End of training

README.md

@@ -3,6 +3,8 @@ library_name: transformers
 license: apache-2.0
 base_model: timm/resnet18.a1_in1k
 tags:
+- image-classification
+- vision
 - generated_from_trainer
 metrics:
 - accuracy
@@ -16,7 +18,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # vit-base-beans
 
-This model is a fine-tuned version of [timm/resnet18.a1_in1k](https://huggingface.co/timm/resnet18.a1_in1k) on an unknown dataset.
+This model is a fine-tuned version of [timm/resnet18.a1_in1k](https://huggingface.co/timm/resnet18.a1_in1k) on the beans dataset.
 It achieves the following results on the evaluation set:
 - Loss: 0.7380
 - Accuracy: 0.8195

all_results.json

@@ -1,13 +1,13 @@
 {
     "epoch": 15.0,
-    "eval_accuracy": 0.8045112781954887,
-    "eval_loss": 0.7388833165168762,
-    "eval_runtime": 0.856,
-    "eval_samples_per_second": 155.372,
-    "eval_steps_per_second": 19.86,
+    "eval_accuracy": 0.8195488721804511,
+    "eval_loss": 0.738013505935669,
+    "eval_runtime": 0.7577,
+    "eval_samples_per_second": 175.539,
+    "eval_steps_per_second": 22.437,
     "total_flos": 1.5658365504595968e+17,
-    "train_loss": 0.923483537771763,
-    "train_runtime": 151.8082,
-    "train_samples_per_second": 102.168,
-    "train_steps_per_second": 12.845
+    "train_loss": 0.9225189507313264,
+    "train_runtime": 140.2,
+    "train_samples_per_second": 110.628,
+    "train_steps_per_second": 13.909
 }

eval_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 15.0,
-    "eval_accuracy": 0.8045112781954887,
-    "eval_loss": 0.7388833165168762,
-    "eval_runtime": 0.856,
-    "eval_samples_per_second": 155.372,
-    "eval_steps_per_second": 19.86
+    "eval_accuracy": 0.8195488721804511,
+    "eval_loss": 0.738013505935669,
+    "eval_runtime": 0.7577,
+    "eval_samples_per_second": 175.539,
+    "eval_steps_per_second": 22.437
 }

runs/Nov19_21-27-50_ip-10-90-1-182/events.out.tfevents.1732051817.ip-10-90-1-182.3342435.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a2ba9d94ca3ff68e13f36ca6cc36d1506b4c47f847c9abc343728c4942b877b9
+size 411

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 15.0,
     "total_flos": 1.5658365504595968e+17,
-    "train_loss": 0.923483537771763,
-    "train_runtime": 151.8082,
-    "train_samples_per_second": 102.168,
-    "train_steps_per_second": 12.845
+    "train_loss": 0.9225189507313264,
+    "train_runtime": 140.2,
+    "train_samples_per_second": 110.628,
+    "train_steps_per_second": 13.909
 }

trainer_state.json

@@ -1,5 +1,5 @@
 {
-  "best_metric": 0.7388833165168762,
+  "best_metric": 0.738013505935669,
   "best_model_checkpoint": "./beans_outputs/checkpoint-1950",
   "epoch": 15.0,
   "eval_steps": 500,
@@ -10,1512 +10,1512 @@
   "log_history": [
     {
       "epoch": 0.07692307692307693,
-      "grad_norm": 2.1216182708740234,
+      "grad_norm": 2.099634885787964,
       "learning_rate": 1.98974358974359e-05,
       "loss": 1.1239,
       "step": 10
     },
     {
       "epoch": 0.15384615384615385,
-      "grad_norm": 1.8308407068252563,
+      "grad_norm": 1.7766097784042358,
       "learning_rate": 1.9794871794871798e-05,
       "loss": 1.1221,
       "step": 20
     },
     {
       "epoch": 0.23076923076923078,
-      "grad_norm": 1.9811038970947266,
+      "grad_norm": 1.9635370969772339,
       "learning_rate": 1.9692307692307696e-05,
-      "loss": 1.1163,
+      "loss": 1.1164,
       "step": 30
     },
     {
       "epoch": 0.3076923076923077,
-      "grad_norm": 2.4690616130828857,
+      "grad_norm": 2.459880828857422,
       "learning_rate": 1.958974358974359e-05,
-      "loss": 1.1046,
+      "loss": 1.1045,
       "step": 40
     },
     {
       "epoch": 0.38461538461538464,
-      "grad_norm": 1.6778465509414673,
+      "grad_norm": 1.7044235467910767,
       "learning_rate": 1.9487179487179488e-05,
-      "loss": 1.1082,
+      "loss": 1.1083,
       "step": 50
     },
     {
       "epoch": 0.46153846153846156,
-      "grad_norm": 1.4839682579040527,
+      "grad_norm": 1.4914859533309937,
       "learning_rate": 1.9384615384615386e-05,
-      "loss": 1.1043,
+      "loss": 1.1044,
       "step": 60
     },
     {
       "epoch": 0.5384615384615384,
-      "grad_norm": 2.6265501976013184,
+      "grad_norm": 2.6082077026367188,
       "learning_rate": 1.9282051282051284e-05,
-      "loss": 1.1028,
+      "loss": 1.1027,
       "step": 70
     },
     {
       "epoch": 0.6153846153846154,
-      "grad_norm": 2.565593719482422,
+      "grad_norm": 2.722012996673584,
       "learning_rate": 1.9179487179487182e-05,
-      "loss": 1.0909,
+      "loss": 1.0908,
       "step": 80
     },
     {
       "epoch": 0.6923076923076923,
-      "grad_norm": 2.020240545272827,
+      "grad_norm": 2.029851198196411,
       "learning_rate": 1.907692307692308e-05,
-      "loss": 1.0939,
+      "loss": 1.0941,
       "step": 90
     },
     {
       "epoch": 0.7692307692307693,
-      "grad_norm": 2.49810791015625,
+      "grad_norm": 2.4984819889068604,
       "learning_rate": 1.8974358974358975e-05,
-      "loss": 1.0917,
+      "loss": 1.0913,
       "step": 100
     },
     {
       "epoch": 0.8461538461538461,
-      "grad_norm": 1.8923050165176392,
+      "grad_norm": 1.7019662857055664,
       "learning_rate": 1.8871794871794873e-05,
-      "loss": 1.0937,
+      "loss": 1.0945,
       "step": 110
     },
     {
       "epoch": 0.9230769230769231,
-      "grad_norm": 2.2869341373443604,
+      "grad_norm": 2.2935667037963867,
       "learning_rate": 1.876923076923077e-05,
-      "loss": 1.1001,
+      "loss": 1.0994,
       "step": 120
     },
     {
       "epoch": 1.0,
-      "grad_norm": 4.069129467010498,
+      "grad_norm": 3.9286272525787354,
       "learning_rate": 1.866666666666667e-05,
-      "loss": 1.0863,
+      "loss": 1.0861,
       "step": 130
     },
     {
       "epoch": 1.0,
       "eval_accuracy": 0.42857142857142855,
-      "eval_loss": 1.0881553888320923,
-      "eval_runtime": 0.9817,
-      "eval_samples_per_second": 135.479,
-      "eval_steps_per_second": 17.317,
+      "eval_loss": 1.0880533456802368,
+      "eval_runtime": 0.7946,
+      "eval_samples_per_second": 167.381,
+      "eval_steps_per_second": 21.395,
       "step": 130
     },
     {
       "epoch": 1.0769230769230769,
-      "grad_norm": 1.7320090532302856,
+      "grad_norm": 2.083320140838623,
       "learning_rate": 1.8564102564102567e-05,
-      "loss": 1.0792,
+      "loss": 1.0772,
       "step": 140
     },
     {
       "epoch": 1.1538461538461537,
-      "grad_norm": 2.5494987964630127,
+      "grad_norm": 2.711829423904419,
       "learning_rate": 1.8461538461538465e-05,
       "loss": 1.072,
       "step": 150
     },
     {
       "epoch": 1.2307692307692308,
-      "grad_norm": 1.4740803241729736,
+      "grad_norm": 1.5575768947601318,
       "learning_rate": 1.835897435897436e-05,
-      "loss": 1.0773,
+      "loss": 1.0787,
       "step": 160
     },
     {
       "epoch": 1.3076923076923077,
-      "grad_norm": 2.3478026390075684,
+      "grad_norm": 2.3275976181030273,
