End of training

README.md

@@ -3,6 +3,8 @@ library_name: transformers
 license: apache-2.0
 base_model: google/efficientnet-b0
 tags:
+- image-classification
+- vision
 - generated_from_trainer
 metrics:
 - accuracy
@@ -16,10 +18,10 @@ should probably proofread and complete it, then remove this comment. -->
 
 # efficientnet-b0-cocoa
 
-This model is a fine-tuned version of [google/efficientnet-b0](https://huggingface.co/google/efficientnet-b0) on an unknown dataset.
+This model is a fine-tuned version of [google/efficientnet-b0](https://huggingface.co/google/efficientnet-b0) on the SemilleroCV/Cocoa-dataset dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.3624
-- Accuracy: 0.8809
+- Loss: 0.2657
+- Accuracy: 0.9097
 
 ## Model description
 

all_results.json

@@ -2,12 +2,12 @@
     "epoch": 100.0,
     "eval_accuracy": 0.9097472924187726,
     "eval_loss": 0.26566287875175476,
-    "eval_runtime": 0.9695,
-    "eval_samples_per_second": 285.718,
-    "eval_steps_per_second": 36.102,
+    "eval_runtime": 0.9793,
+    "eval_samples_per_second": 282.857,
+    "eval_steps_per_second": 35.74,
     "total_flos": 5.678990728814592e+17,
     "train_loss": 0.30055975163317455,
-    "train_runtime": 936.3063,
-    "train_samples_per_second": 167.253,
-    "train_steps_per_second": 20.933
+    "train_runtime": 949.6972,
+    "train_samples_per_second": 164.895,
+    "train_steps_per_second": 20.638
 }

eval_results.json

@@ -2,7 +2,7 @@
     "epoch": 100.0,
     "eval_accuracy": 0.9097472924187726,
     "eval_loss": 0.26566287875175476,
-    "eval_runtime": 0.9695,
-    "eval_samples_per_second": 285.718,
-    "eval_steps_per_second": 36.102
+    "eval_runtime": 0.9793,
+    "eval_samples_per_second": 282.857,
+    "eval_steps_per_second": 35.74
 }

train_results.json

@@ -2,7 +2,7 @@
     "epoch": 100.0,
     "total_flos": 5.678990728814592e+17,
     "train_loss": 0.30055975163317455,
-    "train_runtime": 936.3063,
-    "train_samples_per_second": 167.253,
-    "train_steps_per_second": 20.933
+    "train_runtime": 949.6972,
+    "train_samples_per_second": 164.895,
+    "train_steps_per_second": 20.638
 }

trainer_state.json

@@ -1,6 +1,6 @@
 {
   "best_metric": 0.26566287875175476,
-  "best_model_checkpoint": "./cocoa_outputs_efficientnet/checkpoint-6468",
+  "best_model_checkpoint": "./cocoa_outputs_resnet/checkpoint-6468",
   "epoch": 100.0,
   "eval_steps": 500,
   "global_step": 19600,
@@ -145,9 +145,9 @@
       "epoch": 1.0,
       "eval_accuracy": 0.5595667870036101,
       "eval_loss": 1.5222723484039307,
-      "eval_runtime": 0.9623,
-      "eval_samples_per_second": 287.86,
-      "eval_steps_per_second": 36.372,
+      "eval_runtime": 0.9569,
+      "eval_samples_per_second": 289.486,
+      "eval_steps_per_second": 36.578,
       "step": 196
     },
     {
@@ -294,9 +294,9 @@
       "epoch": 2.0,
       "eval_accuracy": 0.8303249097472925,
       "eval_loss": 0.8974758386611938,
-      "eval_runtime": 0.9254,
-      "eval_samples_per_second": 299.328,
-      "eval_steps_per_second": 37.821,
+      "eval_runtime": 0.934,
+      "eval_samples_per_second": 296.569,
+      "eval_steps_per_second": 37.473,
       "step": 392
     },
     {
@@ -436,9 +436,9 @@
       "epoch": 3.0,
       "eval_accuracy": 0.8772563176895307,
       "eval_loss": 0.6563544273376465,
-      "eval_runtime": 0.9258,
-      "eval_samples_per_second": 299.212,
-      "eval_steps_per_second": 37.807,
+      "eval_runtime": 0.936,
+      "eval_samples_per_second": 295.927,
+      "eval_steps_per_second": 37.391,
       "step": 588
     },
     {
@@ -585,9 +585,9 @@
       "epoch": 4.0,
       "eval_accuracy": 0.8339350180505415,
       "eval_loss": 0.6841569542884827,
-      "eval_runtime": 0.9246,
-      "eval_samples_per_second": 299.576,
-      "eval_steps_per_second": 37.853,
+      "eval_runtime": 0.9273,
+      "eval_samples_per_second": 298.728,
+      "eval_steps_per_second": 37.745,
       "step": 784
     },
     {
@@ -734,9 +734,9 @@
       "epoch": 5.0,
       "eval_accuracy": 0.8808664259927798,
       "eval_loss": 0.48941850662231445,
-      "eval_runtime": 0.9211,
-      "eval_samples_per_second": 300.733,
-      "eval_steps_per_second": 37.999,
+      "eval_runtime": 0.935,
+      "eval_samples_per_second": 296.252,
+      "eval_steps_per_second": 37.433,
       "step": 980
     },
     {
@@ -876,9 +876,9 @@
       "epoch": 6.0,
       "eval_accuracy": 0.8736462093862816,
       "eval_loss": 0.41603580117225647,
-      "eval_runtime": 0.9268,
-      "eval_samples_per_second": 298.89,
-      "eval_steps_per_second": 37.766,
+      "eval_runtime": 0.9296,
+      "eval_samples_per_second": 297.99,
+      "eval_steps_per_second": 37.652,
       "step": 1176
     },
     {
@@ -1025,9 +1025,9 @@
       "epoch": 7.0,
       "eval_accuracy": 0.8844765342960289,
       "eval_loss": 0.40221381187438965,
-      "eval_runtime": 0.9961,
-      "eval_samples_per_second": 278.09,
-      "eval_steps_per_second": 35.138,
+      "eval_runtime": 0.9298,
+      "eval_samples_per_second": 297.902,
+      "eval_steps_per_second": 37.641,
       "step": 1372
     },
     {
@@ -1167,9 +1167,9 @@
       "epoch": 8.0,
       "eval_accuracy": 0.8411552346570397,
