch": 61.51,
+      "learning_rate": 4.231110074626866e-05,
+      "loss": 0.0,
+      "step": 65940
+    },
+    {
+      "epoch": 61.51,
+      "learning_rate": 4.231063432835821e-05,
+      "loss": 0.0,
+      "step": 65944
+    },
+    {
+      "epoch": 61.52,
+      "learning_rate": 4.231016791044776e-05,
+      "loss": 0.0005,
+      "step": 65948
+    },
+    {
+      "epoch": 61.52,
+      "learning_rate": 4.230970149253732e-05,
+      "loss": 0.0,
+      "step": 65952
+    },
+    {
+      "epoch": 61.53,
+      "learning_rate": 4.230923507462687e-05,
+      "loss": 0.0,
+      "step": 65956
+    },
+    {
+      "epoch": 61.53,
+      "learning_rate": 4.230876865671642e-05,
+      "loss": 0.0,
+      "step": 65960
+    },
+    {
+      "epoch": 61.53,
+      "learning_rate": 4.230830223880597e-05,
+      "loss": 0.0,
+      "step": 65964
+    },
+    {
+      "epoch": 61.54,
+      "learning_rate": 4.230783582089553e-05,
+      "loss": 0.0,
+      "step": 65968
+    },
+    {
+      "epoch": 61.54,
+      "learning_rate": 4.2307369402985075e-05,
+      "loss": 0.0,
+      "step": 65972
+    },
+    {
+      "epoch": 61.54,
+      "learning_rate": 4.230690298507463e-05,
+      "loss": 0.0,
+      "step": 65976
+    },
+    {
+      "epoch": 61.55,
+      "learning_rate": 4.230643656716418e-05,
+      "loss": 0.0,
+      "step": 65980
+    },
+    {
+      "epoch": 61.55,
+      "learning_rate": 4.230597014925373e-05,
+      "loss": 0.001,
+      "step": 65984
+    },
+    {
+      "epoch": 61.56,
+      "learning_rate": 4.230550373134329e-05,
+      "loss": 0.0,
+      "step": 65988
+    },
+    {
+      "epoch": 61.56,
+      "learning_rate": 4.2305037313432836e-05,
+      "loss": 0.0,
+      "step": 65992
+    },
+    {
+      "epoch": 61.56,
+      "learning_rate": 4.230457089552239e-05,
+      "loss": 0.0001,
+      "step": 65996
+    },
+    {
+      "epoch": 61.57,
+      "learning_rate": 4.2304104477611946e-05,
+      "loss": 0.0,
+      "step": 66000
+    },
+    {
+      "epoch": 61.57,
+      "eval_exact_match": 0.7446808510638298,
+      "eval_exec": 0.7775628626692457,
+      "eval_loss": 0.4943001866340637,
+      "eval_runtime": 1085.5598,
+      "eval_samples_per_second": 0.953,
+      "step": 66000
+    },
+    {
+      "epoch": 61.57,
+      "learning_rate": 4.2303638059701494e-05,
+      "loss": 0.0,
+      "step": 66004
+    },
+    {
+      "epoch": 61.57,
+      "learning_rate": 4.230317164179104e-05,
+      "loss": 0.0,
+      "step": 66008
+    },
+    {
+      "epoch": 61.58,
+      "learning_rate": 4.2302705223880604e-05,
+      "loss": 0.0,
+      "step": 66012
+    },
+    {
+      "epoch": 61.58,
+      "learning_rate": 4.230223880597015e-05,
+      "loss": 0.0,
+      "step": 66016
+    },
+    {
+      "epoch": 61.59,
+      "learning_rate": 4.23017723880597e-05,
+      "loss": 0.002,
+      "step": 66020
+    },
+    {
+      "epoch": 61.59,
+      "learning_rate": 4.2301305970149255e-05,
+      "loss": 0.0001,
+      "step": 66024
+    },
+    {
+      "epoch": 61.59,
