diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,4249 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.18047073354834284,
+  "eval_steps": 10,
+  "global_step": 2630,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 6.862005077883757e-05,
+      "grad_norm": 1.9216967821121216,
+      "learning_rate": 0.0,
+      "loss": 1.6865,
+      "step": 1
+    },
+    {
+      "epoch": 0.0006862005077883758,
+      "grad_norm": 1.8931514024734497,
+      "learning_rate": 8.234217749313815e-07,
+      "loss": 1.7993,
+      "step": 10
+    },
+    {
+      "epoch": 0.0006862005077883758,
+      "eval_accuracy": 0.5898158335459887,
+      "eval_loss": 1.8085482120513916,
+      "eval_runtime": 2190.4949,
+      "eval_samples_per_second": 13.441,
+      "eval_steps_per_second": 1.68,
+      "step": 10
+    },
+    {
+      "epoch": 0.0013724010155767516,
+      "grad_norm": 1.8920422792434692,
+      "learning_rate": 1.7383348581884722e-06,
+      "loss": 1.8431,
+      "step": 20
+    },
+    {
+      "epoch": 0.0013724010155767516,
+      "eval_accuracy": 0.5915482852347239,
+      "eval_loss": 1.7978498935699463,
+      "eval_runtime": 2191.0867,
+      "eval_samples_per_second": 13.437,
+      "eval_steps_per_second": 1.68,
+      "step": 20
+    },
+    {
+      "epoch": 0.0020586015233651275,
+      "grad_norm": 1.7306218147277832,
+      "learning_rate": 2.653247941445563e-06,
+      "loss": 1.7786,
+      "step": 30
+    },
+    {
+      "epoch": 0.0020586015233651275,
+      "eval_accuracy": 0.5931407264065678,
+      "eval_loss": 1.7860997915267944,
+      "eval_runtime": 2192.9273,
+      "eval_samples_per_second": 13.426,
+      "eval_steps_per_second": 1.679,
+      "step": 30
+    },
+    {
+      "epoch": 0.002744802031153503,
+      "grad_norm": 1.678896427154541,
+      "learning_rate": 3.5681610247026537e-06,
+      "loss": 1.7845,
+      "step": 40
+    },
+    {
+      "epoch": 0.002744802031153503,
+      "eval_accuracy": 0.5940313221156256,
+      "eval_loss": 1.777963638305664,
+      "eval_runtime": 2193.6313,
+      "eval_samples_per_second": 13.422,
+      "eval_steps_per_second": 1.678,
+      "step": 40
+    },
+    {
+      "epoch": 0.003431002538941879,
+      "grad_norm": 1.608749270439148,
+      "learning_rate": 4.483074107959744e-06,
+      "loss": 1.7561,
+      "step": 50
+    },
+    {
+      "epoch": 0.003431002538941879,
+      "eval_accuracy": 0.5950869622352757,
+      "eval_loss": 1.7716691493988037,
+      "eval_runtime": 2195.0575,
+      "eval_samples_per_second": 13.413,
+      "eval_steps_per_second": 1.677,
+      "step": 50
+    },
+    {
+      "epoch": 0.004117203046730255,
+      "grad_norm": 1.5839513540267944,
+      "learning_rate": 5.397987191216835e-06,
+      "loss": 1.7809,
+      "step": 60
+    },
+    {
+      "epoch": 0.004117203046730255,
+      "eval_accuracy": 0.5955689808940925,
+      "eval_loss": 1.767722725868225,
+      "eval_runtime": 2194.6298,
+      "eval_samples_per_second": 13.415,
+      "eval_steps_per_second": 1.677,
+      "step": 60
+    },
+    {
+      "epoch": 0.00480340355451863,
+      "grad_norm": 1.6457411050796509,
+      "learning_rate": 6.312900274473925e-06,
+      "loss": 1.7554,
+      "step": 70
+    },
+    {
+      "epoch": 0.00480340355451863,
+      "eval_accuracy": 0.5958804437015288,
+      "eval_loss": 1.765299916267395,
+      "eval_runtime": 2194.3885,
+      "eval_samples_per_second": 13.417,
+      "eval_steps_per_second": 1.677,
+      "step": 70
+    },
+    {
+      "epoch": 0.005489604062307006,
+      "grad_norm": 1.5799423456192017,
+      "learning_rate": 7.227813357731015e-06,
+      "loss": 1.7361,
+      "step": 80
+    },
+    {
+      "epoch": 0.005489604062307006,
+      "eval_accuracy": 0.5960319751217547,
+      "eval_loss": 1.7639434337615967,
+      "eval_runtime": 2195.254,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 80
+    },
+    {
+      "epoch": 0.0061758045700953815,
+      "grad_norm": 1.7162026166915894,
+      "learning_rate": 8.142726440988106e-06,
+      "loss": 1.7427,
+      "step": 90
+    },
+    {
+      "epoch": 0.0061758045700953815,
+      "eval_accuracy": 0.5961102309406575,
+      "eval_loss": 1.7636356353759766,
+      "eval_runtime": 2195.6998,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.676,
+      "step": 90
+    },
+    {
+      "epoch": 0.006862005077883758,
+      "grad_norm": 1.703099012374878,
+      "learning_rate": 9.057639524245198e-06,
+      "loss": 1.7479,
+      "step": 100
+    },
+    {
+      "epoch": 0.006862005077883758,
+      "eval_accuracy": 0.5962347363801508,
+      "eval_loss": 1.7628475427627563,
+      "eval_runtime": 2195.4873,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 100
+    },
+    {
+      "epoch": 0.007548205585672134,
+      "grad_norm": 1.5629897117614746,
+      "learning_rate": 9.972552607502288e-06,
+      "loss": 1.7511,
+      "step": 110
+    },
+    {
+      "epoch": 0.007548205585672134,
+      "eval_accuracy": 0.5963783990566007,
+      "eval_loss": 1.761244535446167,
+      "eval_runtime": 2194.8861,
+      "eval_samples_per_second": 13.414,
+      "eval_steps_per_second": 1.677,
+      "step": 110
+    },
+    {
+      "epoch": 0.00823440609346051,
+      "grad_norm": 1.8800327777862549,
+      "learning_rate": 1.0887465690759378e-05,
+      "loss": 1.7918,
+      "step": 120
+    },
+    {
+      "epoch": 0.00823440609346051,
+      "eval_accuracy": 0.595912250691138,
+      "eval_loss": 1.7631102800369263,
+      "eval_runtime": 2195.8634,
+      "eval_samples_per_second": 13.408,
+      "eval_steps_per_second": 1.676,
+      "step": 120
+    },
+    {
+      "epoch": 0.008920606601248885,
+      "grad_norm": 1.6483738422393799,
+      "learning_rate": 1.180237877401647e-05,
+      "loss": 1.7451,
+      "step": 130
+    },
+    {
+      "epoch": 0.008920606601248885,
+      "eval_accuracy": 0.5959484734736679,
+      "eval_loss": 1.7631044387817383,
+      "eval_runtime": 2194.7206,
+      "eval_samples_per_second": 13.415,
+      "eval_steps_per_second": 1.677,
+      "step": 130
+    },
+    {
+      "epoch": 0.00960680710903726,
+      "grad_norm": 1.5531351566314697,
+      "learning_rate": 1.2717291857273558e-05,
+      "loss": 1.8069,
+      "step": 140
+    },
+    {
+      "epoch": 0.00960680710903726,
+      "eval_accuracy": 0.595846910236492,
+      "eval_loss": 1.7640652656555176,
+      "eval_runtime": 2194.5306,
+      "eval_samples_per_second": 13.416,
+      "eval_steps_per_second": 1.677,
+      "step": 140
+    },
+    {
+      "epoch": 0.010293007616825637,
+      "grad_norm": 1.653224229812622,
+      "learning_rate": 1.3632204940530649e-05,
+      "loss": 1.774,
+      "step": 150
+    },
+    {
+      "epoch": 0.010293007616825637,
+      "eval_accuracy": 0.5955285083258945,
+      "eval_loss": 1.76512610912323,
+      "eval_runtime": 2194.8618,
+      "eval_samples_per_second": 13.414,
+      "eval_steps_per_second": 1.677,
+      "step": 150
+    },
+    {
+      "epoch": 0.010979208124614013,
+      "grad_norm": 1.4969476461410522,
+      "learning_rate": 1.454711802378774e-05,
+      "loss": 1.7293,
+      "step": 160
+    },
+    {
+      "epoch": 0.010979208124614013,
+      "eval_accuracy": 0.5952186723895343,
+      "eval_loss": 1.7661479711532593,
+      "eval_runtime": 2195.9262,
+      "eval_samples_per_second": 13.408,
+      "eval_steps_per_second": 1.676,
+      "step": 160
+    },
+    {
+      "epoch": 0.011665408632402388,
+      "grad_norm": 1.5510554313659668,
+      "learning_rate": 1.546203110704483e-05,
+      "loss": 1.8144,
+      "step": 170
+    },
+    {
+      "epoch": 0.011665408632402388,
+      "eval_accuracy": 0.5952723923364943,
+      "eval_loss": 1.7658929824829102,
+      "eval_runtime": 2195.7289,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.676,
+      "step": 170
+    },
+    {
+      "epoch": 0.012351609140190763,
+      "grad_norm": 1.5824214220046997,
+      "learning_rate": 1.6376944190301923e-05,
+      "loss": 1.762,
+      "step": 180
+    },
+    {
+      "epoch": 0.012351609140190763,
+      "eval_accuracy": 0.5950958270225676,
+      "eval_loss": 1.7668511867523193,
+      "eval_runtime": 2196.0319,
+      "eval_samples_per_second": 13.407,
+      "eval_steps_per_second": 1.676,
+      "step": 180
+    },
+    {
+      "epoch": 0.01303780964797914,
+      "grad_norm": 1.536055088043213,
+      "learning_rate": 1.7291857273559013e-05,
+      "loss": 1.7647,
+      "step": 190
+    },
+    {
+      "epoch": 0.01303780964797914,
+      "eval_accuracy": 0.5946427600286144,
+      "eval_loss": 1.7685387134552002,
+      "eval_runtime": 2194.9602,
+      "eval_samples_per_second": 13.413,
+      "eval_steps_per_second": 1.677,
+      "step": 190
+    },
+    {
+      "epoch": 0.013724010155767515,
+      "grad_norm": 1.5057470798492432,
+      "learning_rate": 1.8206770356816103e-05,
+      "loss": 1.7745,
+      "step": 200
+    },
+    {
+      "epoch": 0.013724010155767515,
+      "eval_accuracy": 0.5946046447634041,
+      "eval_loss": 1.7696959972381592,
+      "eval_runtime": 2194.9716,
+      "eval_samples_per_second": 13.413,
+      "eval_steps_per_second": 1.677,
+      "step": 200
+    },
+    {
+      "epoch": 0.01441021066355589,
+      "grad_norm": 1.3596829175949097,
+      "learning_rate": 1.9121683440073193e-05,
+      "loss": 1.7936,
+      "step": 210
+    },
+    {
+      "epoch": 0.01441021066355589,
+      "eval_accuracy": 0.5943018475345566,
+      "eval_loss": 1.7707691192626953,
+      "eval_runtime": 2195.7035,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.676,
+      "step": 210
+    },
+    {
+      "epoch": 0.015096411171344267,
+      "grad_norm": 1.5187965631484985,
+      "learning_rate": 2.0036596523330283e-05,
+      "loss": 1.7726,
+      "step": 220
+    },
+    {
+      "epoch": 0.015096411171344267,
+      "eval_accuracy": 0.5939556892113148,
+      "eval_loss": 1.7723795175552368,
+      "eval_runtime": 2196.0461,
+      "eval_samples_per_second": 13.407,
+      "eval_steps_per_second": 1.676,
+      "step": 220
+    },
+    {
+      "epoch": 0.01578261167913264,
+      "grad_norm": 1.4719396829605103,
+      "learning_rate": 2.0951509606587374e-05,
+      "loss": 1.7646,
+      "step": 230
+    },
+    {
+      "epoch": 0.01578261167913264,
+      "eval_accuracy": 0.5935164340207827,
+      "eval_loss": 1.7753576040267944,
+      "eval_runtime": 2196.1999,
+      "eval_samples_per_second": 13.406,
+      "eval_steps_per_second": 1.676,
+      "step": 230
+    },
+    {
+      "epoch": 0.01646881218692102,
+      "grad_norm": 1.4845266342163086,
+      "learning_rate": 2.1866422689844464e-05,
+      "loss": 1.7424,
+      "step": 240
+    },
+    {
+      "epoch": 0.01646881218692102,
+      "eval_accuracy": 0.5930875044813658,
+      "eval_loss": 1.777539849281311,
+      "eval_runtime": 2196.2424,
+      "eval_samples_per_second": 13.406,
+      "eval_steps_per_second": 1.676,
+      "step": 240
+    },
+    {
+      "epoch": 0.017155012694709395,
+      "grad_norm": 1.3525185585021973,
+      "learning_rate": 2.2781335773101558e-05,
+      "loss": 1.7129,
+      "step": 250
+    },
+    {
+      "epoch": 0.017155012694709395,
+      "eval_accuracy": 0.5927156814368632,
+      "eval_loss": 1.7804397344589233,
+      "eval_runtime": 2195.161,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 250
+    },
+    {
+      "epoch": 0.01784121320249777,
+      "grad_norm": 1.3116035461425781,
+      "learning_rate": 2.3696248856358648e-05,
+      "loss": 1.7057,
+      "step": 260
+    },
+    {
+      "epoch": 0.01784121320249777,
+      "eval_accuracy": 0.5927907831179655,
+      "eval_loss": 1.77951180934906,
+      "eval_runtime": 2195.0811,
+      "eval_samples_per_second": 13.413,
+      "eval_steps_per_second": 1.677,
+      "step": 260
+    },
+    {
+      "epoch": 0.018527413710286145,
+      "grad_norm": 1.427150011062622,
+      "learning_rate": 2.4611161939615738e-05,
+      "loss": 1.7832,
+      "step": 270
+    },
+    {
+      "epoch": 0.018527413710286145,
+      "eval_accuracy": 0.5925692298385685,
+      "eval_loss": 1.78168523311615,
+      "eval_runtime": 2195.7308,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.676,
+      "step": 270
+    },
+    {
+      "epoch": 0.01921361421807452,
+      "grad_norm": 1.4132215976715088,
+      "learning_rate": 2.552607502287283e-05,
+      "loss": 1.7742,
+      "step": 280
+    },
+    {
+      "epoch": 0.01921361421807452,
+      "eval_accuracy": 0.5916866688805393,
+      "eval_loss": 1.7839642763137817,
+      "eval_runtime": 2194.9088,
+      "eval_samples_per_second": 13.414,
+      "eval_steps_per_second": 1.677,
+      "step": 280
+    },
+    {
+      "epoch": 0.019899814725862896,
+      "grad_norm": 1.3926459550857544,
+      "learning_rate": 2.6440988106129922e-05,
+      "loss": 1.7713,
+      "step": 290
+    },
+    {
+      "epoch": 0.019899814725862896,
+      "eval_accuracy": 0.592083725027446,
+      "eval_loss": 1.784454345703125,
+      "eval_runtime": 2195.1648,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 290
+    },
+    {
+      "epoch": 0.020586015233651275,
+      "grad_norm": 1.2985360622406006,
+      "learning_rate": 2.7355901189387012e-05,
+      "loss": 1.8099,
+      "step": 300
+    },
+    {
+      "epoch": 0.020586015233651275,
+      "eval_accuracy": 0.5915102363724148,
+      "eval_loss": 1.7870558500289917,
+      "eval_runtime": 2196.189,
+      "eval_samples_per_second": 13.406,
+      "eval_steps_per_second": 1.676,
+      "step": 300
+    },
+    {
+      "epoch": 0.02127221574143965,
+      "grad_norm": 1.3202718496322632,
+      "learning_rate": 2.8270814272644102e-05,
+      "loss": 1.8158,
+      "step": 310
+    },
+    {
+      "epoch": 0.02127221574143965,
+      "eval_accuracy": 0.5911451532223702,
+      "eval_loss": 1.789017677307129,
+      "eval_runtime": 2195.7623,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.676,
+      "step": 310
+    },
+    {
+      "epoch": 0.021958416249228025,
+      "grad_norm": 1.3661465644836426,
+      "learning_rate": 2.9185727355901192e-05,
+      "loss": 1.7867,
+      "step": 320
+    },
+    {
+      "epoch": 0.021958416249228025,
+      "eval_accuracy": 0.5906047664135189,
+      "eval_loss": 1.7920600175857544,
