End of training

README.md

@@ -18,7 +18,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [google/vit-base-patch16-224-in21k](https://huggingface.co/google/vit-base-patch16-224-in21k) on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.1325
+- Loss: 0.1092
 - Accuracy: 0.9732
 
 ## Model description

all_results.json

@@ -1,8 +1,13 @@
 {
-    "epoch": 46.15384615384615,
-    "total_flos": 4.3781443993328026e+18,
-    "train_loss": 0.4221388864517212,
-    "train_runtime": 14616.4414,
-    "train_samples_per_second": 4.187,
-    "train_steps_per_second": 0.031
+    "epoch": 50.0,
+    "eval_accuracy": 0.9731993299832495,
+    "eval_loss": 0.10922118276357651,
+    "eval_runtime": 99.3108,
+    "eval_samples_per_second": 6.011,
+    "eval_steps_per_second": 0.191,
+    "total_flos": 2.0803097508518707e+19,
+    "train_loss": 0.1354110169055916,
+    "train_runtime": 52600.7464,
+    "train_samples_per_second": 5.104,
+    "train_steps_per_second": 0.04
 }

eval_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 46.15384615384615,
-    "eval_accuracy": 0.9338235294117647,
-    "eval_loss": 0.33108076453208923,
-    "eval_runtime": 29.2681,
-    "eval_samples_per_second": 4.647,
-    "eval_steps_per_second": 0.171
+    "epoch": 50.0,
+    "eval_accuracy": 0.9731993299832495,
+    "eval_loss": 0.10922118276357651,
+    "eval_runtime": 99.3108,
+    "eval_samples_per_second": 6.011,
+    "eval_steps_per_second": 0.191
 }

runs/Oct15_14-56-41_c59c7e1de817/events.out.tfevents.1729056918.c59c7e1de817.961.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db49938b6368770d38de49330925fdb6ba26c5f5d5c79b848b7c8a9ca7c76c34
+size 411

train_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 46.15384615384615,
-    "total_flos": 4.3781443993328026e+18,
-    "train_loss": 0.4221388864517212,
-    "train_runtime": 14616.4414,
-    "train_samples_per_second": 4.187,
-    "train_steps_per_second": 0.031
+    "epoch": 50.0,
+    "total_flos": 2.0803097508518707e+19,
+    "train_loss": 0.1354110169055916,
+    "train_runtime": 52600.7464,
+    "train_samples_per_second": 5.104,
+    "train_steps_per_second": 0.04
 }

trainer_state.json

@@ -1,763 +1,1945 @@
 {
-  "best_metric": 0.9044117647058824,
-  "best_model_checkpoint": "vit-base-patch16-224-in21k-finetuned-papsmear/checkpoint-360",
-  "epoch": 46.15384615384615,
+  "best_metric": 0.9731993299832495,
+  "best_model_checkpoint": "vit-base-patch16-224-in21k-finetuned-papsmear/checkpoint-1680",
+  "epoch": 50.0,
   "eval_steps": 500,
-  "global_step": 450,
+  "global_step": 2100,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.9230769230769231,
-      "eval_accuracy": 0.2426470588235294,
-      "eval_loss": 1.7588815689086914,
-      "eval_runtime": 30.041,
-      "eval_samples_per_second": 4.527,
-      "eval_steps_per_second": 0.166,
-      "step": 9
+      "epoch": 0.23809523809523808,
+      "grad_norm": 1.9611310958862305,
+      "learning_rate": 2.3809523809523808e-06,
+      "loss": 1.4059,
+      "step": 10
     },
     {
-      "epoch": 1.0256410256410255,
-      "grad_norm": 1.042138695716858,
-      "learning_rate": 1.1111111111111112e-05,
-      "loss": 1.7862,
-      "step": 10
+      "epoch": 0.47619047619047616,
+      "grad_norm": 1.8307807445526123,
+      "learning_rate": 4.7619047619047615e-06,
+      "loss": 1.3406,
+      "step": 20
     },
     {
-      "epoch": 1.9487179487179487,
-      "eval_accuracy": 0.38235294117647056,
-      "eval_loss": 1.58797025680542,
-      "eval_runtime": 32.3153,
-      "eval_samples_per_second": 4.209,
-      "eval_steps_per_second": 0.155,
-      "step": 19
+      "epoch": 0.7142857142857143,
+      "grad_norm": 1.8257180452346802,
+      "learning_rate": 7.142857142857143e-06,
+      "loss": 1.2533,
+      "step": 30
     },
     {
-      "epoch": 2.051282051282051,
-      "grad_norm": 0.996979296207428,
-      "learning_rate": 2.2222222222222223e-05,
-      "loss": 1.6727,
-      "step": 20
+      "epoch": 0.9523809523809523,
+      "grad_norm": 1.2919795513153076,
+      "learning_rate": 9.523809523809523e-06,
+      "loss": 1.1553,
+      "step": 40
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.5477386934673367,
+      "eval_loss": 1.0950442552566528,
+      "eval_runtime": 104.5,
+      "eval_samples_per_second": 5.713,
+      "eval_steps_per_second": 0.182,
+      "step": 42
+    },
+    {
+      "epoch": 1.1904761904761905,
+      "grad_norm": 0.9992073178291321,
+      "learning_rate": 1.1904761904761905e-05,
+      "loss": 1.0508,
+      "step": 50
+    },
+    {
+      "epoch": 1.4285714285714286,
+      "grad_norm": 0.8946818113327026,
+      "learning_rate": 1.4285714285714285e-05,
+      "loss": 0.952,
+      "step": 60
+    },
+    {
+      "epoch": 1.6666666666666665,
+      "grad_norm": 1.0831001996994019,
+      "learning_rate": 1.6666666666666667e-05,
+      "loss": 0.8539,
+      "step": 70
     },
     {
-      "epoch": 2.9743589743589745,
-      "eval_accuracy": 0.4264705882352941,
-      "eval_loss": 1.4212044477462769,
-      "eval_runtime": 31.937,
-      "eval_samples_per_second": 4.258,
-      "eval_steps_per_second": 0.157,
-      "step": 29
+      "epoch": 1.9047619047619047,
+      "grad_norm": 1.271881103515625,
+      "learning_rate": 1.9047619047619046e-05,
+      "loss": 0.7791,
+      "step": 80
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.8525963149078727,
+      "eval_loss": 0.6485655307769775,
+      "eval_runtime": 107.1905,
+      "eval_samples_per_second": 5.57,
+      "eval_steps_per_second": 0.177,
+      "step": 84
     },
     {
-      "epoch": 3.076923076923077,
-      "grad_norm": 1.1018966436386108,
+      "epoch": 2.142857142857143,
+      "grad_norm": 1.7645951509475708,
+      "learning_rate": 2.1428571428571428e-05,
+      "loss": 0.6112,
+      "step": 90
+    },
+    {
+      "epoch": 2.380952380952381,
+      "grad_norm": 2.271763563156128,
+      "learning_rate": 2.380952380952381e-05,
+      "loss": 0.5029,
+      "step": 100
+    },
+    {
+      "epoch": 2.619047619047619,
+      "grad_norm": 2.248612403869629,
+      "learning_rate": 2.6190476190476192e-05,
+      "loss": 0.4809,
+      "step": 110
+    },
+    {
+      "epoch": 2.857142857142857,
+      "grad_norm": 2.1449034214019775,
+      "learning_rate": 2.857142857142857e-05,
+      "loss": 0.433,
+      "step": 120
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.9128978224455612,
+      "eval_loss": 0.3716076910495758,
+      "eval_runtime": 107.5442,
+      "eval_samples_per_second": 5.551,
+      "eval_steps_per_second": 0.177,
+      "step": 126
+    },
+    {
+      "epoch": 3.0952380952380953,
+      "grad_norm": 2.3766279220581055,
+      "learning_rate": 3.095238095238095e-05,
+      "loss": 0.3777,
+      "step": 130
+    },
+    {
+      "epoch": 3.3333333333333335,
+      "grad_norm": 1.3414802551269531,
       "learning_rate": 3.3333333333333335e-05,
-      "loss": 1.5102,
-      "step": 30
+      "loss": 0.3424,
+      "step": 140
+    },
+    {
+      "epoch": 3.571428571428571,
+      "grad_norm": 7.162674903869629,
+      "learning_rate": 3.571428571428572e-05,
+      "loss": 0.3719,
+      "step": 150
+    },
+    {
+      "epoch": 3.8095238095238093,
+      "grad_norm": 2.170807123184204,
+      "learning_rate": 3.809523809523809e-05,
+      "loss": 0.3495,
+      "step": 160
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.5808823529411765,
-      "eval_loss": 1.2241116762161255,
-      "eval_runtime": 30.9919,
-      "eval_samples_per_second": 4.388,
-      "eval_steps_per_second": 0.161,
-      "step": 39
+      "eval_accuracy": 0.9346733668341709,
+      "eval_loss": 0.28690266609191895,
+      "eval_runtime": 106.4197,
+      "eval_samples_per_second": 5.61,
+      "eval_steps_per_second": 0.179,
+      "step": 168
+    },
+    {
+      "epoch": 4.0476190476190474,
+      "grad_norm": 1.2484022378921509,
+      "learning_rate": 4.047619047619048e-05,
+      "loss": 0.2892,
+      "step": 170
     },
     {
-      "epoch": 4.102564102564102,
-      "grad_norm": 1.025856614112854,
-      "learning_rate": 4.4444444444444447e-05,
-      "loss": 1.3247,
-      "step": 40
+      "epoch": 4.285714285714286,
+      "grad_norm": 7.085641860961914,
+      "learning_rate": 4.2857142857142856e-05,
+      "loss": 0.3386,
+      "step": 180
     },
     {
-      "epoch": 4.923076923076923,
-      "eval_accuracy": 0.6102941176470589,
-      "eval_loss": 1.0905669927597046,
-      "eval_runtime": 31.2612,
-      "eval_samples_per_second": 4.35,
-      "eval_steps_per_second": 0.16,
-      "step": 48
+      "epoch": 4.523809523809524,
+      "grad_norm": 1.4148013591766357,
+      "learning_rate": 4.523809523809524e-05,
+      "loss": 0.3225,
+      "step": 190
     },
     {
-      "epoch": 5.128205128205128,
-      "grad_norm": 3.468245029449463,
-      "learning_rate": 4.938271604938271e-05,
-      "loss": 1.1047,
-      "step": 50
+      "epoch": 4.761904761904762,
+      "grad_norm": 2.7628612518310547,