+      "learning_rate": 4.230083955223881e-05,
+      "loss": 0.0,
+      "step": 66028
+    },
+    {
+      "epoch": 61.6,
+      "learning_rate": 4.230037313432836e-05,
+      "loss": 0.0,
+      "step": 66032
+    },
+    {
+      "epoch": 61.6,
+      "learning_rate": 4.229990671641791e-05,
+      "loss": 0.0005,
+      "step": 66036
+    },
+    {
+      "epoch": 61.6,
+      "learning_rate": 4.229944029850746e-05,
+      "loss": 0.0,
+      "step": 66040
+    },
+    {
+      "epoch": 61.61,
+      "learning_rate": 4.2298973880597016e-05,
+      "loss": 0.0,
+      "step": 66044
+    },
+    {
+      "epoch": 61.61,
+      "learning_rate": 4.229850746268657e-05,
+      "loss": 0.0,
+      "step": 66048
+    },
+    {
+      "epoch": 61.62,
+      "learning_rate": 4.229804104477612e-05,
+      "loss": 0.0,
+      "step": 66052
+    },
+    {
+      "epoch": 61.62,
+      "learning_rate": 4.2297574626865674e-05,
+      "loss": 0.0,
+      "step": 66056
+    },
+    {
+      "epoch": 61.62,
+      "learning_rate": 4.229710820895523e-05,
+      "loss": 0.0,
+      "step": 66060
+    },
+    {
+      "epoch": 61.63,
+      "learning_rate": 4.229664179104478e-05,
+      "loss": 0.0002,
+      "step": 66064
+    },
+    {
+      "epoch": 61.63,
+      "learning_rate": 4.2296175373134325e-05,
+      "loss": 0.0,
+      "step": 66068
+    },
+    {
+      "epoch": 61.63,
+      "learning_rate": 4.229570895522389e-05,
+      "loss": 0.0,
+      "step": 66072
+    },
+    {
+      "epoch": 61.64,
+      "learning_rate": 4.2295242537313435e-05,
+      "loss": 0.0,
+      "step": 66076
+    },
+    {
+      "epoch": 61.64,
+      "learning_rate": 4.229477611940298e-05,
+      "loss": 0.0,
+      "step": 66080
+    },
+    {
+      "epoch": 61.65,
+      "learning_rate": 4.229430970149254e-05,
+      "loss": 0.0,
+      "step": 66084
+    },
+    {
+      "epoch": 61.65,
+      "learning_rate": 4.229384328358209e-05,
+      "loss": 0.0,
+      "step": 66088
+    },
+    {
+      "epoch": 61.65,
+      "learning_rate": 4.229337686567164e-05,
+      "loss": 0.0,
+      "step": 66092
+    },
+    {
+      "epoch": 61.66,
+      "learning_rate": 4.2292910447761196e-05,
+      "loss": 0.0,
+      "step": 66096
+    },
+    {
+      "epoch": 61.66,
+      "learning_rate": 4.2292444029850744e-05,
+      "loss": 0.0,
+      "step": 66100
+    },
+    {
+      "epoch": 61.66,
+      "learning_rate": 4.2291977611940306e-05,
+      "loss": 0.005,
+      "step": 66104
+    },
+    {
+      "epoch": 61.67,
+      "learning_rate": 4.2291511194029854e-05,
+      "loss": 0.0,
+      "step": 66108
+    },
+    {
+      "epoch": 61.67,
+      "learning_rate": 4.22910447761194e-05,
+      "loss": 0.0,
+      "step": 66112
+    },
+    {
+      "epoch": 61.68,
+      "learning_rate": 4.229057835820896e-05,
+      "loss": 0.0,
+      "step": 66116
+    },
+    {
+      "epoch": 61.68,
+      "learning_rate": 4.229011194029851e-05,
+      "loss": 0.0,
+      "step": 66120
+    },
+    {
+      "epoch": 61.68,
+      "learning_rate": 4.228964552238806e-05,
+      "loss": 0.0,