+      "eval_runtime": 2196.3235,
+      "eval_samples_per_second": 13.405,
+      "eval_steps_per_second": 1.676,
+      "step": 320
+    },
+    {
+      "epoch": 0.0226446167570164,
+      "grad_norm": 1.2746013402938843,
+      "learning_rate": 3.0100640439158283e-05,
+      "loss": 1.7356,
+      "step": 330
+    },
+    {
+      "epoch": 0.0226446167570164,
+      "eval_accuracy": 0.5906917210124609,
+      "eval_loss": 1.7915929555892944,
+      "eval_runtime": 2196.2755,
+      "eval_samples_per_second": 13.405,
+      "eval_steps_per_second": 1.676,
+      "step": 330
+    },
+    {
+      "epoch": 0.023330817264804776,
+      "grad_norm": 1.3848775625228882,
+      "learning_rate": 3.101555352241537e-05,
+      "loss": 1.7845,
+      "step": 340
+    },
+    {
+      "epoch": 0.023330817264804776,
+      "eval_accuracy": 0.5902880577768986,
+      "eval_loss": 1.7954319715499878,
+      "eval_runtime": 2195.6581,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.676,
+      "step": 340
+    },
+    {
+      "epoch": 0.02401701777259315,
+      "grad_norm": 1.3434277772903442,
+      "learning_rate": 3.193046660567246e-05,
+      "loss": 1.8831,
+      "step": 350
+    },
+    {
+      "epoch": 0.02401701777259315,
+      "eval_accuracy": 0.5896593219081829,
+      "eval_loss": 1.7972683906555176,
+      "eval_runtime": 2196.3435,
+      "eval_samples_per_second": 13.405,
+      "eval_steps_per_second": 1.676,
+      "step": 350
+    },
+    {
+      "epoch": 0.024703218280381526,
+      "grad_norm": 1.266026258468628,
+      "learning_rate": 3.284537968892955e-05,
+      "loss": 1.7802,
+      "step": 360
+    },
+    {
+      "epoch": 0.024703218280381526,
+      "eval_accuracy": 0.5896452112917071,
+      "eval_loss": 1.7977614402770996,
+      "eval_runtime": 2194.9145,
+      "eval_samples_per_second": 13.414,
+      "eval_steps_per_second": 1.677,
+      "step": 360
+    },
+    {
+      "epoch": 0.025389418788169905,
+      "grad_norm": 1.2292088270187378,
+      "learning_rate": 3.3760292772186643e-05,
+      "loss": 1.8164,
+      "step": 370
+    },
+    {
+      "epoch": 0.025389418788169905,
+      "eval_accuracy": 0.5890495108662703,
+      "eval_loss": 1.8011078834533691,
+      "eval_runtime": 2195.3229,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 370
+    },
+    {
+      "epoch": 0.02607561929595828,
+      "grad_norm": 1.2202467918395996,
+      "learning_rate": 3.467520585544373e-05,
+      "loss": 1.8117,
+      "step": 380
+    },
+    {
+      "epoch": 0.02607561929595828,
+      "eval_accuracy": 0.5890773336818158,
+      "eval_loss": 1.8025044202804565,
+      "eval_runtime": 2195.1721,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 380
+    },
+    {
+      "epoch": 0.026761819803746655,
+      "grad_norm": 1.2751961946487427,
+      "learning_rate": 3.5590118938700824e-05,
+      "loss": 1.8681,
+      "step": 390
+    },
+    {
+      "epoch": 0.026761819803746655,
+      "eval_accuracy": 0.5889283255718303,
+      "eval_loss": 1.8029353618621826,
+      "eval_runtime": 2194.9883,
+      "eval_samples_per_second": 13.413,
+      "eval_steps_per_second": 1.677,
+      "step": 390
+    },
+    {
+      "epoch": 0.02744802031153503,
+      "grad_norm": 1.1804041862487793,
+      "learning_rate": 3.650503202195791e-05,
+      "loss": 1.7925,
+      "step": 400
+    },
+    {
+      "epoch": 0.02744802031153503,
+      "eval_accuracy": 0.5888150754240672,
+      "eval_loss": 1.8044122457504272,
+      "eval_runtime": 2195.7663,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.676,
+      "step": 400
+    },
+    {
+      "epoch": 0.028134220819323406,
+      "grad_norm": 1.2452822923660278,
+      "learning_rate": 3.7419945105215004e-05,
+      "loss": 1.8606,
+      "step": 410
+    },
+    {
+      "epoch": 0.028134220819323406,
+      "eval_accuracy": 0.5877631206654261,
+      "eval_loss": 1.8087084293365479,
+      "eval_runtime": 2195.4238,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 410
+    },
+    {
+      "epoch": 0.02882042132711178,
+      "grad_norm": 1.3155659437179565,
+      "learning_rate": 3.83348581884721e-05,
+      "loss": 1.7495,
+      "step": 420
+    },
+    {
+      "epoch": 0.02882042132711178,
+      "eval_accuracy": 0.5877042876950842,
+      "eval_loss": 1.8115092515945435,
+      "eval_runtime": 2195.0116,
+      "eval_samples_per_second": 13.413,
+      "eval_steps_per_second": 1.677,
+      "step": 420
+    },
+    {
+      "epoch": 0.029506621834900156,
+      "grad_norm": 1.2945083379745483,
+      "learning_rate": 3.9249771271729185e-05,
+      "loss": 1.8207,
+      "step": 430
+    },
+    {
+      "epoch": 0.029506621834900156,
+      "eval_accuracy": 0.5871699103487793,
+      "eval_loss": 1.812555193901062,
+      "eval_runtime": 2196.4551,
+      "eval_samples_per_second": 13.404,
+      "eval_steps_per_second": 1.676,
+      "step": 430
+    },
+    {
+      "epoch": 0.030192822342688535,
+      "grad_norm": 1.2076581716537476,
+      "learning_rate": 4.016468435498628e-05,
+      "loss": 1.8009,
+      "step": 440
+    },
+    {
+      "epoch": 0.030192822342688535,
+      "eval_accuracy": 0.5872769186238423,
+      "eval_loss": 1.8119513988494873,
+      "eval_runtime": 2195.9383,
+      "eval_samples_per_second": 13.407,
+      "eval_steps_per_second": 1.676,
+      "step": 440
+    },
+    {
+      "epoch": 0.03087902285047691,
+      "grad_norm": 1.2269212007522583,
+      "learning_rate": 4.1079597438243365e-05,
+      "loss": 1.8264,
+      "step": 450
+    },
+    {
+      "epoch": 0.03087902285047691,
+      "eval_accuracy": 0.5867487167473362,
+      "eval_loss": 1.815908670425415,
+      "eval_runtime": 2195.2403,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 450
+    },
+    {
+      "epoch": 0.03156522335826528,
+      "grad_norm": 1.2208211421966553,
+      "learning_rate": 4.199451052150046e-05,
+      "loss": 1.8048,
+      "step": 460
+    },
+    {
+      "epoch": 0.03156522335826528,
+      "eval_accuracy": 0.5865989450039885,
+      "eval_loss": 1.8151869773864746,
+      "eval_runtime": 2196.2404,
+      "eval_samples_per_second": 13.406,
+      "eval_steps_per_second": 1.676,
+      "step": 460
+    },
+    {
+      "epoch": 0.03225142386605366,
+      "grad_norm": 1.1229742765426636,
+      "learning_rate": 4.2909423604757546e-05,
+      "loss": 1.8115,
+      "step": 470
+    },
+    {
+      "epoch": 0.03225142386605366,
+      "eval_accuracy": 0.5854921082476188,
+      "eval_loss": 1.8196277618408203,
+      "eval_runtime": 2196.4836,
+      "eval_samples_per_second": 13.404,
+      "eval_steps_per_second": 1.676,
+      "step": 470
+    },
+    {
+      "epoch": 0.03293762437384204,
+      "grad_norm": 1.1489847898483276,
+      "learning_rate": 4.382433668801464e-05,
+      "loss": 1.8195,
+      "step": 480
+    },
+    {
+      "epoch": 0.03293762437384204,
+      "eval_accuracy": 0.5857865387109324,
+      "eval_loss": 1.8216103315353394,
+      "eval_runtime": 2195.3397,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 480
+    },
+    {
+      "epoch": 0.03362382488163041,
+      "grad_norm": 1.2842459678649902,
+      "learning_rate": 4.473924977127173e-05,
+      "loss": 1.7904,
+      "step": 490
+    },
+    {
+      "epoch": 0.03362382488163041,
+      "eval_accuracy": 0.5854642190291723,
+      "eval_loss": 1.823735237121582,
+      "eval_runtime": 2195.5222,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 490
+    },
+    {
+      "epoch": 0.03431002538941879,
+      "grad_norm": 1.1588455438613892,
+      "learning_rate": 4.565416285452882e-05,
+      "loss": 1.8076,
+      "step": 500
+    },
+    {
+      "epoch": 0.03431002538941879,
+      "eval_accuracy": 0.584858989787433,
+      "eval_loss": 1.82528555393219,
+      "eval_runtime": 2196.6782,
+      "eval_samples_per_second": 13.403,
+      "eval_steps_per_second": 1.676,
+      "step": 500
+    },
+    {
+      "epoch": 0.03499622589720716,
+      "grad_norm": 1.0610827207565308,
+      "learning_rate": 4.656907593778591e-05,
+      "loss": 1.7858,
+      "step": 510
+    },
+    {
+      "epoch": 0.03499622589720716,
+      "eval_accuracy": 0.5846732608731596,
+      "eval_loss": 1.8272473812103271,
+      "eval_runtime": 2195.3692,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 510
+    },
+    {
+      "epoch": 0.03568242640499554,
+      "grad_norm": 1.1871790885925293,
+      "learning_rate": 4.7483989021043e-05,
+      "loss": 1.8204,
+      "step": 520
+    },
+    {
+      "epoch": 0.03568242640499554,
+      "eval_accuracy": 0.5842074113207517,
+      "eval_loss": 1.829712986946106,
+      "eval_runtime": 2195.2548,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 520
+    },
+    {
+      "epoch": 0.03636862691278391,
+      "grad_norm": 1.1551142930984497,
+      "learning_rate": 4.8398902104300094e-05,
+      "loss": 1.8692,
+      "step": 530
+    },
+    {
+      "epoch": 0.03636862691278391,
+      "eval_accuracy": 0.5840751699432846,
+      "eval_loss": 1.8313241004943848,
+      "eval_runtime": 2196.5372,
+      "eval_samples_per_second": 13.404,
+      "eval_steps_per_second": 1.676,
+      "step": 530
+    },
+    {
+      "epoch": 0.03705482742057229,
+      "grad_norm": 1.1456727981567383,
+      "learning_rate": 4.931381518755718e-05,
+      "loss": 1.8427,
+      "step": 540
+    },
+    {
+      "epoch": 0.03705482742057229,
+      "eval_accuracy": 0.5839454850775084,
+      "eval_loss": 1.8317539691925049,
+      "eval_runtime": 2197.0175,
+      "eval_samples_per_second": 13.401,
+      "eval_steps_per_second": 1.675,
+      "step": 540
+    },
+    {
+      "epoch": 0.03774102792836067,
+      "grad_norm": 1.045682668685913,
+      "learning_rate": 5.0228728270814274e-05,
+      "loss": 1.8425,
+      "step": 550
+    },
+    {
+      "epoch": 0.03774102792836067,
+      "eval_accuracy": 0.5831079120849495,
+      "eval_loss": 1.83717679977417,
+      "eval_runtime": 2195.5341,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 550
+    },
+    {
+      "epoch": 0.03842722843614904,
+      "grad_norm": 1.0730020999908447,
+      "learning_rate": 5.114364135407137e-05,
+      "loss": 1.7764,
+      "step": 560
+    },
+    {
+      "epoch": 0.03842722843614904,
+      "eval_accuracy": 0.5835356463721473,
+      "eval_loss": 1.835798978805542,
+      "eval_runtime": 2195.2526,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 560
+    },
+    {
+      "epoch": 0.03911342894393742,
+      "grad_norm": 1.0931340456008911,
+      "learning_rate": 5.2058554437328455e-05,
+      "loss": 1.8409,
+      "step": 570
+    },
+    {
+      "epoch": 0.03911342894393742,
+      "eval_accuracy": 0.5838007599546415,
+      "eval_loss": 1.8357644081115723,
+      "eval_runtime": 2196.814,
+      "eval_samples_per_second": 13.402,
+      "eval_steps_per_second": 1.676,
+      "step": 570
+    },
+    {
+      "epoch": 0.03979962945172579,
+      "grad_norm": 1.0866916179656982,
+      "learning_rate": 5.297346752058555e-05,
+      "loss": 1.8691,
+      "step": 580
+    },
+    {
+      "epoch": 0.03979962945172579,
+      "eval_accuracy": 0.5829323428145388,
+      "eval_loss": 1.8371539115905762,
+      "eval_runtime": 2195.2154,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 580
+    },
+    {
+      "epoch": 0.04048582995951417,
+      "grad_norm": 1.0397108793258667,
+      "learning_rate": 5.3888380603842635e-05,
+      "loss": 1.8297,
+      "step": 590
+    },
+    {
+      "epoch": 0.04048582995951417,
+      "eval_accuracy": 0.582625661015979,
+      "eval_loss": 1.8405317068099976,
+      "eval_runtime": 2196.6833,
+      "eval_samples_per_second": 13.403,
+      "eval_steps_per_second": 1.676,
+      "step": 590
+    },
+    {
+      "epoch": 0.04117203046730255,
+      "grad_norm": 1.045896291732788,
+      "learning_rate": 5.4803293687099735e-05,
+      "loss": 1.846,
+      "step": 600
+    },
+    {
+      "epoch": 0.04117203046730255,
+      "eval_accuracy": 0.582842034869093,
+      "eval_loss": 1.83939790725708,
+      "eval_runtime": 2195.4627,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 600
+    },
+    {
+      "epoch": 0.04185823097509092,
+      "grad_norm": 1.009037733078003,
+      "learning_rate": 5.5718206770356815e-05,
+      "loss": 1.8152,
+      "step": 610
+    },
+    {
+      "epoch": 0.04185823097509092,
+      "eval_accuracy": 0.5820976251467255,
+      "eval_loss": 1.8433318138122559,
+      "eval_runtime": 2195.2269,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 610
+    },
+    {
+      "epoch": 0.0425444314828793,
+      "grad_norm": 1.0480693578720093,
+      "learning_rate": 5.6633119853613916e-05,
+      "loss": 1.8718,
+      "step": 620
+    },
+    {
+      "epoch": 0.0425444314828793,
+      "eval_accuracy": 0.5822570585121779,
+      "eval_loss": 1.8436963558197021,
+      "eval_runtime": 2195.2658,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 620
+    },
+    {
+      "epoch": 0.04323063199066767,
+      "grad_norm": 1.1000436544418335,
+      "learning_rate": 5.7548032936870996e-05,
+      "loss": 1.8646,
+      "step": 630
+    },
+    {
+      "epoch": 0.04323063199066767,
+      "eval_accuracy": 0.5819806564365029,
+      "eval_loss": 1.8459354639053345,
+      "eval_runtime": 2195.2676,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 630
+    },
+    {
+      "epoch": 0.04391683249845605,
+      "grad_norm": 1.154166579246521,
+      "learning_rate": 5.8462946020128096e-05,
+      "loss": 1.8593,
+      "step": 640
+    },
+    {
+      "epoch": 0.04391683249845605,
+      "eval_accuracy": 0.582069735928279,
+      "eval_loss": 1.847158670425415,
+      "eval_runtime": 2196.3163,
+      "eval_samples_per_second": 13.405,
+      "eval_steps_per_second": 1.676,
+      "step": 640
+    },
+    {
+      "epoch": 0.04460303300624442,
+      "grad_norm": 1.0694483518600464,
+      "learning_rate": 5.937785910338518e-05,
+      "loss": 1.8511,
+      "step": 650
+    },
+    {
+      "epoch": 0.04460303300624442,
+      "eval_accuracy": 0.5816468822543094,
+      "eval_loss": 1.8503717184066772,
+      "eval_runtime": 2195.3338,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 650
+    },
+    {
+      "epoch": 0.0452892335140328,
+      "grad_norm": 1.0069674253463745,