       "eval_loss": 0.4939042627811432,
-      "eval_runtime": 0.9269,
-      "eval_samples_per_second": 298.835,
-      "eval_steps_per_second": 37.759,
+      "eval_runtime": 0.9311,
+      "eval_samples_per_second": 297.494,
+      "eval_steps_per_second": 37.589,
       "step": 1568
     },
     {
@@ -1316,9 +1316,9 @@
       "epoch": 9.0,
       "eval_accuracy": 0.8772563176895307,
       "eval_loss": 0.35403916239738464,
-      "eval_runtime": 0.9961,
-      "eval_samples_per_second": 278.089,
-      "eval_steps_per_second": 35.138,
+      "eval_runtime": 0.9315,
+      "eval_samples_per_second": 297.362,
+      "eval_steps_per_second": 37.573,
       "step": 1764
     },
     {
@@ -1465,9 +1465,9 @@
       "epoch": 10.0,
       "eval_accuracy": 0.8953068592057761,
       "eval_loss": 0.349290668964386,
-      "eval_runtime": 0.9261,
-      "eval_samples_per_second": 299.092,
-      "eval_steps_per_second": 37.791,
+      "eval_runtime": 0.9365,
+      "eval_samples_per_second": 295.793,
+      "eval_steps_per_second": 37.375,
       "step": 1960
     },
     {
@@ -1607,9 +1607,9 @@
       "epoch": 11.0,
       "eval_accuracy": 0.776173285198556,
       "eval_loss": 0.7928438186645508,
-      "eval_runtime": 0.9258,
-      "eval_samples_per_second": 299.201,
-      "eval_steps_per_second": 37.805,
+      "eval_runtime": 0.9321,
+      "eval_samples_per_second": 297.189,
+      "eval_steps_per_second": 37.551,
       "step": 2156
     },
     {
@@ -1756,9 +1756,9 @@
       "epoch": 12.0,
       "eval_accuracy": 0.296028880866426,
       "eval_loss": 2.06593918800354,
-      "eval_runtime": 0.9316,
-      "eval_samples_per_second": 297.347,
-      "eval_steps_per_second": 37.571,
+      "eval_runtime": 0.9947,
+      "eval_samples_per_second": 278.482,
+      "eval_steps_per_second": 35.187,
       "step": 2352
     },
     {
@@ -1898,9 +1898,9 @@
       "epoch": 13.0,
       "eval_accuracy": 0.851985559566787,
       "eval_loss": 0.44439464807510376,
-      "eval_runtime": 0.9238,
-      "eval_samples_per_second": 299.863,
-      "eval_steps_per_second": 37.889,
+      "eval_runtime": 0.9289,
+      "eval_samples_per_second": 298.212,
+      "eval_steps_per_second": 37.68,
       "step": 2548
     },
     {
@@ -2047,9 +2047,9 @@
       "epoch": 14.0,
       "eval_accuracy": 0.8736462093862816,
       "eval_loss": 0.46400919556617737,
-      "eval_runtime": 0.9277,
-      "eval_samples_per_second": 298.595,
-      "eval_steps_per_second": 37.729,
+      "eval_runtime": 0.9937,
+      "eval_samples_per_second": 278.754,
+      "eval_steps_per_second": 35.222,
       "step": 2744
     },
     {
@@ -2196,9 +2196,9 @@
       "epoch": 15.0,
       "eval_accuracy": 0.8772563176895307,
       "eval_loss": 0.37753719091415405,
-      "eval_runtime": 0.9263,
-      "eval_samples_per_second": 299.04,
-      "eval_steps_per_second": 37.785,
+      "eval_runtime": 0.9317,
+      "eval_samples_per_second": 297.316,
+      "eval_steps_per_second": 37.567,
       "step": 2940
     },
     {
@@ -2338,9 +2338,9 @@
       "epoch": 16.0,
       "eval_accuracy": 0.8880866425992779,
       "eval_loss": 0.3519056737422943,
-      "eval_runtime": 0.9249,
-      "eval_samples_per_second": 299.483,
-      "eval_steps_per_second": 37.841,
+      "eval_runtime": 0.9334,
+      "eval_samples_per_second": 296.766,
+      "eval_steps_per_second": 37.498,
       "step": 3136
     },
     {
@@ -2487,9 +2487,9 @@
       "epoch": 17.0,
       "eval_accuracy": 0.9061371841155235,
       "eval_loss": 0.33103030920028687,
-      "eval_runtime": 0.9321,
-      "eval_samples_per_second": 297.184,
-      "eval_steps_per_second": 37.55,
+      "eval_runtime": 0.9376,
+      "eval_samples_per_second": 295.446,
+      "eval_steps_per_second": 37.331,
       "step": 3332
     },
     {
@@ -2629,9 +2629,9 @@
       "epoch": 18.0,
       "eval_accuracy": 0.8916967509025271,
       "eval_loss": 0.34751203656196594,
-      "eval_runtime": 0.9902,
-      "eval_samples_per_second": 279.742,
-      "eval_steps_per_second": 35.346,
+      "eval_runtime": 0.9366,
+      "eval_samples_per_second": 295.765,
+      "eval_steps_per_second": 37.371,
       "step": 3528
     },
     {
@@ -2778,9 +2778,9 @@
       "epoch": 19.0,
       "eval_accuracy": 0.8592057761732852,
       "eval_loss": 0.38391339778900146,
-      "eval_runtime": 0.9252,
-      "eval_samples_per_second": 299.384,
-      "eval_steps_per_second": 37.828,
+      "eval_runtime": 0.9352,
+      "eval_samples_per_second": 296.2,
+      "eval_steps_per_second": 37.426,
       "step": 3724
     },
     {
@@ -2927,9 +2927,9 @@
       "epoch": 20.0,
       "eval_accuracy": 0.9133574007220217,
       "eval_loss": 0.3031846880912781,
-      "eval_runtime": 0.9294,
-      "eval_samples_per_second": 298.047,
-      "eval_steps_per_second": 37.659,
+      "eval_runtime": 0.9346,
+      "eval_samples_per_second": 296.374,
+      "eval_steps_per_second": 37.448,
       "step": 3920
     },
     {
@@ -3069,9 +3069,9 @@
       "epoch": 21.0,
       "eval_accuracy": 0.851985559566787,
       "eval_loss": 0.4246116876602173,
-      "eval_runtime": 0.9793,
-      "eval_samples_per_second": 282.866,
-      "eval_steps_per_second": 35.741,
+      "eval_runtime": 0.9314,