+      "step": 66124
+    },
+    {
+      "epoch": 61.69,
+      "learning_rate": 4.2289179104477615e-05,
+      "loss": 0.0,
+      "step": 66128
+    },
+    {
+      "epoch": 61.69,
+      "learning_rate": 4.228871268656717e-05,
+      "loss": 0.0,
+      "step": 66132
+    },
+    {
+      "epoch": 61.69,
+      "learning_rate": 4.228824626865672e-05,
+      "loss": 0.0309,
+      "step": 66136
+    },
+    {
+      "epoch": 61.7,
+      "learning_rate": 4.228777985074627e-05,
+      "loss": 0.0,
+      "step": 66140
+    },
+    {
+      "epoch": 61.7,
+      "learning_rate": 4.228731343283582e-05,
+      "loss": 0.0,
+      "step": 66144
+    },
+    {
+      "epoch": 61.71,
+      "learning_rate": 4.2286847014925376e-05,
+      "loss": 0.0,
+      "step": 66148
+    },
+    {
+      "epoch": 61.71,
+      "learning_rate": 4.228638059701493e-05,
+      "loss": 0.0,
+      "step": 66152
+    },
+    {
+      "epoch": 61.71,
+      "learning_rate": 4.228591417910448e-05,
+      "loss": 0.0,
+      "step": 66156
+    },
+    {
+      "epoch": 61.72,
+      "learning_rate": 4.228544776119403e-05,
+      "loss": 0.0,
+      "step": 66160
+    },
+    {
+      "epoch": 61.72,
+      "learning_rate": 4.228498134328359e-05,
+      "loss": 0.0,
+      "step": 66164
+    },
+    {
+      "epoch": 61.72,
+      "learning_rate": 4.228451492537314e-05,
+      "loss": 0.0,
+      "step": 66168
+    },
+    {
+      "epoch": 61.73,
+      "learning_rate": 4.2284048507462685e-05,
+      "loss": 0.0,
+      "step": 66172
+    },
+    {
+      "epoch": 61.73,
+      "learning_rate": 4.228358208955224e-05,
+      "loss": 0.0,
+      "step": 66176
+    },
+    {
+      "epoch": 61.73,
+      "learning_rate": 4.2283115671641795e-05,
+      "loss": 0.0,
+      "step": 66180
+    },
+    {
+      "epoch": 61.74,
+      "learning_rate": 4.228264925373134e-05,
+      "loss": 0.0,
+      "step": 66184
+    },
+    {
+      "epoch": 61.74,
+      "learning_rate": 4.22821828358209e-05,
+      "loss": 0.0,
+      "step": 66188
+    },
+    {
+      "epoch": 61.75,
+      "learning_rate": 4.228171641791045e-05,
+      "loss": 0.0,
+      "step": 66192
+    },
+    {
+      "epoch": 61.75,
+      "learning_rate": 4.228125e-05,
+      "loss": 0.0001,
+      "step": 66196
+    },
+    {
+      "epoch": 61.75,
+      "learning_rate": 4.2280783582089556e-05,
+      "loss": 0.0,
+      "step": 66200
+    },
+    {
+      "epoch": 61.76,
+      "learning_rate": 4.2280317164179104e-05,
+      "loss": 0.0,
+      "step": 66204
+    },
+    {
+      "epoch": 61.76,
+      "learning_rate": 4.227985074626866e-05,
+      "loss": 0.0,
+      "step": 66208
+    },
+    {
+      "epoch": 61.76,
+      "learning_rate": 4.2279384328358214e-05,
+      "loss": 0.0,
+      "step": 66212
+    },
+    {
+      "epoch": 61.77,
+      "learning_rate": 4.227891791044776e-05,
+      "loss": 0.0,
+      "step": 66216
+    },
+    {
+      "epoch": 61.77,
+      "learning_rate": 4.227845149253731e-05,
+      "loss": 0.0,
+      "step": 66220
+    },
+    {
+      "epoch": 61.78,