+      "learning_rate": 6.029277218664228e-05,
+      "loss": 1.8305,
+      "step": 660
+    },
+    {
+      "epoch": 0.0452892335140328,
+      "eval_accuracy": 0.5812435178318018,
+      "eval_loss": 1.851124882698059,
+      "eval_runtime": 2196.3471,
+      "eval_samples_per_second": 13.405,
+      "eval_steps_per_second": 1.676,
+      "step": 660
+    },
+    {
+      "epoch": 0.04597543402182118,
+      "grad_norm": 0.93955397605896,
+      "learning_rate": 6.120768526989936e-05,
+      "loss": 1.8284,
+      "step": 670
+    },
+    {
+      "epoch": 0.04597543402182118,
+      "eval_accuracy": 0.5811864777397887,
+      "eval_loss": 1.8510762453079224,
+      "eval_runtime": 2196.11,
+      "eval_samples_per_second": 13.406,
+      "eval_steps_per_second": 1.676,
+      "step": 670
+    },
+    {
+      "epoch": 0.04666163452960955,
+      "grad_norm": 0.9924960732460022,
+      "learning_rate": 6.212259835315646e-05,
+      "loss": 1.8664,
+      "step": 680
+    },
+    {
+      "epoch": 0.04666163452960955,
+      "eval_accuracy": 0.5802622489613424,
+      "eval_loss": 1.854859709739685,
+      "eval_runtime": 2196.6759,
+      "eval_samples_per_second": 13.403,
+      "eval_steps_per_second": 1.676,
+      "step": 680
+    },
+    {
+      "epoch": 0.04734783503739793,
+      "grad_norm": 1.0502827167510986,
+      "learning_rate": 6.303751143641354e-05,
+      "loss": 1.8518,
+      "step": 690
+    },
+    {
+      "epoch": 0.04734783503739793,
+      "eval_accuracy": 0.5797041657793579,
+      "eval_loss": 1.8587929010391235,
+      "eval_runtime": 2195.6371,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.677,
+      "step": 690
+    },
+    {
+      "epoch": 0.0480340355451863,
+      "grad_norm": 0.9641822576522827,
+      "learning_rate": 6.395242451967064e-05,
+      "loss": 1.8469,
+      "step": 700
+    },
+    {
+      "epoch": 0.0480340355451863,
+      "eval_accuracy": 0.5796691714504977,
+      "eval_loss": 1.859383225440979,
+      "eval_runtime": 2195.5862,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 700
+    },
+    {
+      "epoch": 0.04872023605297468,
+      "grad_norm": 1.0197440385818481,
+      "learning_rate": 6.486733760292772e-05,
+      "loss": 1.871,
+      "step": 710
+    },
+    {
+      "epoch": 0.04872023605297468,
+      "eval_accuracy": 0.57873611108621,
+      "eval_loss": 1.8629941940307617,
+      "eval_runtime": 2195.7373,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.676,
+      "step": 710
+    },
+    {
+      "epoch": 0.04940643656076305,
+      "grad_norm": 1.0116243362426758,
+      "learning_rate": 6.578225068618482e-05,
+      "loss": 1.8012,
+      "step": 720
+    },
+    {
+      "epoch": 0.04940643656076305,
+      "eval_accuracy": 0.5800620774160878,
+      "eval_loss": 1.8581833839416504,
+      "eval_runtime": 2195.6346,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.677,
+      "step": 720
+    },
+    {
+      "epoch": 0.05009263706855143,
+      "grad_norm": 1.015205979347229,
+      "learning_rate": 6.669716376944191e-05,
+      "loss": 1.8966,
+      "step": 730
+    },
+    {
+      "epoch": 0.05009263706855143,
+      "eval_accuracy": 0.5792581042914667,
+      "eval_loss": 1.8612667322158813,
+      "eval_runtime": 2195.4078,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 730
+    },
+    {
+      "epoch": 0.05077883757633981,
+      "grad_norm": 1.031814694404602,
+      "learning_rate": 6.7612076852699e-05,
+      "loss": 1.8505,
+      "step": 740
+    },
+    {
+      "epoch": 0.05077883757633981,
+      "eval_accuracy": 0.5786211344630193,
+      "eval_loss": 1.8662667274475098,
+      "eval_runtime": 2195.1361,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 740
+    },
+    {
+      "epoch": 0.05146503808412818,
+      "grad_norm": 0.9193501472473145,
+      "learning_rate": 6.852698993595609e-05,
+      "loss": 1.8415,
+      "step": 750
+    },
+    {
+      "epoch": 0.05146503808412818,
+      "eval_accuracy": 0.5790159329112897,
+      "eval_loss": 1.8643760681152344,
+      "eval_runtime": 2196.5278,
+      "eval_samples_per_second": 13.404,
+      "eval_steps_per_second": 1.676,
+      "step": 750
+    },
+    {
+      "epoch": 0.05215123859191656,
+      "grad_norm": 0.9670405387878418,
+      "learning_rate": 6.944190301921317e-05,
+      "loss": 1.9061,
+      "step": 760
+    },
+    {
+      "epoch": 0.05215123859191656,
+      "eval_accuracy": 0.5785704026466072,
+      "eval_loss": 1.8654682636260986,
+      "eval_runtime": 2196.4101,
+      "eval_samples_per_second": 13.405,
+      "eval_steps_per_second": 1.676,
+      "step": 760
+    },
+    {
+      "epoch": 0.05283743909970493,
+      "grad_norm": 0.9446126222610474,
+      "learning_rate": 7.035681610247027e-05,
+      "loss": 1.8452,
+      "step": 770
+    },
+    {
+      "epoch": 0.05283743909970493,
+      "eval_accuracy": 0.5786606109877013,
+      "eval_loss": 1.8661186695098877,
+      "eval_runtime": 2196.1052,
+      "eval_samples_per_second": 13.406,
+      "eval_steps_per_second": 1.676,
+      "step": 770
+    },
+    {
+      "epoch": 0.05352363960749331,
+      "grad_norm": 0.9109969139099121,
+      "learning_rate": 7.127172918572736e-05,
+      "loss": 1.8865,
+      "step": 780
+    },
+    {
+      "epoch": 0.05352363960749331,
+      "eval_accuracy": 0.5784766749517566,
+      "eval_loss": 1.8670191764831543,
+      "eval_runtime": 2196.1816,
+      "eval_samples_per_second": 13.406,
+      "eval_steps_per_second": 1.676,
+      "step": 780
+    },
+    {
+      "epoch": 0.05420984011528168,
+      "grad_norm": 1.0019862651824951,
+      "learning_rate": 7.218664226898445e-05,
+      "loss": 1.8181,
+      "step": 790
+    },
+    {
+      "epoch": 0.05420984011528168,
+      "eval_accuracy": 0.5783223214082356,
+      "eval_loss": 1.8682514429092407,
+      "eval_runtime": 2196.2402,
+      "eval_samples_per_second": 13.406,
+      "eval_steps_per_second": 1.676,
+      "step": 790
+    },
+    {
+      "epoch": 0.05489604062307006,
+      "grad_norm": 0.9845278859138489,
+      "learning_rate": 7.310155535224153e-05,
+      "loss": 1.8274,
+      "step": 800
+    },
+    {
+      "epoch": 0.05489604062307006,
+      "eval_accuracy": 0.5783366312334146,
+      "eval_loss": 1.8687405586242676,
+      "eval_runtime": 2197.5837,
+      "eval_samples_per_second": 13.397,
+      "eval_steps_per_second": 1.675,
+      "step": 800
+    },
+    {
+      "epoch": 0.05558224113085844,
+      "grad_norm": 0.9345710873603821,
+      "learning_rate": 7.401646843549863e-05,
+      "loss": 1.8175,
+      "step": 810
+    },
+    {
+      "epoch": 0.05558224113085844,
+      "eval_accuracy": 0.578313921441251,
+      "eval_loss": 1.870363712310791,
+      "eval_runtime": 2195.3953,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 810
+    },
+    {
+      "epoch": 0.05626844163864681,
+      "grad_norm": 0.9270123839378357,
+      "learning_rate": 7.493138151875572e-05,
+      "loss": 1.8407,
+      "step": 820
+    },
+    {
+      "epoch": 0.05626844163864681,
+      "eval_accuracy": 0.5773186415586673,
+      "eval_loss": 1.8754173517227173,
+      "eval_runtime": 2197.5914,
+      "eval_samples_per_second": 13.397,
+      "eval_steps_per_second": 1.675,
+      "step": 820
+    },
+    {
+      "epoch": 0.05695464214643519,
+      "grad_norm": 0.9666882753372192,
+      "learning_rate": 7.584629460201281e-05,
+      "loss": 1.8148,
+      "step": 830
+    },
+    {
+      "epoch": 0.05695464214643519,
+      "eval_accuracy": 0.5775507196978827,
+      "eval_loss": 1.8739426136016846,
+      "eval_runtime": 2198.1195,
+      "eval_samples_per_second": 13.394,
+      "eval_steps_per_second": 1.675,
+      "step": 830
+    },
+    {
+      "epoch": 0.05764084265422356,
+      "grad_norm": 0.9308087229728699,
+      "learning_rate": 7.67612076852699e-05,
+      "loss": 1.8673,
+      "step": 840
+    },
+    {
+      "epoch": 0.05764084265422356,
+      "eval_accuracy": 0.5774692101368278,
+      "eval_loss": 1.8749210834503174,
+      "eval_runtime": 2197.1707,
+      "eval_samples_per_second": 13.4,
+      "eval_steps_per_second": 1.675,
+      "step": 840
+    },
+    {
+      "epoch": 0.05832704316201194,
+      "grad_norm": 1.0570358037948608,
+      "learning_rate": 7.7676120768527e-05,
+      "loss": 1.8628,
+      "step": 850
+    },
+    {
+      "epoch": 0.05832704316201194,
+      "eval_accuracy": 0.577259841789776,
+      "eval_loss": 1.87544846534729,
+      "eval_runtime": 2196.2567,
+      "eval_samples_per_second": 13.406,
+      "eval_steps_per_second": 1.676,
+      "step": 850
+    },
+    {
+      "epoch": 0.05901324366980031,
+      "grad_norm": 0.9226791262626648,
+      "learning_rate": 7.859103385178408e-05,
+      "loss": 1.8726,
+      "step": 860
+    },
+    {
+      "epoch": 0.05901324366980031,
+      "eval_accuracy": 0.5769537908187763,
+      "eval_loss": 1.8771995306015015,
+      "eval_runtime": 2196.3678,
+      "eval_samples_per_second": 13.405,
+      "eval_steps_per_second": 1.676,
+      "step": 860
+    },
+    {
+      "epoch": 0.05969944417758869,
+      "grad_norm": 0.8840520977973938,
+      "learning_rate": 7.950594693504118e-05,
+      "loss": 1.8312,
+      "step": 870
+    },
+    {
+      "epoch": 0.05969944417758869,
+      "eval_accuracy": 0.576975570970325,
+      "eval_loss": 1.8782720565795898,
+      "eval_runtime": 2197.3135,
+      "eval_samples_per_second": 13.399,
+      "eval_steps_per_second": 1.675,
+      "step": 870
+    },
+    {
+      "epoch": 0.06038564468537707,
+      "grad_norm": 0.9905896782875061,
+      "learning_rate": 8.042086001829825e-05,
+      "loss": 1.9177,
+      "step": 880
+    },
+    {
+      "epoch": 0.06038564468537707,
+      "eval_accuracy": 0.5767803464411996,
+      "eval_loss": 1.877514123916626,
+      "eval_runtime": 2195.3275,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 880
+    },
+    {
+      "epoch": 0.06107184519316544,
+      "grad_norm": 1.012568712234497,
+      "learning_rate": 8.133577310155535e-05,
+      "loss": 1.9192,
+      "step": 890
+    },
+    {
+      "epoch": 0.06107184519316544,
+      "eval_accuracy": 0.5767326359567858,
+      "eval_loss": 1.8775851726531982,
+      "eval_runtime": 2196.0616,
+      "eval_samples_per_second": 13.407,
+      "eval_steps_per_second": 1.676,
+      "step": 890
+    },
+    {
+      "epoch": 0.06175804570095382,
+      "grad_norm": 0.9244153499603271,
+      "learning_rate": 8.225068618481244e-05,
+      "loss": 1.8696,
+      "step": 900
+    },
+    {
+      "epoch": 0.06175804570095382,
+      "eval_accuracy": 0.5768811460450133,
+      "eval_loss": 1.8783646821975708,
+      "eval_runtime": 2195.2053,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 900
+    },
+    {
+      "epoch": 0.06244424620874219,
+      "grad_norm": 0.8586702346801758,
+      "learning_rate": 8.316559926806954e-05,
+      "loss": 1.902,
+      "step": 910
+    },
+    {
+      "epoch": 0.06244424620874219,
+      "eval_accuracy": 0.5765674255389409,
+      "eval_loss": 1.8812183141708374,
+      "eval_runtime": 2196.1137,
+      "eval_samples_per_second": 13.406,
+      "eval_steps_per_second": 1.676,
+      "step": 910
+    },
+    {
+      "epoch": 0.06313044671653056,
+      "grad_norm": 0.9296178221702576,
+      "learning_rate": 8.408051235132663e-05,
+      "loss": 1.8953,
+      "step": 920
+    },
+    {
+      "epoch": 0.06313044671653056,
+      "eval_accuracy": 0.5759858689314306,
+      "eval_loss": 1.882770299911499,
+      "eval_runtime": 2196.7044,
+      "eval_samples_per_second": 13.403,
+      "eval_steps_per_second": 1.676,
+      "step": 920
+    },
+    {
+      "epoch": 0.06381664722431894,
+      "grad_norm": 0.8966830968856812,
+      "learning_rate": 8.499542543458372e-05,
+      "loss": 1.8931,
+      "step": 930
+    },
+    {
+      "epoch": 0.06381664722431894,
+      "eval_accuracy": 0.5755770926724864,
+      "eval_loss": 1.8848005533218384,
+      "eval_runtime": 2195.3041,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 930
+    },
+    {
+      "epoch": 0.06450284773210732,
+      "grad_norm": 1.0561726093292236,
+      "learning_rate": 8.59103385178408e-05,
+      "loss": 1.8726,
+      "step": 940
+    },
+    {
+      "epoch": 0.06450284773210732,
+      "eval_accuracy": 0.5759230185855744,
+      "eval_loss": 1.8830986022949219,
+      "eval_runtime": 2196.3236,
+      "eval_samples_per_second": 13.405,
+      "eval_steps_per_second": 1.676,
+      "step": 940
+    },
+    {
+      "epoch": 0.0651890482398957,
+      "grad_norm": 0.8631952404975891,
+      "learning_rate": 8.68252516010979e-05,
+      "loss": 1.8391,
+      "step": 950
+    },
+    {
+      "epoch": 0.0651890482398957,
+      "eval_accuracy": 0.5761208660292918,
+      "eval_loss": 1.8824081420898438,
+      "eval_runtime": 2195.5043,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 950
+    },
+    {
+      "epoch": 0.06587524874768408,
+      "grad_norm": 0.8912535905838013,
+      "learning_rate": 8.774016468435499e-05,
+      "loss": 1.8492,
+      "step": 960
+    },
+    {
+      "epoch": 0.06587524874768408,
+      "eval_accuracy": 0.575300856604064,
+      "eval_loss": 1.8885464668273926,
+      "eval_runtime": 2195.4665,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 960
+    },
+    {
+      "epoch": 0.06656144925547244,
+      "grad_norm": 0.941037654876709,
+      "learning_rate": 8.865507776761208e-05,
+      "loss": 1.9497,
+      "step": 970
+    },
+    {
+      "epoch": 0.06656144925547244,
+      "eval_accuracy": 0.5762686456856077,
+      "eval_loss": 1.8846741914749146,
+      "eval_runtime": 2195.3271,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 970
+    },
+    {
+      "epoch": 0.06724764976326082,
+      "grad_norm": 0.9157822728157043,
+      "learning_rate": 8.956999085086916e-05,
+      "loss": 1.8173,
+      "step": 980
+    },
+    {
+      "epoch": 0.06724764976326082,
+      "eval_accuracy": 0.5756494718346451,
+      "eval_loss": 1.8873111009597778,
+      "eval_runtime": 2195.0363,
+      "eval_samples_per_second": 13.413,
+      "eval_steps_per_second": 1.677,
+      "step": 980
+    },
+    {
+      "epoch": 0.0679338502710492,
+      "grad_norm": 0.7956060767173767,
+      "learning_rate": 9.048490393412626e-05,
+      "loss": 1.8227,