       "learning_rate": 1.8256410256410257e-05,
-      "loss": 1.0907,
+      "loss": 1.0935,
       "step": 170
     },
     {
       "epoch": 1.3846153846153846,
-      "grad_norm": 2.1786012649536133,
+      "grad_norm": 2.1124396324157715,
       "learning_rate": 1.8153846153846155e-05,
-      "loss": 1.08,
+      "loss": 1.0806,
       "step": 180
     },
     {
       "epoch": 1.4615384615384617,
-      "grad_norm": 2.1593339443206787,
+      "grad_norm": 2.1334290504455566,
       "learning_rate": 1.8051282051282053e-05,
-      "loss": 1.0789,
+      "loss": 1.0796,
       "step": 190
     },
     {
       "epoch": 1.5384615384615383,
-      "grad_norm": 2.1174795627593994,
+      "grad_norm": 1.7488656044006348,
       "learning_rate": 1.794871794871795e-05,
-      "loss": 1.0779,
+      "loss": 1.0762,
       "step": 200
     },
     {
       "epoch": 1.6153846153846154,
-      "grad_norm": 1.8251662254333496,
+      "grad_norm": 1.7498658895492554,
       "learning_rate": 1.784615384615385e-05,
-      "loss": 1.067,
+      "loss": 1.0682,
       "step": 210
     },
     {
       "epoch": 1.6923076923076923,
-      "grad_norm": 3.535820245742798,
+      "grad_norm": 3.5284996032714844,
       "learning_rate": 1.7743589743589744e-05,
-      "loss": 1.0701,
+      "loss": 1.0699,
       "step": 220
     },
     {
       "epoch": 1.7692307692307692,
-      "grad_norm": 2.0242528915405273,
+      "grad_norm": 2.088862895965576,
       "learning_rate": 1.7641025641025642e-05,
-      "loss": 1.0714,
+      "loss": 1.0707,
       "step": 230
     },
     {
       "epoch": 1.8461538461538463,
-      "grad_norm": 2.1979262828826904,
+      "grad_norm": 1.9373514652252197,
       "learning_rate": 1.753846153846154e-05,
-      "loss": 1.0868,
+      "loss": 1.0864,
       "step": 240
     },
     {
       "epoch": 1.9230769230769231,
-      "grad_norm": 2.096327543258667,
+      "grad_norm": 2.00813889503479,
       "learning_rate": 1.7435897435897438e-05,
-      "loss": 1.0682,
+      "loss": 1.0681,
       "step": 250
     },
     {
       "epoch": 2.0,
-      "grad_norm": 3.008599042892456,
+      "grad_norm": 3.0624523162841797,
       "learning_rate": 1.7333333333333336e-05,
-      "loss": 1.063,
+      "loss": 1.0631,
       "step": 260
     },
     {
       "epoch": 2.0,
       "eval_accuracy": 0.5413533834586466,
-      "eval_loss": 1.0589852333068848,
-      "eval_runtime": 0.7866,
-      "eval_samples_per_second": 169.088,
-      "eval_steps_per_second": 21.613,
+      "eval_loss": 1.0597343444824219,
+      "eval_runtime": 0.7394,
+      "eval_samples_per_second": 179.866,
+      "eval_steps_per_second": 22.99,
       "step": 260
     },
     {
       "epoch": 2.076923076923077,
-      "grad_norm": 2.03446888923645,
+      "grad_norm": 2.6268935203552246,
       "learning_rate": 1.7230769230769234e-05,
-      "loss": 1.0712,
+      "loss": 1.0716,
       "step": 270
     },
     {
       "epoch": 2.1538461538461537,
-      "grad_norm": 2.4047350883483887,
+      "grad_norm": 2.3874154090881348,
       "learning_rate": 1.7128205128205128e-05,
-      "loss": 1.0593,
+      "loss": 1.0599,
       "step": 280
     },
     {
       "epoch": 2.230769230769231,
-      "grad_norm": 2.37992262840271,
+      "grad_norm": 2.3503990173339844,
       "learning_rate": 1.7025641025641026e-05,
-      "loss": 1.0577,
+      "loss": 1.054,
       "step": 290
     },
     {
       "epoch": 2.3076923076923075,
-      "grad_norm": 2.940575361251831,
+      "grad_norm": 2.3312108516693115,
       "learning_rate": 1.6923076923076924e-05,
-      "loss": 1.044,
+      "loss": 1.0436,
       "step": 300
     },
     {
       "epoch": 2.3846153846153846,
-      "grad_norm": 1.983780026435852,
+      "grad_norm": 1.9090198278427124,
       "learning_rate": 1.6820512820512822e-05,
-      "loss": 1.0573,
+      "loss": 1.0569,
       "step": 310
     },
     {
       "epoch": 2.4615384615384617,
-      "grad_norm": 1.9737600088119507,
+      "grad_norm": 1.9505618810653687,
       "learning_rate": 1.671794871794872e-05,
-      "loss": 1.048,
+      "loss": 1.0479,
       "step": 320
     },
     {
       "epoch": 2.5384615384615383,
-      "grad_norm": 1.4561364650726318,
+      "grad_norm": 1.4548966884613037,
       "learning_rate": 1.6615384615384618e-05,
-      "loss": 1.0317,
+      "loss": 1.0304,
       "step": 330
     },
     {
       "epoch": 2.6153846153846154,
-      "grad_norm": 2.187842845916748,
+      "grad_norm": 2.1505134105682373,
       "learning_rate": 1.6512820512820513e-05,
-      "loss": 1.0482,
+      "loss": 1.0467,
       "step": 340
     },
     {
       "epoch": 2.6923076923076925,
-      "grad_norm": 1.9264109134674072,
+      "grad_norm": 2.8420169353485107,
       "learning_rate": 1.641025641025641e-05,
-      "loss": 1.0491,
+      "loss": 1.0509,
       "step": 350
     },
     {
       "epoch": 2.769230769230769,
-      "grad_norm": 1.8492430448532104,
+      "grad_norm": 1.8315626382827759,
       "learning_rate": 1.630769230769231e-05,
-      "loss": 1.0409,
+      "loss": 1.0407,
       "step": 360
     },
     {
       "epoch": 2.8461538461538463,
-      "grad_norm": 2.1464972496032715,
+      "grad_norm": 1.8499083518981934,
       "learning_rate": 1.6205128205128207e-05,
-      "loss": 1.0359,
+      "loss": 1.0344,
       "step": 370
     },
     {
       "epoch": 2.9230769230769234,
-      "grad_norm": 1.8184483051300049,
+      "grad_norm": 1.9010175466537476,
       "learning_rate": 1.6102564102564105e-05,
-      "loss": 1.0237,
+      "loss": 1.0219,
       "step": 380
     },
     {
       "epoch": 3.0,
-      "grad_norm": 4.10977029800415,
+      "grad_norm": 4.809950828552246,
       "learning_rate": 1.6000000000000003e-05,
-      "loss": 1.0447,
+      "loss": 1.0443,
       "step": 390
     },
     {
       "epoch": 3.0,
-      "eval_accuracy": 0.6992481203007519,
-      "eval_loss": 1.022922396659851,
-      "eval_runtime": 0.831,
-      "eval_samples_per_second": 160.057,
-      "eval_steps_per_second": 20.458,
+      "eval_accuracy": 0.6691729323308271,
+      "eval_loss": 1.0225275754928589,
+      "eval_runtime": 0.7405,
+      "eval_samples_per_second": 179.608,
+      "eval_steps_per_second": 22.957,
       "step": 390
     },
     {
       "epoch": 3.076923076923077,
-      "grad_norm": 2.4164669513702393,
+      "grad_norm": 2.195730686187744,
       "learning_rate": 1.5897435897435897e-05,
-      "loss": 1.0417,
+      "loss": 1.0414,
       "step": 400
     },
     {
       "epoch": 3.1538461538461537,
-      "grad_norm": 1.9611194133758545,
+      "grad_norm": 2.1667442321777344,
       "learning_rate": 1.5794871794871795e-05,
-      "loss": 1.03,
+      "loss": 1.0302,
       "step": 410
     },
     {
       "epoch": 3.230769230769231,
-      "grad_norm": 2.1152236461639404,
+      "grad_norm": 2.149664878845215,
       "learning_rate": 1.5692307692307693e-05,
-      "loss": 1.038,
+      "loss": 1.0365,
       "step": 420
     },
     {
       "epoch": 3.3076923076923075,