+      "learning_rate": 4.227798507462687e-05,
+      "loss": 0.0,
+      "step": 66224
+    },
+    {
+      "epoch": 61.78,
+      "learning_rate": 4.227751865671642e-05,
+      "loss": 0.0,
+      "step": 66228
+    },
+    {
+      "epoch": 61.78,
+      "learning_rate": 4.227705223880597e-05,
+      "loss": 0.0,
+      "step": 66232
+    },
+    {
+      "epoch": 61.79,
+      "learning_rate": 4.227658582089552e-05,
+      "loss": 0.0,
+      "step": 66236
+    },
+    {
+      "epoch": 61.79,
+      "learning_rate": 4.227611940298508e-05,
+      "loss": 0.0,
+      "step": 66240
+    },
+    {
+      "epoch": 61.79,
+      "learning_rate": 4.2275652985074626e-05,
+      "loss": 0.0,
+      "step": 66244
+    },
+    {
+      "epoch": 61.8,
+      "learning_rate": 4.227518656716418e-05,
+      "loss": 0.0011,
+      "step": 66248
+    },
+    {
+      "epoch": 61.8,
+      "learning_rate": 4.2274720149253736e-05,
+      "loss": 0.0,
+      "step": 66252
+    },
+    {
+      "epoch": 61.81,
+      "learning_rate": 4.2274253731343284e-05,
+      "loss": 0.0,
+      "step": 66256
+    },
+    {
+      "epoch": 61.81,
+      "learning_rate": 4.227378731343284e-05,
+      "loss": 0.0001,
+      "step": 66260
+    },
+    {
+      "epoch": 61.81,
+      "learning_rate": 4.227332089552239e-05,
+      "loss": 0.0,
+      "step": 66264
+    },
+    {
+      "epoch": 61.82,
+      "learning_rate": 4.227285447761194e-05,
+      "loss": 0.0,
+      "step": 66268
+    },
+    {
+      "epoch": 61.82,
+      "learning_rate": 4.22723880597015e-05,
+      "loss": 0.0,
+      "step": 66272
+    },
+    {
+      "epoch": 61.82,
+      "learning_rate": 4.2271921641791045e-05,
+      "loss": 0.0007,
+      "step": 66276
+    },
+    {
+      "epoch": 61.83,
+      "learning_rate": 4.22714552238806e-05,
+      "loss": 0.0,
+      "step": 66280
+    },
+    {
+      "epoch": 61.83,
+      "learning_rate": 4.2270988805970155e-05,
+      "loss": 0.0111,
+      "step": 66284
+    },
+    {
+      "epoch": 61.84,
+      "learning_rate": 4.22705223880597e-05,
+      "loss": 0.0006,
+      "step": 66288
+    },
+    {
+      "epoch": 61.84,
+      "learning_rate": 4.227005597014926e-05,
+      "loss": 0.0,
+      "step": 66292
+    },
+    {
+      "epoch": 61.84,
+      "learning_rate": 4.2269589552238806e-05,
+      "loss": 0.0,
+      "step": 66296
+    },
+    {
+      "epoch": 61.85,
+      "learning_rate": 4.226912313432836e-05,
+      "loss": 0.0,
+      "step": 66300
+    },
+    {
+      "epoch": 61.85,
+      "learning_rate": 4.2268656716417916e-05,
+      "loss": 0.0,
+      "step": 66304
+    },
+    {
+      "epoch": 61.85,
+      "learning_rate": 4.2268190298507464e-05,
+      "loss": 0.0001,
+      "step": 66308
+    },
+    {
+      "epoch": 61.86,
+      "learning_rate": 4.226772388059701e-05,
+      "loss": 0.0029,
+      "step": 66312
+    },
+    {
+      "epoch": 61.86,
+      "learning_rate": 4.2267257462686574e-05,
+      "loss": 0.0003,
+      "step": 66316
+    },
+    {
+      "epoch": 61.87,
+      "learning_rate": 4.226679104477612e-05,