+      "step": 990
+    },
+    {
+      "epoch": 0.0679338502710492,
+      "eval_accuracy": 0.5754663658349637,
+      "eval_loss": 1.8905543088912964,
+      "eval_runtime": 2196.6945,
+      "eval_samples_per_second": 13.403,
+      "eval_steps_per_second": 1.676,
+      "step": 990
+    },
+    {
+      "epoch": 0.06862005077883758,
+      "grad_norm": 0.8911742568016052,
+      "learning_rate": 9.139981701738335e-05,
+      "loss": 1.8767,
+      "step": 1000
+    },
+    {
+      "epoch": 0.06862005077883758,
+      "eval_accuracy": 0.5748883285812098,
+      "eval_loss": 1.8892529010772705,
+      "eval_runtime": 2195.3255,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 1000
+    },
+    {
+      "epoch": 0.06930625128662596,
+      "grad_norm": 0.8083513379096985,
+      "learning_rate": 9.231473010064045e-05,
+      "loss": 1.9234,
+      "step": 1010
+    },
+    {
+      "epoch": 0.06930625128662596,
+      "eval_accuracy": 0.5748815222838508,
+      "eval_loss": 1.8903882503509521,
+      "eval_runtime": 2195.4835,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 1010
+    },
+    {
+      "epoch": 0.06999245179441432,
+      "grad_norm": 0.8500548005104065,
+      "learning_rate": 9.322964318389752e-05,
+      "loss": 1.8992,
+      "step": 1020
+    },
+    {
+      "epoch": 0.06999245179441432,
+      "eval_accuracy": 0.5743459496853266,
+      "eval_loss": 1.8903559446334839,
+      "eval_runtime": 2195.1026,
+      "eval_samples_per_second": 13.413,
+      "eval_steps_per_second": 1.677,
+      "step": 1020
+    },
+    {
+      "epoch": 0.0706786523022027,
+      "grad_norm": 0.8529781699180603,
+      "learning_rate": 9.414455626715462e-05,
+      "loss": 1.9097,
+      "step": 1030
+    },
+    {
+      "epoch": 0.0706786523022027,
+      "eval_accuracy": 0.574473210845214,
+      "eval_loss": 1.8921775817871094,
+      "eval_runtime": 2196.6371,
+      "eval_samples_per_second": 13.403,
+      "eval_steps_per_second": 1.676,
+      "step": 1030
+    },
+    {
+      "epoch": 0.07136485280999108,
+      "grad_norm": 0.985618531703949,
+      "learning_rate": 9.505946935041171e-05,
+      "loss": 1.8252,
+      "step": 1040
+    },
+    {
+      "epoch": 0.07136485280999108,
+      "eval_accuracy": 0.5746204260768708,
+      "eval_loss": 1.8922775983810425,
+      "eval_runtime": 2195.2982,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 1040
+    },
+    {
+      "epoch": 0.07205105331777946,
+      "grad_norm": 0.8113195896148682,
+      "learning_rate": 9.597438243366881e-05,
+      "loss": 1.8873,
+      "step": 1050
+    },
+    {
+      "epoch": 0.07205105331777946,
+      "eval_accuracy": 0.5742100893497516,
+      "eval_loss": 1.8929826021194458,
+      "eval_runtime": 2195.2211,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 1050
+    },
+    {
+      "epoch": 0.07273725382556782,
+      "grad_norm": 0.848109245300293,
+      "learning_rate": 9.68892955169259e-05,
+      "loss": 1.8595,
+      "step": 1060
+    },
+    {
+      "epoch": 0.07273725382556782,
+      "eval_accuracy": 0.5738744890877788,
+      "eval_loss": 1.896013855934143,
+      "eval_runtime": 2195.349,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 1060
+    },
+    {
+      "epoch": 0.0734234543333562,
+      "grad_norm": 0.8576720356941223,
+      "learning_rate": 9.780420860018299e-05,
+      "loss": 1.8415,
+      "step": 1070
+    },
+    {
+      "epoch": 0.0734234543333562,
+      "eval_accuracy": 0.573847263898343,
+      "eval_loss": 1.896903395652771,
+      "eval_runtime": 2195.4292,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 1070
+    },
+    {
+      "epoch": 0.07410965484114458,
+      "grad_norm": 0.8684922456741333,
+      "learning_rate": 9.871912168344007e-05,
+      "loss": 1.8104,
+      "step": 1080
+    },
+    {
+      "epoch": 0.07410965484114458,
+      "eval_accuracy": 0.5738449397968058,
+      "eval_loss": 1.895707607269287,
+      "eval_runtime": 2195.3809,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 1080
+    },
+    {
+      "epoch": 0.07479585534893296,
+      "grad_norm": 0.8526385426521301,
+      "learning_rate": 9.963403476669717e-05,
+      "loss": 1.8955,
+      "step": 1090
+    },
+    {
+      "epoch": 0.07479585534893296,
+      "eval_accuracy": 0.5738943767566472,
+      "eval_loss": 1.895605444908142,
+      "eval_runtime": 2195.5152,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 1090
+    },
+    {
+      "epoch": 0.07548205585672134,
+      "grad_norm": 0.7912049889564514,
+      "learning_rate": 0.00010054894784995426,
+      "loss": 1.8411,
+      "step": 1100
+    },
+    {
+      "epoch": 0.07548205585672134,
+      "eval_accuracy": 0.5737731582607566,
+      "eval_loss": 1.8975555896759033,
+      "eval_runtime": 2196.2407,
+      "eval_samples_per_second": 13.406,
+      "eval_steps_per_second": 1.676,
+      "step": 1100
+    },
+    {
+      "epoch": 0.0761682563645097,
+      "grad_norm": 0.8417245149612427,
+      "learning_rate": 0.00010146386093321135,
+      "loss": 1.8934,
+      "step": 1110
+    },
+    {
+      "epoch": 0.0761682563645097,
+      "eval_accuracy": 0.5735296588225584,
+      "eval_loss": 1.8995074033737183,
+      "eval_runtime": 2195.4727,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 1110
+    },
+    {
+      "epoch": 0.07685445687229808,
+      "grad_norm": 0.8422715663909912,
+      "learning_rate": 0.00010237877401646845,
+      "loss": 1.872,
+      "step": 1120
+    },
+    {
+      "epoch": 0.07685445687229808,
+      "eval_accuracy": 0.5736483208067581,
+      "eval_loss": 1.8992631435394287,
+      "eval_runtime": 2195.5393,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 1120
+    },
+    {
+      "epoch": 0.07754065738008646,
+      "grad_norm": 0.924197793006897,
+      "learning_rate": 0.00010329368709972553,
+      "loss": 1.8788,
+      "step": 1130
+    },
+    {
+      "epoch": 0.07754065738008646,
+      "eval_accuracy": 0.573535568680753,
+      "eval_loss": 1.8997005224227905,
+      "eval_runtime": 2195.1245,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 1130
+    },
+    {
+      "epoch": 0.07822685788787484,
+      "grad_norm": 0.8739652037620544,
+      "learning_rate": 0.00010420860018298262,
+      "loss": 1.8611,
+      "step": 1140
+    },
+    {
+      "epoch": 0.07822685788787484,
+      "eval_accuracy": 0.5737993874066766,
+      "eval_loss": 1.897829532623291,
+      "eval_runtime": 2194.8865,
+      "eval_samples_per_second": 13.414,
+      "eval_steps_per_second": 1.677,
+      "step": 1140
+    },
+    {
+      "epoch": 0.07891305839566322,
+      "grad_norm": 0.9717639684677124,
+      "learning_rate": 0.00010512351326623971,
+      "loss": 1.9323,
+      "step": 1150
+    },
+    {
+      "epoch": 0.07891305839566322,
+      "eval_accuracy": 0.5726149588604147,
+      "eval_loss": 1.9035004377365112,
+      "eval_runtime": 2195.2968,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 1150
+    },
+    {
+      "epoch": 0.07959925890345158,
+      "grad_norm": 0.9229555130004883,
+      "learning_rate": 0.00010603842634949681,
+      "loss": 1.892,
+      "step": 1160
+    },
+    {
+      "epoch": 0.07959925890345158,
+      "eval_accuracy": 0.5725945399683378,
+      "eval_loss": 1.903469204902649,
+      "eval_runtime": 2195.7477,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.676,
+      "step": 1160
+    },
+    {
+      "epoch": 0.08028545941123996,
+      "grad_norm": 0.8387541174888611,
+      "learning_rate": 0.0001069533394327539,
+      "loss": 1.8585,
+      "step": 1170
+    },
+    {
+      "epoch": 0.08028545941123996,
+      "eval_accuracy": 0.5729163616283399,
+      "eval_loss": 1.9036465883255005,
+      "eval_runtime": 2195.4236,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 1170
+    },
+    {
+      "epoch": 0.08097165991902834,
+      "grad_norm": 0.8355444073677063,
+      "learning_rate": 0.00010786825251601098,
+      "loss": 1.8718,
+      "step": 1180
+    },
+    {
+      "epoch": 0.08097165991902834,
+      "eval_accuracy": 0.5726727293843396,
+      "eval_loss": 1.9039238691329956,
+      "eval_runtime": 2195.5829,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 1180
+    },
+    {
+      "epoch": 0.08165786042681672,
+      "grad_norm": 0.8180125951766968,
+      "learning_rate": 0.00010878316559926807,
+      "loss": 1.9032,
+      "step": 1190
+    },
+    {
+      "epoch": 0.08165786042681672,
+      "eval_accuracy": 0.5721860957238989,
+      "eval_loss": 1.906183123588562,
+      "eval_runtime": 2195.35,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 1190
+    },
+    {
+      "epoch": 0.0823440609346051,
+      "grad_norm": 0.7993365526199341,
+      "learning_rate": 0.00010969807868252517,
+      "loss": 1.8526,
+      "step": 1200
+    },
+    {
+      "epoch": 0.0823440609346051,
+      "eval_accuracy": 0.5722786613679808,
+      "eval_loss": 1.9059474468231201,
+      "eval_runtime": 2195.3016,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 1200
+    },
+    {
+      "epoch": 0.08303026144239346,
+      "grad_norm": 1.0333908796310425,
+      "learning_rate": 0.00011061299176578226,
+      "loss": 1.8649,
+      "step": 1210
+    },
+    {
+      "epoch": 0.08303026144239346,
+      "eval_accuracy": 0.570262403680102,
+      "eval_loss": 1.922554850578308,
+      "eval_runtime": 2195.4851,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 1210
+    },
+    {
+      "epoch": 0.08371646195018184,
+      "grad_norm": 1.1051403284072876,
+      "learning_rate": 0.00011143641354071363,
+      "loss": 1.9549,
+      "step": 1220
+    },
+    {
+      "epoch": 0.08371646195018184,
+      "eval_accuracy": 0.5712613025207935,
+      "eval_loss": 1.9148966073989868,
+      "eval_runtime": 2195.4835,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 1220
+    },
+    {
+      "epoch": 0.08440266245797022,
+      "grad_norm": 0.8354151248931885,
+      "learning_rate": 0.00011235132662397072,
+      "loss": 1.9199,
+      "step": 1230
+    },
+    {
+      "epoch": 0.08440266245797022,
+      "eval_accuracy": 0.5723187355187723,
+      "eval_loss": 1.9058703184127808,
+      "eval_runtime": 2195.1846,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 1230
+    },
+    {
+      "epoch": 0.0850888629657586,
+      "grad_norm": 0.7311375737190247,
+      "learning_rate": 0.00011317474839890212,
+      "loss": 1.8916,
+      "step": 1240
+    },
+    {
+      "epoch": 0.0850888629657586,
+      "eval_accuracy": 0.572127262753557,
+      "eval_loss": 1.9079400300979614,
+      "eval_runtime": 2195.7032,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.676,
+      "step": 1240
+    },
+    {
+      "epoch": 0.08577506347354696,
+      "grad_norm": 0.8696947693824768,
+      "learning_rate": 0.00011408966148215921,
+      "loss": 1.9645,
+      "step": 1250
+    },
+    {
+      "epoch": 0.08577506347354696,
+      "eval_accuracy": 0.5722400480810126,
+      "eval_loss": 1.9090999364852905,
+      "eval_runtime": 2195.3879,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 1250
+    },
+    {
+      "epoch": 0.08646126398133534,
+      "grad_norm": 0.9006031155586243,
+      "learning_rate": 0.0001150045745654163,
+      "loss": 1.919,
+      "step": 1260
+    },
+    {
+      "epoch": 0.08646126398133534,
+      "eval_accuracy": 0.5701970632254558,
+      "eval_loss": 1.9196306467056274,
+      "eval_runtime": 2195.6859,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.676,
+      "step": 1260
+    },
+    {
+      "epoch": 0.08714746448912372,
+      "grad_norm": 1.0175645351409912,
+      "learning_rate": 0.00011591948764867338,
+      "loss": 1.8906,
+      "step": 1270
+    },
+    {
+      "epoch": 0.08714746448912372,
+      "eval_accuracy": 0.571257019533675,
+      "eval_loss": 1.9163960218429565,
+      "eval_runtime": 2195.6701,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.676,
+      "step": 1270
+    },
+    {
+      "epoch": 0.0878336649969121,
+      "grad_norm": 0.8517598509788513,
+      "learning_rate": 0.00011683440073193048,
+      "loss": 1.9111,
+      "step": 1280
+    },
+    {
+      "epoch": 0.0878336649969121,
+      "eval_accuracy": 0.5712075161709325,
+      "eval_loss": 1.9151411056518555,
+      "eval_runtime": 2196.2166,
+      "eval_samples_per_second": 13.406,
+      "eval_steps_per_second": 1.676,
+      "step": 1280
+    },
+    {
+      "epoch": 0.08851986550470048,
+      "grad_norm": 0.8764553070068359,
+      "learning_rate": 0.00011774931381518757,
+      "loss": 1.9597,
+      "step": 1290
+    },
+    {
+      "epoch": 0.08851986550470048,
+      "eval_accuracy": 0.5705109497387809,
+      "eval_loss": 1.9204518795013428,
+      "eval_runtime": 2196.4778,
+      "eval_samples_per_second": 13.404,
+      "eval_steps_per_second": 1.676,
+      "step": 1290
+    },
+    {
+      "epoch": 0.08920606601248884,
+      "grad_norm": 0.8551300764083862,
+      "learning_rate": 0.00011866422689844465,
+      "loss": 1.885,
+      "step": 1300
+    },
+    {
+      "epoch": 0.08920606601248884,
+      "eval_accuracy": 0.5686897173713243,
+      "eval_loss": 1.9303290843963623,
+      "eval_runtime": 2195.3977,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 1300
+    },
+    {
+      "epoch": 0.08989226652027722,
+      "grad_norm": 0.8597418069839478,
+      "learning_rate": 0.00011957913998170174,
+      "loss": 1.9086,
+      "step": 1310
+    },
+    {
+      "epoch": 0.08989226652027722,
+      "eval_accuracy": 0.5694771229721302,
+      "eval_loss": 1.9275456666946411,
+      "eval_runtime": 2195.6823,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.676,
+      "step": 1310
+    },
+    {
+      "epoch": 0.0905784670280656,
+      "grad_norm": 1.0071135759353638,
+      "learning_rate": 0.00012049405306495884,
+      "loss": 1.9019,
+      "step": 1320
+    },
+    {
+      "epoch": 0.0905784670280656,
+      "eval_accuracy": 0.5700387587093215,
+      "eval_loss": 1.92341947555542,
+      "eval_runtime": 2195.6668,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.676,
+      "step": 1320
+    },
+    {
+      "epoch": 0.09126466753585398,
+      "grad_norm": 0.7982167601585388,
+      "learning_rate": 0.00012140896614821593,
+      "loss": 1.9072,
+      "step": 1330
+    },
+    {
+      "epoch": 0.09126466753585398,