+      "eval_samples_per_second": 297.413,
+      "eval_steps_per_second": 37.579,
       "step": 4116
     },
     {
@@ -3218,9 +3218,9 @@
       "epoch": 22.0,
       "eval_accuracy": 0.8989169675090253,
       "eval_loss": 0.29509276151657104,
-      "eval_runtime": 0.9468,
-      "eval_samples_per_second": 292.564,
-      "eval_steps_per_second": 36.967,
+      "eval_runtime": 0.9307,
+      "eval_samples_per_second": 297.612,
+      "eval_steps_per_second": 37.604,
       "step": 4312
     },
     {
@@ -3360,9 +3360,9 @@
       "epoch": 23.0,
       "eval_accuracy": 0.851985559566787,
       "eval_loss": 0.4356595277786255,
-      "eval_runtime": 0.9291,
-      "eval_samples_per_second": 298.141,
-      "eval_steps_per_second": 37.671,
+      "eval_runtime": 0.9603,
+      "eval_samples_per_second": 288.437,
+      "eval_steps_per_second": 36.445,
       "step": 4508
     },
     {
@@ -3509,9 +3509,9 @@
       "epoch": 24.0,
       "eval_accuracy": 0.8916967509025271,
       "eval_loss": 0.399036169052124,
-      "eval_runtime": 0.9832,
-      "eval_samples_per_second": 281.719,
-      "eval_steps_per_second": 35.596,
+      "eval_runtime": 0.9342,
+      "eval_samples_per_second": 296.5,
+      "eval_steps_per_second": 37.464,
       "step": 4704
     },
     {
@@ -3658,9 +3658,9 @@
       "epoch": 25.0,
       "eval_accuracy": 0.9097472924187726,
       "eval_loss": 0.35359257459640503,
-      "eval_runtime": 0.9278,
-      "eval_samples_per_second": 298.55,
-      "eval_steps_per_second": 37.723,
+      "eval_runtime": 0.9646,
+      "eval_samples_per_second": 287.175,
+      "eval_steps_per_second": 36.286,
       "step": 4900
     },
     {
@@ -3800,9 +3800,9 @@
       "epoch": 26.0,
       "eval_accuracy": 0.8592057761732852,
       "eval_loss": 0.4166140854358673,
-      "eval_runtime": 0.9508,
-      "eval_samples_per_second": 291.327,
-      "eval_steps_per_second": 36.81,
+      "eval_runtime": 0.9729,
+      "eval_samples_per_second": 284.728,
+      "eval_steps_per_second": 35.976,
       "step": 5096
     },
     {
@@ -3949,9 +3949,9 @@
       "epoch": 27.0,
       "eval_accuracy": 0.851985559566787,
       "eval_loss": 0.4353993237018585,
-      "eval_runtime": 0.9437,
-      "eval_samples_per_second": 293.541,
-      "eval_steps_per_second": 37.09,
+      "eval_runtime": 0.9398,
+      "eval_samples_per_second": 294.757,
+      "eval_steps_per_second": 37.244,
       "step": 5292
     },
     {
@@ -4091,9 +4091,9 @@
       "epoch": 28.0,
       "eval_accuracy": 0.8592057761732852,
       "eval_loss": 0.4088672697544098,
-      "eval_runtime": 0.9296,
-      "eval_samples_per_second": 297.977,
-      "eval_steps_per_second": 37.651,
+      "eval_runtime": 0.9406,
+      "eval_samples_per_second": 294.489,
+      "eval_steps_per_second": 37.21,
       "step": 5488
     },
     {
@@ -4240,9 +4240,9 @@
       "epoch": 29.0,
       "eval_accuracy": 0.8231046931407943,
       "eval_loss": 0.4787161946296692,
-      "eval_runtime": 0.9706,
-      "eval_samples_per_second": 285.392,
-      "eval_steps_per_second": 36.06,
+      "eval_runtime": 0.933,
+      "eval_samples_per_second": 296.9,
+      "eval_steps_per_second": 37.514,
       "step": 5684
     },
     {
@@ -4389,9 +4389,9 @@
       "epoch": 30.0,
       "eval_accuracy": 0.9097472924187726,
       "eval_loss": 0.3165305256843567,
-      "eval_runtime": 0.9312,
-      "eval_samples_per_second": 297.455,
-      "eval_steps_per_second": 37.585,
+      "eval_runtime": 0.9292,
+      "eval_samples_per_second": 298.12,
+      "eval_steps_per_second": 37.669,
       "step": 5880
     },
     {
@@ -4531,9 +4531,9 @@
       "epoch": 31.0,
       "eval_accuracy": 0.9133574007220217,
       "eval_loss": 0.31053850054740906,
-      "eval_runtime": 0.9255,
-      "eval_samples_per_second": 299.294,
-      "eval_steps_per_second": 37.817,
+      "eval_runtime": 0.9345,
+      "eval_samples_per_second": 296.427,
+      "eval_steps_per_second": 37.455,
       "step": 6076
     },
     {
@@ -4680,9 +4680,9 @@
       "epoch": 32.0,
       "eval_accuracy": 0.9061371841155235,
       "eval_loss": 0.3559906780719757,
-      "eval_runtime": 0.9627,
-      "eval_samples_per_second": 287.745,
-      "eval_steps_per_second": 36.358,
+      "eval_runtime": 0.9353,
+      "eval_samples_per_second": 296.174,
+      "eval_steps_per_second": 37.423,
       "step": 6272
     },
     {
@@ -4822,9 +4822,9 @@
       "epoch": 33.0,
       "eval_accuracy": 0.9097472924187726,
       "eval_loss": 0.26566287875175476,
-      "eval_runtime": 0.928,
-      "eval_samples_per_second": 298.475,
-      "eval_steps_per_second": 37.714,
+      "eval_runtime": 0.9334,
+      "eval_samples_per_second": 296.762,
+      "eval_steps_per_second": 37.497,
       "step": 6468
     },
     {
@@ -4971,9 +4971,9 @@
       "epoch": 34.0,
       "eval_accuracy": 0.9133574007220217,
       "eval_loss": 0.2679595947265625,
-      "eval_runtime": 0.9366,
-      "eval_samples_per_second": 295.757,
-      "eval_steps_per_second": 37.37,
+      "eval_runtime": 0.936,
+      "eval_samples_per_second": 295.944,
+      "eval_steps_per_second": 37.394,
       "step": 6664
     },
     {
@@ -5120,9 +5120,9 @@
       "epoch": 35.0,