+      "learning_rate": 4.761904761904762e-05,
+      "loss": 0.2979,
+      "step": 200
+    },
+    {
+      "epoch": 5.0,
+      "grad_norm": 1.5106964111328125,
+      "learning_rate": 5e-05,
+      "loss": 0.2556,
+      "step": 210
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.9279731993299832,
+      "eval_loss": 0.27220866084098816,
+      "eval_runtime": 106.2256,
+      "eval_samples_per_second": 5.62,
+      "eval_steps_per_second": 0.179,
+      "step": 210
+    },
+    {
+      "epoch": 5.238095238095238,
+      "grad_norm": 4.107568740844727,
+      "learning_rate": 4.973544973544973e-05,
+      "loss": 0.2635,
+      "step": 220
+    },
+    {
+      "epoch": 5.476190476190476,
+      "grad_norm": 2.1257033348083496,
+      "learning_rate": 4.9470899470899475e-05,
+      "loss": 0.2581,
+      "step": 230
+    },
+    {
+      "epoch": 5.714285714285714,
+      "grad_norm": 2.4218671321868896,
+      "learning_rate": 4.9206349206349204e-05,
+      "loss": 0.2303,
+      "step": 240
+    },
+    {
+      "epoch": 5.9523809523809526,
+      "grad_norm": 2.0036144256591797,
+      "learning_rate": 4.894179894179895e-05,
+      "loss": 0.2791,
+      "step": 250
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.932998324958124,
+      "eval_loss": 0.2610774338245392,
+      "eval_runtime": 106.375,
+      "eval_samples_per_second": 5.612,
+      "eval_steps_per_second": 0.179,
+      "step": 252
+    },
+    {
+      "epoch": 6.190476190476191,
+      "grad_norm": 2.8978521823883057,
+      "learning_rate": 4.8677248677248676e-05,
+      "loss": 0.224,
+      "step": 260
     },
     {
-      "epoch": 5.948717948717949,
-      "eval_accuracy": 0.6764705882352942,
-      "eval_loss": 0.9746549129486084,
-      "eval_runtime": 29.9981,
-      "eval_samples_per_second": 4.534,
-      "eval_steps_per_second": 0.167,
-      "step": 58
+      "epoch": 6.428571428571429,
+      "grad_norm": 1.3699370622634888,
+      "learning_rate": 4.841269841269841e-05,
+      "loss": 0.2208,
+      "step": 270
     },
     {
-      "epoch": 6.153846153846154,
-      "grad_norm": 1.6158350706100464,
+      "epoch": 6.666666666666667,
+      "grad_norm": 1.7077748775482178,
       "learning_rate": 4.814814814814815e-05,
-      "loss": 0.9405,
-      "step": 60
+      "loss": 0.2294,
+      "step": 280
     },
     {
-      "epoch": 6.9743589743589745,
-      "eval_accuracy": 0.7426470588235294,
-      "eval_loss": 0.8744558691978455,
-      "eval_runtime": 29.8899,
-      "eval_samples_per_second": 4.55,
-      "eval_steps_per_second": 0.167,
-      "step": 68
+      "epoch": 6.904761904761905,
+      "grad_norm": 2.22580885887146,
+      "learning_rate": 4.7883597883597884e-05,
+      "loss": 0.2343,
+      "step": 290
     },
     {
-      "epoch": 7.17948717948718,
-      "grad_norm": 2.7250912189483643,
-      "learning_rate": 4.691358024691358e-05,
-      "loss": 0.823,
-      "step": 70
+      "epoch": 7.0,
+      "eval_accuracy": 0.9380234505862647,
+      "eval_loss": 0.2376711070537567,
+      "eval_runtime": 104.2609,
+      "eval_samples_per_second": 5.726,
+      "eval_steps_per_second": 0.182,
+      "step": 294
+    },
+    {
+      "epoch": 7.142857142857143,
+      "grad_norm": 1.8247556686401367,
+      "learning_rate": 4.761904761904762e-05,
+      "loss": 0.2295,
+      "step": 300
+    },
+    {
+      "epoch": 7.380952380952381,
+      "grad_norm": 2.389512538909912,
+      "learning_rate": 4.7354497354497356e-05,
+      "loss": 0.2059,
+      "step": 310
+    },
+    {
+      "epoch": 7.619047619047619,
+      "grad_norm": 1.6845765113830566,
+      "learning_rate": 4.708994708994709e-05,
+      "loss": 0.1938,
+      "step": 320
+    },
+    {
+      "epoch": 7.857142857142857,
+      "grad_norm": 1.345927119255066,
+      "learning_rate": 4.682539682539683e-05,
+      "loss": 0.186,
+      "step": 330
     },
     {
       "epoch": 8.0,
-      "eval_accuracy": 0.7426470588235294,
-      "eval_loss": 0.7832698822021484,
-      "eval_runtime": 29.3695,
-      "eval_samples_per_second": 4.631,
-      "eval_steps_per_second": 0.17,
-      "step": 78
-    },
-    {
-      "epoch": 8.205128205128204,
-      "grad_norm": 1.1131880283355713,
-      "learning_rate": 4.567901234567901e-05,
-      "loss": 0.7244,
-      "step": 80
+      "eval_accuracy": 0.9396984924623115,
+      "eval_loss": 0.2157616764307022,
+      "eval_runtime": 104.0067,
+      "eval_samples_per_second": 5.74,
+      "eval_steps_per_second": 0.183,
+      "step": 336
+    },
+    {
+      "epoch": 8.095238095238095,
+      "grad_norm": 1.470657467842102,
+      "learning_rate": 4.656084656084656e-05,
+      "loss": 0.2077,
+      "step": 340
     },
     {
-      "epoch": 8.923076923076923,
-      "eval_accuracy": 0.7794117647058824,
-      "eval_loss": 0.7159935235977173,
-      "eval_runtime": 30.5505,
-      "eval_samples_per_second": 4.452,
-      "eval_steps_per_second": 0.164,
-      "step": 87
+      "epoch": 8.333333333333334,
+      "grad_norm": 1.4709361791610718,
+      "learning_rate": 4.62962962962963e-05,
+      "loss": 0.1809,
+      "step": 350
+    },
+    {
+      "epoch": 8.571428571428571,
+      "grad_norm": 1.731719732284546,
+      "learning_rate": 4.603174603174603e-05,
+      "loss": 0.2163,
+      "step": 360
+    },
+    {
+      "epoch": 8.80952380952381,
+      "grad_norm": 1.8033312559127808,
+      "learning_rate": 4.576719576719577e-05,
+      "loss": 0.1984,
+      "step": 370
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": 0.9346733668341709,
+      "eval_loss": 0.22224725782871246,
+      "eval_runtime": 104.8934,
+      "eval_samples_per_second": 5.691,
+      "eval_steps_per_second": 0.181,
+      "step": 378
+    },
+    {
+      "epoch": 9.047619047619047,
+      "grad_norm": 1.9741649627685547,
+      "learning_rate": 4.55026455026455e-05,
+      "loss": 0.2093,
+      "step": 380
+    },
+    {
+      "epoch": 9.285714285714286,
+      "grad_norm": 1.6360487937927246,
+      "learning_rate": 4.523809523809524e-05,
+      "loss": 0.1586,
+      "step": 390
+    },
+    {
+      "epoch": 9.523809523809524,
+      "grad_norm": 2.772472858428955,
+      "learning_rate": 4.4973544973544974e-05,
+      "loss": 0.1423,
+      "step": 400
+    },
+    {
+      "epoch": 9.761904761904763,
+      "grad_norm": 1.9369553327560425,
+      "learning_rate": 4.470899470899471e-05,
+      "loss": 0.1458,
+      "step": 410
     },
     {
-      "epoch": 9.23076923076923,
-      "grad_norm": 1.162032961845398,
+      "epoch": 10.0,
+      "grad_norm": 2.552593231201172,
       "learning_rate": 4.4444444444444447e-05,
-      "loss": 0.6367,
-      "step": 90
+      "loss": 0.1751,
+      "step": 420
     },
     {
-      "epoch": 9.948717948717949,
-      "eval_accuracy": 0.7794117647058824,
-      "eval_loss": 0.7327755093574524,
-      "eval_runtime": 31.3159,
-      "eval_samples_per_second": 4.343,
-      "eval_steps_per_second": 0.16,
-      "step": 97
+      "epoch": 10.0,
+      "eval_accuracy": 0.9514237855946399,
+      "eval_loss": 0.19929908215999603,
+      "eval_runtime": 106.1844,
+      "eval_samples_per_second": 5.622,
+      "eval_steps_per_second": 0.179,
+      "step": 420
     },
     {
-      "epoch": 10.256410256410255,
-      "grad_norm": 2.753892421722412,
-      "learning_rate": 4.3209876543209875e-05,
-      "loss": 0.5537,
-      "step": 100
+      "epoch": 10.238095238095237,
+      "grad_norm": 2.9462811946868896,
+      "learning_rate": 4.417989417989418e-05,
+      "loss": 0.1305,
+      "step": 430
     },
     {
-      "epoch": 10.974358974358974,
-      "eval_accuracy": 0.7867647058823529,
-      "eval_loss": 0.6572667956352234,
-      "eval_runtime": 29.9075,
-      "eval_samples_per_second": 4.547,
-      "eval_steps_per_second": 0.167,
-      "step": 107
+      "epoch": 10.476190476190476,
+      "grad_norm": 2.5522372722625732,
+      "learning_rate": 4.391534391534391e-05,
+      "loss": 0.1555,
+      "step": 440
     },
     {
-      "epoch": 11.282051282051283,
-      "grad_norm": 1.8267817497253418,
-      "learning_rate": 4.197530864197531e-05,
-      "loss": 0.484,
-      "step": 110
+      "epoch": 10.714285714285714,
+      "grad_norm": 1.2257236242294312,
+      "learning_rate": 4.3650793650793655e-05,
+      "loss": 0.1575,
+      "step": 450
+    },
+    {
+      "epoch": 10.952380952380953,
+      "grad_norm": 0.9242783188819885,
+      "learning_rate": 4.3386243386243384e-05,
+      "loss": 0.1529,
+      "step": 460
+    },
+    {
+      "epoch": 11.0,
+      "eval_accuracy": 0.9430485762144054,
+      "eval_loss": 0.2100822627544403,
+      "eval_runtime": 102.6058,
+      "eval_samples_per_second": 5.818,
+      "eval_steps_per_second": 0.185,
+      "step": 462
+    },
+    {
+      "epoch": 11.19047619047619,
+      "grad_norm": 1.2224242687225342,
+      "learning_rate": 4.312169312169313e-05,
+      "loss": 0.1171,
+      "step": 470
+    },
+    {
+      "epoch": 11.428571428571429,