-      "grad_norm": 2.2656948566436768,
+      "grad_norm": 2.2559187412261963,
       "learning_rate": 1.558974358974359e-05,
-      "loss": 1.0163,
+      "loss": 1.0159,
       "step": 430
     },
     {
       "epoch": 3.3846153846153846,
-      "grad_norm": 2.1608102321624756,
+      "grad_norm": 2.165260076522827,
       "learning_rate": 1.548717948717949e-05,
-      "loss": 1.0277,
+      "loss": 1.0274,
       "step": 440
     },
     {
       "epoch": 3.4615384615384617,
-      "grad_norm": 1.7433905601501465,
+      "grad_norm": 1.799578309059143,
       "learning_rate": 1.5384615384615387e-05,
-      "loss": 1.0379,
+      "loss": 1.0371,
       "step": 450
     },
     {
       "epoch": 3.5384615384615383,
-      "grad_norm": 2.625889778137207,
+      "grad_norm": 2.441862106323242,
       "learning_rate": 1.5282051282051282e-05,
-      "loss": 1.0314,
+      "loss": 1.0322,
       "step": 460
     },
     {
       "epoch": 3.6153846153846154,
-      "grad_norm": 1.6532872915267944,
+      "grad_norm": 1.571476936340332,
       "learning_rate": 1.517948717948718e-05,
-      "loss": 1.0052,
+      "loss": 1.0046,
       "step": 470
     },
     {
       "epoch": 3.6923076923076925,
-      "grad_norm": 2.715625524520874,
+      "grad_norm": 2.725184440612793,
       "learning_rate": 1.5076923076923078e-05,
-      "loss": 1.0101,
+      "loss": 1.0081,
       "step": 480
     },
     {
       "epoch": 3.769230769230769,
-      "grad_norm": 1.6237256526947021,
+      "grad_norm": 1.5710386037826538,
       "learning_rate": 1.4974358974358976e-05,
-      "loss": 0.9981,
+      "loss": 0.9977,
       "step": 490
     },
     {
       "epoch": 3.8461538461538463,
-      "grad_norm": 1.582321286201477,
+      "grad_norm": 1.6186603307724,
       "learning_rate": 1.4871794871794874e-05,
-      "loss": 1.0226,
+      "loss": 1.0217,
       "step": 500
     },
     {
       "epoch": 3.9230769230769234,
-      "grad_norm": 1.7656166553497314,
+      "grad_norm": 1.8194940090179443,
       "learning_rate": 1.4769230769230772e-05,
-      "loss": 0.9989,
+      "loss": 0.997,
       "step": 510
     },
     {
       "epoch": 4.0,
-      "grad_norm": 5.592857837677002,
+      "grad_norm": 5.478014945983887,
       "learning_rate": 1.4666666666666666e-05,
-      "loss": 1.0223,
+      "loss": 1.0218,
       "step": 520
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.6917293233082706,
-      "eval_loss": 0.9968266487121582,
-      "eval_runtime": 0.8013,
-      "eval_samples_per_second": 165.971,
-      "eval_steps_per_second": 21.214,
+      "eval_accuracy": 0.6842105263157895,
+      "eval_loss": 0.9959980845451355,
+      "eval_runtime": 0.7638,
+      "eval_samples_per_second": 174.124,
+      "eval_steps_per_second": 22.256,
       "step": 520
     },
     {
       "epoch": 4.076923076923077,
-      "grad_norm": 2.7708559036254883,
+      "grad_norm": 2.7807815074920654,
       "learning_rate": 1.4564102564102564e-05,
-      "loss": 1.0136,
+      "loss": 1.0126,
       "step": 530
     },
     {
       "epoch": 4.153846153846154,
-      "grad_norm": 1.880313515663147,
+      "grad_norm": 1.8810123205184937,
       "learning_rate": 1.4461538461538462e-05,
-      "loss": 1.0166,
+      "loss": 1.0142,
       "step": 540
     },
     {
       "epoch": 4.230769230769231,
-      "grad_norm": 2.6722023487091064,
+      "grad_norm": 2.737741708755493,
       "learning_rate": 1.435897435897436e-05,
-      "loss": 1.0238,
+      "loss": 1.0241,
       "step": 550
     },
     {
       "epoch": 4.3076923076923075,
-      "grad_norm": 2.3504159450531006,
+      "grad_norm": 2.183462381362915,
       "learning_rate": 1.4256410256410258e-05,
-      "loss": 1.0338,
+      "loss": 1.0324,
       "step": 560
     },
     {
       "epoch": 4.384615384615385,
-      "grad_norm": 2.2466416358947754,
+      "grad_norm": 1.9445362091064453,
       "learning_rate": 1.4153846153846156e-05,
-      "loss": 0.9838,
+      "loss": 0.9823,
       "step": 570
     },
     {
       "epoch": 4.461538461538462,
-      "grad_norm": 2.590592861175537,
+      "grad_norm": 2.8362057209014893,
       "learning_rate": 1.405128205128205e-05,
-      "loss": 0.9915,
+      "loss": 0.9919,
       "step": 580
     },
     {
       "epoch": 4.538461538461538,
-      "grad_norm": 2.3943469524383545,
+      "grad_norm": 2.5588510036468506,
       "learning_rate": 1.3948717948717949e-05,
-      "loss": 0.9885,
+      "loss": 0.987,
       "step": 590
     },
     {
       "epoch": 4.615384615384615,
-      "grad_norm": 1.8981530666351318,
+      "grad_norm": 1.740641474723816,
       "learning_rate": 1.3846153846153847e-05,
-      "loss": 0.9802,
+      "loss": 0.9778,
       "step": 600
     },
     {
       "epoch": 4.6923076923076925,
-      "grad_norm": 3.2020835876464844,
+      "grad_norm": 3.247788429260254,
       "learning_rate": 1.3743589743589745e-05,
-      "loss": 0.9993,
+      "loss": 0.9978,
       "step": 610
     },
     {
       "epoch": 4.769230769230769,
-      "grad_norm": 2.171642541885376,
+      "grad_norm": 2.136213779449463,
       "learning_rate": 1.3641025641025643e-05,
-      "loss": 1.0033,
+      "loss": 1.0067,
       "step": 620
     },
     {
       "epoch": 4.846153846153846,
-      "grad_norm": 1.8594719171524048,
+      "grad_norm": 1.8435287475585938,
       "learning_rate": 1.353846153846154e-05,
-      "loss": 0.9689,
+      "loss": 0.9691,
       "step": 630
     },
     {
       "epoch": 4.923076923076923,
-      "grad_norm": 2.3601503372192383,
+      "grad_norm": 2.3936328887939453,
       "learning_rate": 1.3435897435897435e-05,
-      "loss": 0.9703,
+      "loss": 0.9692,
       "step": 640
     },
     {
       "epoch": 5.0,
-      "grad_norm": 5.1482744216918945,
+      "grad_norm": 4.921462535858154,
       "learning_rate": 1.3333333333333333e-05,
-      "loss": 1.0,
+      "loss": 1.0028,
       "step": 650
     },
     {
       "epoch": 5.0,
-      "eval_accuracy": 0.7518796992481203,
-      "eval_loss": 0.9575085639953613,
-      "eval_runtime": 0.8114,
-      "eval_samples_per_second": 163.92,
-      "eval_steps_per_second": 20.952,
+      "eval_accuracy": 0.7293233082706767,
+      "eval_loss": 0.9568173289299011,
+      "eval_runtime": 0.7715,
+      "eval_samples_per_second": 172.394,
+      "eval_steps_per_second": 22.035,
       "step": 650
     },
     {
       "epoch": 5.076923076923077,
-      "grad_norm": 1.8753620386123657,
+      "grad_norm": 2.051286458969116,
       "learning_rate": 1.3230769230769231e-05,
-      "loss": 0.9664,
+      "loss": 0.9638,
       "step": 660
     },
     {
       "epoch": 5.153846153846154,
-      "grad_norm": 2.668313503265381,
+      "grad_norm": 2.7295546531677246,
       "learning_rate": 1.312820512820513e-05,
-      "loss": 1.0055,
+      "loss": 1.0056,
       "step": 670
     },
     {
       "epoch": 5.230769230769231,
-      "grad_norm": 2.544369697570801,
+      "grad_norm": 2.3378870487213135,
       "learning_rate": 1.3025641025641027e-05,
-      "loss": 0.9936,
+      "loss": 0.9916,
