🍻 cheers

README.md

@@ -2,6 +2,7 @@
 license: apache-2.0
 base_model: google/vit-base-patch16-224-in21k
 tags:
+- image-classification
 - generated_from_trainer
 datasets:
 - image_folder
@@ -14,7 +15,7 @@ model-index:
       name: Image Classification
       type: image-classification
     dataset:
-      name: image_folder
+      name: agent_action_class
       type: image_folder
       config: hierarchical-action-agent
       split: train
@@ -22,7 +23,7 @@ model-index:
     metrics:
     - name: Accuracy
       type: accuracy
-      value: 0.8330935251798561
+      value: 0.8402877697841726
 ---
 
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
@@ -30,10 +31,10 @@ should probably proofread and complete it, then remove this comment. -->
 
 # Hierarchical_Agent_Action
 
-This model is a fine-tuned version of [google/vit-base-patch16-224-in21k](https://huggingface.co/google/vit-base-patch16-224-in21k) on the image_folder dataset.
+This model is a fine-tuned version of [google/vit-base-patch16-224-in21k](https://huggingface.co/google/vit-base-patch16-224-in21k) on the agent_action_class dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.6262
-- Accuracy: 0.8331
+- Loss: 0.5942
+- Accuracy: 0.8403
 
 ## Model description
 

all_results.json

@@ -1,13 +1,13 @@
 {
-    "epoch": 20.0,
-    "eval_accuracy": 0.8215827338129497,
-    "eval_loss": 0.6021825075149536,
-    "eval_runtime": 12.5627,
-    "eval_samples_per_second": 55.322,
-    "eval_steps_per_second": 6.925,
-    "total_flos": 6.103469271205233e+18,
-    "train_loss": 0.6948060337574251,
-    "train_runtime": 2796.6359,
-    "train_samples_per_second": 28.155,
-    "train_steps_per_second": 0.887
+    "epoch": 30.0,
+    "eval_accuracy": 0.8402877697841726,
+    "eval_loss": 0.5942012071609497,
+    "eval_runtime": 12.2205,
+    "eval_samples_per_second": 56.872,
+    "eval_steps_per_second": 7.119,
+    "total_flos": 9.155203906807849e+18,
+    "train_loss": 0.5316014153983003,
+    "train_runtime": 4018.8325,
+    "train_samples_per_second": 29.389,
+    "train_steps_per_second": 0.926
 }

eval_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 20.0,
-    "eval_accuracy": 0.8215827338129497,
-    "eval_loss": 0.6021825075149536,
-    "eval_runtime": 12.5627,
-    "eval_samples_per_second": 55.322,
-    "eval_steps_per_second": 6.925
+    "epoch": 30.0,
+    "eval_accuracy": 0.8402877697841726,
+    "eval_loss": 0.5942012071609497,
+    "eval_runtime": 12.2205,
+    "eval_samples_per_second": 56.872,
+    "eval_steps_per_second": 7.119
 }

runs/Dec27_07-48-13_fc6b8939f5d8/events.out.tfevents.1703667338.fc6b8939f5d8.47.3

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5cccf719961c28b8e75b354a3fac564d928f8a8e2592a378edba4b45e26b9969
+size 411

train_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 20.0,
-    "total_flos": 6.103469271205233e+18,
-    "train_loss": 0.6948060337574251,
-    "train_runtime": 2796.6359,
-    "train_samples_per_second": 28.155,
-    "train_steps_per_second": 0.887
+    "epoch": 30.0,
+    "total_flos": 9.155203906807849e+18,
+    "train_loss": 0.5316014153983003,
+    "train_runtime": 4018.8325,
+    "train_samples_per_second": 29.389,
+    "train_steps_per_second": 0.926
 }