+      "eval_accuracy": 0.5695152714387909,
+      "eval_loss": 1.9324274063110352,
+      "eval_runtime": 2196.4255,
+      "eval_samples_per_second": 13.405,
+      "eval_steps_per_second": 1.676,
+      "step": 1330
+    },
+    {
+      "epoch": 0.09195086804364236,
+      "grad_norm": 0.8429548144340515,
+      "learning_rate": 0.000122323879231473,
+      "loss": 1.9473,
+      "step": 1340
+    },
+    {
+      "epoch": 0.09195086804364236,
+      "eval_accuracy": 0.5696431634262383,
+      "eval_loss": 1.923658013343811,
+      "eval_runtime": 2196.8488,
+      "eval_samples_per_second": 13.402,
+      "eval_steps_per_second": 1.676,
+      "step": 1340
+    },
+    {
+      "epoch": 0.09263706855143072,
+      "grad_norm": 0.8137506246566772,
+      "learning_rate": 0.0001232387923147301,
+      "loss": 1.9171,
+      "step": 1350
+    },
+    {
+      "epoch": 0.09263706855143072,
+      "eval_accuracy": 0.5698193635242091,
+      "eval_loss": 1.9207695722579956,
+      "eval_runtime": 2195.6978,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.676,
+      "step": 1350
+    },
+    {
+      "epoch": 0.0933232690592191,
+      "grad_norm": 0.7542211413383484,
+      "learning_rate": 0.0001241537053979872,
+      "loss": 1.9515,
+      "step": 1360
+    },
+    {
+      "epoch": 0.0933232690592191,
+      "eval_accuracy": 0.5703601819519173,
+      "eval_loss": 1.9201329946517944,
+      "eval_runtime": 2196.079,
+      "eval_samples_per_second": 13.407,
+      "eval_steps_per_second": 1.676,
+      "step": 1360
+    },
+    {
+      "epoch": 0.09400946956700748,
+      "grad_norm": 0.7821772694587708,
+      "learning_rate": 0.0001250686184812443,
+      "loss": 1.9078,
+      "step": 1370
+    },
+    {
+      "epoch": 0.09400946956700748,
+      "eval_accuracy": 0.5695221109376003,
+      "eval_loss": 1.9211959838867188,
+      "eval_runtime": 2195.4801,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 1370
+    },
+    {
+      "epoch": 0.09469567007479586,
+      "grad_norm": 0.7843822240829468,
+      "learning_rate": 0.00012598353156450138,
+      "loss": 1.9052,
+      "step": 1380
+    },
+    {
+      "epoch": 0.09469567007479586,
+      "eval_accuracy": 0.5696008647782611,
+      "eval_loss": 1.9253789186477661,
+      "eval_runtime": 2195.2622,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 1380
+    },
+    {
+      "epoch": 0.09538187058258422,
+      "grad_norm": 0.7415681481361389,
+      "learning_rate": 0.00012689844464775846,
+      "loss": 1.9027,
+      "step": 1390
+    },
+    {
+      "epoch": 0.09538187058258422,
+      "eval_accuracy": 0.5689813257113427,
+      "eval_loss": 1.9291355609893799,
+      "eval_runtime": 2196.3808,
+      "eval_samples_per_second": 13.405,
+      "eval_steps_per_second": 1.676,
+      "step": 1390
+    },
+    {
+      "epoch": 0.0960680710903726,
+      "grad_norm": 0.7906999588012695,
+      "learning_rate": 0.00012781335773101558,
+      "loss": 1.9165,
+      "step": 1400
+    },
+    {
+      "epoch": 0.0960680710903726,
+      "eval_accuracy": 0.5686311168111361,
+      "eval_loss": 1.929186463356018,
+      "eval_runtime": 2195.3777,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 1400
+    },
+    {
+      "epoch": 0.09675427159816098,
+      "grad_norm": 0.7665316462516785,
+      "learning_rate": 0.00012872827081427266,
+      "loss": 1.8921,
+      "step": 1410
+    },
+    {
+      "epoch": 0.09675427159816098,
+      "eval_accuracy": 0.5691787415362032,
+      "eval_loss": 1.9266636371612549,
+      "eval_runtime": 2196.5486,
+      "eval_samples_per_second": 13.404,
+      "eval_steps_per_second": 1.676,
+      "step": 1410
+    },
+    {
+      "epoch": 0.09744047210594936,
+      "grad_norm": 0.7428321242332458,
+      "learning_rate": 0.00012964318389752975,
+      "loss": 2.0197,
+      "step": 1420
+    },
+    {
+      "epoch": 0.09744047210594936,
+      "eval_accuracy": 0.5681328958444599,
+      "eval_loss": 1.9324249029159546,
+      "eval_runtime": 2199.1199,
+      "eval_samples_per_second": 13.388,
+      "eval_steps_per_second": 1.674,
+      "step": 1420
+    },
+    {
+      "epoch": 0.09812667261373774,
+      "grad_norm": 0.7564303874969482,
+      "learning_rate": 0.00013055809698078684,
+      "loss": 1.9049,
+      "step": 1430
+    },
+    {
+      "epoch": 0.09812667261373774,
+      "eval_accuracy": 0.5684988754336691,
+      "eval_loss": 1.9301478862762451,
+      "eval_runtime": 2196.3543,
+      "eval_samples_per_second": 13.405,
+      "eval_steps_per_second": 1.676,
+      "step": 1430
+    },
+    {
+      "epoch": 0.0988128731215261,
+      "grad_norm": 0.9298701286315918,
+      "learning_rate": 0.00013147301006404392,
+      "loss": 1.9086,
+      "step": 1440
+    },
+    {
+      "epoch": 0.0988128731215261,
+      "eval_accuracy": 0.5689413179634523,
+      "eval_loss": 1.9311435222625732,
+      "eval_runtime": 2195.5784,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 1440
+    },
+    {
+      "epoch": 0.09949907362931448,
+      "grad_norm": 0.8423255681991577,
+      "learning_rate": 0.000132387923147301,
+      "loss": 1.902,
+      "step": 1450
+    },
+    {
+      "epoch": 0.09949907362931448,
+      "eval_accuracy": 0.5683401725001284,
+      "eval_loss": 1.9306671619415283,
+      "eval_runtime": 2195.5481,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 1450
+    },
+    {
+      "epoch": 0.10018527413710286,
+      "grad_norm": 0.8051897883415222,
+      "learning_rate": 0.0001333028362305581,
+      "loss": 1.9863,
+      "step": 1460
+    },
+    {
+      "epoch": 0.10018527413710286,
+      "eval_accuracy": 0.568742474476219,
+      "eval_loss": 1.9322450160980225,
+      "eval_runtime": 2196.6508,
+      "eval_samples_per_second": 13.403,
+      "eval_steps_per_second": 1.676,
+      "step": 1460
+    },
+    {
+      "epoch": 0.10087147464489124,
+      "grad_norm": 0.826680600643158,
+      "learning_rate": 0.00013421774931381518,
+      "loss": 1.8864,
+      "step": 1470
+    },
+    {
+      "epoch": 0.10087147464489124,
+      "eval_accuracy": 0.5676466606014257,
+      "eval_loss": 1.9348758459091187,
+      "eval_runtime": 2196.1461,
+      "eval_samples_per_second": 13.406,
+      "eval_steps_per_second": 1.676,
+      "step": 1470
+    },
+    {
+      "epoch": 0.10155767515267962,
+      "grad_norm": 0.8538419604301453,
+      "learning_rate": 0.00013513266239707227,
+      "loss": 1.8823,
+      "step": 1480
+    },
+    {
+      "epoch": 0.10155767515267962,
+      "eval_accuracy": 0.5668430858942111,
+      "eval_loss": 1.942825436592102,
+      "eval_runtime": 2195.3646,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 1480
+    },
+    {
+      "epoch": 0.10224387566046798,
+      "grad_norm": 0.7539196014404297,
+      "learning_rate": 0.00013604757548032939,
+      "loss": 1.9582,
+      "step": 1490
+    },
+    {
+      "epoch": 0.10224387566046798,
+      "eval_accuracy": 0.5680551380473151,
+      "eval_loss": 1.9357633590698242,
+      "eval_runtime": 2195.6893,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.676,
+      "step": 1490
+    },
+    {
+      "epoch": 0.10293007616825636,
+      "grad_norm": 0.8606800436973572,
+      "learning_rate": 0.00013696248856358647,
+      "loss": 1.9724,
+      "step": 1500
+    },
+    {
+      "epoch": 0.10293007616825636,
+      "eval_accuracy": 0.5682361191541625,
+      "eval_loss": 1.9344924688339233,
+      "eval_runtime": 2196.3487,
+      "eval_samples_per_second": 13.405,
+      "eval_steps_per_second": 1.676,
+      "step": 1500
+    },
+    {
+      "epoch": 0.10361627667604474,
+      "grad_norm": 1.327166199684143,
+      "learning_rate": 0.00013787740164684356,
+      "loss": 1.8999,
+      "step": 1510
+    },
+    {
+      "epoch": 0.10361627667604474,
+      "eval_accuracy": 0.5677454681182075,
+      "eval_loss": 1.9374973773956299,
+      "eval_runtime": 2196.3215,
+      "eval_samples_per_second": 13.405,
+      "eval_steps_per_second": 1.676,
+      "step": 1510
+    },
+    {
+      "epoch": 0.10430247718383312,
+      "grad_norm": 0.7882415056228638,
+      "learning_rate": 0.00013879231473010065,
+      "loss": 1.939,
+      "step": 1520
+    },
+    {
+      "epoch": 0.10430247718383312,
+      "eval_accuracy": 0.5680506226500428,
+      "eval_loss": 1.9378398656845093,
+      "eval_runtime": 2195.9406,
+      "eval_samples_per_second": 13.407,
+      "eval_steps_per_second": 1.676,
+      "step": 1520
+    },
+    {
+      "epoch": 0.1049886776916215,
+      "grad_norm": 0.8347277045249939,
+      "learning_rate": 0.00013970722781335773,
+      "loss": 1.9144,
+      "step": 1530
+    },
+    {
+      "epoch": 0.1049886776916215,
+      "eval_accuracy": 0.566825787938484,
+      "eval_loss": 1.9458295106887817,
+      "eval_runtime": 2196.2125,
+      "eval_samples_per_second": 13.406,
+      "eval_steps_per_second": 1.676,
+      "step": 1530
+    },
+    {
+      "epoch": 0.10567487819940986,
+      "grad_norm": 0.9527395963668823,
+      "learning_rate": 0.00014062214089661482,
+      "loss": 1.9098,
+      "step": 1540
+    },
+    {
+      "epoch": 0.10567487819940986,
+      "eval_accuracy": 0.5678965015166755,
+      "eval_loss": 1.940508484840393,
+      "eval_runtime": 2195.4239,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 1540
+    },
+    {
+      "epoch": 0.10636107870719824,
+      "grad_norm": 0.8285815715789795,
+      "learning_rate": 0.0001415370539798719,
+      "loss": 1.9036,
+      "step": 1550
+    },
+    {
+      "epoch": 0.10636107870719824,
+      "eval_accuracy": 0.567786471909614,
+      "eval_loss": 1.9401248693466187,
+      "eval_runtime": 2196.2023,
+      "eval_samples_per_second": 13.406,
+      "eval_steps_per_second": 1.676,
+      "step": 1550
+    },
+    {
+      "epoch": 0.10704727921498662,
+      "grad_norm": 0.9443630576133728,
+      "learning_rate": 0.000142451967063129,
+      "loss": 2.0052,
+      "step": 1560
+    },
+    {
+      "epoch": 0.10704727921498662,
+      "eval_accuracy": 0.5666123026115664,
+      "eval_loss": 1.9421296119689941,
+      "eval_runtime": 2196.6048,
+      "eval_samples_per_second": 13.403,
+      "eval_steps_per_second": 1.676,
+      "step": 1560
+    },
+    {
+      "epoch": 0.107733479722775,
+      "grad_norm": 0.8364240527153015,
+      "learning_rate": 0.0001433668801463861,
+      "loss": 1.8997,
+      "step": 1570
+    },
+    {
+      "epoch": 0.107733479722775,
+      "eval_accuracy": 0.5669286460322308,
+      "eval_loss": 1.941907286643982,
+      "eval_runtime": 2193.9823,
+      "eval_samples_per_second": 13.419,
+      "eval_steps_per_second": 1.678,
+      "step": 1570
+    },
+    {
+      "epoch": 0.10841968023056336,
+      "grad_norm": 0.7851582765579224,
+      "learning_rate": 0.0001442817932296432,
+      "loss": 1.9388,
+      "step": 1580
+    },
+    {
+      "epoch": 0.10841968023056336,
+      "eval_accuracy": 0.5667885691124382,
+      "eval_loss": 1.9420804977416992,
+      "eval_runtime": 2193.8438,
+      "eval_samples_per_second": 13.42,
+      "eval_steps_per_second": 1.678,
+      "step": 1580
+    },
+    {
+      "epoch": 0.10910588073835174,
+      "grad_norm": 1.1334728002548218,
+      "learning_rate": 0.00014519670631290028,
+      "loss": 1.9436,
+      "step": 1590
+    },
+    {
+      "epoch": 0.10910588073835174,
+      "eval_accuracy": 0.5664039303080305,
+      "eval_loss": 1.9469659328460693,
+      "eval_runtime": 2194.723,
+      "eval_samples_per_second": 13.415,
+      "eval_steps_per_second": 1.677,
+      "step": 1590
+    },
+    {
+      "epoch": 0.10979208124614012,
+      "grad_norm": 3.6574079990386963,
+      "learning_rate": 0.00014611161939615737,
+      "loss": 1.9346,
+      "step": 1600
+    },
+    {
+      "epoch": 0.10979208124614012,
+      "eval_accuracy": 0.5647517265252299,
+      "eval_loss": 1.9564319849014282,
+      "eval_runtime": 2193.9968,
+      "eval_samples_per_second": 13.419,
+      "eval_steps_per_second": 1.678,
+      "step": 1600
+    },
+    {
+      "epoch": 0.1104782817539285,
+      "grad_norm": 0.9866987466812134,
+      "learning_rate": 0.00014702653247941448,
+      "loss": 1.9583,
+      "step": 1610
+    },
+    {
+      "epoch": 0.1104782817539285,
+      "eval_accuracy": 0.5648301815528358,
+      "eval_loss": 1.9578943252563477,
+      "eval_runtime": 2195.1234,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 1610
+    },
+    {
+      "epoch": 0.11116448226171688,
+      "grad_norm": 0.7200923562049866,
+      "learning_rate": 0.00014794144556267154,
+      "loss": 1.9934,
+      "step": 1620
+    },
+    {
+      "epoch": 0.11116448226171688,
+      "eval_accuracy": 0.5669039109515848,
+      "eval_loss": 1.94541335105896,
+      "eval_runtime": 2195.0872,
+      "eval_samples_per_second": 13.413,
+      "eval_steps_per_second": 1.677,
+      "step": 1620
+    },
+    {
+      "epoch": 0.11185068276950524,
+      "grad_norm": 0.8922938108444214,
+      "learning_rate": 0.00014885635864592863,
+      "loss": 1.9672,
+      "step": 1630
+    },
+    {
+      "epoch": 0.11185068276950524,
+      "eval_accuracy": 0.5670405681219726,
+      "eval_loss": 1.943771243095398,
+      "eval_runtime": 2195.4513,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 1630
+    },
+    {
+      "epoch": 0.11253688327729362,
+      "grad_norm": 0.8861325979232788,
+      "learning_rate": 0.00014977127172918572,
+      "loss": 1.9451,
+      "step": 1640
+    },
+    {
+      "epoch": 0.11253688327729362,
+      "eval_accuracy": 0.5672741403264618,
+      "eval_loss": 1.9424519538879395,
+      "eval_runtime": 2195.3731,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 1640
+    },
+    {
+      "epoch": 0.113223083785082,
+      "grad_norm": 0.7567145824432373,
+      "learning_rate": 0.00015068618481244283,
+      "loss": 1.9691,
+      "step": 1650
+    },
+    {
+      "epoch": 0.113223083785082,
+      "eval_accuracy": 0.565372195232763,
+      "eval_loss": 1.9486464262008667,
+      "eval_runtime": 2195.4507,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 1650
+    },
+    {
+      "epoch": 0.11390928429287038,
+      "grad_norm": 0.7195202708244324,
+      "learning_rate": 0.00015160109789569992,
+      "loss": 1.9507,
+      "step": 1660
+    },
+    {