+      "grad_norm": 1.332038164138794,
+      "learning_rate": 4.2857142857142856e-05,
+      "loss": 0.1399,
+      "step": 480
+    },
+    {
+      "epoch": 11.666666666666666,
+      "grad_norm": 2.5912814140319824,
+      "learning_rate": 4.259259259259259e-05,
+      "loss": 0.129,
+      "step": 490
+    },
+    {
+      "epoch": 11.904761904761905,
+      "grad_norm": 3.0691795349121094,
+      "learning_rate": 4.232804232804233e-05,
+      "loss": 0.1616,
+      "step": 500
     },
     {
       "epoch": 12.0,
-      "eval_accuracy": 0.8088235294117647,
-      "eval_loss": 0.5988054275512695,
-      "eval_runtime": 29.8243,
-      "eval_samples_per_second": 4.56,
-      "eval_steps_per_second": 0.168,
-      "step": 117
+      "eval_accuracy": 0.9296482412060302,
+      "eval_loss": 0.2542562484741211,
+      "eval_runtime": 107.198,
+      "eval_samples_per_second": 5.569,
+      "eval_steps_per_second": 0.177,
+      "step": 504
     },
     {
-      "epoch": 12.307692307692308,
-      "grad_norm": 3.6334152221679688,
-      "learning_rate": 4.074074074074074e-05,
-      "loss": 0.4642,
-      "step": 120
+      "epoch": 12.142857142857142,
+      "grad_norm": 1.015551209449768,
+      "learning_rate": 4.2063492063492065e-05,
+      "loss": 0.1374,
+      "step": 510
     },
     {
-      "epoch": 12.923076923076923,
-      "eval_accuracy": 0.7941176470588235,
-      "eval_loss": 0.626797080039978,
-      "eval_runtime": 29.8057,
-      "eval_samples_per_second": 4.563,
-      "eval_steps_per_second": 0.168,
-      "step": 126
+      "epoch": 12.380952380952381,
+      "grad_norm": 1.5744130611419678,
+      "learning_rate": 4.17989417989418e-05,
+      "loss": 0.1219,
+      "step": 520
+    },
+    {
+      "epoch": 12.619047619047619,
+      "grad_norm": 2.7706940174102783,
+      "learning_rate": 4.153439153439154e-05,
+      "loss": 0.1159,
+      "step": 530
+    },
+    {
+      "epoch": 12.857142857142858,
+      "grad_norm": 0.9542893767356873,
+      "learning_rate": 4.126984126984127e-05,
+      "loss": 0.1404,
+      "step": 540
+    },
+    {
+      "epoch": 13.0,
+      "eval_accuracy": 0.9396984924623115,
+      "eval_loss": 0.2028820812702179,
+      "eval_runtime": 106.2276,
+      "eval_samples_per_second": 5.62,
+      "eval_steps_per_second": 0.179,
+      "step": 546
+    },
+    {
+      "epoch": 13.095238095238095,
+      "grad_norm": 1.9553595781326294,
+      "learning_rate": 4.100529100529101e-05,
+      "loss": 0.1128,
+      "step": 550
     },
     {
       "epoch": 13.333333333333334,
-      "grad_norm": 2.4481348991394043,
-      "learning_rate": 3.950617283950617e-05,
-      "loss": 0.4166,
-      "step": 130
+      "grad_norm": 0.4872543513774872,
+      "learning_rate": 4.074074074074074e-05,
+      "loss": 0.1133,
+      "step": 560
     },
     {
-      "epoch": 13.948717948717949,
-      "eval_accuracy": 0.7794117647058824,
-      "eval_loss": 0.6549181342124939,
-      "eval_runtime": 29.9078,
-      "eval_samples_per_second": 4.547,
-      "eval_steps_per_second": 0.167,
-      "step": 136
+      "epoch": 13.571428571428571,
+      "grad_norm": 1.1375516653060913,
+      "learning_rate": 4.047619047619048e-05,
+      "loss": 0.1195,
+      "step": 570
     },
     {
-      "epoch": 14.35897435897436,
-      "grad_norm": 2.9035937786102295,
-      "learning_rate": 3.82716049382716e-05,
-      "loss": 0.4106,
-      "step": 140
+      "epoch": 13.80952380952381,
+      "grad_norm": 2.1001851558685303,
+      "learning_rate": 4.021164021164021e-05,
+      "loss": 0.1078,
+      "step": 580
     },
     {
-      "epoch": 14.974358974358974,
-      "eval_accuracy": 0.8529411764705882,
-      "eval_loss": 0.5330095887184143,
-      "eval_runtime": 29.4929,
-      "eval_samples_per_second": 4.611,
-      "eval_steps_per_second": 0.17,
-      "step": 146
+      "epoch": 14.0,
+      "eval_accuracy": 0.9413735343383585,
+      "eval_loss": 0.20870448648929596,
+      "eval_runtime": 103.1036,
+      "eval_samples_per_second": 5.79,
+      "eval_steps_per_second": 0.184,
+      "step": 588
     },
     {
-      "epoch": 15.384615384615385,
-      "grad_norm": 3.0346601009368896,
-      "learning_rate": 3.7037037037037037e-05,
-      "loss": 0.3947,
-      "step": 150
+      "epoch": 14.047619047619047,
+      "grad_norm": 1.273701786994934,
+      "learning_rate": 3.9947089947089946e-05,
+      "loss": 0.1078,
+      "step": 590
+    },
+    {
+      "epoch": 14.285714285714286,
+      "grad_norm": 2.2141387462615967,
+      "learning_rate": 3.968253968253968e-05,
+      "loss": 0.1005,
+      "step": 600
+    },
+    {
+      "epoch": 14.523809523809524,
+      "grad_norm": 2.147643566131592,
+      "learning_rate": 3.941798941798942e-05,
+      "loss": 0.1195,
+      "step": 610
+    },
+    {
+      "epoch": 14.761904761904763,
+      "grad_norm": 1.8408890962600708,
+      "learning_rate": 3.9153439153439155e-05,
+      "loss": 0.0981,
+      "step": 620
+    },
+    {
+      "epoch": 15.0,
+      "grad_norm": 2.7218682765960693,
+      "learning_rate": 3.888888888888889e-05,
+      "loss": 0.1109,
+      "step": 630
+    },
+    {
+      "epoch": 15.0,
+      "eval_accuracy": 0.9614740368509213,
+      "eval_loss": 0.1381397545337677,
+      "eval_runtime": 102.743,
+      "eval_samples_per_second": 5.811,
+      "eval_steps_per_second": 0.185,
+      "step": 630
+    },
+    {
+      "epoch": 15.238095238095237,
+      "grad_norm": 2.087874412536621,
+      "learning_rate": 3.862433862433863e-05,
+      "loss": 0.0932,
+      "step": 640
+    },
+    {
+      "epoch": 15.476190476190476,
+      "grad_norm": 1.6582282781600952,
+      "learning_rate": 3.835978835978836e-05,
+      "loss": 0.12,
+      "step": 650
+    },
+    {
+      "epoch": 15.714285714285714,
+      "grad_norm": 1.8671879768371582,
+      "learning_rate": 3.809523809523809e-05,
+      "loss": 0.0981,
+      "step": 660
+    },
+    {
+      "epoch": 15.952380952380953,
+      "grad_norm": 1.7882401943206787,
+      "learning_rate": 3.7830687830687835e-05,
+      "loss": 0.1072,
+      "step": 670
     },
     {
       "epoch": 16.0,
-      "eval_accuracy": 0.8382352941176471,
-      "eval_loss": 0.5133553147315979,
-      "eval_runtime": 29.8094,
-      "eval_samples_per_second": 4.562,
-      "eval_steps_per_second": 0.168,
-      "step": 156
-    },
-    {
-      "epoch": 16.41025641025641,
-      "grad_norm": 2.665196657180786,
-      "learning_rate": 3.580246913580247e-05,
-      "loss": 0.3469,
-      "step": 160
+      "eval_accuracy": 0.9413735343383585,
+      "eval_loss": 0.18952256441116333,
+      "eval_runtime": 104.7109,
+      "eval_samples_per_second": 5.701,
+      "eval_steps_per_second": 0.181,
+      "step": 672
     },
     {
-      "epoch": 16.923076923076923,
-      "eval_accuracy": 0.7794117647058824,
-      "eval_loss": 0.5879342555999756,
-      "eval_runtime": 29.7747,
-      "eval_samples_per_second": 4.568,
-      "eval_steps_per_second": 0.168,
-      "step": 165
+      "epoch": 16.19047619047619,
+      "grad_norm": 3.6218953132629395,
+      "learning_rate": 3.7566137566137564e-05,
+      "loss": 0.0818,
+      "step": 680
     },
     {
-      "epoch": 17.435897435897434,
-      "grad_norm": 4.382056713104248,
-      "learning_rate": 3.45679012345679e-05,
-      "loss": 0.3151,
-      "step": 170
+      "epoch": 16.428571428571427,
+      "grad_norm": 1.1185909509658813,
+      "learning_rate": 3.730158730158731e-05,
+      "loss": 0.0951,
+      "step": 690
     },
     {
-      "epoch": 17.94871794871795,
-      "eval_accuracy": 0.8382352941176471,
-      "eval_loss": 0.5682740211486816,
-      "eval_runtime": 29.9811,
-      "eval_samples_per_second": 4.536,
-      "eval_steps_per_second": 0.167,
-      "step": 175
+      "epoch": 16.666666666666668,
+      "grad_norm": 0.9545117616653442,
+      "learning_rate": 3.7037037037037037e-05,
+      "loss": 0.1064,
+      "step": 700
     },
     {
-      "epoch": 18.46153846153846,
-      "grad_norm": 1.3831549882888794,
-      "learning_rate": 3.3333333333333335e-05,
-      "loss": 0.2946,
-      "step": 180
+      "epoch": 16.904761904761905,
+      "grad_norm": 0.7912140488624573,
+      "learning_rate": 3.677248677248677e-05,
+      "loss": 0.0949,
+      "step": 710
     },
     {
-      "epoch": 18.974358974358974,
-      "eval_accuracy": 0.8161764705882353,
-      "eval_loss": 0.5382511615753174,
-      "eval_runtime": 29.6021,
-      "eval_samples_per_second": 4.594,
-      "eval_steps_per_second": 0.169,
-      "step": 185
+      "epoch": 17.0,
+      "eval_accuracy": 0.9396984924623115,
+      "eval_loss": 0.19812369346618652,
+      "eval_runtime": 103.2395,
+      "eval_samples_per_second": 5.783,
+      "eval_steps_per_second": 0.184,
+      "step": 714
     },
     {
-      "epoch": 19.487179487179485,
-      "grad_norm": 2.7299916744232178,
-      "learning_rate": 3.209876543209876e-05,
-      "loss": 0.2927,
-      "step": 190
+      "epoch": 17.142857142857142,