       "step": 680
     },
     {
       "epoch": 5.3076923076923075,
-      "grad_norm": 2.7921175956726074,
+      "grad_norm": 2.754603862762451,
       "learning_rate": 1.2923076923076925e-05,
-      "loss": 0.9852,
+      "loss": 0.9869,
       "step": 690
     },
     {
       "epoch": 5.384615384615385,
-      "grad_norm": 2.6936609745025635,
+      "grad_norm": 2.6367554664611816,
       "learning_rate": 1.2820512820512823e-05,
-      "loss": 0.9725,
+      "loss": 0.9743,
       "step": 700
     },
     {
       "epoch": 5.461538461538462,
-      "grad_norm": 2.4392247200012207,
+      "grad_norm": 2.152855396270752,
       "learning_rate": 1.2717948717948718e-05,
-      "loss": 0.9653,
+      "loss": 0.9669,
       "step": 710
     },
     {
       "epoch": 5.538461538461538,
-      "grad_norm": 1.9628013372421265,
+      "grad_norm": 1.970173954963684,
       "learning_rate": 1.2615384615384616e-05,
-      "loss": 0.9497,
+      "loss": 0.9496,
       "step": 720
     },
     {
       "epoch": 5.615384615384615,
-      "grad_norm": 2.3558847904205322,
+      "grad_norm": 2.49542498588562,
       "learning_rate": 1.2512820512820514e-05,
-      "loss": 0.9646,
+      "loss": 0.9644,
       "step": 730
     },
     {
       "epoch": 5.6923076923076925,
-      "grad_norm": 2.4973325729370117,
+      "grad_norm": 2.5210750102996826,
       "learning_rate": 1.2410256410256412e-05,
-      "loss": 0.9458,
+      "loss": 0.9453,
       "step": 740
     },
     {
       "epoch": 5.769230769230769,
-      "grad_norm": 2.091280221939087,
+      "grad_norm": 2.0949254035949707,
       "learning_rate": 1.230769230769231e-05,
-      "loss": 0.9506,
+      "loss": 0.95,
       "step": 750
     },
     {
       "epoch": 5.846153846153846,
-      "grad_norm": 2.1584551334381104,
+      "grad_norm": 2.167081594467163,
       "learning_rate": 1.2205128205128208e-05,
-      "loss": 0.9517,
+      "loss": 0.9519,
       "step": 760
     },
     {
       "epoch": 5.923076923076923,
-      "grad_norm": 3.8231847286224365,
+      "grad_norm": 3.6734519004821777,
       "learning_rate": 1.2102564102564102e-05,
-      "loss": 0.9375,
+      "loss": 0.9376,
       "step": 770
     },
     {
       "epoch": 6.0,
-      "grad_norm": 4.746103286743164,
+      "grad_norm": 4.938449859619141,
       "learning_rate": 1.2e-05,
-      "loss": 0.9726,
+      "loss": 0.9752,
       "step": 780
     },
     {
       "epoch": 6.0,
-      "eval_accuracy": 0.7744360902255639,
-      "eval_loss": 0.92984938621521,
-      "eval_runtime": 0.8125,
-      "eval_samples_per_second": 163.686,
-      "eval_steps_per_second": 20.922,
+      "eval_accuracy": 0.7669172932330827,
+      "eval_loss": 0.9278604984283447,
+      "eval_runtime": 0.7739,
+      "eval_samples_per_second": 171.864,
+      "eval_steps_per_second": 21.968,
       "step": 780
     },
     {
       "epoch": 6.076923076923077,
-      "grad_norm": 2.8764736652374268,
+      "grad_norm": 2.8171768188476562,
       "learning_rate": 1.1897435897435898e-05,
-      "loss": 0.9424,
+      "loss": 0.9423,
       "step": 790
     },
     {
       "epoch": 6.153846153846154,
-      "grad_norm": 2.512113332748413,
+      "grad_norm": 2.5053982734680176,
       "learning_rate": 1.1794871794871796e-05,
-      "loss": 0.9378,
+      "loss": 0.9382,
       "step": 800
     },
     {
       "epoch": 6.230769230769231,
-      "grad_norm": 1.8056284189224243,
+      "grad_norm": 1.875543475151062,
       "learning_rate": 1.1692307692307694e-05,
-      "loss": 0.9309,
+      "loss": 0.9318,
       "step": 810
     },
     {
       "epoch": 6.3076923076923075,
-      "grad_norm": 2.3125550746917725,
+      "grad_norm": 2.222604513168335,
       "learning_rate": 1.1589743589743592e-05,
-      "loss": 0.9316,
+      "loss": 0.9313,
       "step": 820
     },
     {
       "epoch": 6.384615384615385,
-      "grad_norm": 2.485017776489258,
+      "grad_norm": 2.5009353160858154,
       "learning_rate": 1.1487179487179487e-05,
-      "loss": 0.9359,
+      "loss": 0.9354,
       "step": 830
     },
     {
       "epoch": 6.461538461538462,
-      "grad_norm": 3.3460640907287598,
+      "grad_norm": 3.2472548484802246,
       "learning_rate": 1.1384615384615385e-05,
-      "loss": 0.9316,
+      "loss": 0.9335,
       "step": 840
     },
     {
       "epoch": 6.538461538461538,
-      "grad_norm": 2.9212265014648438,
+      "grad_norm": 2.944819927215576,
       "learning_rate": 1.1282051282051283e-05,
-      "loss": 0.9417,
+      "loss": 0.9403,
       "step": 850
     },
     {
       "epoch": 6.615384615384615,
-      "grad_norm": 1.9060611724853516,
+      "grad_norm": 1.9217369556427002,
       "learning_rate": 1.117948717948718e-05,
-      "loss": 0.9127,
+      "loss": 0.9109,
       "step": 860
     },
     {
       "epoch": 6.6923076923076925,
-      "grad_norm": 2.729116201400757,
+      "grad_norm": 2.2056405544281006,
       "learning_rate": 1.1076923076923079e-05,
-      "loss": 0.9581,
+      "loss": 0.9578,
       "step": 870
     },
     {
       "epoch": 6.769230769230769,
-      "grad_norm": 2.170494794845581,
+      "grad_norm": 2.6266028881073,
       "learning_rate": 1.0974358974358977e-05,
-      "loss": 0.9267,
+      "loss": 0.9249,
       "step": 880
     },
     {
       "epoch": 6.846153846153846,
-      "grad_norm": 2.36336350440979,
+      "grad_norm": 2.48388409614563,
       "learning_rate": 1.0871794871794871e-05,
-      "loss": 0.9209,
+      "loss": 0.9198,
       "step": 890
     },
     {
       "epoch": 6.923076923076923,
-      "grad_norm": 3.067629098892212,
+      "grad_norm": 3.0685908794403076,
       "learning_rate": 1.076923076923077e-05,
-      "loss": 0.9425,
+      "loss": 0.9424,
       "step": 900
     },
     {
       "epoch": 7.0,
-      "grad_norm": 4.193312168121338,
+      "grad_norm": 4.163443088531494,
       "learning_rate": 1.0666666666666667e-05,
-      "loss": 0.9258,
+      "loss": 0.924,
       "step": 910
     },
     {
       "epoch": 7.0,
       "eval_accuracy": 0.8045112781954887,
-      "eval_loss": 0.8871035575866699,
-      "eval_runtime": 0.9311,
-      "eval_samples_per_second": 142.845,
-      "eval_steps_per_second": 18.258,
+      "eval_loss": 0.8851932287216187,
+      "eval_runtime": 0.7436,
+      "eval_samples_per_second": 178.849,
+      "eval_steps_per_second": 22.86,
       "step": 910
     },
     {
       "epoch": 7.076923076923077,
-      "grad_norm": 2.743016004562378,
+      "grad_norm": 2.746605634689331,
       "learning_rate": 1.0564102564102565e-05,
-      "loss": 0.9111,
+      "loss": 0.9102,
       "step": 920
     },
     {
       "epoch": 7.153846153846154,
-      "grad_norm": 3.4307682514190674,
+      "grad_norm": 3.5535519123077393,
       "learning_rate": 1.0461538461538463e-05,
-      "loss": 0.9108,
+      "loss": 0.9097,
       "step": 930
     },
     {
       "epoch": 7.230769230769231,
-      "grad_norm": 3.426872968673706,
+      "grad_norm": 3.44745135307312,
       "learning_rate": 1.0358974358974361e-05,
-      "loss": 0.9011,