trainer_state.json

@@ -1,988 +1,1477 @@
 {
-  "best_metric": 0.6021825075149536,
-  "best_model_checkpoint": "Hierarchical_Agent_Action/checkpoint-2200",
-  "epoch": 20.0,
+  "best_metric": 0.5942012071609497,
+  "best_model_checkpoint": "Hierarchical_Agent_Action/checkpoint-3100",
+  "epoch": 30.0,
   "eval_steps": 100,
-  "global_step": 2480,
+  "global_step": 3720,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 0.16,
-      "learning_rate": 9.919354838709678e-05,
-      "loss": 3.4603,
+      "learning_rate": 9.946236559139786e-05,
+      "loss": 3.4378,
       "step": 20
     },
     {
       "epoch": 0.32,
-      "learning_rate": 9.838709677419355e-05,
-      "loss": 3.2338,
+      "learning_rate": 9.892473118279571e-05,
+      "loss": 3.1863,
       "step": 40
     },
     {
       "epoch": 0.48,
-      "learning_rate": 9.758064516129033e-05,
-      "loss": 2.9478,
+      "learning_rate": 9.838709677419355e-05,
+      "loss": 2.9075,
       "step": 60
     },
     {
       "epoch": 0.65,
-      "learning_rate": 9.677419354838711e-05,
-      "loss": 2.7149,
+      "learning_rate": 9.78494623655914e-05,
+      "loss": 2.6661,
       "step": 80
     },
     {
       "epoch": 0.81,
-      "learning_rate": 9.596774193548387e-05,
-      "loss": 2.4733,
+      "learning_rate": 9.731182795698925e-05,
+      "loss": 2.4407,
       "step": 100
     },
     {
       "epoch": 0.81,
-      "eval_accuracy": 0.6172661870503597,
-      "eval_loss": 2.2958028316497803,
-      "eval_runtime": 12.3246,
-      "eval_samples_per_second": 56.391,
-      "eval_steps_per_second": 7.059,
+      "eval_accuracy": 0.6057553956834533,
+      "eval_loss": 2.271561622619629,
+      "eval_runtime": 12.2022,
+      "eval_samples_per_second": 56.957,
+      "eval_steps_per_second": 7.13,
       "step": 100
     },
     {
       "epoch": 0.97,
-      "learning_rate": 9.516129032258065e-05,
-      "loss": 2.308,
+      "learning_rate": 9.677419354838711e-05,
+      "loss": 2.2877,
       "step": 120
     },
     {
       "epoch": 1.13,
-      "learning_rate": 9.439516129032259e-05,
-      "loss": 2.0956,
+      "learning_rate": 9.626344086021506e-05,
+      "loss": 2.0861,
       "step": 140
     },
     {
       "epoch": 1.29,
-      "learning_rate": 9.358870967741936e-05,
-      "loss": 2.0524,
+      "learning_rate": 9.572580645161291e-05,
+      "loss": 2.0179,
       "step": 160
     },
     {
       "epoch": 1.45,
-      "learning_rate": 9.278225806451613e-05,
-      "loss": 1.8984,
+      "learning_rate": 9.518817204301076e-05,
+      "loss": 1.8839,
       "step": 180
     },
     {
       "epoch": 1.61,
-      "learning_rate": 9.197580645161291e-05,
-      "loss": 1.7952,
+      "learning_rate": 9.465053763440861e-05,
+      "loss": 1.7756,
       "step": 200
     },
     {
       "epoch": 1.61,
-      "eval_accuracy": 0.7251798561151079,
-      "eval_loss": 1.6266770362854004,
-      "eval_runtime": 12.444,
-      "eval_samples_per_second": 55.85,
-      "eval_steps_per_second": 6.991,
+      "eval_accuracy": 0.7064748201438849,
+      "eval_loss": 1.6162313222885132,
+      "eval_runtime": 13.1922,
+      "eval_samples_per_second": 52.683,
+      "eval_steps_per_second": 6.595,
       "step": 200
     },
     {
       "epoch": 1.77,
-      "learning_rate": 9.116935483870969e-05,
-      "loss": 1.7496,
+      "learning_rate": 9.411290322580645e-05,
+      "loss": 1.7271,
       "step": 220
     },
     {
       "epoch": 1.94,