+      "grad_norm": 2.5102455615997314,
+      "learning_rate": 3.650793650793651e-05,
+      "loss": 0.0942,
+      "step": 720
+    },
+    {
+      "epoch": 17.38095238095238,
+      "grad_norm": 0.907739520072937,
+      "learning_rate": 3.6243386243386245e-05,
+      "loss": 0.081,
+      "step": 730
+    },
+    {
+      "epoch": 17.61904761904762,
+      "grad_norm": 0.9580628275871277,
+      "learning_rate": 3.597883597883598e-05,
+      "loss": 0.0939,
+      "step": 740
+    },
+    {
+      "epoch": 17.857142857142858,
+      "grad_norm": 0.4892265498638153,
+      "learning_rate": 3.571428571428572e-05,
+      "loss": 0.0908,
+      "step": 750
+    },
+    {
+      "epoch": 18.0,
+      "eval_accuracy": 0.9581239530988275,
+      "eval_loss": 0.16083765029907227,
+      "eval_runtime": 103.0208,
+      "eval_samples_per_second": 5.795,
+      "eval_steps_per_second": 0.184,
+      "step": 756
+    },
+    {
+      "epoch": 18.095238095238095,
+      "grad_norm": 0.6354324221611023,
+      "learning_rate": 3.5449735449735446e-05,
+      "loss": 0.0674,
+      "step": 760
+    },
+    {
+      "epoch": 18.333333333333332,
+      "grad_norm": 1.7310459613800049,
+      "learning_rate": 3.518518518518519e-05,
+      "loss": 0.0963,
+      "step": 770
+    },
+    {
+      "epoch": 18.571428571428573,
+      "grad_norm": 1.0339725017547607,
+      "learning_rate": 3.492063492063492e-05,
+      "loss": 0.0846,
+      "step": 780
+    },
+    {
+      "epoch": 18.80952380952381,
+      "grad_norm": 2.478813648223877,
+      "learning_rate": 3.465608465608466e-05,
+      "loss": 0.0809,
+      "step": 790
+    },
+    {
+      "epoch": 19.0,
+      "eval_accuracy": 0.9581239530988275,
+      "eval_loss": 0.1764398217201233,
+      "eval_runtime": 104.4696,
+      "eval_samples_per_second": 5.715,
+      "eval_steps_per_second": 0.182,
+      "step": 798
+    },
+    {
+      "epoch": 19.047619047619047,
+      "grad_norm": 0.9190245270729065,
+      "learning_rate": 3.439153439153439e-05,
+      "loss": 0.0831,
+      "step": 800
+    },
+    {
+      "epoch": 19.285714285714285,
+      "grad_norm": 1.1679134368896484,
+      "learning_rate": 3.412698412698413e-05,
+      "loss": 0.0737,
+      "step": 810
+    },
+    {
+      "epoch": 19.523809523809526,
+      "grad_norm": 1.52895987033844,
+      "learning_rate": 3.386243386243386e-05,
+      "loss": 0.0695,
+      "step": 820
+    },
+    {
+      "epoch": 19.761904761904763,
+      "grad_norm": 1.6142102479934692,
+      "learning_rate": 3.35978835978836e-05,
+      "loss": 0.0696,
+      "step": 830
     },
     {
       "epoch": 20.0,
-      "eval_accuracy": 0.8161764705882353,
-      "eval_loss": 0.568187952041626,
-      "eval_runtime": 29.6931,
-      "eval_samples_per_second": 4.58,
-      "eval_steps_per_second": 0.168,
-      "step": 195
-    },
-    {
-      "epoch": 20.51282051282051,
-      "grad_norm": 2.4772286415100098,
-      "learning_rate": 3.08641975308642e-05,
-      "loss": 0.2879,
-      "step": 200
+      "grad_norm": 2.2493298053741455,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 0.0708,
+      "step": 840
     },
     {
-      "epoch": 20.923076923076923,
-      "eval_accuracy": 0.8602941176470589,
-      "eval_loss": 0.4721927046775818,
-      "eval_runtime": 29.6838,
-      "eval_samples_per_second": 4.582,
-      "eval_steps_per_second": 0.168,
-      "step": 204
+      "epoch": 20.0,
+      "eval_accuracy": 0.9530988274706867,
+      "eval_loss": 0.15123647451400757,
+      "eval_runtime": 102.2975,
+      "eval_samples_per_second": 5.836,
+      "eval_steps_per_second": 0.186,
+      "step": 840
     },
     {
-      "epoch": 21.53846153846154,
-      "grad_norm": 1.078134536743164,
-      "learning_rate": 2.962962962962963e-05,
-      "loss": 0.2512,
-      "step": 210
+      "epoch": 20.238095238095237,
+      "grad_norm": 4.499787330627441,
+      "learning_rate": 3.306878306878307e-05,
+      "loss": 0.0936,
+      "step": 850
     },
     {
-      "epoch": 21.94871794871795,
-      "eval_accuracy": 0.8455882352941176,
-      "eval_loss": 0.48056113719940186,
-      "eval_runtime": 29.966,
-      "eval_samples_per_second": 4.538,
-      "eval_steps_per_second": 0.167,
-      "step": 214
+      "epoch": 20.476190476190474,
+      "grad_norm": 5.423216819763184,
+      "learning_rate": 3.280423280423281e-05,
+      "loss": 0.0712,
+      "step": 860
     },
     {
-      "epoch": 22.564102564102566,
-      "grad_norm": 1.4218604564666748,
-      "learning_rate": 2.839506172839506e-05,
-      "loss": 0.2633,
-      "step": 220
+      "epoch": 20.714285714285715,
+      "grad_norm": 1.0831531286239624,
+      "learning_rate": 3.253968253968254e-05,
+      "loss": 0.0817,
+      "step": 870
     },
     {
-      "epoch": 22.974358974358974,
-      "eval_accuracy": 0.8455882352941176,
-      "eval_loss": 0.4712737500667572,
-      "eval_runtime": 30.7829,
-      "eval_samples_per_second": 4.418,
-      "eval_steps_per_second": 0.162,
-      "step": 224
+      "epoch": 20.952380952380953,
+      "grad_norm": 1.6317639350891113,
+      "learning_rate": 3.227513227513227e-05,
+      "loss": 0.0757,
+      "step": 880
     },
     {
-      "epoch": 23.58974358974359,
-      "grad_norm": 1.4475338459014893,
-      "learning_rate": 2.7160493827160493e-05,
-      "loss": 0.2286,
-      "step": 230
+      "epoch": 21.0,
+      "eval_accuracy": 0.948073701842546,
+      "eval_loss": 0.20271137356758118,
+      "eval_runtime": 102.4721,
+      "eval_samples_per_second": 5.826,
+      "eval_steps_per_second": 0.185,
+      "step": 882
+    },
+    {
+      "epoch": 21.19047619047619,
+      "grad_norm": 2.1182172298431396,
+      "learning_rate": 3.2010582010582015e-05,
+      "loss": 0.0882,
+      "step": 890
+    },
+    {
+      "epoch": 21.428571428571427,
+      "grad_norm": 0.5835541486740112,
+      "learning_rate": 3.1746031746031745e-05,
+      "loss": 0.0607,
+      "step": 900
+    },
+    {
+      "epoch": 21.666666666666668,
+      "grad_norm": 1.441300392150879,
+      "learning_rate": 3.148148148148148e-05,
+      "loss": 0.0859,
+      "step": 910
+    },
+    {
+      "epoch": 21.904761904761905,
+      "grad_norm": 0.6337174773216248,
+      "learning_rate": 3.121693121693122e-05,
+      "loss": 0.0919,
+      "step": 920
+    },
+    {
+      "epoch": 22.0,
+      "eval_accuracy": 0.9614740368509213,
+      "eval_loss": 0.14867298305034637,
+      "eval_runtime": 103.3409,
+      "eval_samples_per_second": 5.777,
+      "eval_steps_per_second": 0.184,
+      "step": 924
+    },
+    {
+      "epoch": 22.142857142857142,
+      "grad_norm": 1.7783087491989136,
+      "learning_rate": 3.095238095238095e-05,
+      "loss": 0.0591,
+      "step": 930
+    },
+    {
+      "epoch": 22.38095238095238,
+      "grad_norm": 0.4258907735347748,
+      "learning_rate": 3.068783068783069e-05,
+      "loss": 0.0722,
+      "step": 940
+    },
+    {
+      "epoch": 22.61904761904762,
+      "grad_norm": 0.975234866142273,
+      "learning_rate": 3.0423280423280425e-05,
+      "loss": 0.0582,
+      "step": 950
+    },
+    {
+      "epoch": 22.857142857142858,
+      "grad_norm": 0.9665831327438354,
+      "learning_rate": 3.0158730158730158e-05,
+      "loss": 0.07,
+      "step": 960
+    },
+    {
+      "epoch": 23.0,
+      "eval_accuracy": 0.9614740368509213,
+      "eval_loss": 0.16668196022510529,
+      "eval_runtime": 103.2831,
+      "eval_samples_per_second": 5.78,
+      "eval_steps_per_second": 0.184,
+      "step": 966
+    },
+    {
+      "epoch": 23.095238095238095,
+      "grad_norm": 1.7929288148880005,
+      "learning_rate": 2.9894179894179897e-05,
+      "loss": 0.0582,
+      "step": 970
+    },
+    {
+      "epoch": 23.333333333333332,
+      "grad_norm": 0.43466824293136597,
+      "learning_rate": 2.962962962962963e-05,
+      "loss": 0.061,
+      "step": 980
+    },
+    {
+      "epoch": 23.571428571428573,
+      "grad_norm": 2.2438175678253174,
+      "learning_rate": 2.9365079365079366e-05,
+      "loss": 0.0644,
+      "step": 990
+    },
+    {
+      "epoch": 23.80952380952381,
+      "grad_norm": 0.8747345805168152,
+      "learning_rate": 2.91005291005291e-05,
+      "loss": 0.0629,
+      "step": 1000
     },
     {
       "epoch": 24.0,
-      "eval_accuracy": 0.8382352941176471,
-      "eval_loss": 0.5166775584220886,
-      "eval_runtime": 29.6495,
-      "eval_samples_per_second": 4.587,
-      "eval_steps_per_second": 0.169,
-      "step": 234
+      "eval_accuracy": 0.9530988274706867,
+      "eval_loss": 0.19044645130634308,
+      "eval_runtime": 103.4446,
+      "eval_samples_per_second": 5.771,
+      "eval_steps_per_second": 0.184,
+      "step": 1008
     },
     {
-      "epoch": 24.615384615384617,
-      "grad_norm": 2.2939038276672363,
-      "learning_rate": 2.5925925925925925e-05,
-      "loss": 0.2265,
-      "step": 240
+      "epoch": 24.047619047619047,
+      "grad_norm": 1.7004871368408203,