       "eval_accuracy": 0.9061371841155235,
       "eval_loss": 0.3797638416290283,
-      "eval_runtime": 0.9257,
-      "eval_samples_per_second": 299.246,
-      "eval_steps_per_second": 37.811,
+      "eval_runtime": 0.9298,
+      "eval_samples_per_second": 297.901,
+      "eval_steps_per_second": 37.641,
       "step": 6860
     },
     {
@@ -5262,9 +5262,9 @@
       "epoch": 36.0,
       "eval_accuracy": 0.855595667870036,
       "eval_loss": 0.5098311305046082,
-      "eval_runtime": 0.9275,
-      "eval_samples_per_second": 298.64,
-      "eval_steps_per_second": 37.734,
+      "eval_runtime": 0.9428,
+      "eval_samples_per_second": 293.818,
+      "eval_steps_per_second": 37.125,
       "step": 7056
     },
     {
@@ -5411,9 +5411,9 @@
       "epoch": 37.0,
       "eval_accuracy": 0.8808664259927798,
       "eval_loss": 0.42190420627593994,
-      "eval_runtime": 0.9289,
-      "eval_samples_per_second": 298.203,
-      "eval_steps_per_second": 37.679,
+      "eval_runtime": 0.9304,
+      "eval_samples_per_second": 297.732,
+      "eval_steps_per_second": 37.62,
       "step": 7252
     },
     {
@@ -5553,9 +5553,9 @@
       "epoch": 38.0,
       "eval_accuracy": 0.9133574007220217,
       "eval_loss": 0.2851579189300537,
-      "eval_runtime": 0.9232,
-      "eval_samples_per_second": 300.035,
-      "eval_steps_per_second": 37.911,
+      "eval_runtime": 0.9753,
+      "eval_samples_per_second": 284.019,
+      "eval_steps_per_second": 35.887,
       "step": 7448
     },
     {
@@ -5702,9 +5702,9 @@
       "epoch": 39.0,
       "eval_accuracy": 0.8989169675090253,
       "eval_loss": 0.36030298471450806,
-      "eval_runtime": 0.9253,
-      "eval_samples_per_second": 299.352,
-      "eval_steps_per_second": 37.824,
+      "eval_runtime": 0.9403,
+      "eval_samples_per_second": 294.598,
+      "eval_steps_per_second": 37.224,
       "step": 7644
     },
     {
@@ -5851,9 +5851,9 @@
       "epoch": 40.0,
       "eval_accuracy": 0.8736462093862816,
       "eval_loss": 0.351894736289978,
-      "eval_runtime": 0.9281,
-      "eval_samples_per_second": 298.457,
-      "eval_steps_per_second": 37.711,
+      "eval_runtime": 0.9415,
+      "eval_samples_per_second": 294.212,
+      "eval_steps_per_second": 37.175,
       "step": 7840
     },
     {
@@ -5993,9 +5993,9 @@
       "epoch": 41.0,
       "eval_accuracy": 0.9025270758122743,
       "eval_loss": 0.3343473970890045,
-      "eval_runtime": 0.928,
-      "eval_samples_per_second": 298.486,
-      "eval_steps_per_second": 37.715,
+      "eval_runtime": 0.9379,
+      "eval_samples_per_second": 295.341,
+      "eval_steps_per_second": 37.317,
       "step": 8036
     },
     {
@@ -6142,9 +6142,9 @@
       "epoch": 42.0,
       "eval_accuracy": 0.9169675090252708,
       "eval_loss": 0.3383941352367401,
-      "eval_runtime": 0.9259,
-      "eval_samples_per_second": 299.182,
-      "eval_steps_per_second": 37.803,
+      "eval_runtime": 0.9348,
+      "eval_samples_per_second": 296.313,
+      "eval_steps_per_second": 37.44,
       "step": 8232
     },
     {
@@ -6284,9 +6284,9 @@
       "epoch": 43.0,
       "eval_accuracy": 0.8772563176895307,
       "eval_loss": 0.40458711981773376,
-      "eval_runtime": 0.9234,
-      "eval_samples_per_second": 299.99,
-      "eval_steps_per_second": 37.905,
+      "eval_runtime": 0.9383,
+      "eval_samples_per_second": 295.207,
+      "eval_steps_per_second": 37.301,
       "step": 8428
     },
     {
@@ -6433,9 +6433,9 @@
       "epoch": 44.0,
       "eval_accuracy": 0.8916967509025271,
       "eval_loss": 0.39525121450424194,
-      "eval_runtime": 0.9242,
-      "eval_samples_per_second": 299.71,
-      "eval_steps_per_second": 37.87,
+      "eval_runtime": 0.9341,
+      "eval_samples_per_second": 296.557,
+      "eval_steps_per_second": 37.471,
       "step": 8624
     },
     {
@@ -6582,9 +6582,9 @@
       "epoch": 45.0,
       "eval_accuracy": 0.8592057761732852,
       "eval_loss": 0.5002943873405457,
-      "eval_runtime": 0.9317,
-      "eval_samples_per_second": 297.301,
-      "eval_steps_per_second": 37.565,
+      "eval_runtime": 0.9313,
+      "eval_samples_per_second": 297.43,
+      "eval_steps_per_second": 37.581,
       "step": 8820
     },
     {
@@ -6724,9 +6724,9 @@
       "epoch": 46.0,
       "eval_accuracy": 0.9169675090252708,
       "eval_loss": 0.3231888711452484,
-      "eval_runtime": 0.925,
-      "eval_samples_per_second": 299.445,
-      "eval_steps_per_second": 37.836,
+      "eval_runtime": 0.9353,
+      "eval_samples_per_second": 296.162,
+      "eval_steps_per_second": 37.421,
       "step": 9016
     },
     {
@@ -6873,9 +6873,9 @@
       "epoch": 47.0,
       "eval_accuracy": 0.9169675090252708,
       "eval_loss": 0.3112559914588928,
-      "eval_runtime": 0.9224,
-      "eval_samples_per_second": 300.307,
-      "eval_steps_per_second": 37.945,
+      "eval_runtime": 0.9287,
+      "eval_samples_per_second": 298.262,
+      "eval_steps_per_second": 37.687,
       "step": 9212
     },
     {
@@ -7015,9 +7015,9 @@
       "epoch": 48.0,
       "eval_accuracy": 0.855595667870036,
       "eval_loss": 0.4494384229183197,
-      "eval_runtime": 0.9397,
-      "eval_samples_per_second": 294.784,
-      "eval_steps_per_second": 37.247,