+      "epoch": 0.11390928429287038,
+      "eval_accuracy": 0.5668069627160327,
+      "eval_loss": 1.94415283203125,
+      "eval_runtime": 2195.4023,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 1660
+    },
+    {
+      "epoch": 0.11459548480065876,
+      "grad_norm": 0.7501398324966431,
+      "learning_rate": 0.000152516010978957,
+      "loss": 1.9337,
+      "step": 1670
+    },
+    {
+      "epoch": 0.11459548480065876,
+      "eval_accuracy": 0.565814803769799,
+      "eval_loss": 1.9502369165420532,
+      "eval_runtime": 2195.5008,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 1670
+    },
+    {
+      "epoch": 0.11528168530844712,
+      "grad_norm": 5.4457292556762695,
+      "learning_rate": 0.0001534309240622141,
+      "loss": 1.964,
+      "step": 1680
+    },
+    {
+      "epoch": 0.11528168530844712,
+      "eval_accuracy": 0.5647974781240622,
+      "eval_loss": 1.9561764001846313,
+      "eval_runtime": 2195.9745,
+      "eval_samples_per_second": 13.407,
+      "eval_steps_per_second": 1.676,
+      "step": 1680
+    },
+    {
+      "epoch": 0.1159678858162355,
+      "grad_norm": 0.7148681282997131,
+      "learning_rate": 0.0001543458371454712,
+      "loss": 1.9242,
+      "step": 1690
+    },
+    {
+      "epoch": 0.1159678858162355,
+      "eval_accuracy": 0.5624542193498984,
+      "eval_loss": 1.985127329826355,
+      "eval_runtime": 2192.8067,
+      "eval_samples_per_second": 13.427,
+      "eval_steps_per_second": 1.679,
+      "step": 1690
+    },
+    {
+      "epoch": 0.11665408632402388,
+      "grad_norm": 0.8697217702865601,
+      "learning_rate": 0.0001552607502287283,
+      "loss": 1.9982,
+      "step": 1700
+    },
+    {
+      "epoch": 0.11665408632402388,
+      "eval_accuracy": 0.5627469897406854,
+      "eval_loss": 1.975067377090454,
+      "eval_runtime": 2192.9867,
+      "eval_samples_per_second": 13.426,
+      "eval_steps_per_second": 1.679,
+      "step": 1700
+    },
+    {
+      "epoch": 0.11734028683181226,
+      "grad_norm": 0.8448213338851929,
+      "learning_rate": 0.00015617566331198538,
+      "loss": 2.0176,
+      "step": 1710
+    },
+    {
+      "epoch": 0.11734028683181226,
+      "eval_accuracy": 0.5627390877954589,
+      "eval_loss": 1.9716612100601196,
+      "eval_runtime": 2193.5512,
+      "eval_samples_per_second": 13.422,
+      "eval_steps_per_second": 1.678,
+      "step": 1710
+    },
+    {
+      "epoch": 0.11802648733960062,
+      "grad_norm": 3.248958110809326,
+      "learning_rate": 0.00015709057639524244,
+      "loss": 1.9987,
+      "step": 1720
+    },
+    {
+      "epoch": 0.11802648733960062,
+      "eval_accuracy": 0.5634107531397118,
+      "eval_loss": 1.962895154953003,
+      "eval_runtime": 2193.484,
+      "eval_samples_per_second": 13.422,
+      "eval_steps_per_second": 1.678,
+      "step": 1720
+    },
+    {
+      "epoch": 0.118712687847389,
+      "grad_norm": 0.709474503993988,
+      "learning_rate": 0.00015800548947849955,
+      "loss": 1.9584,
+      "step": 1730
+    },
+    {
+      "epoch": 0.118712687847389,
+      "eval_accuracy": 0.5648748043023502,
+      "eval_loss": 1.955528736114502,
+      "eval_runtime": 2192.7873,
+      "eval_samples_per_second": 13.427,
+      "eval_steps_per_second": 1.679,
+      "step": 1730
+    },
+    {
+      "epoch": 0.11939888835517738,
+      "grad_norm": 0.7776058912277222,
+      "learning_rate": 0.00015892040256175664,
+      "loss": 1.9028,
+      "step": 1740
+    },
+    {
+      "epoch": 0.11939888835517738,
+      "eval_accuracy": 0.5636472470718479,
+      "eval_loss": 1.9644140005111694,
+      "eval_runtime": 2193.2656,
+      "eval_samples_per_second": 13.424,
+      "eval_steps_per_second": 1.678,
+      "step": 1740
+    },
+    {
+      "epoch": 0.12008508886296576,
+      "grad_norm": 0.9440386891365051,
+      "learning_rate": 0.00015983531564501372,
+      "loss": 1.9697,
+      "step": 1750
+    },
+    {
+      "epoch": 0.12008508886296576,
+      "eval_accuracy": 0.5645730695199196,
+      "eval_loss": 1.9598476886749268,
+      "eval_runtime": 2192.0993,
+      "eval_samples_per_second": 13.431,
+      "eval_steps_per_second": 1.679,
+      "step": 1750
+    },
+    {
+      "epoch": 0.12077128937075414,
+      "grad_norm": 0.7116687297821045,
+      "learning_rate": 0.0001607502287282708,
+      "loss": 1.971,
+      "step": 1760
+    },
+    {
+      "epoch": 0.12077128937075414,
+      "eval_accuracy": 0.5648594652322046,
+      "eval_loss": 1.957185983657837,
+      "eval_runtime": 2191.728,
+      "eval_samples_per_second": 13.433,
+      "eval_steps_per_second": 1.679,
+      "step": 1760
+    },
+    {
+      "epoch": 0.1214574898785425,
+      "grad_norm": 1.1787147521972656,
+      "learning_rate": 0.00016166514181152793,
+      "loss": 1.9428,
+      "step": 1770
+    },
+    {
+      "epoch": 0.1214574898785425,
+      "eval_accuracy": 0.5637377542259968,
+      "eval_loss": 1.9630035161972046,
+      "eval_runtime": 2191.7408,
+      "eval_samples_per_second": 13.433,
+      "eval_steps_per_second": 1.679,
+      "step": 1770
+    },
+    {
+      "epoch": 0.12214369038633088,
+      "grad_norm": 0.7775124907493591,
+      "learning_rate": 0.000162580054894785,
+      "loss": 1.9251,
+      "step": 1780
+    },
+    {
+      "epoch": 0.12214369038633088,
+      "eval_accuracy": 0.5636699568640114,
+      "eval_loss": 1.962775468826294,
+      "eval_runtime": 2191.7029,
+      "eval_samples_per_second": 13.433,
+      "eval_steps_per_second": 1.68,
+      "step": 1780
+    },
+    {
+      "epoch": 0.12282989089411926,
+      "grad_norm": 0.7894787788391113,
+      "learning_rate": 0.0001634949679780421,
+      "loss": 1.9413,
+      "step": 1790
+    },
+    {
+      "epoch": 0.12282989089411926,
+      "eval_accuracy": 0.5643811651358474,
+      "eval_loss": 1.9617409706115723,
+      "eval_runtime": 2191.672,
+      "eval_samples_per_second": 13.434,
+      "eval_steps_per_second": 1.68,
+      "step": 1790
+    },
+    {
+      "epoch": 0.12351609140190764,
+      "grad_norm": 0.7652885317802429,
+      "learning_rate": 0.00016440988106129919,
+      "loss": 1.9768,
+      "step": 1800
+    },
+    {
+      "epoch": 0.12351609140190764,
+      "eval_accuracy": 0.5625105622114503,
+      "eval_loss": 1.9699786901474,
+      "eval_runtime": 2192.2587,
+      "eval_samples_per_second": 13.43,
+      "eval_steps_per_second": 1.679,
+      "step": 1800
+    },
+    {
+      "epoch": 0.12420229190969602,
+      "grad_norm": 0.7860700488090515,
+      "learning_rate": 0.00016532479414455627,
+      "loss": 1.9325,
+      "step": 1810
+    },
+    {
+      "epoch": 0.12420229190969602,
+      "eval_accuracy": 0.5627916456916503,
+      "eval_loss": 1.96902596950531,
+      "eval_runtime": 2194.8971,
+      "eval_samples_per_second": 13.414,
+      "eval_steps_per_second": 1.677,
+      "step": 1810
+    },
+    {
+      "epoch": 0.12488849241748438,
+      "grad_norm": 0.6445093154907227,
+      "learning_rate": 0.00016623970722781336,
+      "loss": 1.8726,
+      "step": 1820
+    },
+    {
+      "epoch": 0.12488849241748438,
+      "eval_accuracy": 0.5629212641545254,
+      "eval_loss": 1.9668747186660767,
+      "eval_runtime": 2195.3596,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 1820
+    },
+    {
+      "epoch": 0.12557469292527276,
+      "grad_norm": 0.7446492314338684,
+      "learning_rate": 0.00016715462031107045,
+      "loss": 1.9153,
+      "step": 1830
+    },
+    {
+      "epoch": 0.12557469292527276,
+      "eval_accuracy": 0.5622602564758932,
+      "eval_loss": 1.9712408781051636,
+      "eval_runtime": 2195.4086,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 1830
+    },
+    {
+      "epoch": 0.12626089343306113,
+      "grad_norm": 0.6903027296066284,
+      "learning_rate": 0.00016806953339432753,
+      "loss": 1.9477,
+      "step": 1840
+    },
+    {
+      "epoch": 0.12626089343306113,
+      "eval_accuracy": 0.5628142226780117,
+      "eval_loss": 1.9705018997192383,
+      "eval_runtime": 2195.3561,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 1840
+    },
+    {
+      "epoch": 0.12694709394084952,
+      "grad_norm": 0.6859719753265381,
+      "learning_rate": 0.00016898444647758465,
+      "loss": 1.956,
+      "step": 1850
+    },
+    {
+      "epoch": 0.12694709394084952,
+      "eval_accuracy": 0.5628691378771908,
+      "eval_loss": 1.9665277004241943,
+      "eval_runtime": 2196.295,
+      "eval_samples_per_second": 13.405,
+      "eval_steps_per_second": 1.676,
+      "step": 1850
+    },
+    {
+      "epoch": 0.12763329444863789,
+      "grad_norm": 0.6945796012878418,
+      "learning_rate": 0.00016989935956084173,
+      "loss": 1.9526,
+      "step": 1860
+    },
+    {
+      "epoch": 0.12763329444863789,
+      "eval_accuracy": 0.5613948274663382,
+      "eval_loss": 1.9775313138961792,
+      "eval_runtime": 2195.3271,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 1860
+    },
+    {
+      "epoch": 0.12831949495642628,
+      "grad_norm": 0.692680299282074,
+      "learning_rate": 0.00017081427264409882,
+      "loss": 1.9693,
+      "step": 1870
+    },
+    {
+      "epoch": 0.12831949495642628,
+      "eval_accuracy": 0.5631267811333156,
+      "eval_loss": 1.9690190553665161,
+      "eval_runtime": 2196.3618,
+      "eval_samples_per_second": 13.405,
+      "eval_steps_per_second": 1.676,
+      "step": 1870
+    },
+    {
+      "epoch": 0.12900569546421464,
+      "grad_norm": 0.9683026671409607,
+      "learning_rate": 0.0001717291857273559,
+      "loss": 1.9857,
+      "step": 1880
+    },
+    {
+      "epoch": 0.12900569546421464,
+      "eval_accuracy": 0.5625857302954538,
+      "eval_loss": 1.9720834493637085,
+      "eval_runtime": 2194.9732,
+      "eval_samples_per_second": 13.413,
+      "eval_steps_per_second": 1.677,
+      "step": 1880
+    },
+    {
+      "epoch": 0.129691895972003,
+      "grad_norm": 0.9663318991661072,
+      "learning_rate": 0.00017264409881061302,
+      "loss": 1.982,
+      "step": 1890
+    },
+    {
+      "epoch": 0.129691895972003,
+      "eval_accuracy": 0.5605411517702714,
+      "eval_loss": 1.980215311050415,
+      "eval_runtime": 2195.0705,
+      "eval_samples_per_second": 13.413,
+      "eval_steps_per_second": 1.677,
+      "step": 1890
+    },
+    {
+      "epoch": 0.1303780964797914,
+      "grad_norm": 0.9997875690460205,
+      "learning_rate": 0.00017355901189387008,
+      "loss": 1.9939,
+      "step": 1900
+    },
+    {
+      "epoch": 0.1303780964797914,
+      "eval_accuracy": 0.5622947527829953,
+      "eval_loss": 1.9729753732681274,
+      "eval_runtime": 2195.0496,
+      "eval_samples_per_second": 13.413,
+      "eval_steps_per_second": 1.677,
+      "step": 1900
+    },
+    {
+      "epoch": 0.13106429698757976,
+      "grad_norm": 1.0369981527328491,
+      "learning_rate": 0.00017447392497712717,
+      "loss": 1.9709,
+      "step": 1910
+    },
+    {
+      "epoch": 0.13106429698757976,
+      "eval_accuracy": 0.5614108305654945,
+      "eval_loss": 1.9774330854415894,
+      "eval_runtime": 2195.0243,
+      "eval_samples_per_second": 13.413,
+      "eval_steps_per_second": 1.677,
+      "step": 1910
+    },
+    {
+      "epoch": 0.13175049749536816,
+      "grad_norm": 0.7212566137313843,
+      "learning_rate": 0.00017538883806038425,
+      "loss": 2.0167,
+      "step": 1920
+    },
+    {
+      "epoch": 0.13175049749536816,
+      "eval_accuracy": 0.5607967697379137,
+      "eval_loss": 1.984305739402771,
+      "eval_runtime": 2194.9072,
+      "eval_samples_per_second": 13.414,
+      "eval_steps_per_second": 1.677,
+      "step": 1920
+    },
+    {
+      "epoch": 0.13243669800315652,
+      "grad_norm": 0.709562361240387,
+      "learning_rate": 0.00017630375114364137,
+      "loss": 1.9576,
+      "step": 1930
+    },
+    {
+      "epoch": 0.13243669800315652,
+      "eval_accuracy": 0.5579354355296557,
+      "eval_loss": 2.008889675140381,
+      "eval_runtime": 2195.0187,
+      "eval_samples_per_second": 13.413,
+      "eval_steps_per_second": 1.677,
+      "step": 1930
+    },
+    {
+      "epoch": 0.1331228985109449,
+      "grad_norm": 5.758970260620117,
+      "learning_rate": 0.00017721866422689846,
+      "loss": 2.0477,
+      "step": 1940
+    },
+    {
+      "epoch": 0.1331228985109449,
+      "eval_accuracy": 0.5608241941360528,
+      "eval_loss": 1.9867639541625977,
+      "eval_runtime": 2193.1512,
+      "eval_samples_per_second": 13.425,
+      "eval_steps_per_second": 1.678,
+      "step": 1940
+    },
+    {
+      "epoch": 0.13380909901873328,
+      "grad_norm": 0.6865282654762268,
+      "learning_rate": 0.00017813357731015554,
+      "loss": 2.0013,
+      "step": 1950
+    },
+    {
+      "epoch": 0.13380909901873328,
+      "eval_accuracy": 0.5611121171150623,
+      "eval_loss": 1.9813686609268188,
+      "eval_runtime": 2191.7839,
+      "eval_samples_per_second": 13.433,
+      "eval_steps_per_second": 1.679,
+      "step": 1950
+    },
+    {
+      "epoch": 0.13449529952652164,
+      "grad_norm": 1.0006988048553467,
+      "learning_rate": 0.00017904849039341263,
+      "loss": 1.9151,
+      "step": 1960
+    },
+    {
+      "epoch": 0.13449529952652164,
+      "eval_accuracy": 0.5620602177364407,
+      "eval_loss": 1.9793142080307007,
+      "eval_runtime": 2191.7062,
+      "eval_samples_per_second": 13.433,
+      "eval_steps_per_second": 1.68,
+      "step": 1960
+    },
+    {
+      "epoch": 0.13518150003431004,
+      "grad_norm": 25.0100040435791,
+      "learning_rate": 0.00017996340347666974,
+      "loss": 2.0365,
+      "step": 1970
+    },
+    {
+      "epoch": 0.13518150003431004,
+      "eval_accuracy": 0.5600259980638242,
+      "eval_loss": 1.9854369163513184,
+      "eval_runtime": 2191.9684,
+      "eval_samples_per_second": 13.432,
+      "eval_steps_per_second": 1.679,
+      "step": 1970
+    },
+    {
+      "epoch": 0.1358677005420984,
+      "grad_norm": 0.6420261859893799,
+      "learning_rate": 0.00018087831655992683,
+      "loss": 2.0143,
+      "step": 1980
+    },
+    {
+      "epoch": 0.1358677005420984,
+      "eval_accuracy": 0.5613296530189448,
+      "eval_loss": 1.9770421981811523,
+      "eval_runtime": 2191.2212,
+      "eval_samples_per_second": 13.436,