+      "learning_rate": 2.8835978835978838e-05,
+      "loss": 0.0496,
+      "step": 1010
     },
     {
-      "epoch": 24.923076923076923,
-      "eval_accuracy": 0.8823529411764706,
-      "eval_loss": 0.3885728120803833,
-      "eval_runtime": 29.8244,
-      "eval_samples_per_second": 4.56,
-      "eval_steps_per_second": 0.168,
-      "step": 243
+      "epoch": 24.285714285714285,
+      "grad_norm": 1.3886641263961792,
+      "learning_rate": 2.857142857142857e-05,
+      "loss": 0.0468,
+      "step": 1020
     },
     {
-      "epoch": 25.641025641025642,
-      "grad_norm": 2.014761209487915,
-      "learning_rate": 2.4691358024691357e-05,
-      "loss": 0.2107,
-      "step": 250
+      "epoch": 24.523809523809526,
+      "grad_norm": 2.0595200061798096,
+      "learning_rate": 2.830687830687831e-05,
+      "loss": 0.0657,
+      "step": 1030
+    },
+    {
+      "epoch": 24.761904761904763,
+      "grad_norm": 0.9633322954177856,
+      "learning_rate": 2.8042328042328043e-05,
+      "loss": 0.0626,
+      "step": 1040
+    },
+    {
+      "epoch": 25.0,
+      "grad_norm": 0.629084050655365,
+      "learning_rate": 2.777777777777778e-05,
+      "loss": 0.0584,
+      "step": 1050
+    },
+    {
+      "epoch": 25.0,
+      "eval_accuracy": 0.9631490787269682,
+      "eval_loss": 0.15212486684322357,
+      "eval_runtime": 104.2052,
+      "eval_samples_per_second": 5.729,
+      "eval_steps_per_second": 0.182,
+      "step": 1050
+    },
+    {
+      "epoch": 25.238095238095237,
+      "grad_norm": 1.9510831832885742,
+      "learning_rate": 2.7513227513227512e-05,
+      "loss": 0.0514,
+      "step": 1060
     },
     {
-      "epoch": 25.94871794871795,
-      "eval_accuracy": 0.8676470588235294,
-      "eval_loss": 0.4395664930343628,
-      "eval_runtime": 30.087,
-      "eval_samples_per_second": 4.52,
-      "eval_steps_per_second": 0.166,
-      "step": 253
+      "epoch": 25.476190476190474,
+      "grad_norm": 0.6461337208747864,
+      "learning_rate": 2.724867724867725e-05,
+      "loss": 0.0619,
+      "step": 1070
+    },
+    {
+      "epoch": 25.714285714285715,
+      "grad_norm": 0.8791431784629822,
+      "learning_rate": 2.6984126984126984e-05,
+      "loss": 0.0626,
+      "step": 1080
+    },
+    {
+      "epoch": 25.952380952380953,
+      "grad_norm": 1.83372163772583,
+      "learning_rate": 2.6719576719576723e-05,
+      "loss": 0.0666,
+      "step": 1090
+    },
+    {
+      "epoch": 26.0,
+      "eval_accuracy": 0.966499162479062,
+      "eval_loss": 0.1326070874929428,
+      "eval_runtime": 103.6677,
+      "eval_samples_per_second": 5.759,
+      "eval_steps_per_second": 0.183,
+      "step": 1092
+    },
+    {
+      "epoch": 26.19047619047619,
+      "grad_norm": 3.1547927856445312,
+      "learning_rate": 2.6455026455026456e-05,
+      "loss": 0.0607,
+      "step": 1100
+    },
+    {
+      "epoch": 26.428571428571427,
+      "grad_norm": 0.8336120247840881,
+      "learning_rate": 2.6190476190476192e-05,
+      "loss": 0.0458,
+      "step": 1110
     },
     {
       "epoch": 26.666666666666668,
-      "grad_norm": 1.4246245622634888,
-      "learning_rate": 2.345679012345679e-05,
-      "loss": 0.2044,
-      "step": 260
+      "grad_norm": 0.5386803150177002,
+      "learning_rate": 2.5925925925925925e-05,
+      "loss": 0.0638,
+      "step": 1120
     },
     {
-      "epoch": 26.974358974358974,
-      "eval_accuracy": 0.8455882352941176,
-      "eval_loss": 0.47336432337760925,
-      "eval_runtime": 30.2689,
-      "eval_samples_per_second": 4.493,
-      "eval_steps_per_second": 0.165,
-      "step": 263
+      "epoch": 26.904761904761905,
+      "grad_norm": 0.8411057591438293,
+      "learning_rate": 2.5661375661375664e-05,
+      "loss": 0.062,
+      "step": 1130
     },
     {
-      "epoch": 27.692307692307693,
-      "grad_norm": 1.1730149984359741,
-      "learning_rate": 2.2222222222222223e-05,
-      "loss": 0.1925,
-      "step": 270
+      "epoch": 27.0,
+      "eval_accuracy": 0.9564489112227805,
+      "eval_loss": 0.17715045809745789,
+      "eval_runtime": 102.5942,
+      "eval_samples_per_second": 5.819,
+      "eval_steps_per_second": 0.185,
+      "step": 1134
+    },
+    {
+      "epoch": 27.142857142857142,
+      "grad_norm": 1.1352622509002686,
+      "learning_rate": 2.5396825396825397e-05,
+      "loss": 0.0396,
+      "step": 1140
+    },
+    {
+      "epoch": 27.38095238095238,
+      "grad_norm": 1.9047770500183105,
+      "learning_rate": 2.5132275132275137e-05,
+      "loss": 0.0383,
+      "step": 1150
+    },
+    {
+      "epoch": 27.61904761904762,
+      "grad_norm": 2.154599666595459,
+      "learning_rate": 2.4867724867724866e-05,
+      "loss": 0.0728,
+      "step": 1160
+    },
+    {
+      "epoch": 27.857142857142858,
+      "grad_norm": 1.5056850910186768,
+      "learning_rate": 2.4603174603174602e-05,
+      "loss": 0.0568,
+      "step": 1170
     },
     {
       "epoch": 28.0,
-      "eval_accuracy": 0.8529411764705882,
-      "eval_loss": 0.4605894684791565,
-      "eval_runtime": 29.9687,
-      "eval_samples_per_second": 4.538,
-      "eval_steps_per_second": 0.167,
-      "step": 273
-    },
-    {
-      "epoch": 28.71794871794872,
-      "grad_norm": 1.8061479330062866,
-      "learning_rate": 2.0987654320987655e-05,
-      "loss": 0.1866,
-      "step": 280
+      "eval_accuracy": 0.9564489112227805,
+      "eval_loss": 0.14654366672039032,
+      "eval_runtime": 103.1379,
+      "eval_samples_per_second": 5.788,
+      "eval_steps_per_second": 0.184,
+      "step": 1176
     },
     {
-      "epoch": 28.923076923076923,
-      "eval_accuracy": 0.8308823529411765,
-      "eval_loss": 0.506081223487854,
-      "eval_runtime": 29.7747,
-      "eval_samples_per_second": 4.568,
-      "eval_steps_per_second": 0.168,
-      "step": 282
+      "epoch": 28.095238095238095,
+      "grad_norm": 1.634865641593933,
+      "learning_rate": 2.4338624338624338e-05,
+      "loss": 0.0663,
+      "step": 1180
     },
     {
-      "epoch": 29.743589743589745,
-      "grad_norm": 3.999681234359741,
-      "learning_rate": 1.9753086419753087e-05,
-      "loss": 0.1928,
-      "step": 290
+      "epoch": 28.333333333333332,
+      "grad_norm": 1.265386939048767,
+      "learning_rate": 2.4074074074074074e-05,
+      "loss": 0.0487,
+      "step": 1190
     },
     {
-      "epoch": 29.94871794871795,
-      "eval_accuracy": 0.8823529411764706,
-      "eval_loss": 0.42022156715393066,
-      "eval_runtime": 31.5903,
-      "eval_samples_per_second": 4.305,
-      "eval_steps_per_second": 0.158,
-      "step": 292
+      "epoch": 28.571428571428573,
+      "grad_norm": 0.6159355044364929,
+      "learning_rate": 2.380952380952381e-05,
+      "loss": 0.0596,
+      "step": 1200
     },
     {
-      "epoch": 30.76923076923077,
-      "grad_norm": 1.7130581140518188,
-      "learning_rate": 1.8518518518518518e-05,
-      "loss": 0.1907,
-      "step": 300
+      "epoch": 28.80952380952381,
+      "grad_norm": 1.0339206457138062,
+      "learning_rate": 2.3544973544973546e-05,
+      "loss": 0.0453,
+      "step": 1210
     },
     {
-      "epoch": 30.974358974358974,
-      "eval_accuracy": 0.8308823529411765,
-      "eval_loss": 0.5120359659194946,
-      "eval_runtime": 29.2951,
-      "eval_samples_per_second": 4.642,
-      "eval_steps_per_second": 0.171,
-      "step": 302
+      "epoch": 29.0,
+      "eval_accuracy": 0.9681742043551089,
+      "eval_loss": 0.13472113013267517,
+      "eval_runtime": 103.1249,
+      "eval_samples_per_second": 5.789,
+      "eval_steps_per_second": 0.184,
+      "step": 1218
     },
     {
-      "epoch": 31.794871794871796,
-      "grad_norm": 2.6331541538238525,
-      "learning_rate": 1.728395061728395e-05,
-      "loss": 0.1631,
-      "step": 310
+      "epoch": 29.047619047619047,
+      "grad_norm": 0.8177947998046875,
+      "learning_rate": 2.328042328042328e-05,
+      "loss": 0.055,
+      "step": 1220
+    },
+    {
+      "epoch": 29.285714285714285,
+      "grad_norm": 1.7629382610321045,
+      "learning_rate": 2.3015873015873015e-05,
+      "loss": 0.0476,
+      "step": 1230
+    },
+    {
+      "epoch": 29.523809523809526,
+      "grad_norm": 1.8531335592269897,
+      "learning_rate": 2.275132275132275e-05,
+      "loss": 0.0431,
+      "step": 1240
+    },
+    {
+      "epoch": 29.761904761904763,
+      "grad_norm": 0.961283802986145,
+      "learning_rate": 2.2486772486772487e-05,
+      "loss": 0.0579,
+      "step": 1250
+    },
+    {
+      "epoch": 30.0,
+      "grad_norm": 0.18748821318149567,
+      "learning_rate": 2.2222222222222223e-05,
+      "loss": 0.0469,
+      "step": 1260
+    },
+    {
+      "epoch": 30.0,
+      "eval_accuracy": 0.9631490787269682,
+      "eval_loss": 0.16871798038482666,
+      "eval_runtime": 105.6517,
+      "eval_samples_per_second": 5.651,
+      "eval_steps_per_second": 0.18,
+      "step": 1260
+    },
+    {
+      "epoch": 30.238095238095237,
+      "grad_norm": 0.8756251931190491,