+      "eval_runtime": 0.928,
+      "eval_samples_per_second": 298.494,
+      "eval_steps_per_second": 37.716,
       "step": 9408
     },
     {
@@ -7164,9 +7164,9 @@
       "epoch": 49.0,
       "eval_accuracy": 0.9205776173285198,
       "eval_loss": 0.37217509746551514,
-      "eval_runtime": 0.9345,
-      "eval_samples_per_second": 296.412,
-      "eval_steps_per_second": 37.453,
+      "eval_runtime": 0.9297,
+      "eval_samples_per_second": 297.952,
+      "eval_steps_per_second": 37.647,
       "step": 9604
     },
     {
@@ -7313,9 +7313,9 @@
       "epoch": 50.0,
       "eval_accuracy": 0.9097472924187726,
       "eval_loss": 0.2852139472961426,
-      "eval_runtime": 0.9274,
-      "eval_samples_per_second": 298.696,
-      "eval_steps_per_second": 37.741,
+      "eval_runtime": 0.9306,
+      "eval_samples_per_second": 297.668,
+      "eval_steps_per_second": 37.611,
       "step": 9800
     },
     {
@@ -7455,9 +7455,9 @@
       "epoch": 51.0,
       "eval_accuracy": 0.8628158844765343,
       "eval_loss": 0.4137951135635376,
-      "eval_runtime": 0.9394,
-      "eval_samples_per_second": 294.867,
-      "eval_steps_per_second": 37.258,
+      "eval_runtime": 0.9371,
+      "eval_samples_per_second": 295.604,
+      "eval_steps_per_second": 37.351,
       "step": 9996
     },
     {
@@ -7604,9 +7604,9 @@
       "epoch": 52.0,
       "eval_accuracy": 0.8953068592057761,
       "eval_loss": 0.34130236506462097,
-      "eval_runtime": 0.95,
-      "eval_samples_per_second": 291.574,
-      "eval_steps_per_second": 36.841,
+      "eval_runtime": 0.9387,
+      "eval_samples_per_second": 295.092,
+      "eval_steps_per_second": 37.286,
       "step": 10192
     },
     {
@@ -7746,9 +7746,9 @@
       "epoch": 53.0,
       "eval_accuracy": 0.9169675090252708,
       "eval_loss": 0.30462250113487244,
-      "eval_runtime": 0.9555,
-      "eval_samples_per_second": 289.887,
-      "eval_steps_per_second": 36.628,
+      "eval_runtime": 0.9357,
+      "eval_samples_per_second": 296.019,
+      "eval_steps_per_second": 37.403,
       "step": 10388
     },
     {
@@ -7895,9 +7895,9 @@
       "epoch": 54.0,
       "eval_accuracy": 0.8989169675090253,
       "eval_loss": 0.2849152684211731,
-      "eval_runtime": 0.9234,
-      "eval_samples_per_second": 299.993,
-      "eval_steps_per_second": 37.905,
+      "eval_runtime": 0.9334,
+      "eval_samples_per_second": 296.775,
+      "eval_steps_per_second": 37.499,
       "step": 10584
     },
     {
@@ -8044,9 +8044,9 @@
       "epoch": 55.0,
       "eval_accuracy": 0.9133574007220217,
       "eval_loss": 0.35168057680130005,
-      "eval_runtime": 0.9261,
-      "eval_samples_per_second": 299.089,
-      "eval_steps_per_second": 37.791,
+      "eval_runtime": 0.9361,
+      "eval_samples_per_second": 295.907,
+      "eval_steps_per_second": 37.389,
       "step": 10780
     },
     {
@@ -8186,9 +8186,9 @@
       "epoch": 56.0,
       "eval_accuracy": 0.9025270758122743,
       "eval_loss": 0.45381447672843933,
-      "eval_runtime": 0.9559,
-      "eval_samples_per_second": 289.776,
-      "eval_steps_per_second": 36.614,
+      "eval_runtime": 0.9344,
+      "eval_samples_per_second": 296.45,
+      "eval_steps_per_second": 37.458,
       "step": 10976
     },
     {
@@ -8335,9 +8335,9 @@
       "epoch": 57.0,
       "eval_accuracy": 0.8592057761732852,
       "eval_loss": 0.44916799664497375,
-      "eval_runtime": 0.9386,
-      "eval_samples_per_second": 295.131,
-      "eval_steps_per_second": 37.291,
+      "eval_runtime": 0.9373,
+      "eval_samples_per_second": 295.53,
+      "eval_steps_per_second": 37.341,
       "step": 11172
     },
     {
@@ -8477,9 +8477,9 @@
       "epoch": 58.0,
       "eval_accuracy": 0.8628158844765343,
       "eval_loss": 0.4738902747631073,
-      "eval_runtime": 0.9256,
-      "eval_samples_per_second": 299.274,
-      "eval_steps_per_second": 37.814,
+      "eval_runtime": 0.9332,
+      "eval_samples_per_second": 296.824,
+      "eval_steps_per_second": 37.505,
       "step": 11368
     },
     {
@@ -8626,9 +8626,9 @@
       "epoch": 59.0,
       "eval_accuracy": 0.9205776173285198,
       "eval_loss": 0.2831103205680847,
-      "eval_runtime": 0.9267,
-      "eval_samples_per_second": 298.914,
-      "eval_steps_per_second": 37.769,
+      "eval_runtime": 0.9316,
+      "eval_samples_per_second": 297.324,
+      "eval_steps_per_second": 37.568,
       "step": 11564
     },
     {
@@ -8775,9 +8775,9 @@
       "epoch": 60.0,
       "eval_accuracy": 0.9205776173285198,
       "eval_loss": 0.34654152393341064,
-      "eval_runtime": 0.924,
-      "eval_samples_per_second": 299.794,
-      "eval_steps_per_second": 37.88,
+      "eval_runtime": 0.9298,
+      "eval_samples_per_second": 297.899,
+      "eval_steps_per_second": 37.641,
       "step": 11760
     },
     {
@@ -8917,9 +8917,9 @@
       "epoch": 61.0,
       "eval_accuracy": 0.9314079422382672,
       "eval_loss": 0.28843551874160767,
-      "eval_runtime": 0.9387,
-      "eval_samples_per_second": 295.083,
-      "eval_steps_per_second": 37.285,
+      "eval_runtime": 0.938,
+      "eval_samples_per_second": 295.31,
+      "eval_steps_per_second": 37.314,
       "step": 11956
     },