-      "learning_rate": 9.036290322580646e-05,
-      "loss": 1.5949,
+      "learning_rate": 9.357526881720431e-05,
+      "loss": 1.5882,
       "step": 240
     },
     {
       "epoch": 2.1,
-      "learning_rate": 8.955645161290322e-05,
-      "loss": 1.4567,
+      "learning_rate": 9.303763440860216e-05,
+      "loss": 1.4613,
       "step": 260
     },
     {
       "epoch": 2.26,
-      "learning_rate": 8.875e-05,
-      "loss": 1.4255,
+      "learning_rate": 9.250000000000001e-05,
+      "loss": 1.4197,
       "step": 280
     },
     {
       "epoch": 2.42,
-      "learning_rate": 8.794354838709677e-05,
-      "loss": 1.409,
+      "learning_rate": 9.196236559139786e-05,
+      "loss": 1.3948,
       "step": 300
     },
     {
       "epoch": 2.42,
-      "eval_accuracy": 0.7525179856115108,
-      "eval_loss": 1.2853118181228638,
-      "eval_runtime": 12.6438,
-      "eval_samples_per_second": 54.968,
-      "eval_steps_per_second": 6.881,
+      "eval_accuracy": 0.7697841726618705,
+      "eval_loss": 1.2199510335922241,
+      "eval_runtime": 12.2033,
+      "eval_samples_per_second": 56.952,
+      "eval_steps_per_second": 7.129,
       "step": 300
     },
     {
       "epoch": 2.58,
-      "learning_rate": 8.713709677419355e-05,
-      "loss": 1.3729,
+      "learning_rate": 9.14247311827957e-05,
+      "loss": 1.3314,
       "step": 320
     },
     {
       "epoch": 2.74,
-      "learning_rate": 8.633064516129033e-05,
-      "loss": 1.3318,
+      "learning_rate": 9.088709677419354e-05,
+      "loss": 1.3175,
       "step": 340
     },
     {
       "epoch": 2.9,
-      "learning_rate": 8.55241935483871e-05,
-      "loss": 1.246,
+      "learning_rate": 9.03494623655914e-05,
+      "loss": 1.2503,
       "step": 360
     },
     {
       "epoch": 3.06,
-      "learning_rate": 8.471774193548387e-05,
-      "loss": 1.1057,
+      "learning_rate": 8.983870967741936e-05,
+      "loss": 1.1104,
       "step": 380
     },
     {
       "epoch": 3.23,
-      "learning_rate": 8.391129032258065e-05,
-      "loss": 1.1321,
+      "learning_rate": 8.930107526881721e-05,
+      "loss": 1.131,
       "step": 400
     },
     {
       "epoch": 3.23,
-      "eval_accuracy": 0.7971223021582734,
-      "eval_loss": 0.9899287819862366,
-      "eval_runtime": 12.5295,
-      "eval_samples_per_second": 55.469,
-      "eval_steps_per_second": 6.944,
+      "eval_accuracy": 0.7856115107913669,
+      "eval_loss": 1.0011802911758423,
+      "eval_runtime": 12.1771,
+      "eval_samples_per_second": 57.074,
+      "eval_steps_per_second": 7.145,
       "step": 400
     },
     {
       "epoch": 3.39,
-      "learning_rate": 8.310483870967743e-05,
-      "loss": 1.0894,
+      "learning_rate": 8.876344086021506e-05,
+      "loss": 1.1084,
       "step": 420
     },
     {
       "epoch": 3.55,
-      "learning_rate": 8.22983870967742e-05,
-      "loss": 1.0102,
+      "learning_rate": 8.82258064516129e-05,
+      "loss": 0.9978,
       "step": 440
     },
     {
       "epoch": 3.71,
-      "learning_rate": 8.149193548387097e-05,
-      "loss": 1.0241,
+      "learning_rate": 8.768817204301076e-05,
+      "loss": 1.0633,
       "step": 460
     },
     {
       "epoch": 3.87,
-      "learning_rate": 8.068548387096775e-05,
-      "loss": 0.9724,
+      "learning_rate": 8.715053763440861e-05,
+      "loss": 0.9801,
       "step": 480
     },
     {
       "epoch": 4.03,
-      "learning_rate": 7.987903225806452e-05,