+      "learning_rate": 2.1957671957671956e-05,
+      "loss": 0.0536,
+      "step": 1270
+    },
+    {
+      "epoch": 30.476190476190474,
+      "grad_norm": 0.7314756512641907,
+      "learning_rate": 2.1693121693121692e-05,
+      "loss": 0.0394,
+      "step": 1280
+    },
+    {
+      "epoch": 30.714285714285715,
+      "grad_norm": 1.9777828454971313,
+      "learning_rate": 2.1428571428571428e-05,
+      "loss": 0.0346,
+      "step": 1290
+    },
+    {
+      "epoch": 30.952380952380953,
+      "grad_norm": 1.4753316640853882,
+      "learning_rate": 2.1164021164021164e-05,
+      "loss": 0.0541,
+      "step": 1300
+    },
+    {
+      "epoch": 31.0,
+      "eval_accuracy": 0.9715242881072027,
+      "eval_loss": 0.13902144134044647,
+      "eval_runtime": 104.4849,
+      "eval_samples_per_second": 5.714,
+      "eval_steps_per_second": 0.182,
+      "step": 1302
+    },
+    {
+      "epoch": 31.19047619047619,
+      "grad_norm": 1.706081748008728,
+      "learning_rate": 2.08994708994709e-05,
+      "loss": 0.0613,
+      "step": 1310
+    },
+    {
+      "epoch": 31.428571428571427,
+      "grad_norm": 0.12419818341732025,
+      "learning_rate": 2.0634920634920636e-05,
+      "loss": 0.075,
+      "step": 1320
+    },
+    {
+      "epoch": 31.666666666666668,
+      "grad_norm": 0.7071540951728821,
+      "learning_rate": 2.037037037037037e-05,
+      "loss": 0.0468,
+      "step": 1330
+    },
+    {
+      "epoch": 31.904761904761905,
+      "grad_norm": 0.10454891622066498,
+      "learning_rate": 2.0105820105820105e-05,
+      "loss": 0.0602,
+      "step": 1340
     },
     {
       "epoch": 32.0,
-      "eval_accuracy": 0.8676470588235294,
-      "eval_loss": 0.41645094752311707,
-      "eval_runtime": 29.9412,
-      "eval_samples_per_second": 4.542,
-      "eval_steps_per_second": 0.167,
-      "step": 312
-    },
-    {
-      "epoch": 32.82051282051282,
-      "grad_norm": 2.0035409927368164,
-      "learning_rate": 1.604938271604938e-05,
-      "loss": 0.1654,
-      "step": 320
+      "eval_accuracy": 0.9614740368509213,
+      "eval_loss": 0.16181902587413788,
+      "eval_runtime": 102.8803,
+      "eval_samples_per_second": 5.803,
+      "eval_steps_per_second": 0.185,
+      "step": 1344
     },
     {
-      "epoch": 32.92307692307692,
-      "eval_accuracy": 0.8676470588235294,
-      "eval_loss": 0.45997411012649536,
-      "eval_runtime": 29.5665,
-      "eval_samples_per_second": 4.6,
-      "eval_steps_per_second": 0.169,
-      "step": 321
+      "epoch": 32.142857142857146,
+      "grad_norm": 0.773094654083252,
+      "learning_rate": 1.984126984126984e-05,
+      "loss": 0.0555,
+      "step": 1350
     },
     {
-      "epoch": 33.84615384615385,
-      "grad_norm": 0.8273878693580627,
-      "learning_rate": 1.4814814814814815e-05,
-      "loss": 0.154,
-      "step": 330
+      "epoch": 32.38095238095238,
+      "grad_norm": 1.865349292755127,
+      "learning_rate": 1.9576719576719577e-05,
+      "loss": 0.0518,
+      "step": 1360
     },
     {
-      "epoch": 33.94871794871795,
-      "eval_accuracy": 0.8970588235294118,
-      "eval_loss": 0.3834398686885834,
-      "eval_runtime": 29.8196,
-      "eval_samples_per_second": 4.561,
-      "eval_steps_per_second": 0.168,
-      "step": 331
+      "epoch": 32.61904761904762,
+      "grad_norm": 0.4416976571083069,
+      "learning_rate": 1.9312169312169313e-05,
+      "loss": 0.049,
+      "step": 1370
     },
     {
-      "epoch": 34.87179487179487,
-      "grad_norm": 1.8872778415679932,
-      "learning_rate": 1.3580246913580247e-05,
-      "loss": 0.1459,
-      "step": 340
+      "epoch": 32.857142857142854,
+      "grad_norm": 2.0832931995391846,
+      "learning_rate": 1.9047619047619046e-05,
+      "loss": 0.0497,
+      "step": 1380
     },
     {
-      "epoch": 34.97435897435897,
-      "eval_accuracy": 0.8897058823529411,
-      "eval_loss": 0.36863845586776733,
-      "eval_runtime": 29.8029,
-      "eval_samples_per_second": 4.563,
-      "eval_steps_per_second": 0.168,
-      "step": 341
+      "epoch": 33.0,
+      "eval_accuracy": 0.9614740368509213,
+      "eval_loss": 0.1414780616760254,
+      "eval_runtime": 103.4955,
+      "eval_samples_per_second": 5.768,
+      "eval_steps_per_second": 0.184,
+      "step": 1386
     },
     {
-      "epoch": 35.8974358974359,
-      "grad_norm": 1.480739712715149,
-      "learning_rate": 1.2345679012345678e-05,
-      "loss": 0.1452,
-      "step": 350
+      "epoch": 33.095238095238095,
+      "grad_norm": 0.9570621848106384,
+      "learning_rate": 1.8783068783068782e-05,
+      "loss": 0.0532,
+      "step": 1390
+    },
+    {
+      "epoch": 33.333333333333336,
+      "grad_norm": 1.0575621128082275,
+      "learning_rate": 1.8518518518518518e-05,
+      "loss": 0.0555,
+      "step": 1400
+    },
+    {
+      "epoch": 33.57142857142857,
+      "grad_norm": 0.6880443096160889,
+      "learning_rate": 1.8253968253968254e-05,
+      "loss": 0.0454,
+      "step": 1410
+    },
+    {
+      "epoch": 33.80952380952381,
+      "grad_norm": 1.0119032859802246,
+      "learning_rate": 1.798941798941799e-05,
+      "loss": 0.0493,
+      "step": 1420
+    },
+    {
+      "epoch": 34.0,
+      "eval_accuracy": 0.9631490787269682,
+      "eval_loss": 0.1520875245332718,
+      "eval_runtime": 104.0599,
+      "eval_samples_per_second": 5.737,
+      "eval_steps_per_second": 0.183,
+      "step": 1428
+    },
+    {
+      "epoch": 34.04761904761905,
+      "grad_norm": 1.220492959022522,
+      "learning_rate": 1.7724867724867723e-05,
+      "loss": 0.0361,
+      "step": 1430
+    },
+    {
+      "epoch": 34.285714285714285,
+      "grad_norm": 0.5547053217887878,
+      "learning_rate": 1.746031746031746e-05,
+      "loss": 0.0412,
+      "step": 1440
+    },
+    {
+      "epoch": 34.523809523809526,
+      "grad_norm": 0.7015855312347412,
+      "learning_rate": 1.7195767195767195e-05,
+      "loss": 0.0425,
+      "step": 1450
+    },
+    {
+      "epoch": 34.76190476190476,
+      "grad_norm": 1.388316035270691,
+      "learning_rate": 1.693121693121693e-05,
+      "loss": 0.0342,
+      "step": 1460
+    },
+    {
+      "epoch": 35.0,
+      "grad_norm": 1.6578996181488037,
+      "learning_rate": 1.6666666666666667e-05,
+      "loss": 0.0606,
+      "step": 1470
+    },
+    {
+      "epoch": 35.0,
+      "eval_accuracy": 0.9698492462311558,
+      "eval_loss": 0.14287406206130981,
+      "eval_runtime": 103.938,
+      "eval_samples_per_second": 5.744,
+      "eval_steps_per_second": 0.183,
+      "step": 1470
+    },
+    {
+      "epoch": 35.23809523809524,
+      "grad_norm": 1.2321009635925293,
+      "learning_rate": 1.6402116402116404e-05,
+      "loss": 0.036,
+      "step": 1480
+    },
+    {
+      "epoch": 35.476190476190474,
+      "grad_norm": 1.9735788106918335,
+      "learning_rate": 1.6137566137566136e-05,
+      "loss": 0.0485,
+      "step": 1490
+    },
+    {
+      "epoch": 35.714285714285715,
+      "grad_norm": 0.7221766114234924,
+      "learning_rate": 1.5873015873015872e-05,
+      "loss": 0.0518,
+      "step": 1500
+    },
+    {
+      "epoch": 35.95238095238095,
+      "grad_norm": 1.484800100326538,
+      "learning_rate": 1.560846560846561e-05,
+      "loss": 0.0332,
+      "step": 1510
     },
     {
       "epoch": 36.0,
-      "eval_accuracy": 0.8676470588235294,
-      "eval_loss": 0.41742780804634094,
-      "eval_runtime": 30.4904,
-      "eval_samples_per_second": 4.46,
-      "eval_steps_per_second": 0.164,
-      "step": 351
+      "eval_accuracy": 0.964824120603015,
+      "eval_loss": 0.16714587807655334,
+      "eval_runtime": 103.1161,
+      "eval_samples_per_second": 5.79,
+      "eval_steps_per_second": 0.184,
+      "step": 1512
     },
     {
-      "epoch": 36.92307692307692,
-      "grad_norm": 2.4121947288513184,
-      "learning_rate": 1.1111111111111112e-05,
-      "loss": 0.1548,
-      "step": 360
+      "epoch": 36.19047619047619,
+      "grad_norm": 2.1778924465179443,
+      "learning_rate": 1.5343915343915344e-05,
+      "loss": 0.058,
+      "step": 1520
     },
     {
-      "epoch": 36.92307692307692,
-      "eval_accuracy": 0.9044117647058824,
-      "eval_loss": 0.379115492105484,
-      "eval_runtime": 31.2755,
-      "eval_samples_per_second": 4.348,
-      "eval_steps_per_second": 0.16,
-      "step": 360
+      "epoch": 36.42857142857143,
+      "grad_norm": 0.6789590120315552,
+      "learning_rate": 1.5079365079365079e-05,
+      "loss": 0.0341,
+      "step": 1530
     },
     {
-      "epoch": 37.94871794871795,
-      "grad_norm": 1.7541086673736572,
-      "learning_rate": 9.876543209876543e-06,
-      "loss": 0.1395,
-      "step": 370
+      "epoch": 36.666666666666664,
+      "grad_norm": 1.6879972219467163,
+      "learning_rate": 1.4814814814814815e-05,
+      "loss": 0.0395,
+      "step": 1540
     },
     {
-      "epoch": 37.94871794871795,
-      "eval_accuracy": 0.8529411764705882,