     {
@@ -9066,9 +9066,9 @@
       "epoch": 62.0,
       "eval_accuracy": 0.8447653429602888,
       "eval_loss": 0.4874745309352875,
-      "eval_runtime": 0.9245,
-      "eval_samples_per_second": 299.628,
-      "eval_steps_per_second": 37.859,
+      "eval_runtime": 0.935,
+      "eval_samples_per_second": 296.251,
+      "eval_steps_per_second": 37.432,
       "step": 12152
     },
     {
@@ -9208,9 +9208,9 @@
       "epoch": 63.0,
       "eval_accuracy": 0.9061371841155235,
       "eval_loss": 0.3988655209541321,
-      "eval_runtime": 0.9265,
-      "eval_samples_per_second": 298.961,
-      "eval_steps_per_second": 37.775,
+      "eval_runtime": 0.9339,
+      "eval_samples_per_second": 296.617,
+      "eval_steps_per_second": 37.479,
       "step": 12348
     },
     {
@@ -9357,9 +9357,9 @@
       "epoch": 64.0,
       "eval_accuracy": 0.851985559566787,
       "eval_loss": 0.5952923893928528,
-      "eval_runtime": 0.9245,
-      "eval_samples_per_second": 299.616,
-      "eval_steps_per_second": 37.858,
+      "eval_runtime": 0.9334,
+      "eval_samples_per_second": 296.767,
+      "eval_steps_per_second": 37.498,
       "step": 12544
     },
     {
@@ -9506,9 +9506,9 @@
       "epoch": 65.0,
       "eval_accuracy": 0.927797833935018,
       "eval_loss": 0.2953914701938629,
-      "eval_runtime": 0.921,
-      "eval_samples_per_second": 300.758,
-      "eval_steps_per_second": 38.002,
+      "eval_runtime": 0.9948,
+      "eval_samples_per_second": 278.444,
+      "eval_steps_per_second": 35.183,
       "step": 12740
     },
     {
@@ -9648,9 +9648,9 @@
       "epoch": 66.0,
       "eval_accuracy": 0.851985559566787,
       "eval_loss": 0.5033073425292969,
-      "eval_runtime": 0.9234,
-      "eval_samples_per_second": 299.974,
-      "eval_steps_per_second": 37.903,
+      "eval_runtime": 0.9324,
+      "eval_samples_per_second": 297.09,
+      "eval_steps_per_second": 37.538,
       "step": 12936
     },
     {
@@ -9797,9 +9797,9 @@
       "epoch": 67.0,
       "eval_accuracy": 0.8808664259927798,
       "eval_loss": 0.41321179270744324,
-      "eval_runtime": 0.9749,
-      "eval_samples_per_second": 284.139,
-      "eval_steps_per_second": 35.902,
+      "eval_runtime": 0.9337,
+      "eval_samples_per_second": 296.681,
+      "eval_steps_per_second": 37.487,
       "step": 13132
     },
     {
@@ -9939,9 +9939,9 @@
       "epoch": 68.0,
       "eval_accuracy": 0.9169675090252708,
       "eval_loss": 0.37430626153945923,
-      "eval_runtime": 0.927,
-      "eval_samples_per_second": 298.802,
-      "eval_steps_per_second": 37.755,
+      "eval_runtime": 0.9326,
+      "eval_samples_per_second": 297.033,
+      "eval_steps_per_second": 37.531,
       "step": 13328
     },
     {
@@ -10088,9 +10088,9 @@
       "epoch": 69.0,
       "eval_accuracy": 0.776173285198556,
       "eval_loss": 0.6423271298408508,
-      "eval_runtime": 0.9196,
-      "eval_samples_per_second": 301.217,
-      "eval_steps_per_second": 38.06,
+      "eval_runtime": 0.9344,
+      "eval_samples_per_second": 296.457,
+      "eval_steps_per_second": 37.458,
       "step": 13524
     },
     {
@@ -10237,9 +10237,9 @@
       "epoch": 70.0,
       "eval_accuracy": 0.8014440433212996,
       "eval_loss": 0.5979345440864563,
-      "eval_runtime": 0.9248,
-      "eval_samples_per_second": 299.519,
-      "eval_steps_per_second": 37.845,
+      "eval_runtime": 0.9413,
+      "eval_samples_per_second": 294.271,
+      "eval_steps_per_second": 37.182,
       "step": 13720
     },
     {
@@ -10379,9 +10379,9 @@
       "epoch": 71.0,
       "eval_accuracy": 0.8989169675090253,
       "eval_loss": 0.43981537222862244,
-      "eval_runtime": 0.9267,
-      "eval_samples_per_second": 298.919,
-      "eval_steps_per_second": 37.77,
+      "eval_runtime": 0.9369,
+      "eval_samples_per_second": 295.663,
+      "eval_steps_per_second": 37.358,
       "step": 13916
     },
     {
@@ -10528,9 +10528,9 @@
       "epoch": 72.0,
       "eval_accuracy": 0.8231046931407943,
       "eval_loss": 0.5633912086486816,
-      "eval_runtime": 0.9302,
-      "eval_samples_per_second": 297.785,
-      "eval_steps_per_second": 37.626,
+      "eval_runtime": 0.9331,
+      "eval_samples_per_second": 296.846,
+      "eval_steps_per_second": 37.508,
       "step": 14112
     },
     {
@@ -10670,9 +10670,9 @@
       "epoch": 73.0,
       "eval_accuracy": 0.776173285198556,
       "eval_loss": 0.620398223400116,
-      "eval_runtime": 0.9237,
-      "eval_samples_per_second": 299.897,
-      "eval_steps_per_second": 37.893,
+      "eval_runtime": 0.9369,
+      "eval_samples_per_second": 295.645,
+      "eval_steps_per_second": 37.356,
       "step": 14308
     },
     {
@@ -10819,9 +10819,9 @@
       "epoch": 74.0,
       "eval_accuracy": 0.8953068592057761,
       "eval_loss": 0.36853671073913574,
-      "eval_runtime": 0.9295,
-      "eval_samples_per_second": 298.006,
-      "eval_steps_per_second": 37.654,
+      "eval_runtime": 0.9373,
+      "eval_samples_per_second": 295.515,
+      "eval_steps_per_second": 37.339,
       "step": 14504
     },
     {
@@ -10968,9 +10968,9 @@
       "epoch": 75.0,
       "eval_accuracy": 0.8158844765342961,
       "eval_loss": 0.7026161551475525,