+      "loss": 0.899,
       "step": 940
     },
     {
       "epoch": 7.3076923076923075,
-      "grad_norm": 2.8081014156341553,
+      "grad_norm": 2.7519822120666504,
       "learning_rate": 1.0256410256410256e-05,
-      "loss": 0.9076,
+      "loss": 0.9072,
       "step": 950
     },
     {
       "epoch": 7.384615384615385,
-      "grad_norm": 2.9387893676757812,
+      "grad_norm": 2.7004168033599854,
       "learning_rate": 1.0153846153846154e-05,
-      "loss": 0.9236,
+      "loss": 0.9231,
       "step": 960
     },
     {
       "epoch": 7.461538461538462,
-      "grad_norm": 2.192082643508911,
+      "grad_norm": 2.188138961791992,
       "learning_rate": 1.0051282051282052e-05,
-      "loss": 0.8889,
+      "loss": 0.8897,
       "step": 970
     },
     {
       "epoch": 7.538461538461538,
-      "grad_norm": 2.60426926612854,
+      "grad_norm": 2.8544998168945312,
       "learning_rate": 9.94871794871795e-06,
-      "loss": 0.9243,
+      "loss": 0.9233,
       "step": 980
     },
     {
       "epoch": 7.615384615384615,
-      "grad_norm": 2.2395219802856445,
+      "grad_norm": 2.217107057571411,
       "learning_rate": 9.846153846153848e-06,
-      "loss": 0.8859,
+      "loss": 0.8854,
       "step": 990
     },
     {
       "epoch": 7.6923076923076925,
-      "grad_norm": 2.0519468784332275,
+      "grad_norm": 2.083819627761841,
       "learning_rate": 9.743589743589744e-06,
-      "loss": 0.8821,
+      "loss": 0.8808,
       "step": 1000
     },
     {
       "epoch": 7.769230769230769,
-      "grad_norm": 3.1931142807006836,
+      "grad_norm": 2.864044427871704,
       "learning_rate": 9.641025641025642e-06,
-      "loss": 0.8825,
+      "loss": 0.8818,
       "step": 1010
     },
     {
       "epoch": 7.846153846153846,
-      "grad_norm": 2.4902703762054443,
+      "grad_norm": 2.2901501655578613,
       "learning_rate": 9.53846153846154e-06,
-      "loss": 0.9106,
+      "loss": 0.9066,
       "step": 1020
     },
     {
       "epoch": 7.923076923076923,
-      "grad_norm": 2.7248635292053223,
+      "grad_norm": 2.5113565921783447,
       "learning_rate": 9.435897435897436e-06,
-      "loss": 0.8597,
+      "loss": 0.8577,
       "step": 1030
     },
     {
       "epoch": 8.0,
-      "grad_norm": 4.114623069763184,
+      "grad_norm": 4.092830181121826,
       "learning_rate": 9.333333333333334e-06,
-      "loss": 0.9203,
+      "loss": 0.9179,
       "step": 1040
     },
     {
       "epoch": 8.0,
-      "eval_accuracy": 0.8345864661654135,
-      "eval_loss": 0.8486866354942322,
-      "eval_runtime": 0.8033,
-      "eval_samples_per_second": 165.565,
-      "eval_steps_per_second": 21.162,
+      "eval_accuracy": 0.8120300751879699,
+      "eval_loss": 0.8504595160484314,
+      "eval_runtime": 0.7701,
+      "eval_samples_per_second": 172.703,
+      "eval_steps_per_second": 22.075,
       "step": 1040
     },
     {
       "epoch": 8.076923076923077,
-      "grad_norm": 1.7978770732879639,
+      "grad_norm": 1.9615790843963623,
       "learning_rate": 9.230769230769232e-06,
-      "loss": 0.8922,
+      "loss": 0.8931,
       "step": 1050
     },
     {
       "epoch": 8.153846153846153,
-      "grad_norm": 2.1388049125671387,
+      "grad_norm": 2.1320254802703857,
       "learning_rate": 9.128205128205129e-06,
-      "loss": 0.8696,
+      "loss": 0.8652,
       "step": 1060
     },
     {
       "epoch": 8.23076923076923,
-      "grad_norm": 3.935715913772583,
+      "grad_norm": 2.0207359790802,
       "learning_rate": 9.025641025641027e-06,
-      "loss": 0.9006,
+      "loss": 0.8992,
       "step": 1070
     },
     {
       "epoch": 8.307692307692308,
-      "grad_norm": 2.129542112350464,
+      "grad_norm": 1.9664427042007446,
       "learning_rate": 8.923076923076925e-06,
-      "loss": 0.8603,
+      "loss": 0.8552,
       "step": 1080
     },
     {
       "epoch": 8.384615384615385,
-      "grad_norm": 2.4084651470184326,
+      "grad_norm": 2.321737766265869,
       "learning_rate": 8.820512820512821e-06,
-      "loss": 0.8989,
+      "loss": 0.8973,
       "step": 1090
     },
     {
       "epoch": 8.461538461538462,
-      "grad_norm": 2.7610435485839844,
+      "grad_norm": 2.780322313308716,
       "learning_rate": 8.717948717948719e-06,
-      "loss": 0.8732,
+      "loss": 0.8695,
       "step": 1100
     },
     {
       "epoch": 8.538461538461538,
-      "grad_norm": 1.6848160028457642,
+      "grad_norm": 1.664422631263733,
       "learning_rate": 8.615384615384617e-06,
-      "loss": 0.8715,
+      "loss": 0.8711,
       "step": 1110
     },
     {
       "epoch": 8.615384615384615,
-      "grad_norm": 3.277689218521118,
+      "grad_norm": 4.4292097091674805,
       "learning_rate": 8.512820512820513e-06,
-      "loss": 0.8989,
+      "loss": 0.8983,
       "step": 1120
     },
     {
       "epoch": 8.692307692307692,
-      "grad_norm": 3.7184839248657227,
+      "grad_norm": 3.7464237213134766,
       "learning_rate": 8.410256410256411e-06,
-      "loss": 0.8822,
+      "loss": 0.8831,
       "step": 1130
     },
     {
       "epoch": 8.76923076923077,
-      "grad_norm": 2.3165555000305176,
+      "grad_norm": 1.9225133657455444,
       "learning_rate": 8.307692307692309e-06,
-      "loss": 0.889,
+      "loss": 0.8851,
       "step": 1140
     },
     {
       "epoch": 8.846153846153847,
-      "grad_norm": 2.629028558731079,
+      "grad_norm": 2.6193883419036865,
       "learning_rate": 8.205128205128205e-06,
-      "loss": 0.8753,
+      "loss": 0.8718,
       "step": 1150
     },
     {
       "epoch": 8.923076923076923,
-      "grad_norm": 2.8620731830596924,
+      "grad_norm": 3.2045023441314697,
       "learning_rate": 8.102564102564103e-06,
-      "loss": 0.8855,
+      "loss": 0.8843,
       "step": 1160
     },
     {
       "epoch": 9.0,
-      "grad_norm": 3.924490451812744,
+      "grad_norm": 4.040080547332764,
       "learning_rate": 8.000000000000001e-06,
-      "loss": 0.9038,
+      "loss": 0.9,
       "step": 1170
     },
     {
       "epoch": 9.0,
-      "eval_accuracy": 0.8120300751879699,
-      "eval_loss": 0.8329855799674988,
-      "eval_runtime": 0.781,
-      "eval_samples_per_second": 170.303,
-      "eval_steps_per_second": 21.768,
+      "eval_accuracy": 0.8045112781954887,
+      "eval_loss": 0.8327888250350952,
+      "eval_runtime": 0.764,
+      "eval_samples_per_second": 174.095,
+      "eval_steps_per_second": 22.253,
       "step": 1170
     },
     {
       "epoch": 9.076923076923077,
-      "grad_norm": 1.9303114414215088,
+      "grad_norm": 1.9053655862808228,
       "learning_rate": 7.897435897435898e-06,
-      "loss": 0.8952,
+      "loss": 0.8961,
       "step": 1180
     },
     {
       "epoch": 9.153846153846153,
-      "grad_norm": 2.1960716247558594,
+      "grad_norm": 2.056864023208618,
       "learning_rate": 7.794871794871796e-06,
-      "loss": 0.8401,
+      "loss": 0.8369,
       "step": 1190
     },
     {
       "epoch": 9.23076923076923,
-      "grad_norm": 2.7040772438049316,
+      "grad_norm": 2.1695327758789062,