+      "eval_steps_per_second": 1.68,
+      "step": 1980
+    },
+    {
+      "epoch": 0.13655390104988677,
+      "grad_norm": 0.7176883816719055,
+      "learning_rate": 0.00018179322964318392,
+      "loss": 1.9782,
+      "step": 1990
+    },
+    {
+      "epoch": 0.13655390104988677,
+      "eval_accuracy": 0.5617598442134819,
+      "eval_loss": 1.980700135231018,
+      "eval_runtime": 2191.7861,
+      "eval_samples_per_second": 13.433,
+      "eval_steps_per_second": 1.679,
+      "step": 1990
+    },
+    {
+      "epoch": 0.13724010155767516,
+      "grad_norm": 0.6866204738616943,
+      "learning_rate": 0.00018270814272644098,
+      "loss": 1.9449,
+      "step": 2000
+    },
+    {
+      "epoch": 0.13724010155767516,
+      "eval_accuracy": 0.5615476869445853,
+      "eval_loss": 1.9827600717544556,
+      "eval_runtime": 2193.8402,
+      "eval_samples_per_second": 13.42,
+      "eval_steps_per_second": 1.678,
+      "step": 2000
+    },
+    {
+      "epoch": 0.13792630206546352,
+      "grad_norm": 0.6652859449386597,
+      "learning_rate": 0.0001836230558096981,
+      "loss": 1.9812,
+      "step": 2010
+    },
+    {
+      "epoch": 0.13792630206546352,
+      "eval_accuracy": 0.5610583639666517,
+      "eval_loss": 1.9824488162994385,
+      "eval_runtime": 2198.5601,
+      "eval_samples_per_second": 13.391,
+      "eval_steps_per_second": 1.674,
+      "step": 2010
+    },
+    {
+      "epoch": 0.13861250257325192,
+      "grad_norm": 0.712452232837677,
+      "learning_rate": 0.00018453796889295518,
+      "loss": 2.0223,
+      "step": 2020
+    },
+    {
+      "epoch": 0.13861250257325192,
+      "eval_accuracy": 0.5598986372995852,
+      "eval_loss": 1.9870091676712036,
+      "eval_runtime": 2199.0783,
+      "eval_samples_per_second": 13.388,
+      "eval_steps_per_second": 1.674,
+      "step": 2020
+    },
+    {
+      "epoch": 0.13929870308104028,
+      "grad_norm": 0.6659247875213623,
+      "learning_rate": 0.00018545288197621226,
+      "loss": 2.0113,
+      "step": 2030
+    },
+    {
+      "epoch": 0.13929870308104028,
+      "eval_accuracy": 0.560590721535915,
+      "eval_loss": 1.983353614807129,
+      "eval_runtime": 2195.4289,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 2030
+    },
+    {
+      "epoch": 0.13998490358882865,
+      "grad_norm": 0.7091332674026489,
+      "learning_rate": 0.00018636779505946935,
+      "loss": 2.0452,
+      "step": 2040
+    },
+    {
+      "epoch": 0.13998490358882865,
+      "eval_accuracy": 0.5607067938069733,
+      "eval_loss": 1.9845926761627197,
+      "eval_runtime": 2196.3562,
+      "eval_samples_per_second": 13.405,
+      "eval_steps_per_second": 1.676,
+      "step": 2040
+    },
+    {
+      "epoch": 0.14067110409661704,
+      "grad_norm": 0.6729557514190674,
+      "learning_rate": 0.00018728270814272646,
+      "loss": 1.9922,
+      "step": 2050
+    },
+    {
+      "epoch": 0.14067110409661704,
+      "eval_accuracy": 0.5608056013237551,
+      "eval_loss": 1.9840885400772095,
+      "eval_runtime": 2204.0084,
+      "eval_samples_per_second": 13.358,
+      "eval_steps_per_second": 1.67,
+      "step": 2050
+    },
+    {
+      "epoch": 0.1413573046044054,
+      "grad_norm": 0.6952201128005981,
+      "learning_rate": 0.00018819762122598355,
+      "loss": 1.9517,
+      "step": 2060
+    },
+    {
+      "epoch": 0.1413573046044054,
+      "eval_accuracy": 0.5593980922313718,
+      "eval_loss": 1.9891562461853027,
+      "eval_runtime": 2195.6228,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.677,
+      "step": 2060
+    },
+    {
+      "epoch": 0.1420435051121938,
+      "grad_norm": 0.7625614404678345,
+      "learning_rate": 0.00018911253430924064,
+      "loss": 1.9531,
+      "step": 2070
+    },
+    {
+      "epoch": 0.1420435051121938,
+      "eval_accuracy": 0.5598096242107102,
+      "eval_loss": 1.9889034032821655,
+      "eval_runtime": 2200.6199,
+      "eval_samples_per_second": 13.379,
+      "eval_steps_per_second": 1.673,
+      "step": 2070
+    },
+    {
+      "epoch": 0.14272970561998216,
+      "grad_norm": 0.7365880608558655,
+      "learning_rate": 0.00019002744739249773,
+      "loss": 1.9819,
+      "step": 2080
+    },
+    {
+      "epoch": 0.14272970561998216,
+      "eval_accuracy": 0.55980836255559,
+      "eval_loss": 1.9877856969833374,
+      "eval_runtime": 2198.0873,
+      "eval_samples_per_second": 13.394,
+      "eval_steps_per_second": 1.675,
+      "step": 2080
+    },
+    {
+      "epoch": 0.14341590612777053,
+      "grad_norm": 0.7005255818367004,
+      "learning_rate": 0.0001909423604757548,
+      "loss": 1.9977,
+      "step": 2090
+    },
+    {
+      "epoch": 0.14341590612777053,
+      "eval_accuracy": 0.5598768903494871,
+      "eval_loss": 1.9907660484313965,
+      "eval_runtime": 2194.9399,
+      "eval_samples_per_second": 13.414,
+      "eval_steps_per_second": 1.677,
+      "step": 2090
+    },
+    {
+      "epoch": 0.14410210663555892,
+      "grad_norm": 0.6998113393783569,
+      "learning_rate": 0.0001918572735590119,
+      "loss": 1.983,
+      "step": 2100
+    },
+    {
+      "epoch": 0.14410210663555892,
+      "eval_accuracy": 0.5593210980675892,
+      "eval_loss": 1.9914487600326538,
+      "eval_runtime": 2195.2308,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 2100
+    },
+    {
+      "epoch": 0.14478830714334728,
+      "grad_norm": 1.0305057764053345,
+      "learning_rate": 0.00019277218664226899,
+      "loss": 2.0584,
+      "step": 2110
+    },
+    {
+      "epoch": 0.14478830714334728,
+      "eval_accuracy": 0.5584136028202109,
+      "eval_loss": 1.9976584911346436,
+      "eval_runtime": 2195.1184,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 2110
+    },
+    {
+      "epoch": 0.14547450765113565,
+      "grad_norm": 0.6542972922325134,
+      "learning_rate": 0.00019368709972552607,
+      "loss": 1.9884,
+      "step": 2120
+    },
+    {
+      "epoch": 0.14547450765113565,
+      "eval_accuracy": 0.5590180684285879,
+      "eval_loss": 1.9945608377456665,
+      "eval_runtime": 2196.4694,
+      "eval_samples_per_second": 13.404,
+      "eval_steps_per_second": 1.676,
+      "step": 2120
+    },
+    {
+      "epoch": 0.14616070815892404,
+      "grad_norm": 0.6551515460014343,
+      "learning_rate": 0.0001946020128087832,
+      "loss": 1.9907,
+      "step": 2130
+    },
+    {
+      "epoch": 0.14616070815892404,
+      "eval_accuracy": 0.5585770535611776,
+      "eval_loss": 1.996436357498169,
+      "eval_runtime": 2195.8503,
+      "eval_samples_per_second": 13.408,
+      "eval_steps_per_second": 1.676,
+      "step": 2130
+    },
+    {
+      "epoch": 0.1468469086667124,
+      "grad_norm": 0.7233726382255554,
+      "learning_rate": 0.00019551692589204027,
+      "loss": 1.8951,
+      "step": 2140
+    },
+    {
+      "epoch": 0.1468469086667124,
+      "eval_accuracy": 0.5590956602184801,
+      "eval_loss": 1.994140625,
+      "eval_runtime": 2198.6021,
+      "eval_samples_per_second": 13.391,
+      "eval_steps_per_second": 1.674,
+      "step": 2140
+    },
+    {
+      "epoch": 0.1475331091745008,
+      "grad_norm": 0.6772713661193848,
+      "learning_rate": 0.00019643183897529736,
+      "loss": 1.9488,
+      "step": 2150
+    },
+    {
+      "epoch": 0.1475331091745008,
+      "eval_accuracy": 0.558802757021891,
+      "eval_loss": 1.9978493452072144,
+      "eval_runtime": 2198.8115,
+      "eval_samples_per_second": 13.39,
+      "eval_steps_per_second": 1.674,
+      "step": 2150
+    },
+    {
+      "epoch": 0.14821930968228916,
+      "grad_norm": NaN,
+      "learning_rate": 0.00019734675205855445,
+      "loss": 2.0301,
+      "step": 2160
+    },
+    {
+      "epoch": 0.14821930968228916,
+      "eval_accuracy": 0.5583828582770187,
+      "eval_loss": 2.0025622844696045,
+      "eval_runtime": 2195.3189,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 2160
+    },
+    {
+      "epoch": 0.14890551019007753,
+      "grad_norm": 0.7715523838996887,
+      "learning_rate": 0.00019817017383348584,
+      "loss": 2.0215,
+      "step": 2170
+    },
+    {
+      "epoch": 0.14890551019007753,
+      "eval_accuracy": 0.5581144909523723,
+      "eval_loss": 2.007227659225464,
+      "eval_runtime": 2201.5716,
+      "eval_samples_per_second": 13.373,
+      "eval_steps_per_second": 1.672,
+      "step": 2170
+    },
+    {
+      "epoch": 0.14959171069786592,
+      "grad_norm": 0.7113980650901794,
+      "learning_rate": 0.00019908508691674292,
+      "loss": 1.9645,
+      "step": 2180
+    },
+    {
+      "epoch": 0.14959171069786592,
+      "eval_accuracy": 0.5589439627910016,
+      "eval_loss": 1.9947305917739868,
+      "eval_runtime": 2196.025,
+      "eval_samples_per_second": 13.407,
+      "eval_steps_per_second": 1.676,
+      "step": 2180
+    },
+    {
+      "epoch": 0.15027791120565429,
+      "grad_norm": 0.6457582712173462,
+      "learning_rate": 0.0002,
+      "loss": 1.9758,
+      "step": 2190
+    },
+    {
+      "epoch": 0.15027791120565429,
+      "eval_accuracy": 0.5590743780886894,
+      "eval_loss": 1.9934245347976685,
+      "eval_runtime": 2196.674,
+      "eval_samples_per_second": 13.403,
+      "eval_steps_per_second": 1.676,
+      "step": 2190
+    },
+    {
+      "epoch": 0.15096411171344268,
+      "grad_norm": 0.5914057493209839,
+      "learning_rate": 0.00019995184552042955,
+      "loss": 1.9647,
+      "step": 2200
+    },
+    {
+      "epoch": 0.15096411171344268,
+      "eval_accuracy": 0.5594601789438658,
+      "eval_loss": 1.9928838014602661,
+      "eval_runtime": 2200.2943,
+      "eval_samples_per_second": 13.381,
+      "eval_steps_per_second": 1.673,
+      "step": 2200
+    },
+    {
+      "epoch": 0.15165031222123104,
+      "grad_norm": 0.6294174194335938,
+      "learning_rate": 0.00019990369104085908,
+      "loss": 1.9752,
+      "step": 2210
+    },
+    {
+      "epoch": 0.15165031222123104,
+      "eval_accuracy": 0.5588302478229311,
+      "eval_loss": 1.9968582391738892,
+      "eval_runtime": 2195.8847,
+      "eval_samples_per_second": 13.408,
+      "eval_steps_per_second": 1.676,
+      "step": 2210
+    },
+    {
+      "epoch": 0.1523365127290194,
+      "grad_norm": 0.6448546051979065,
+      "learning_rate": 0.00019985553656128862,
+      "loss": 1.9605,
+      "step": 2220
+    },
+    {
+      "epoch": 0.1523365127290194,
+      "eval_accuracy": 0.5588568089833563,
+      "eval_loss": 1.9973441362380981,
+      "eval_runtime": 2200.5474,
+      "eval_samples_per_second": 13.379,
+      "eval_steps_per_second": 1.673,
+      "step": 2220
+    },
+    {
+      "epoch": 0.1530227132368078,
+      "grad_norm": 0.8129657506942749,
+      "learning_rate": 0.00019980738208171816,
+      "loss": 2.0269,
+      "step": 2230
+    },
+    {
+      "epoch": 0.1530227132368078,
+      "eval_accuracy": 0.5584346525398479,
+      "eval_loss": 1.997729778289795,
+      "eval_runtime": 2196.0138,
+      "eval_samples_per_second": 13.407,
+      "eval_steps_per_second": 1.676,
+      "step": 2230
+    },
+    {
+      "epoch": 0.15370891374459617,
+      "grad_norm": 0.5907541513442993,
+      "learning_rate": 0.0001997592276021477,
+      "loss": 1.9763,
+      "step": 2240
+    },
+    {
+      "epoch": 0.15370891374459617,
+      "eval_accuracy": 0.5583808329885362,
+      "eval_loss": 2.0000393390655518,
+      "eval_runtime": 2195.0414,
+      "eval_samples_per_second": 13.413,
+      "eval_steps_per_second": 1.677,
+      "step": 2240
+    },
+    {
+      "epoch": 0.15439511425238456,
+      "grad_norm": 0.6172444820404053,
+      "learning_rate": 0.00019971107312257723,
+      "loss": 1.961,
+      "step": 2250
+    },
+    {
+      "epoch": 0.15439511425238456,
+      "eval_accuracy": 0.5588485086207234,
+      "eval_loss": 1.9951170682907104,
+      "eval_runtime": 2195.2092,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 2250
+    },
+    {
+      "epoch": 0.15508131476017292,
+      "grad_norm": 1.708965539932251,
+      "learning_rate": 0.00019966291864300677,
+      "loss": 1.9676,
+      "step": 2260
+    },
+    {
+      "epoch": 0.15508131476017292,
+      "eval_accuracy": 0.5570037696262905,
+      "eval_loss": 2.0055410861968994,
+      "eval_runtime": 2195.1746,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 2260
+    },
+    {
+      "epoch": 0.1557675152679613,
+      "grad_norm": 0.7572776079177856,
+      "learning_rate": 0.0001996147641634363,
+      "loss": 1.9993,
+      "step": 2270
+    },
+    {
+      "epoch": 0.1557675152679613,
+      "eval_accuracy": 0.5573890724597089,
+      "eval_loss": 1.9998122453689575,
+      "eval_runtime": 2195.7893,
+      "eval_samples_per_second": 13.408,
+      "eval_steps_per_second": 1.676,
+      "step": 2270
+    },
+    {
+      "epoch": 0.15645371577574968,
+      "grad_norm": 0.6893654465675354,
+      "learning_rate": 0.00019956660968386585,
+      "loss": 1.949,
+      "step": 2280
+    },
+    {
+      "epoch": 0.15645371577574968,
+      "eval_accuracy": 0.5586115498682799,
+      "eval_loss": 1.9959582090377808,
+      "eval_runtime": 2195.3478,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 2280
+    },
+    {
+      "epoch": 0.15713991628353804,
+      "grad_norm": 0.6196193695068359,
+      "learning_rate": 0.00019951845520429538,
+      "loss": 2.0126,
+      "step": 2290
+    },
+    {
+      "epoch": 0.15713991628353804,
+      "eval_accuracy": 0.5588276249083391,
+      "eval_loss": 1.9977123737335205,
+      "eval_runtime": 2196.787,
+      "eval_samples_per_second": 13.402,
+      "eval_steps_per_second": 1.676,
+      "step": 2290
+    },
+    {
+      "epoch": 0.15782611679132644,
+      "grad_norm": 5.14415979385376,
+      "learning_rate": 0.00019947030072472492,
+      "loss": 2.0028,
+      "step": 2300
+    },
+    {
+      "epoch": 0.15782611679132644,
+      "eval_accuracy": 0.5582278739059375,
+      "eval_loss": 2.0048232078552246,
+      "eval_runtime": 2196.1102,
+      "eval_samples_per_second": 13.406,
+      "eval_steps_per_second": 1.676,
+      "step": 2300
+    },
+    {
+      "epoch": 0.1585123172991148,
+      "grad_norm": 0.6127263307571411,
+      "learning_rate": 0.00019942214624515448,
+      "loss": 2.0287,
+      "step": 2310
+    },
+    {
+      "epoch": 0.1585123172991148,