-      "eval_runtime": 0.9261,
-      "eval_samples_per_second": 299.112,
-      "eval_steps_per_second": 37.794,
+      "eval_runtime": 0.9279,
+      "eval_samples_per_second": 298.519,
+      "eval_steps_per_second": 37.719,
       "step": 14700
     },
     {
@@ -11110,9 +11110,9 @@
       "epoch": 76.0,
       "eval_accuracy": 0.9169675090252708,
       "eval_loss": 0.40481600165367126,
-      "eval_runtime": 0.9284,
-      "eval_samples_per_second": 298.348,
-      "eval_steps_per_second": 37.697,
+      "eval_runtime": 0.9401,
+      "eval_samples_per_second": 294.635,
+      "eval_steps_per_second": 37.228,
       "step": 14896
     },
     {
@@ -11259,9 +11259,9 @@
       "epoch": 77.0,
       "eval_accuracy": 0.8844765342960289,
       "eval_loss": 0.48912858963012695,
-      "eval_runtime": 0.9337,
-      "eval_samples_per_second": 296.677,
-      "eval_steps_per_second": 37.486,
+      "eval_runtime": 0.9257,
+      "eval_samples_per_second": 299.23,
+      "eval_steps_per_second": 37.809,
       "step": 15092
     },
     {
@@ -11401,9 +11401,9 @@
       "epoch": 78.0,
       "eval_accuracy": 0.8772563176895307,
       "eval_loss": 0.42159637808799744,
-      "eval_runtime": 0.9262,
-      "eval_samples_per_second": 299.086,
-      "eval_steps_per_second": 37.791,
+      "eval_runtime": 0.9931,
+      "eval_samples_per_second": 278.916,
+      "eval_steps_per_second": 35.242,
       "step": 15288
     },
     {
@@ -11550,9 +11550,9 @@
       "epoch": 79.0,
       "eval_accuracy": 0.8953068592057761,
       "eval_loss": 0.2720978558063507,
-      "eval_runtime": 0.9227,
-      "eval_samples_per_second": 300.203,
-      "eval_steps_per_second": 37.932,
+      "eval_runtime": 0.9347,
+      "eval_samples_per_second": 296.358,
+      "eval_steps_per_second": 37.446,
       "step": 15484
     },
     {
@@ -11699,9 +11699,9 @@
       "epoch": 80.0,
       "eval_accuracy": 0.8483754512635379,
       "eval_loss": 0.42701977491378784,
-      "eval_runtime": 0.9196,
-      "eval_samples_per_second": 301.231,
-      "eval_steps_per_second": 38.062,
+      "eval_runtime": 0.9306,
+      "eval_samples_per_second": 297.657,
+      "eval_steps_per_second": 37.61,
       "step": 15680
     },
     {
@@ -11841,9 +11841,9 @@
       "epoch": 81.0,
       "eval_accuracy": 0.7833935018050542,
       "eval_loss": 0.763225793838501,
-      "eval_runtime": 0.9257,
-      "eval_samples_per_second": 299.248,
-      "eval_steps_per_second": 37.811,
+      "eval_runtime": 0.9357,
+      "eval_samples_per_second": 296.038,
+      "eval_steps_per_second": 37.405,
       "step": 15876
     },
     {
@@ -11990,9 +11990,9 @@
       "epoch": 82.0,
       "eval_accuracy": 0.9025270758122743,
       "eval_loss": 0.3249270021915436,
-      "eval_runtime": 0.9713,
-      "eval_samples_per_second": 285.18,
-      "eval_steps_per_second": 36.034,
+      "eval_runtime": 0.9334,
+      "eval_samples_per_second": 296.754,
+      "eval_steps_per_second": 37.496,
       "step": 16072
     },
     {
@@ -12132,9 +12132,9 @@
       "epoch": 83.0,
       "eval_accuracy": 0.8989169675090253,
       "eval_loss": 0.38384127616882324,
-      "eval_runtime": 0.9259,
-      "eval_samples_per_second": 299.169,
-      "eval_steps_per_second": 37.801,
+      "eval_runtime": 0.9338,
+      "eval_samples_per_second": 296.64,
+      "eval_steps_per_second": 37.482,
       "step": 16268
     },
     {
@@ -12281,9 +12281,9 @@
       "epoch": 84.0,
       "eval_accuracy": 0.9205776173285198,
       "eval_loss": 0.3329784572124481,
-      "eval_runtime": 0.9245,
-      "eval_samples_per_second": 299.612,
-      "eval_steps_per_second": 37.857,
+      "eval_runtime": 0.9338,
+      "eval_samples_per_second": 296.638,
+      "eval_steps_per_second": 37.481,
       "step": 16464
     },
     {
@@ -12430,9 +12430,9 @@
       "epoch": 85.0,
       "eval_accuracy": 0.8700361010830325,
       "eval_loss": 0.32456666231155396,
-      "eval_runtime": 0.929,
-      "eval_samples_per_second": 298.181,
-      "eval_steps_per_second": 37.676,
+      "eval_runtime": 0.9335,
+      "eval_samples_per_second": 296.719,
+      "eval_steps_per_second": 37.492,
       "step": 16660
     },
     {
@@ -12572,9 +12572,9 @@
       "epoch": 86.0,
       "eval_accuracy": 0.8736462093862816,
       "eval_loss": 0.3710058927536011,
-      "eval_runtime": 0.9876,
-      "eval_samples_per_second": 280.471,
-      "eval_steps_per_second": 35.439,
+      "eval_runtime": 0.9462,
+      "eval_samples_per_second": 292.762,
+      "eval_steps_per_second": 36.992,
       "step": 16856
     },
     {
@@ -12721,9 +12721,9 @@
       "epoch": 87.0,
       "eval_accuracy": 0.9061371841155235,
       "eval_loss": 0.33839184045791626,
-      "eval_runtime": 0.9267,
-      "eval_samples_per_second": 298.923,
-      "eval_steps_per_second": 37.77,
+      "eval_runtime": 0.9931,
+      "eval_samples_per_second": 278.938,
+      "eval_steps_per_second": 35.245,
       "step": 17052
     },
     {
@@ -12863,9 +12863,9 @@
       "epoch": 88.0,
       "eval_accuracy": 0.9097472924187726,
       "eval_loss": 0.3703186511993408,
-      "eval_runtime": 0.9909,
-      "eval_samples_per_second": 279.556,
-      "eval_steps_per_second": 35.323,
+      "eval_runtime": 0.9324,