-      "loss": 0.9094,
+      "learning_rate": 8.661290322580645e-05,
+      "loss": 0.9239,
       "step": 500
     },
     {
       "epoch": 4.03,
-      "eval_accuracy": 0.8028776978417266,
-      "eval_loss": 0.8827974200248718,
-      "eval_runtime": 12.471,
-      "eval_samples_per_second": 55.729,
-      "eval_steps_per_second": 6.976,
+      "eval_accuracy": 0.7827338129496403,
+      "eval_loss": 0.9055307507514954,
+      "eval_runtime": 12.1524,
+      "eval_samples_per_second": 57.19,
+      "eval_steps_per_second": 7.159,
       "step": 500
     },
     {
       "epoch": 4.19,
-      "learning_rate": 7.90725806451613e-05,
-      "loss": 0.8774,
+      "learning_rate": 8.60752688172043e-05,
+      "loss": 0.8721,
       "step": 520
     },
     {
       "epoch": 4.35,
-      "learning_rate": 7.826612903225808e-05,
-      "loss": 0.8719,
+      "learning_rate": 8.553763440860215e-05,
+      "loss": 0.8904,
       "step": 540
     },
     {
       "epoch": 4.52,
-      "learning_rate": 7.745967741935484e-05,
-      "loss": 0.9112,
+      "learning_rate": 8.5e-05,
+      "loss": 0.8903,
       "step": 560
     },
     {
       "epoch": 4.68,
-      "learning_rate": 7.665322580645161e-05,
-      "loss": 0.8573,
+      "learning_rate": 8.446236559139785e-05,
+      "loss": 0.8564,
       "step": 580
     },
     {
       "epoch": 4.84,
-      "learning_rate": 7.584677419354839e-05,
-      "loss": 0.8552,
+      "learning_rate": 8.39247311827957e-05,
+      "loss": 0.8699,
       "step": 600
     },
     {
       "epoch": 4.84,
-      "eval_accuracy": 0.7884892086330936,
-      "eval_loss": 0.8248726725578308,
-      "eval_runtime": 12.3792,
-      "eval_samples_per_second": 56.143,
-      "eval_steps_per_second": 7.028,
+      "eval_accuracy": 0.7827338129496403,
+      "eval_loss": 0.8103253841400146,
+      "eval_runtime": 11.9742,
+      "eval_samples_per_second": 58.041,
+      "eval_steps_per_second": 7.266,
       "step": 600
     },
     {
       "epoch": 5.0,
-      "learning_rate": 7.504032258064517e-05,
-      "loss": 0.8021,
+      "learning_rate": 8.338709677419355e-05,
+      "loss": 0.8503,
       "step": 620
     },
     {
       "epoch": 5.16,
-      "learning_rate": 7.423387096774193e-05,
-      "loss": 0.7215,
+      "learning_rate": 8.28494623655914e-05,
+      "loss": 0.7251,
       "step": 640
     },
     {
       "epoch": 5.32,
-      "learning_rate": 7.342741935483871e-05,
-      "loss": 0.6933,
+      "learning_rate": 8.231182795698926e-05,
+      "loss": 0.6939,
       "step": 660
     },
     {
       "epoch": 5.48,
-      "learning_rate": 7.262096774193549e-05,
-      "loss": 0.7272,
+      "learning_rate": 8.17741935483871e-05,
+      "loss": 0.7263,
       "step": 680
     },
     {
       "epoch": 5.65,
-      "learning_rate": 7.181451612903225e-05,
-      "loss": 0.7082,
+      "learning_rate": 8.123655913978495e-05,
+      "loss": 0.6707,
       "step": 700
     },
     {
       "epoch": 5.65,
-      "eval_accuracy": 0.7755395683453238,
-      "eval_loss": 0.8111026287078857,
-      "eval_runtime": 12.4288,
-      "eval_samples_per_second": 55.919,
-      "eval_steps_per_second": 7.0,
+      "eval_accuracy": 0.7841726618705036,
+      "eval_loss": 0.7609687447547913,
+      "eval_runtime": 12.2276,
+      "eval_samples_per_second": 56.839,
+      "eval_steps_per_second": 7.115,
       "step": 700
     },
     {
       "epoch": 5.81,
-      "learning_rate": 7.100806451612903e-05,