       "learning_rate": 7.692307692307694e-06,
-      "loss": 0.8488,
+      "loss": 0.8496,
       "step": 1200
     },
     {
       "epoch": 9.307692307692308,
-      "grad_norm": 2.097287654876709,
+      "grad_norm": 2.1930296421051025,
       "learning_rate": 7.58974358974359e-06,
-      "loss": 0.8544,
+      "loss": 0.8527,
       "step": 1210
     },
     {
       "epoch": 9.384615384615385,
-      "grad_norm": 2.012044906616211,
+      "grad_norm": 2.0710272789001465,
       "learning_rate": 7.487179487179488e-06,
-      "loss": 0.8679,
+      "loss": 0.867,
       "step": 1220
     },
     {
       "epoch": 9.461538461538462,
-      "grad_norm": 3.2651612758636475,
+      "grad_norm": 3.325305938720703,
       "learning_rate": 7.384615384615386e-06,
-      "loss": 0.8558,
+      "loss": 0.8541,
       "step": 1230
     },
     {
       "epoch": 9.538461538461538,
-      "grad_norm": 2.5597023963928223,
+      "grad_norm": 2.5524134635925293,
       "learning_rate": 7.282051282051282e-06,
-      "loss": 0.8719,
+      "loss": 0.8711,
       "step": 1240
     },
     {
       "epoch": 9.615384615384615,
-      "grad_norm": 2.756129503250122,
+      "grad_norm": 3.508930206298828,
       "learning_rate": 7.17948717948718e-06,
-      "loss": 0.8272,
+      "loss": 0.8263,
       "step": 1250
     },
     {
       "epoch": 9.692307692307692,
-      "grad_norm": 2.7830724716186523,
+      "grad_norm": 2.657597303390503,
       "learning_rate": 7.076923076923078e-06,
-      "loss": 0.854,
+      "loss": 0.8519,
       "step": 1260
     },
     {
       "epoch": 9.76923076923077,
-      "grad_norm": 1.841320514678955,
+      "grad_norm": 1.858414888381958,
       "learning_rate": 6.974358974358974e-06,
-      "loss": 0.8283,
+      "loss": 0.8306,
       "step": 1270
     },
     {
       "epoch": 9.846153846153847,
-      "grad_norm": 2.4447691440582275,
+      "grad_norm": 2.1380136013031006,
       "learning_rate": 6.871794871794872e-06,
-      "loss": 0.8179,
+      "loss": 0.8115,
       "step": 1280
     },
     {
       "epoch": 9.923076923076923,
-      "grad_norm": 2.266535758972168,
+      "grad_norm": 2.1686654090881348,
       "learning_rate": 6.76923076923077e-06,
-      "loss": 0.8368,
+      "loss": 0.8366,
       "step": 1290
     },
     {
       "epoch": 10.0,
-      "grad_norm": 4.2985944747924805,
+      "grad_norm": 4.619501113891602,
       "learning_rate": 6.666666666666667e-06,
-      "loss": 0.8112,
+      "loss": 0.8084,
       "step": 1300
     },
     {
       "epoch": 10.0,
-      "eval_accuracy": 0.8345864661654135,
-      "eval_loss": 0.8083619475364685,
-      "eval_runtime": 0.8269,
-      "eval_samples_per_second": 160.851,
-      "eval_steps_per_second": 20.56,
+      "eval_accuracy": 0.8421052631578947,
+      "eval_loss": 0.807080864906311,
+      "eval_runtime": 0.7501,
+      "eval_samples_per_second": 177.313,
+      "eval_steps_per_second": 22.664,
       "step": 1300
     },
     {
       "epoch": 10.076923076923077,
-      "grad_norm": 1.931143045425415,
+      "grad_norm": 2.3092093467712402,
       "learning_rate": 6.564102564102565e-06,
-      "loss": 0.8378,
+      "loss": 0.8366,
       "step": 1310
     },
     {
       "epoch": 10.153846153846153,
-      "grad_norm": 2.4910976886749268,
+      "grad_norm": 2.2793080806732178,
       "learning_rate": 6.461538461538463e-06,
-      "loss": 0.8909,
+      "loss": 0.89,
       "step": 1320
     },
     {
       "epoch": 10.23076923076923,
-      "grad_norm": 2.3511204719543457,
+      "grad_norm": 2.369584560394287,
       "learning_rate": 6.358974358974359e-06,
-      "loss": 0.8715,
+      "loss": 0.8718,
       "step": 1330
     },
     {
       "epoch": 10.307692307692308,
-      "grad_norm": 2.0618457794189453,
+      "grad_norm": 2.307018995285034,
       "learning_rate": 6.256410256410257e-06,
-      "loss": 0.833,
+      "loss": 0.8299,
       "step": 1340
     },
     {
       "epoch": 10.384615384615385,
-      "grad_norm": 2.4037184715270996,
+      "grad_norm": 2.086519718170166,
       "learning_rate": 6.153846153846155e-06,
-      "loss": 0.809,
+      "loss": 0.8096,
       "step": 1350
     },
     {
       "epoch": 10.461538461538462,
-      "grad_norm": 2.6920347213745117,
+      "grad_norm": 2.7159907817840576,
       "learning_rate": 6.051282051282051e-06,
-      "loss": 0.8441,
+      "loss": 0.8442,
       "step": 1360
     },
     {
       "epoch": 10.538461538461538,
-      "grad_norm": 2.5570709705352783,
+      "grad_norm": 2.477699041366577,
       "learning_rate": 5.948717948717949e-06,
-      "loss": 0.8166,
+      "loss": 0.8133,
       "step": 1370
     },
     {
       "epoch": 10.615384615384615,
-      "grad_norm": 2.6978418827056885,
+      "grad_norm": 4.113278388977051,
       "learning_rate": 5.846153846153847e-06,
-      "loss": 0.8193,
+      "loss": 0.8159,
       "step": 1380
     },
     {
       "epoch": 10.692307692307692,
-      "grad_norm": 2.370861291885376,
+      "grad_norm": 2.264570951461792,
       "learning_rate": 5.743589743589743e-06,
-      "loss": 0.81,
+      "loss": 0.8086,
       "step": 1390
     },
     {
       "epoch": 10.76923076923077,
-      "grad_norm": 3.260789394378662,
+      "grad_norm": 3.312915802001953,
       "learning_rate": 5.641025641025641e-06,
-      "loss": 0.8671,
+      "loss": 0.8668,
       "step": 1400
     },
     {
       "epoch": 10.846153846153847,
-      "grad_norm": 2.27559494972229,
+      "grad_norm": 2.189178228378296,
       "learning_rate": 5.538461538461539e-06,
-      "loss": 0.8321,
+      "loss": 0.8298,
       "step": 1410
     },
     {
       "epoch": 10.923076923076923,
-      "grad_norm": 1.8184912204742432,
+      "grad_norm": 1.8542472124099731,
       "learning_rate": 5.435897435897436e-06,
-      "loss": 0.8388,
+      "loss": 0.8329,
       "step": 1420
     },
     {
       "epoch": 11.0,
-      "grad_norm": 4.08992862701416,
+      "grad_norm": 4.028012752532959,
       "learning_rate": 5.333333333333334e-06,
-      "loss": 0.8335,
+      "loss": 0.8306,
       "step": 1430
     },
     {
       "epoch": 11.0,
       "eval_accuracy": 0.8345864661654135,
-      "eval_loss": 0.7785491943359375,
-      "eval_runtime": 0.8255,
-      "eval_samples_per_second": 161.106,
-      "eval_steps_per_second": 20.592,
+      "eval_loss": 0.7760081887245178,
+      "eval_runtime": 0.7954,
+      "eval_samples_per_second": 167.205,
+      "eval_steps_per_second": 21.372,
       "step": 1430
     },
     {
       "epoch": 11.076923076923077,
-      "grad_norm": 2.0166661739349365,
+      "grad_norm": 1.9505723714828491,
       "learning_rate": 5.230769230769232e-06,
-      "loss": 0.8471,
+      "loss": 0.8461,
       "step": 1440
     },
     {
       "epoch": 11.153846153846153,
-      "grad_norm": 3.2678654193878174,
+      "grad_norm": 3.543419122695923,
       "learning_rate": 5.128205128205128e-06,
-      "loss": 0.8298,
+      "loss": 0.8287,
       "step": 1450
     },
     {
       "epoch": 11.23076923076923,
-      "grad_norm": 2.545058488845825,