+      "eval_samples_per_second": 297.094,
+      "eval_steps_per_second": 37.539,
       "step": 17248
     },
     {
@@ -13012,9 +13012,9 @@
       "epoch": 89.0,
       "eval_accuracy": 0.8592057761732852,
       "eval_loss": 0.5013492703437805,
-      "eval_runtime": 0.9203,
-      "eval_samples_per_second": 300.985,
-      "eval_steps_per_second": 38.031,
+      "eval_runtime": 0.9311,
+      "eval_samples_per_second": 297.495,
+      "eval_steps_per_second": 37.59,
       "step": 17444
     },
     {
@@ -13161,9 +13161,9 @@
       "epoch": 90.0,
       "eval_accuracy": 0.8989169675090253,
       "eval_loss": 0.3456423580646515,
-      "eval_runtime": 1.0017,
-      "eval_samples_per_second": 276.522,
-      "eval_steps_per_second": 34.94,
+      "eval_runtime": 0.9297,
+      "eval_samples_per_second": 297.96,
+      "eval_steps_per_second": 37.648,
       "step": 17640
     },
     {
@@ -13303,9 +13303,9 @@
       "epoch": 91.0,
       "eval_accuracy": 0.8880866425992779,
       "eval_loss": 0.3798539340496063,
-      "eval_runtime": 0.9203,
-      "eval_samples_per_second": 301.002,
-      "eval_steps_per_second": 38.033,
+      "eval_runtime": 0.9367,
+      "eval_samples_per_second": 295.72,
+      "eval_steps_per_second": 37.365,
       "step": 17836
     },
     {
@@ -13452,9 +13452,9 @@
       "epoch": 92.0,
       "eval_accuracy": 0.9061371841155235,
       "eval_loss": 0.32391440868377686,
-      "eval_runtime": 0.9941,
-      "eval_samples_per_second": 278.638,
-      "eval_steps_per_second": 35.207,
+      "eval_runtime": 0.9318,
+      "eval_samples_per_second": 297.278,
+      "eval_steps_per_second": 37.562,
       "step": 18032
     },
     {
@@ -13594,9 +13594,9 @@
       "epoch": 93.0,
       "eval_accuracy": 0.8989169675090253,
       "eval_loss": 0.4861180782318115,
-      "eval_runtime": 0.923,
-      "eval_samples_per_second": 300.096,
-      "eval_steps_per_second": 37.918,
+      "eval_runtime": 0.9969,
+      "eval_samples_per_second": 277.851,
+      "eval_steps_per_second": 35.107,
       "step": 18228
     },
     {
@@ -13743,9 +13743,9 @@
       "epoch": 94.0,
       "eval_accuracy": 0.8736462093862816,
       "eval_loss": 0.4443919062614441,
-      "eval_runtime": 0.9312,
-      "eval_samples_per_second": 297.478,
-      "eval_steps_per_second": 37.587,
+      "eval_runtime": 0.9334,
+      "eval_samples_per_second": 296.762,
+      "eval_steps_per_second": 37.497,
       "step": 18424
     },
     {
@@ -13892,9 +13892,9 @@
       "epoch": 95.0,
       "eval_accuracy": 0.8844765342960289,
       "eval_loss": 0.3712524473667145,
-      "eval_runtime": 0.9321,
-      "eval_samples_per_second": 297.175,
-      "eval_steps_per_second": 37.549,
+      "eval_runtime": 0.9337,
+      "eval_samples_per_second": 296.658,
+      "eval_steps_per_second": 37.484,
       "step": 18620
     },
     {
@@ -14034,9 +14034,9 @@
       "epoch": 96.0,
       "eval_accuracy": 0.8303249097472925,
       "eval_loss": 0.427941232919693,
-      "eval_runtime": 1.0012,
-      "eval_samples_per_second": 276.68,
-      "eval_steps_per_second": 34.96,
+      "eval_runtime": 0.9388,
+      "eval_samples_per_second": 295.043,
+      "eval_steps_per_second": 37.28,
       "step": 18816
     },
     {
@@ -14183,9 +14183,9 @@
       "epoch": 97.0,
       "eval_accuracy": 0.8411552346570397,
       "eval_loss": 0.6587985157966614,
-      "eval_runtime": 0.937,
-      "eval_samples_per_second": 295.619,
-      "eval_steps_per_second": 37.353,
+      "eval_runtime": 0.9391,
+      "eval_samples_per_second": 294.975,
+      "eval_steps_per_second": 37.271,
       "step": 19012
     },
     {
@@ -14325,9 +14325,9 @@
       "epoch": 98.0,
       "eval_accuracy": 0.8267148014440433,
       "eval_loss": 0.66323322057724,
-      "eval_runtime": 0.928,
-      "eval_samples_per_second": 298.496,
-      "eval_steps_per_second": 37.716,
+      "eval_runtime": 0.9369,
+      "eval_samples_per_second": 295.657,
+      "eval_steps_per_second": 37.357,
       "step": 19208
     },
     {
@@ -14474,9 +14474,9 @@
       "epoch": 99.0,
       "eval_accuracy": 0.8808664259927798,
       "eval_loss": 0.4641692638397217,
-      "eval_runtime": 0.9262,
-      "eval_samples_per_second": 299.067,
-      "eval_steps_per_second": 37.788,
+      "eval_runtime": 0.9404,
+      "eval_samples_per_second": 294.555,
+      "eval_steps_per_second": 37.218,
       "step": 19404
     },
     {
@@ -14623,9 +14623,9 @@
       "epoch": 100.0,
       "eval_accuracy": 0.8808664259927798,
       "eval_loss": 0.3624129593372345,
-      "eval_runtime": 0.9461,
-      "eval_samples_per_second": 292.771,
-      "eval_steps_per_second": 36.993,
+      "eval_runtime": 1.0032,
+      "eval_samples_per_second": 276.121,
+      "eval_steps_per_second": 34.889,
       "step": 19600
     },
     {
@@ -14633,9 +14633,9 @@
       "step": 19600,
       "total_flos": 5.678990728814592e+17,
       "train_loss": 0.30055975163317455,
-      "train_runtime": 936.3063,
-      "train_samples_per_second": 167.253,
-      "train_steps_per_second": 20.933
+      "train_runtime": 949.6972,
+      "train_samples_per_second": 164.895,
+      "train_steps_per_second": 20.638
     }
   ],
   "logging_steps": 10,