+      "loss": 0.0002,
+      "step": 66320
+    },
+    {
+      "epoch": 61.87,
+      "learning_rate": 4.226632462686567e-05,
+      "loss": 0.0002,
+      "step": 66324
+    },
+    {
+      "epoch": 61.87,
+      "learning_rate": 4.2265858208955225e-05,
+      "loss": 0.0001,
+      "step": 66328
+    },
+    {
+      "epoch": 61.88,
+      "learning_rate": 4.226539179104478e-05,
+      "loss": 0.0,
+      "step": 66332
+    },
+    {
+      "epoch": 61.88,
+      "learning_rate": 4.226492537313433e-05,
+      "loss": 0.0,
+      "step": 66336
+    },
+    {
+      "epoch": 61.88,
+      "learning_rate": 4.226445895522388e-05,
+      "loss": 0.0,
+      "step": 66340
+    },
+    {
+      "epoch": 61.89,
+      "learning_rate": 4.226399253731344e-05,
+      "loss": 0.0,
+      "step": 66344
+    },
+    {
+      "epoch": 61.89,
+      "learning_rate": 4.2263526119402986e-05,
+      "loss": 0.0002,
+      "step": 66348
+    },
+    {
+      "epoch": 61.9,
+      "learning_rate": 4.226305970149254e-05,
+      "loss": 0.0,
+      "step": 66352
+    },
+    {
+      "epoch": 61.9,
+      "learning_rate": 4.226259328358209e-05,
+      "loss": 0.0,
+      "step": 66356
+    },
+    {
+      "epoch": 61.9,
+      "learning_rate": 4.2262126865671644e-05,
+      "loss": 0.0009,
+      "step": 66360
+    },
+    {
+      "epoch": 61.91,
+      "learning_rate": 4.22616604477612e-05,
+      "loss": 0.0,
+      "step": 66364
+    },
+    {
+      "epoch": 61.91,
+      "learning_rate": 4.226119402985075e-05,
+      "loss": 0.0,
+      "step": 66368
+    },
+    {
+      "epoch": 61.91,
+      "learning_rate": 4.2260727611940295e-05,
+      "loss": 0.0,
+      "step": 66372
+    },
+    {
+      "epoch": 61.92,
+      "learning_rate": 4.226026119402986e-05,
+      "loss": 0.0,
+      "step": 66376
+    },
+    {
+      "epoch": 61.92,
+      "learning_rate": 4.2259794776119405e-05,
+      "loss": 0.0,
+      "step": 66380
+    },
+    {
+      "epoch": 61.93,
+      "learning_rate": 4.225932835820895e-05,
+      "loss": 0.0012,
+      "step": 66384
+    },
+    {
+      "epoch": 61.93,
+      "learning_rate": 4.225886194029851e-05,
+      "loss": 0.0,
+      "step": 66388
+    },
+    {
+      "epoch": 61.93,
+      "learning_rate": 4.225839552238806e-05,
+      "loss": 0.0,
+      "step": 66392
+    },
+    {
+      "epoch": 61.94,
+      "learning_rate": 4.225792910447761e-05,
+      "loss": 0.0,
+      "step": 66396
+    },
+    {
+      "epoch": 61.94,
+      "learning_rate": 4.2257462686567166e-05,
+      "loss": 0.0,
+      "step": 66400
+    },
+    {
+      "epoch": 61.94,
+      "learning_rate": 4.225699626865672e-05,
+      "loss": 0.0,
+      "step": 66404
+    },
+    {
+      "epoch": 61.95,
+      "learning_rate": 4.225652985074627e-05,
+      "loss": 0.0,
+      "step": 66408
+    },
+    {
+      "epoch": 61.95,
+      "learning_rate": 4.2256063432835824e-05,
+      "loss": 0.0005,
+      "step": 66412
+    },
+    {
+      "epoch": 61.96,
+      "learning_rate": 4.225559701492537e-05,
+      "loss": 0.0,