+      "grad_norm": 2.6018764972686768,
       "learning_rate": 5.025641025641026e-06,
-      "loss": 0.7991,
+      "loss": 0.7996,
       "step": 1460
     },
     {
       "epoch": 11.307692307692308,
-      "grad_norm": 2.469082832336426,
+      "grad_norm": 2.7204878330230713,
       "learning_rate": 4.923076923076924e-06,
-      "loss": 0.8256,
+      "loss": 0.8218,
       "step": 1470
     },
     {
       "epoch": 11.384615384615385,
-      "grad_norm": 2.8207569122314453,
+      "grad_norm": 2.6924397945404053,
       "learning_rate": 4.820512820512821e-06,
-      "loss": 0.8172,
+      "loss": 0.8152,
       "step": 1480
     },
     {
       "epoch": 11.461538461538462,
-      "grad_norm": 2.142630100250244,
+      "grad_norm": 2.1294894218444824,
       "learning_rate": 4.717948717948718e-06,
-      "loss": 0.8448,
+      "loss": 0.8452,
       "step": 1490
     },
     {
       "epoch": 11.538461538461538,
-      "grad_norm": 2.3974075317382812,
+      "grad_norm": 2.45246958732605,
       "learning_rate": 4.615384615384616e-06,
-      "loss": 0.8292,
+      "loss": 0.8316,
       "step": 1500
     },
     {
       "epoch": 11.615384615384615,
-      "grad_norm": 3.240954875946045,
+      "grad_norm": 3.2710139751434326,
       "learning_rate": 4.512820512820513e-06,
-      "loss": 0.8275,
+      "loss": 0.8272,
       "step": 1510
     },
     {
       "epoch": 11.692307692307692,
-      "grad_norm": 3.3133740425109863,
+      "grad_norm": 3.3108620643615723,
       "learning_rate": 4.4102564102564104e-06,
-      "loss": 0.8569,
+      "loss": 0.8539,
       "step": 1520
     },
     {
       "epoch": 11.76923076923077,
-      "grad_norm": 2.0835375785827637,
+      "grad_norm": 2.079916000366211,
       "learning_rate": 4.307692307692308e-06,
-      "loss": 0.8086,
+      "loss": 0.8067,
       "step": 1530
     },
     {
       "epoch": 11.846153846153847,
-      "grad_norm": 2.20538067817688,
+      "grad_norm": 2.2175967693328857,
       "learning_rate": 4.2051282051282055e-06,
-      "loss": 0.797,
+      "loss": 0.7943,
       "step": 1540
     },
     {
       "epoch": 11.923076923076923,
-      "grad_norm": 2.1278247833251953,
+      "grad_norm": 2.064258098602295,
       "learning_rate": 4.102564102564103e-06,
-      "loss": 0.8301,
+      "loss": 0.8268,
       "step": 1550
     },
     {
       "epoch": 12.0,
-      "grad_norm": 3.4031107425689697,
+      "grad_norm": 3.4313971996307373,
       "learning_rate": 4.000000000000001e-06,
-      "loss": 0.8062,
+      "loss": 0.8031,
       "step": 1560
     },
     {
       "epoch": 12.0,
       "eval_accuracy": 0.8345864661654135,
-      "eval_loss": 0.7569367289543152,
-      "eval_runtime": 0.799,
-      "eval_samples_per_second": 166.458,
-      "eval_steps_per_second": 21.277,
+      "eval_loss": 0.7563135623931885,
+      "eval_runtime": 0.7817,
+      "eval_samples_per_second": 170.134,
+      "eval_steps_per_second": 21.746,
       "step": 1560
     },
     {
       "epoch": 12.076923076923077,
-      "grad_norm": 2.181518793106079,
+      "grad_norm": 2.1071786880493164,
       "learning_rate": 3.897435897435898e-06,
-      "loss": 0.7903,
+      "loss": 0.7877,
       "step": 1570
     },
     {
       "epoch": 12.153846153846153,
-      "grad_norm": 2.4533674716949463,
+      "grad_norm": 2.6646640300750732,
       "learning_rate": 3.794871794871795e-06,
-      "loss": 0.8228,
+      "loss": 0.8223,
       "step": 1580
     },
     {
       "epoch": 12.23076923076923,
-      "grad_norm": 1.6895153522491455,
+      "grad_norm": 1.7268288135528564,
       "learning_rate": 3.692307692307693e-06,
-      "loss": 0.8242,
+      "loss": 0.8204,
       "step": 1590
     },
     {
       "epoch": 12.307692307692308,
-      "grad_norm": 2.1413424015045166,
+      "grad_norm": 2.982988119125366,
       "learning_rate": 3.58974358974359e-06,
-      "loss": 0.8378,
+      "loss": 0.8336,
       "step": 1600
     },
     {
       "epoch": 12.384615384615385,
-      "grad_norm": 2.416987657546997,
+      "grad_norm": 2.559271812438965,
       "learning_rate": 3.487179487179487e-06,
-      "loss": 0.8088,
+      "loss": 0.8064,
       "step": 1610
     },
     {
       "epoch": 12.461538461538462,
-      "grad_norm": 1.704318881034851,
+      "grad_norm": 2.528869390487671,
       "learning_rate": 3.384615384615385e-06,
-      "loss": 0.7953,
+      "loss": 0.793,
       "step": 1620
     },
     {
       "epoch": 12.538461538461538,
-      "grad_norm": 2.015375852584839,
+      "grad_norm": 2.1543517112731934,
       "learning_rate": 3.2820512820512823e-06,
-      "loss": 0.799,
+      "loss": 0.8031,
       "step": 1630
     },
     {
       "epoch": 12.615384615384615,
-      "grad_norm": 1.851975917816162,
+      "grad_norm": 1.7778912782669067,
       "learning_rate": 3.1794871794871795e-06,
-      "loss": 0.813,
+      "loss": 0.8086,
       "step": 1640
     },
     {
       "epoch": 12.692307692307692,
-      "grad_norm": 3.352241277694702,
+      "grad_norm": 3.2962191104888916,
       "learning_rate": 3.0769230769230774e-06,
-      "loss": 0.763,
+      "loss": 0.7617,
       "step": 1650
     },
     {
       "epoch": 12.76923076923077,
-      "grad_norm": 2.4369592666625977,
+      "grad_norm": 2.336732864379883,
       "learning_rate": 2.9743589743589746e-06,
-      "loss": 0.8583,
+      "loss": 0.8541,
       "step": 1660
     },
     {
       "epoch": 12.846153846153847,
-      "grad_norm": 1.7659847736358643,
+      "grad_norm": 1.9721437692642212,
       "learning_rate": 2.8717948717948717e-06,
-      "loss": 0.8354,
+      "loss": 0.8249,
       "step": 1670
     },
     {
       "epoch": 12.923076923076923,
-      "grad_norm": 1.8726952075958252,
+      "grad_norm": 2.0368008613586426,
       "learning_rate": 2.7692307692307697e-06,
-      "loss": 0.7972,
+      "loss": 0.795,
       "step": 1680
     },
     {
       "epoch": 13.0,
-      "grad_norm": 3.766446352005005,
+      "grad_norm": 3.8102405071258545,
       "learning_rate": 2.666666666666667e-06,
-      "loss": 0.8141,
+      "loss": 0.8138,
       "step": 1690
     },
     {
       "epoch": 13.0,
-      "eval_accuracy": 0.849624060150376,
-      "eval_loss": 0.7535560131072998,
-      "eval_runtime": 0.8236,
-      "eval_samples_per_second": 161.479,
-      "eval_steps_per_second": 20.64,
+      "eval_accuracy": 0.8421052631578947,
+      "eval_loss": 0.753366231918335,
+      "eval_runtime": 0.7711,
+      "eval_samples_per_second": 172.483,
+      "eval_steps_per_second": 22.047,
       "step": 1690
     },
     {
       "epoch": 13.076923076923077,
-      "grad_norm": 3.104832649230957,
+      "grad_norm": 3.0844614505767822,
       "learning_rate": 2.564102564102564e-06,
-      "loss": 0.7937,
+      "loss": 0.7901,
       "step": 1700
     },
     {
       "epoch": 13.153846153846153,
-      "grad_norm": 2.46419095993042,
+      "grad_norm": 2.517000913619995,
       "learning_rate": 2.461538461538462e-06,
-      "loss": 0.822,
+      "loss": 0.8205,
       "step": 1710
     },