-      "eval_loss": 0.4511679708957672,
-      "eval_runtime": 29.0831,
-      "eval_samples_per_second": 4.676,
-      "eval_steps_per_second": 0.172,
-      "step": 370
+      "epoch": 36.904761904761905,
+      "grad_norm": 0.9731617569923401,
+      "learning_rate": 1.455026455026455e-05,
+      "loss": 0.0432,
+      "step": 1550
     },
     {
-      "epoch": 38.97435897435897,
-      "grad_norm": 1.2144207954406738,
-      "learning_rate": 8.641975308641975e-06,
-      "loss": 0.1333,
-      "step": 380
+      "epoch": 37.0,
+      "eval_accuracy": 0.966499162479062,
+      "eval_loss": 0.14411257207393646,
+      "eval_runtime": 103.3621,
+      "eval_samples_per_second": 5.776,
+      "eval_steps_per_second": 0.184,
+      "step": 1554
     },
     {
-      "epoch": 38.97435897435897,
-      "eval_accuracy": 0.8897058823529411,
-      "eval_loss": 0.37747910618782043,
-      "eval_runtime": 29.5567,
-      "eval_samples_per_second": 4.601,
-      "eval_steps_per_second": 0.169,
-      "step": 380
+      "epoch": 37.142857142857146,
+      "grad_norm": 1.3034260272979736,
+      "learning_rate": 1.4285714285714285e-05,
+      "loss": 0.0376,
+      "step": 1560
+    },
+    {
+      "epoch": 37.38095238095238,
+      "grad_norm": 0.8573827743530273,
+      "learning_rate": 1.4021164021164022e-05,
+      "loss": 0.0475,
+      "step": 1570
+    },
+    {
+      "epoch": 37.61904761904762,
+      "grad_norm": 0.7534766793251038,
+      "learning_rate": 1.3756613756613756e-05,
+      "loss": 0.0519,
+      "step": 1580
+    },
+    {
+      "epoch": 37.857142857142854,
+      "grad_norm": 0.7222949266433716,
+      "learning_rate": 1.3492063492063492e-05,
+      "loss": 0.0354,
+      "step": 1590
+    },
+    {
+      "epoch": 38.0,
+      "eval_accuracy": 0.9681742043551089,
+      "eval_loss": 0.15929608047008514,
+      "eval_runtime": 107.453,
+      "eval_samples_per_second": 5.556,
+      "eval_steps_per_second": 0.177,
+      "step": 1596
+    },
+    {
+      "epoch": 38.095238095238095,
+      "grad_norm": 1.8834507465362549,
+      "learning_rate": 1.3227513227513228e-05,
+      "loss": 0.0338,
+      "step": 1600
+    },
+    {
+      "epoch": 38.333333333333336,
+      "grad_norm": 0.6104734539985657,
+      "learning_rate": 1.2962962962962962e-05,
+      "loss": 0.0462,
+      "step": 1610
+    },
+    {
+      "epoch": 38.57142857142857,
+      "grad_norm": 0.7713958024978638,
+      "learning_rate": 1.2698412698412699e-05,
+      "loss": 0.0279,
+      "step": 1620
+    },
+    {
+      "epoch": 38.80952380952381,
+      "grad_norm": 0.2185550034046173,
+      "learning_rate": 1.2433862433862433e-05,
+      "loss": 0.0432,
+      "step": 1630
+    },
+    {
+      "epoch": 39.0,
+      "eval_accuracy": 0.966499162479062,
+      "eval_loss": 0.13952085375785828,
+      "eval_runtime": 107.8325,
+      "eval_samples_per_second": 5.536,
+      "eval_steps_per_second": 0.176,
+      "step": 1638
+    },
+    {
+      "epoch": 39.04761904761905,
+      "grad_norm": 1.2039453983306885,
+      "learning_rate": 1.2169312169312169e-05,
+      "loss": 0.0451,
+      "step": 1640
+    },
+    {
+      "epoch": 39.285714285714285,
+      "grad_norm": 2.02799391746521,
+      "learning_rate": 1.1904761904761905e-05,
+      "loss": 0.0249,
+      "step": 1650
+    },
+    {
+      "epoch": 39.523809523809526,
+      "grad_norm": 0.5916322469711304,
+      "learning_rate": 1.164021164021164e-05,
+      "loss": 0.0408,
+      "step": 1660
+    },
+    {
+      "epoch": 39.76190476190476,
+      "grad_norm": 2.4064204692840576,
+      "learning_rate": 1.1375661375661376e-05,
+      "loss": 0.0439,
+      "step": 1670
     },
     {
       "epoch": 40.0,
-      "grad_norm": 1.966362714767456,
-      "learning_rate": 7.4074074074074075e-06,
-      "loss": 0.1236,
-      "step": 390
+      "grad_norm": 0.05854567885398865,
+      "learning_rate": 1.1111111111111112e-05,
+      "loss": 0.0363,
+      "step": 1680
     },
     {
       "epoch": 40.0,
-      "eval_accuracy": 0.8970588235294118,
-      "eval_loss": 0.3665925860404968,
-      "eval_runtime": 29.5708,
-      "eval_samples_per_second": 4.599,
-      "eval_steps_per_second": 0.169,
-      "step": 390
+      "eval_accuracy": 0.9731993299832495,
+      "eval_loss": 0.10922118276357651,
+      "eval_runtime": 107.5538,
+      "eval_samples_per_second": 5.551,
+      "eval_steps_per_second": 0.177,
+      "step": 1680
     },
     {
-      "epoch": 40.92307692307692,
-      "eval_accuracy": 0.8970588235294118,
-      "eval_loss": 0.38919442892074585,
-      "eval_runtime": 29.9522,
-      "eval_samples_per_second": 4.541,
-      "eval_steps_per_second": 0.167,
-      "step": 399
+      "epoch": 40.23809523809524,
+      "grad_norm": 1.7765388488769531,
+      "learning_rate": 1.0846560846560846e-05,
+      "loss": 0.0337,
+      "step": 1690
     },
     {
-      "epoch": 41.02564102564103,
-      "grad_norm": 1.2910165786743164,
-      "learning_rate": 6.172839506172839e-06,
-      "loss": 0.1314,
-      "step": 400
+      "epoch": 40.476190476190474,
+      "grad_norm": 1.7078924179077148,
+      "learning_rate": 1.0582010582010582e-05,
+      "loss": 0.0272,
+      "step": 1700
     },
     {
-      "epoch": 41.94871794871795,
-      "eval_accuracy": 0.8897058823529411,
-      "eval_loss": 0.3831816613674164,
-      "eval_runtime": 30.006,
-      "eval_samples_per_second": 4.532,
-      "eval_steps_per_second": 0.167,
-      "step": 409
+      "epoch": 40.714285714285715,
+      "grad_norm": 1.3066785335540771,
+      "learning_rate": 1.0317460317460318e-05,
+      "loss": 0.0408,
+      "step": 1710
     },
     {
-      "epoch": 42.05128205128205,
-      "grad_norm": 1.5890743732452393,
-      "learning_rate": 4.938271604938272e-06,
-      "loss": 0.1322,
-      "step": 410
+      "epoch": 40.95238095238095,
+      "grad_norm": 0.261738121509552,
+      "learning_rate": 1.0052910052910053e-05,
+      "loss": 0.0288,
+      "step": 1720
     },
     {
-      "epoch": 42.97435897435897,
-      "eval_accuracy": 0.8823529411764706,
-      "eval_loss": 0.39192765951156616,
-      "eval_runtime": 30.5878,
-      "eval_samples_per_second": 4.446,
-      "eval_steps_per_second": 0.163,
-      "step": 419
+      "epoch": 41.0,
+      "eval_accuracy": 0.966499162479062,
+      "eval_loss": 0.15500447154045105,
+      "eval_runtime": 107.8053,
+      "eval_samples_per_second": 5.538,
+      "eval_steps_per_second": 0.176,
+      "step": 1722
     },
     {
-      "epoch": 43.07692307692308,
-      "grad_norm": 0.9582741260528564,
-      "learning_rate": 3.7037037037037037e-06,
-      "loss": 0.1156,
-      "step": 420
+      "epoch": 41.19047619047619,
+      "grad_norm": 0.7850339412689209,
+      "learning_rate": 9.788359788359789e-06,
+      "loss": 0.031,
+      "step": 1730
+    },
+    {
+      "epoch": 41.42857142857143,
+      "grad_norm": 1.2644939422607422,
+      "learning_rate": 9.523809523809523e-06,
+      "loss": 0.0435,
+      "step": 1740
+    },
+    {
+      "epoch": 41.666666666666664,
+      "grad_norm": 1.832721471786499,
+      "learning_rate": 9.259259259259259e-06,
+      "loss": 0.0275,
+      "step": 1750
+    },
+    {
+      "epoch": 41.904761904761905,
+      "grad_norm": 0.4885420501232147,
+      "learning_rate": 8.994708994708995e-06,
+      "loss": 0.0305,
+      "step": 1760
+    },
+    {
+      "epoch": 42.0,
+      "eval_accuracy": 0.9681742043551089,
+      "eval_loss": 0.14619530737400055,
+      "eval_runtime": 107.0057,
+      "eval_samples_per_second": 5.579,
+      "eval_steps_per_second": 0.178,
+      "step": 1764
+    },
+    {
+      "epoch": 42.142857142857146,
+      "grad_norm": 1.3336294889450073,
+      "learning_rate": 8.73015873015873e-06,
+      "loss": 0.0442,
+      "step": 1770
+    },
+    {
+      "epoch": 42.38095238095238,
+      "grad_norm": 1.4869195222854614,
+      "learning_rate": 8.465608465608466e-06,
+      "loss": 0.0367,
+      "step": 1780
+    },
+    {
+      "epoch": 42.61904761904762,
+      "grad_norm": 0.8099101781845093,
+      "learning_rate": 8.201058201058202e-06,
+      "loss": 0.0457,
+      "step": 1790
+    },
+    {
+      "epoch": 42.857142857142854,
+      "grad_norm": 0.36065608263015747,
+      "learning_rate": 7.936507936507936e-06,
+      "loss": 0.0326,
+      "step": 1800
+    },
+    {
+      "epoch": 43.0,
+      "eval_accuracy": 0.9681742043551089,
+      "eval_loss": 0.1343374401330948,
+      "eval_runtime": 107.1742,
+      "eval_samples_per_second": 5.57,
+      "eval_steps_per_second": 0.177,
+      "step": 1806
+    },
+    {
+      "epoch": 43.095238095238095,
+      "grad_norm": 0.7654176354408264,
+      "learning_rate": 7.671957671957672e-06,
+      "loss": 0.0182,
+      "step": 1810
+    },
+    {
+      "epoch": 43.333333333333336,
+      "grad_norm": 0.9363707900047302,
+      "learning_rate": 7.4074074074074075e-06,
+      "loss": 0.0364,
+      "step": 1820
+    },
+    {
+      "epoch": 43.57142857142857,