+      "eval_accuracy": 0.5586611860368246,
+      "eval_loss": 1.995181679725647,
+      "eval_runtime": 2195.907,
+      "eval_samples_per_second": 13.408,
+      "eval_steps_per_second": 1.676,
+      "step": 2310
+    },
+    {
+      "epoch": 0.15919851780690317,
+      "grad_norm": 0.6055991649627686,
+      "learning_rate": 0.000199373991765584,
+      "loss": 1.9834,
+      "step": 2320
+    },
+    {
+      "epoch": 0.15919851780690317,
+      "eval_accuracy": 0.5592904863301992,
+      "eval_loss": 1.9946677684783936,
+      "eval_runtime": 2195.4572,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 2320
+    },
+    {
+      "epoch": 0.15988471831469156,
+      "grad_norm": 0.6904438734054565,
+      "learning_rate": 0.00019932583728601356,
+      "loss": 1.98,
+      "step": 2330
+    },
+    {
+      "epoch": 0.15988471831469156,
+      "eval_accuracy": 0.5583102135032557,
+      "eval_loss": 1.9992380142211914,
+      "eval_runtime": 2195.7129,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.676,
+      "step": 2330
+    },
+    {
+      "epoch": 0.16057091882247992,
+      "grad_norm": 0.7183199524879456,
+      "learning_rate": 0.00019927768280644307,
+      "loss": 2.0404,
+      "step": 2340
+    },
+    {
+      "epoch": 0.16057091882247992,
+      "eval_accuracy": 0.5585802741018792,
+      "eval_loss": 1.9996939897537231,
+      "eval_runtime": 2195.2538,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 2340
+    },
+    {
+      "epoch": 0.16125711933026832,
+      "grad_norm": 0.6204356551170349,
+      "learning_rate": 0.00019922952832687263,
+      "loss": 1.9546,
+      "step": 2350
+    },
+    {
+      "epoch": 0.16125711933026832,
+      "eval_accuracy": 0.5581378647735465,
+      "eval_loss": 1.9987818002700806,
+      "eval_runtime": 2195.0806,
+      "eval_samples_per_second": 13.413,
+      "eval_steps_per_second": 1.677,
+      "step": 2350
+    },
+    {
+      "epoch": 0.16194331983805668,
+      "grad_norm": 0.6199485063552856,
+      "learning_rate": 0.00019918137384730214,
+      "loss": 1.9747,
+      "step": 2360
+    },
+    {
+      "epoch": 0.16194331983805668,
+      "eval_accuracy": 0.558920688574179,
+      "eval_loss": 1.9964611530303955,
+      "eval_runtime": 2195.364,
+      "eval_samples_per_second": 13.411,
+      "eval_steps_per_second": 1.677,
+      "step": 2360
+    },
+    {
+      "epoch": 0.16262952034584505,
+      "grad_norm": 0.6137880086898804,
+      "learning_rate": 0.00019913321936773168,
+      "loss": 1.9203,
+      "step": 2370
+    },
+    {
+      "epoch": 0.16262952034584505,
+      "eval_accuracy": 0.5578602342442018,
+      "eval_loss": 1.9984673261642456,
+      "eval_runtime": 2195.4886,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 2370
+    },
+    {
+      "epoch": 0.16331572085363344,
+      "grad_norm": 0.6386215090751648,
+      "learning_rate": 0.00019908506488816125,
+      "loss": 2.0383,
+      "step": 2380
+    },
+    {
+      "epoch": 0.16331572085363344,
+      "eval_accuracy": 0.559068036611638,
+      "eval_loss": 1.9952830076217651,
+      "eval_runtime": 2195.2671,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 2380
+    },
+    {
+      "epoch": 0.1640019213614218,
+      "grad_norm": 0.6018114686012268,
+      "learning_rate": 0.00019903691040859076,
+      "loss": 1.9676,
+      "step": 2390
+    },
+    {
+      "epoch": 0.1640019213614218,
+      "eval_accuracy": 0.5586635433398123,
+      "eval_loss": 1.9960980415344238,
+      "eval_runtime": 2197.0618,
+      "eval_samples_per_second": 13.401,
+      "eval_steps_per_second": 1.675,
+      "step": 2390
+    },
+    {
+      "epoch": 0.1646881218692102,
+      "grad_norm": 0.6331018805503845,
+      "learning_rate": 0.00019898875592902032,
+      "loss": 1.9598,
+      "step": 2400
+    },
+    {
+      "epoch": 0.1646881218692102,
+      "eval_accuracy": 0.5588065419872515,
+      "eval_loss": 1.9976948499679565,
+      "eval_runtime": 2195.2367,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 2400
+    },
+    {
+      "epoch": 0.16537432237699856,
+      "grad_norm": 0.6709062457084656,
+      "learning_rate": 0.00019894060144944983,
+      "loss": 1.9985,
+      "step": 2410
+    },
+    {
+      "epoch": 0.16537432237699856,
+      "eval_accuracy": 0.5586920633858189,
+      "eval_loss": 1.9963314533233643,
+      "eval_runtime": 2195.5736,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 2410
+    },
+    {
+      "epoch": 0.16606052288478693,
+      "grad_norm": 0.6212249398231506,
+      "learning_rate": 0.0001988924469698794,
+      "loss": 1.9174,
+      "step": 2420
+    },
+    {
+      "epoch": 0.16606052288478693,
+      "eval_accuracy": 0.5585765223379692,
+      "eval_loss": 1.9955825805664062,
+      "eval_runtime": 2195.9048,
+      "eval_samples_per_second": 13.408,
+      "eval_steps_per_second": 1.676,
+      "step": 2420
+    },
+    {
+      "epoch": 0.16674672339257532,
+      "grad_norm": 0.5908493399620056,
+      "learning_rate": 0.0001988442924903089,
+      "loss": 1.9768,
+      "step": 2430
+    },
+    {
+      "epoch": 0.16674672339257532,
+      "eval_accuracy": 0.5599598607743654,
+      "eval_loss": 1.996363878250122,
+      "eval_runtime": 2196.1829,
+      "eval_samples_per_second": 13.406,
+      "eval_steps_per_second": 1.676,
+      "step": 2430
+    },
+    {
+      "epoch": 0.16743292390036368,
+      "grad_norm": 0.6591713428497314,
+      "learning_rate": 0.00019879613801073847,
+      "loss": 1.9607,
+      "step": 2440
+    },
+    {
+      "epoch": 0.16743292390036368,
+      "eval_accuracy": 0.5579004412007955,
+      "eval_loss": 1.9992504119873047,
+      "eval_runtime": 2195.5161,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 2440
+    },
+    {
+      "epoch": 0.16811912440815205,
+      "grad_norm": 0.6437056660652161,
+      "learning_rate": 0.00019874798353116798,
+      "loss": 2.0276,
+      "step": 2450
+    },
+    {
+      "epoch": 0.16811912440815205,
+      "eval_accuracy": 0.5586110186450713,
+      "eval_loss": 1.9952548742294312,
+      "eval_runtime": 2195.2496,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 2450
+    },
+    {
+      "epoch": 0.16880532491594044,
+      "grad_norm": 0.564586877822876,
+      "learning_rate": 0.00019869982905159754,
+      "loss": 1.9414,
+      "step": 2460
+    },
+    {
+      "epoch": 0.16880532491594044,
+      "eval_accuracy": 0.5588329039389736,
+      "eval_loss": 1.9931731224060059,
+      "eval_runtime": 2195.1623,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 2460
+    },
+    {
+      "epoch": 0.1694915254237288,
+      "grad_norm": 0.6013181805610657,
+      "learning_rate": 0.00019865167457202708,
+      "loss": 1.9564,
+      "step": 2470
+    },
+    {
+      "epoch": 0.1694915254237288,
+      "eval_accuracy": 0.5598632777547692,
+      "eval_loss": 1.9917162656784058,
+      "eval_runtime": 2196.6864,
+      "eval_samples_per_second": 13.403,
+      "eval_steps_per_second": 1.676,
+      "step": 2470
+    },
+    {
+      "epoch": 0.1701777259315172,
+      "grad_norm": 0.6139594912528992,
+      "learning_rate": 0.00019860352009245662,
+      "loss": 1.9678,
+      "step": 2480
+    },
+    {
+      "epoch": 0.1701777259315172,
+      "eval_accuracy": 0.5594104431709697,
+      "eval_loss": 1.9919596910476685,
+      "eval_runtime": 2195.2781,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 2480
+    },
+    {
+      "epoch": 0.17086392643930556,
+      "grad_norm": 0.6118595600128174,
+      "learning_rate": 0.00019855536561288616,
+      "loss": 1.9903,
+      "step": 2490
+    },
+    {
+      "epoch": 0.17086392643930556,
+      "eval_accuracy": 0.5595184142880981,
+      "eval_loss": 1.9930351972579956,
+      "eval_runtime": 2195.0637,
+      "eval_samples_per_second": 13.413,
+      "eval_steps_per_second": 1.677,
+      "step": 2490
+    },
+    {
+      "epoch": 0.17155012694709393,
+      "grad_norm": 0.6277731657028198,
+      "learning_rate": 0.0001985072111333157,
+      "loss": 1.9994,
+      "step": 2500
+    },
+    {
+      "epoch": 0.17155012694709393,
+      "eval_accuracy": 0.5604368328127014,
+      "eval_loss": 1.9892007112503052,
+      "eval_runtime": 2196.3916,
+      "eval_samples_per_second": 13.405,
+      "eval_steps_per_second": 1.676,
+      "step": 2500
+    },
+    {
+      "epoch": 0.17223632745488232,
+      "grad_norm": 0.6022769808769226,
+      "learning_rate": 0.00019845905665374523,
+      "loss": 1.955,
+      "step": 2510
+    },
+    {
+      "epoch": 0.17223632745488232,
+      "eval_accuracy": 0.5598575671052778,
+      "eval_loss": 1.9900598526000977,
+      "eval_runtime": 2195.1324,
+      "eval_samples_per_second": 13.412,
+      "eval_steps_per_second": 1.677,
+      "step": 2510
+    },
+    {
+      "epoch": 0.17292252796267069,
+      "grad_norm": 0.7676475644111633,
+      "learning_rate": 0.00019841090217417477,
+      "loss": 1.9778,
+      "step": 2520
+    },
+    {
+      "epoch": 0.17292252796267069,
+      "eval_accuracy": 0.5599910369364145,
+      "eval_loss": 1.9912259578704834,
+      "eval_runtime": 2195.5014,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 2520
+    },
+    {
+      "epoch": 0.17360872847045908,
+      "grad_norm": 0.6049858331680298,
+      "learning_rate": 0.0001983627476946043,
+      "loss": 2.0018,
+      "step": 2530
+    },
+    {
+      "epoch": 0.17360872847045908,
+      "eval_accuracy": 0.5597790788762212,
+      "eval_loss": 1.9898487329483032,
+      "eval_runtime": 2196.3189,
+      "eval_samples_per_second": 13.405,
+      "eval_steps_per_second": 1.676,
+      "step": 2530
+    },
+    {
+      "epoch": 0.17429492897824744,
+      "grad_norm": 0.597239077091217,
+      "learning_rate": 0.00019831459321503382,
+      "loss": 1.9675,
+      "step": 2540
+    },
+    {
+      "epoch": 0.17429492897824744,
+      "eval_accuracy": 0.5608272154680511,
+      "eval_loss": 1.987794041633606,
+      "eval_runtime": 2196.2383,
+      "eval_samples_per_second": 13.406,
+      "eval_steps_per_second": 1.676,
+      "step": 2540
+    },
+    {
+      "epoch": 0.1749811294860358,
+      "grad_norm": 0.6299741864204407,
+      "learning_rate": 0.00019826643873546338,
+      "loss": 2.0204,
+      "step": 2550
+    },
+    {
+      "epoch": 0.1749811294860358,
+      "eval_accuracy": 0.5596887709307754,
+      "eval_loss": 1.9905835390090942,
+      "eval_runtime": 2196.3707,
+      "eval_samples_per_second": 13.405,
+      "eval_steps_per_second": 1.676,
+      "step": 2550
+    },
+    {
+      "epoch": 0.1756673299938242,
+      "grad_norm": 0.6488902568817139,
+      "learning_rate": 0.00019821828425589292,
+      "loss": 1.9863,
+      "step": 2560
+    },
+    {
+      "epoch": 0.1756673299938242,
+      "eval_accuracy": 0.5600538872822707,
+      "eval_loss": 1.9906458854675293,
+      "eval_runtime": 2195.6961,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.676,
+      "step": 2560
+    },
+    {
+      "epoch": 0.17635353050161257,
+      "grad_norm": 0.6427273154258728,
+      "learning_rate": 0.00019817012977632246,
+      "loss": 1.9403,
+      "step": 2570
+    },
+    {
+      "epoch": 0.17635353050161257,
+      "eval_accuracy": 0.5603546260211854,
+      "eval_loss": 1.989209532737732,
+      "eval_runtime": 2196.6688,
+      "eval_samples_per_second": 13.403,
+      "eval_steps_per_second": 1.676,
+      "step": 2570
+    },
+    {
+      "epoch": 0.17703973100940096,
+      "grad_norm": 0.5979894399642944,
+      "learning_rate": 0.000198121975296752,
+      "loss": 1.9437,
+      "step": 2580
+    },
+    {
+      "epoch": 0.17703973100940096,
+      "eval_accuracy": 0.5593561587993505,
+      "eval_loss": 1.9925955533981323,
+      "eval_runtime": 2195.5236,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 2580
+    },
+    {
+      "epoch": 0.17772593151718932,
+      "grad_norm": 0.6633705496788025,
+      "learning_rate": 0.00019807382081718153,
+      "loss": 1.9804,
+      "step": 2590
+    },
+    {
+      "epoch": 0.17772593151718932,
+      "eval_accuracy": 0.5602276636743527,
+      "eval_loss": 1.988720178604126,
+      "eval_runtime": 2195.5843,
+      "eval_samples_per_second": 13.41,
+      "eval_steps_per_second": 1.677,
+      "step": 2590
+    },
+    {
+      "epoch": 0.1784121320249777,
+      "grad_norm": 0.5960366129875183,
+      "learning_rate": 0.00019802566633761107,
+      "loss": 1.9351,
+      "step": 2600
+    },
+    {
+      "epoch": 0.1784121320249777,
+      "eval_accuracy": 0.559032876275525,
+      "eval_loss": 1.9919941425323486,
+      "eval_runtime": 2195.8734,
+      "eval_samples_per_second": 13.408,
+      "eval_steps_per_second": 1.676,
+      "step": 2600
+    },
+    {
+      "epoch": 0.17909833253276608,
+      "grad_norm": 0.6105952858924866,
+      "learning_rate": 0.0001979775118580406,
+      "loss": 1.9903,
+      "step": 2610
+    },
+    {
+      "epoch": 0.17909833253276608,
+      "eval_accuracy": 0.5600523268140957,
+      "eval_loss": 1.9881607294082642,
+      "eval_runtime": 2196.3376,
+      "eval_samples_per_second": 13.405,
+      "eval_steps_per_second": 1.676,
+      "step": 2610
+    },
+    {
+      "epoch": 0.17978453304055445,
+      "grad_norm": 0.6086561679840088,
+      "learning_rate": 0.00019792935737847014,
+      "loss": 1.9367,
+      "step": 2620
+    },
+    {
+      "epoch": 0.17978453304055445,
+      "eval_accuracy": 0.5591538955627124,
+      "eval_loss": 1.991269588470459,
+      "eval_runtime": 2196.0471,
+      "eval_samples_per_second": 13.407,
+      "eval_steps_per_second": 1.676,
+      "step": 2620
+    },
+    {
+      "epoch": 0.18047073354834284,
+      "grad_norm": 2.8778371810913086,
+      "learning_rate": 0.00019788120289889968,
+      "loss": 1.9686,
+      "step": 2630
+    },
+    {
+      "epoch": 0.18047073354834284,
+      "eval_accuracy": 0.5572145988371657,
+      "eval_loss": 2.003204107284546,
+      "eval_runtime": 2195.7568,
+      "eval_samples_per_second": 13.409,
+      "eval_steps_per_second": 1.676,
+      "step": 2630
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 43719,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 10,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.8121688667193344e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}