     {
       "epoch": 13.23076923076923,
-      "grad_norm": 3.035752296447754,
+      "grad_norm": 3.047574520111084,
       "learning_rate": 2.358974358974359e-06,
-      "loss": 0.8127,
+      "loss": 0.8113,
       "step": 1720
     },
     {
       "epoch": 13.307692307692308,
-      "grad_norm": 1.9470882415771484,
+      "grad_norm": 1.9097496271133423,
       "learning_rate": 2.2564102564102566e-06,
-      "loss": 0.8151,
+      "loss": 0.817,
       "step": 1730
     },
     {
       "epoch": 13.384615384615385,
-      "grad_norm": 2.086946964263916,
+      "grad_norm": 2.553558826446533,
       "learning_rate": 2.153846153846154e-06,
-      "loss": 0.7773,
+      "loss": 0.7794,
       "step": 1740
     },
     {
       "epoch": 13.461538461538462,
-      "grad_norm": 3.7044293880462646,
+      "grad_norm": 3.915072202682495,
       "learning_rate": 2.0512820512820513e-06,
-      "loss": 0.8276,
+      "loss": 0.8238,
       "step": 1750
     },
     {
       "epoch": 13.538461538461538,
-      "grad_norm": 2.735135555267334,
+      "grad_norm": 2.7563774585723877,
       "learning_rate": 1.948717948717949e-06,
-      "loss": 0.8462,
+      "loss": 0.8464,
       "step": 1760
     },
     {
       "epoch": 13.615384615384615,
-      "grad_norm": 2.065619468688965,
+      "grad_norm": 1.9687329530715942,
       "learning_rate": 1.8461538461538465e-06,
-      "loss": 0.8024,
+      "loss": 0.7998,
       "step": 1770
     },
     {
       "epoch": 13.692307692307692,
-      "grad_norm": 2.519625663757324,
+      "grad_norm": 1.9707388877868652,
       "learning_rate": 1.7435897435897436e-06,
-      "loss": 0.7813,
+      "loss": 0.7827,
       "step": 1780
     },
     {
       "epoch": 13.76923076923077,
-      "grad_norm": 1.8556421995162964,
+      "grad_norm": 1.8745958805084229,
       "learning_rate": 1.6410256410256412e-06,
-      "loss": 0.8218,
+      "loss": 0.8211,
       "step": 1790
     },
     {
       "epoch": 13.846153846153847,
-      "grad_norm": 3.1282896995544434,
+      "grad_norm": 3.193314552307129,
       "learning_rate": 1.5384615384615387e-06,
-      "loss": 0.7891,
+      "loss": 0.7874,
       "step": 1800
     },
     {
       "epoch": 13.923076923076923,
-      "grad_norm": 2.363374948501587,
+      "grad_norm": 2.4726743698120117,
       "learning_rate": 1.4358974358974359e-06,
-      "loss": 0.7341,
+      "loss": 0.7323,
       "step": 1810
     },
     {
       "epoch": 14.0,
-      "grad_norm": 4.762740612030029,
+      "grad_norm": 4.2939653396606445,
       "learning_rate": 1.3333333333333334e-06,
-      "loss": 0.8172,
+      "loss": 0.8178,
       "step": 1820
     },
     {
       "epoch": 14.0,
       "eval_accuracy": 0.8270676691729323,
-      "eval_loss": 0.7531744837760925,
-      "eval_runtime": 0.8266,
-      "eval_samples_per_second": 160.91,
-      "eval_steps_per_second": 20.567,
+      "eval_loss": 0.7507623434066772,
+      "eval_runtime": 0.7401,
+      "eval_samples_per_second": 179.698,
+      "eval_steps_per_second": 22.969,
       "step": 1820
     },
     {
       "epoch": 14.076923076923077,
-      "grad_norm": 1.8959217071533203,
+      "grad_norm": 2.51151967048645,
       "learning_rate": 1.230769230769231e-06,
       "loss": 0.8201,
       "step": 1830
     },
     {
       "epoch": 14.153846153846153,
-      "grad_norm": 2.1285548210144043,
+      "grad_norm": 2.035071849822998,
       "learning_rate": 1.1282051282051283e-06,
-      "loss": 0.7906,
+      "loss": 0.7885,
       "step": 1840
     },
     {
       "epoch": 14.23076923076923,
-      "grad_norm": 2.916303873062134,
+      "grad_norm": 2.924006938934326,
       "learning_rate": 1.0256410256410257e-06,
-      "loss": 0.7793,
+      "loss": 0.7802,
       "step": 1850
     },
     {
       "epoch": 14.307692307692308,
-      "grad_norm": 2.9460341930389404,
+      "grad_norm": 3.3788881301879883,
       "learning_rate": 9.230769230769232e-07,
-      "loss": 0.773,
+      "loss": 0.7717,
       "step": 1860
     },
     {
       "epoch": 14.384615384615385,
-      "grad_norm": 2.4907443523406982,
+      "grad_norm": 2.6425819396972656,
       "learning_rate": 8.205128205128206e-07,
-      "loss": 0.7887,
+      "loss": 0.7871,
       "step": 1870
     },
     {
       "epoch": 14.461538461538462,
-      "grad_norm": 2.431607961654663,
+      "grad_norm": 2.781729221343994,
       "learning_rate": 7.179487179487179e-07,
-      "loss": 0.7886,
+      "loss": 0.7859,
       "step": 1880
     },
     {
       "epoch": 14.538461538461538,
-      "grad_norm": 3.005627155303955,
+      "grad_norm": 3.0325913429260254,
       "learning_rate": 6.153846153846155e-07,
-      "loss": 0.7789,
+      "loss": 0.7783,
       "step": 1890
     },
     {
       "epoch": 14.615384615384615,
-      "grad_norm": 2.981405735015869,
+      "grad_norm": 2.7311503887176514,
       "learning_rate": 5.128205128205128e-07,
-      "loss": 0.8262,
+      "loss": 0.8273,
       "step": 1900
     },
     {
       "epoch": 14.692307692307692,
-      "grad_norm": 1.9199621677398682,
+      "grad_norm": 2.3128809928894043,
       "learning_rate": 4.102564102564103e-07,
-      "loss": 0.8015,
+      "loss": 0.7977,
       "step": 1910
     },
     {
       "epoch": 14.76923076923077,
-      "grad_norm": 3.6987714767456055,
+      "grad_norm": 3.4510748386383057,
       "learning_rate": 3.0769230769230774e-07,
-      "loss": 0.7998,
+      "loss": 0.8029,
       "step": 1920
     },
     {
       "epoch": 14.846153846153847,
-      "grad_norm": 3.5848920345306396,
+      "grad_norm": 3.1088345050811768,
       "learning_rate": 2.0512820512820514e-07,
-      "loss": 0.8552,
+      "loss": 0.8509,
       "step": 1930
     },
     {
       "epoch": 14.923076923076923,
-      "grad_norm": 2.4082224369049072,
+      "grad_norm": 2.3365509510040283,
       "learning_rate": 1.0256410256410257e-07,
-      "loss": 0.759,
+      "loss": 0.7569,
       "step": 1940
     },
     {
       "epoch": 15.0,
-      "grad_norm": 4.4640116691589355,
+      "grad_norm": 4.560791015625,
       "learning_rate": 0.0,
-      "loss": 0.7896,
+      "loss": 0.7901,
       "step": 1950
     },
     {
       "epoch": 15.0,
-      "eval_accuracy": 0.8045112781954887,
-      "eval_loss": 0.7388833165168762,
-      "eval_runtime": 0.8692,
-      "eval_samples_per_second": 153.016,
-      "eval_steps_per_second": 19.558,
+      "eval_accuracy": 0.8195488721804511,
+      "eval_loss": 0.738013505935669,
+      "eval_runtime": 0.796,
+      "eval_samples_per_second": 167.085,
+      "eval_steps_per_second": 21.357,
       "step": 1950
     },
     {
       "epoch": 15.0,
       "step": 1950,
       "total_flos": 1.5658365504595968e+17,
-      "train_loss": 0.923483537771763,
-      "train_runtime": 151.8082,
-      "train_samples_per_second": 102.168,
-      "train_steps_per_second": 12.845
+      "train_loss": 0.9225189507313264,
+      "train_runtime": 140.2,
+      "train_samples_per_second": 110.628,
+      "train_steps_per_second": 13.909
     }
   ],
   "logging_steps": 10,