+      "step": 66416
+    },
+    {
+      "epoch": 61.96,
+      "learning_rate": 4.225513059701493e-05,
+      "loss": 0.0,
+      "step": 66420
+    },
+    {
+      "epoch": 61.96,
+      "learning_rate": 4.225466417910448e-05,
+      "loss": 0.0,
+      "step": 66424
+    },
+    {
+      "epoch": 61.97,
+      "learning_rate": 4.225419776119403e-05,
+      "loss": 0.0,
+      "step": 66428
+    },
+    {
+      "epoch": 61.97,
+      "learning_rate": 4.2253731343283585e-05,
+      "loss": 0.0,
+      "step": 66432
+    },
+    {
+      "epoch": 61.97,
+      "learning_rate": 4.225326492537314e-05,
+      "loss": 0.0,
+      "step": 66436
+    },
+    {
+      "epoch": 61.98,
+      "learning_rate": 4.225279850746269e-05,
+      "loss": 0.0,
+      "step": 66440
+    },
+    {
+      "epoch": 61.98,
+      "learning_rate": 4.225233208955224e-05,
+      "loss": 0.0,
+      "step": 66444
+    },
+    {
+      "epoch": 61.98,
+      "learning_rate": 4.225186567164179e-05,
+      "loss": 0.0,
+      "step": 66448
+    },
+    {
+      "epoch": 61.99,
+      "learning_rate": 4.2251399253731346e-05,
+      "loss": 0.0004,
+      "step": 66452
+    },
+    {
+      "epoch": 61.99,
+      "learning_rate": 4.22509328358209e-05,
+      "loss": 0.0001,
+      "step": 66456
+    },
+    {
+      "epoch": 62.0,
+      "learning_rate": 4.225046641791045e-05,
+      "loss": 0.0,
+      "step": 66460
+    },
+    {
+      "epoch": 62.0,
+      "learning_rate": 4.2250000000000004e-05,
+      "loss": 0.0,
+      "step": 66464
+    },
+    {
+      "epoch": 62.0,
+      "learning_rate": 4.224953358208956e-05,
+      "loss": 0.0,
+      "step": 66468
+    },
+    {
+      "epoch": 62.01,
+      "learning_rate": 4.224906716417911e-05,
+      "loss": 0.0,
+      "step": 66472
+    },
+    {
+      "epoch": 62.01,
+      "learning_rate": 4.2248600746268655e-05,
+      "loss": 0.0001,
+      "step": 66476
+    },
+    {
+      "epoch": 62.01,
+      "learning_rate": 4.224813432835821e-05,
+      "loss": 0.0,
+      "step": 66480
+    },
+    {
+      "epoch": 62.02,
+      "learning_rate": 4.2247667910447765e-05,
+      "loss": 0.0019,
+      "step": 66484
+    },
+    {
+      "epoch": 62.02,
+      "learning_rate": 4.224720149253731e-05,
+      "loss": 0.0,
+      "step": 66488
+    },
+    {
+      "epoch": 62.03,
+      "learning_rate": 4.224673507462687e-05,
+      "loss": 0.0,
+      "step": 66492
+    },
+    {
+      "epoch": 62.03,
+      "learning_rate": 4.224626865671642e-05,
+      "loss": 0.0,
+      "step": 66496
+    },
+    {
+      "epoch": 62.03,
+      "learning_rate": 4.224580223880597e-05,
+      "loss": 0.0,
+      "step": 66500
+    },
+    {
+      "epoch": 62.03,
+      "eval_exact_match": 0.7562862669245648,
+      "eval_exec": 0.7823984526112185,
+      "eval_loss": 0.4995794892311096,
+      "eval_runtime": 1097.5777,
+      "eval_samples_per_second": 0.942,
+      "step": 66500
+    }
+  ],
+  "max_steps": 428800,
+  "num_train_epochs": 400,
+  "total_flos": 2.2981425097872384e+18,
+  "trial_name": null,
+  "trial_params": null
+}