-      "loss": 0.6704,
+      "learning_rate": 8.06989247311828e-05,
+      "loss": 0.6822,
       "step": 720
     },
     {
       "epoch": 5.97,
-      "learning_rate": 7.020161290322581e-05,
-      "loss": 0.6963,
+      "learning_rate": 8.016129032258065e-05,
+      "loss": 0.716,
       "step": 740
     },
     {
       "epoch": 6.13,
-      "learning_rate": 6.939516129032259e-05,
-      "loss": 0.6072,
+      "learning_rate": 7.96236559139785e-05,
+      "loss": 0.6162,
       "step": 760
     },
     {
       "epoch": 6.29,
-      "learning_rate": 6.858870967741936e-05,
-      "loss": 0.65,
+      "learning_rate": 7.908602150537635e-05,
+      "loss": 0.6409,
       "step": 780
     },
     {
       "epoch": 6.45,
-      "learning_rate": 6.778225806451614e-05,
-      "loss": 0.5812,
+      "learning_rate": 7.85483870967742e-05,
+      "loss": 0.6206,
       "step": 800
     },
     {
       "epoch": 6.45,
-      "eval_accuracy": 0.8115107913669065,
-      "eval_loss": 0.7086611986160278,
-      "eval_runtime": 12.3594,
-      "eval_samples_per_second": 56.232,
-      "eval_steps_per_second": 7.039,
+      "eval_accuracy": 0.7884892086330936,
+      "eval_loss": 0.7312321066856384,
+      "eval_runtime": 12.0325,
+      "eval_samples_per_second": 57.76,
+      "eval_steps_per_second": 7.23,
       "step": 800
     },
     {
       "epoch": 6.61,
-      "learning_rate": 6.697580645161292e-05,
-      "loss": 0.6067,
+      "learning_rate": 7.801075268817205e-05,
+      "loss": 0.6156,
       "step": 820
     },
     {
       "epoch": 6.77,
-      "learning_rate": 6.616935483870968e-05,
-      "loss": 0.562,
+      "learning_rate": 7.74731182795699e-05,
+      "loss": 0.5887,
       "step": 840
     },
     {
       "epoch": 6.94,
-      "learning_rate": 6.536290322580646e-05,
-      "loss": 0.5473,
+      "learning_rate": 7.693548387096776e-05,
+      "loss": 0.549,
       "step": 860
     },
     {
       "epoch": 7.1,
-      "learning_rate": 6.455645161290323e-05,
-      "loss": 0.6343,
+      "learning_rate": 7.63978494623656e-05,
+      "loss": 0.581,
       "step": 880
     },
     {
       "epoch": 7.26,
-      "learning_rate": 6.375e-05,
-      "loss": 0.5714,
+      "learning_rate": 7.586021505376343e-05,
+      "loss": 0.5795,
       "step": 900
     },
     {
       "epoch": 7.26,
-      "eval_accuracy": 0.8158273381294964,
-      "eval_loss": 0.6596143245697021,
-      "eval_runtime": 12.6005,
-      "eval_samples_per_second": 55.156,
-      "eval_steps_per_second": 6.904,
+      "eval_accuracy": 0.8100719424460432,
+      "eval_loss": 0.6988666653633118,
+      "eval_runtime": 12.1203,
+      "eval_samples_per_second": 57.342,
+      "eval_steps_per_second": 7.178,
       "step": 900
     },
     {
       "epoch": 7.42,
-      "learning_rate": 6.294354838709677e-05,
-      "loss": 0.4871,
+      "learning_rate": 7.532258064516129e-05,
+      "loss": 0.4974,
       "step": 920
     },
     {
       "epoch": 7.58,
-      "learning_rate": 6.213709677419355e-05,
-      "loss": 0.5548,
+      "learning_rate": 7.478494623655914e-05,
+      "loss": 0.5397,
       "step": 940
     },
     {
       "epoch": 7.74,
-      "learning_rate": 6.133064516129033e-05,
-      "loss": 0.6063,
+      "learning_rate": 7.424731182795699e-05,
+      "loss": 0.6128,
       "step": 960
     },
     {
       "epoch": 7.9,
-      "learning_rate": 6.05241935483871e-05,
-      "loss": 0.5082,
+      "learning_rate": 7.370967741935485e-05,