+      "grad_norm": 1.6939359903335571,
+      "learning_rate": 7.142857142857143e-06,
+      "loss": 0.0274,
+      "step": 1830
+    },
+    {
+      "epoch": 43.80952380952381,
+      "grad_norm": 0.285452663898468,
+      "learning_rate": 6.878306878306878e-06,
+      "loss": 0.027,
+      "step": 1840
     },
     {
       "epoch": 44.0,
-      "eval_accuracy": 0.8970588235294118,
-      "eval_loss": 0.369939923286438,
-      "eval_runtime": 30.3858,
-      "eval_samples_per_second": 4.476,
-      "eval_steps_per_second": 0.165,
-      "step": 429
-    },
-    {
-      "epoch": 44.1025641025641,
-      "grad_norm": 2.247335910797119,
-      "learning_rate": 2.469135802469136e-06,
-      "loss": 0.1222,
-      "step": 430
+      "eval_accuracy": 0.9731993299832495,
+      "eval_loss": 0.11093433946371078,
+      "eval_runtime": 107.916,
+      "eval_samples_per_second": 5.532,
+      "eval_steps_per_second": 0.176,
+      "step": 1848
     },
     {
-      "epoch": 44.92307692307692,
-      "eval_accuracy": 0.8970588235294118,
-      "eval_loss": 0.38276419043540955,
-      "eval_runtime": 29.9911,
-      "eval_samples_per_second": 4.535,
-      "eval_steps_per_second": 0.167,
-      "step": 438
+      "epoch": 44.04761904761905,
+      "grad_norm": 1.4939181804656982,
+      "learning_rate": 6.613756613756614e-06,
+      "loss": 0.035,
+      "step": 1850
     },
     {
-      "epoch": 45.12820512820513,
-      "grad_norm": 0.5052188038825989,
-      "learning_rate": 1.234567901234568e-06,
-      "loss": 0.1254,
-      "step": 440
+      "epoch": 44.285714285714285,
+      "grad_norm": 0.328795850276947,
+      "learning_rate": 6.349206349206349e-06,
+      "loss": 0.0328,
+      "step": 1860
     },
     {
-      "epoch": 45.94871794871795,
-      "eval_accuracy": 0.8897058823529411,
-      "eval_loss": 0.38526448607444763,
-      "eval_runtime": 31.3146,
-      "eval_samples_per_second": 4.343,
-      "eval_steps_per_second": 0.16,
-      "step": 448
+      "epoch": 44.523809523809526,
+      "grad_norm": 0.7609361410140991,
+      "learning_rate": 6.0846560846560845e-06,
+      "loss": 0.0383,
+      "step": 1870
     },
     {
-      "epoch": 46.15384615384615,
-      "grad_norm": 0.5415890216827393,
+      "epoch": 44.76190476190476,
+      "grad_norm": 0.4592248201370239,
+      "learning_rate": 5.82010582010582e-06,
+      "loss": 0.0248,
+      "step": 1880
+    },
+    {
+      "epoch": 45.0,
+      "grad_norm": 0.208053857088089,
+      "learning_rate": 5.555555555555556e-06,
+      "loss": 0.0233,
+      "step": 1890
+    },
+    {
+      "epoch": 45.0,
+      "eval_accuracy": 0.9731993299832495,
+      "eval_loss": 0.1315459907054901,
+      "eval_runtime": 105.7748,
+      "eval_samples_per_second": 5.644,
+      "eval_steps_per_second": 0.18,
+      "step": 1890
+    },
+    {
+      "epoch": 45.23809523809524,
+      "grad_norm": 0.4980098605155945,
+      "learning_rate": 5.291005291005291e-06,
+      "loss": 0.0209,
+      "step": 1900
+    },
+    {
+      "epoch": 45.476190476190474,
+      "grad_norm": 1.4569426774978638,
+      "learning_rate": 5.026455026455026e-06,
+      "loss": 0.0249,
+      "step": 1910
+    },
+    {
+      "epoch": 45.714285714285715,
+      "grad_norm": 0.21140669286251068,
+      "learning_rate": 4.7619047619047615e-06,
+      "loss": 0.0274,
+      "step": 1920
+    },
+    {
+      "epoch": 45.95238095238095,
+      "grad_norm": 0.9561221599578857,
+      "learning_rate": 4.497354497354498e-06,
+      "loss": 0.042,
+      "step": 1930
+    },
+    {
+      "epoch": 46.0,
+      "eval_accuracy": 0.9731993299832495,
+      "eval_loss": 0.1261177957057953,
+      "eval_runtime": 106.0727,
+      "eval_samples_per_second": 5.628,
+      "eval_steps_per_second": 0.179,
+      "step": 1932
+    },
+    {
+      "epoch": 46.19047619047619,
+      "grad_norm": 1.0604710578918457,
+      "learning_rate": 4.232804232804233e-06,
+      "loss": 0.022,
+      "step": 1940
+    },
+    {
+      "epoch": 46.42857142857143,
+      "grad_norm": 2.230454683303833,
+      "learning_rate": 3.968253968253968e-06,
+      "loss": 0.0351,
+      "step": 1950
+    },
+    {
+      "epoch": 46.666666666666664,
+      "grad_norm": 0.6946862936019897,
+      "learning_rate": 3.7037037037037037e-06,
+      "loss": 0.0301,
+      "step": 1960
+    },
+    {
+      "epoch": 46.904761904761905,
+      "grad_norm": 0.760875403881073,
+      "learning_rate": 3.439153439153439e-06,
+      "loss": 0.0251,
+      "step": 1970
+    },
+    {
+      "epoch": 47.0,
+      "eval_accuracy": 0.9731993299832495,
+      "eval_loss": 0.13198845088481903,
+      "eval_runtime": 107.949,
+      "eval_samples_per_second": 5.53,
+      "eval_steps_per_second": 0.176,
+      "step": 1974
+    },
+    {
+      "epoch": 47.142857142857146,
+      "grad_norm": 0.19793380796909332,
+      "learning_rate": 3.1746031746031746e-06,
+      "loss": 0.0194,
+      "step": 1980
+    },
+    {
+      "epoch": 47.38095238095238,
+      "grad_norm": 1.171229362487793,
+      "learning_rate": 2.91005291005291e-06,
+      "loss": 0.0207,
+      "step": 1990
+    },
+    {
+      "epoch": 47.61904761904762,
+      "grad_norm": 1.4025137424468994,
+      "learning_rate": 2.6455026455026455e-06,
+      "loss": 0.0264,
+      "step": 2000
+    },
+    {
+      "epoch": 47.857142857142854,
+      "grad_norm": 0.5469716191291809,
+      "learning_rate": 2.3809523809523808e-06,
+      "loss": 0.041,
+      "step": 2010
+    },
+    {
+      "epoch": 48.0,
+      "eval_accuracy": 0.9731993299832495,
+      "eval_loss": 0.12820355594158173,
+      "eval_runtime": 107.9105,
+      "eval_samples_per_second": 5.532,
+      "eval_steps_per_second": 0.176,
+      "step": 2016
+    },
+    {
+      "epoch": 48.095238095238095,
+      "grad_norm": 0.5962417125701904,
+      "learning_rate": 2.1164021164021164e-06,
+      "loss": 0.0257,
+      "step": 2020
+    },
+    {
+      "epoch": 48.333333333333336,
+      "grad_norm": 1.9686793088912964,
+      "learning_rate": 1.8518518518518519e-06,
+      "loss": 0.0297,
+      "step": 2030
+    },
+    {
+      "epoch": 48.57142857142857,
+      "grad_norm": 1.7944121360778809,
+      "learning_rate": 1.5873015873015873e-06,
+      "loss": 0.0272,
+      "step": 2040
+    },
+    {
+      "epoch": 48.80952380952381,
+      "grad_norm": 0.4869058430194855,
+      "learning_rate": 1.3227513227513228e-06,
+      "loss": 0.0445,
+      "step": 2050
+    },
+    {
+      "epoch": 49.0,
+      "eval_accuracy": 0.9731993299832495,
+      "eval_loss": 0.12962684035301208,
+      "eval_runtime": 107.6607,
+      "eval_samples_per_second": 5.545,
+      "eval_steps_per_second": 0.176,
+      "step": 2058
+    },
+    {
+      "epoch": 49.04761904761905,
+      "grad_norm": 0.599247932434082,
+      "learning_rate": 1.0582010582010582e-06,
+      "loss": 0.0289,
+      "step": 2060
+    },
+    {
+      "epoch": 49.285714285714285,
+      "grad_norm": 1.3819619417190552,
+      "learning_rate": 7.936507936507937e-07,
+      "loss": 0.0236,
+      "step": 2070
+    },
+    {
+      "epoch": 49.523809523809526,
+      "grad_norm": 0.9784670472145081,
+      "learning_rate": 5.291005291005291e-07,
+      "loss": 0.0306,
+      "step": 2080
+    },
+    {
+      "epoch": 49.76190476190476,
+      "grad_norm": 0.8563596606254578,
+      "learning_rate": 2.6455026455026455e-07,
+      "loss": 0.0212,
+      "step": 2090
+    },
+    {
+      "epoch": 50.0,
+      "grad_norm": 1.10076904296875,
       "learning_rate": 0.0,
-      "loss": 0.1129,
-      "step": 450
+      "loss": 0.0308,
+      "step": 2100
     },
     {
-      "epoch": 46.15384615384615,
-      "eval_accuracy": 0.8897058823529411,
-      "eval_loss": 0.38528940081596375,
-      "eval_runtime": 30.9585,
-      "eval_samples_per_second": 4.393,
-      "eval_steps_per_second": 0.162,
-      "step": 450
+      "epoch": 50.0,
+      "eval_accuracy": 0.9731993299832495,
+      "eval_loss": 0.13253676891326904,
+      "eval_runtime": 110.4625,
+      "eval_samples_per_second": 5.405,
+      "eval_steps_per_second": 0.172,
+      "step": 2100
     },
     {
-      "epoch": 46.15384615384615,
-      "step": 450,
-      "total_flos": 4.3781443993328026e+18,
-      "train_loss": 0.4221388864517212,
-      "train_runtime": 14616.4414,
-      "train_samples_per_second": 4.187,
-      "train_steps_per_second": 0.031
+      "epoch": 50.0,
+      "step": 2100,
+      "total_flos": 2.0803097508518707e+19,
+      "train_loss": 0.1354110169055916,
+      "train_runtime": 52600.7464,
+      "train_samples_per_second": 5.104,
+      "train_steps_per_second": 0.04
     }
   ],
   "logging_steps": 10,
-  "max_steps": 450,
+  "max_steps": 2100,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 50,
   "save_steps": 500,
@@ -773,7 +1955,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 4.3781443993328026e+18,
+  "total_flos": 2.0803097508518707e+19,
   "train_batch_size": 32,
   "trial_name": null,
   "trial_params": null