+      "loss": 0.5246,
       "step": 980
     },
     {
       "epoch": 8.06,
-      "learning_rate": 5.971774193548387e-05,
-      "loss": 0.4802,
+      "learning_rate": 7.317204301075268e-05,
+      "loss": 0.4914,
       "step": 1000
     },
     {
       "epoch": 8.06,
       "eval_accuracy": 0.781294964028777,
-      "eval_loss": 0.7486134767532349,
-      "eval_runtime": 12.6896,
-      "eval_samples_per_second": 54.769,
-      "eval_steps_per_second": 6.856,
+      "eval_loss": 0.7066284418106079,
+      "eval_runtime": 12.169,
+      "eval_samples_per_second": 57.112,
+      "eval_steps_per_second": 7.149,
       "step": 1000
     },
     {
       "epoch": 8.23,
-      "learning_rate": 5.8911290322580645e-05,
-      "loss": 0.4552,
+      "learning_rate": 7.263440860215054e-05,
+      "loss": 0.4653,
       "step": 1020
     },
     {
       "epoch": 8.39,
-      "learning_rate": 5.8104838709677424e-05,
-      "loss": 0.5083,
+      "learning_rate": 7.209677419354839e-05,
+      "loss": 0.486,
       "step": 1040
     },
     {
       "epoch": 8.55,
-      "learning_rate": 5.7298387096774197e-05,
-      "loss": 0.4695,
+      "learning_rate": 7.155913978494624e-05,
+      "loss": 0.4781,
       "step": 1060
     },
     {
       "epoch": 8.71,
-      "learning_rate": 5.6491935483870976e-05,
-      "loss": 0.5316,
+      "learning_rate": 7.102150537634408e-05,
+      "loss": 0.5277,
       "step": 1080
     },
     {
       "epoch": 8.87,
-      "learning_rate": 5.568548387096775e-05,
-      "loss": 0.5135,
+      "learning_rate": 7.048387096774193e-05,
+      "loss": 0.5087,
       "step": 1100
     },
     {
       "epoch": 8.87,
-      "eval_accuracy": 0.8258992805755395,
-      "eval_loss": 0.6208385229110718,
-      "eval_runtime": 12.6127,
-      "eval_samples_per_second": 55.103,
-      "eval_steps_per_second": 6.898,
+      "eval_accuracy": 0.818705035971223,
+      "eval_loss": 0.6398155689239502,
+      "eval_runtime": 12.2159,
+      "eval_samples_per_second": 56.893,
+      "eval_steps_per_second": 7.122,
       "step": 1100
     },
     {
       "epoch": 9.03,
-      "learning_rate": 5.491935483870968e-05,
-      "loss": 0.4198,
+      "learning_rate": 6.994623655913979e-05,
+      "loss": 0.4671,
       "step": 1120
     },
     {
       "epoch": 9.19,
-      "learning_rate": 5.411290322580646e-05,
-      "loss": 0.4311,
+      "learning_rate": 6.940860215053764e-05,
+      "loss": 0.4841,
       "step": 1140
     },
     {
       "epoch": 9.35,
-      "learning_rate": 5.330645161290323e-05,
-      "loss": 0.467,
+      "learning_rate": 6.887096774193549e-05,
+      "loss": 0.4619,
       "step": 1160
     },
     {
       "epoch": 9.52,
-      "learning_rate": 5.25e-05,
-      "loss": 0.4832,
+      "learning_rate": 6.833333333333333e-05,
+      "loss": 0.479,
       "step": 1180
     },
     {
       "epoch": 9.68,
-      "learning_rate": 5.169354838709678e-05,
-      "loss": 0.4724,
+      "learning_rate": 6.779569892473118e-05,
+      "loss": 0.4373,
       "step": 1200
     },
     {
       "epoch": 9.68,
-      "eval_accuracy": 0.8129496402877698,
-      "eval_loss": 0.6434023976325989,
-      "eval_runtime": 12.4712,
-      "eval_samples_per_second": 55.729,
-      "eval_steps_per_second": 6.976,
+      "eval_accuracy": 0.8043165467625899,
+      "eval_loss": 0.6293413043022156,
+      "eval_runtime": 12.0132,
+      "eval_samples_per_second": 57.853,
+      "eval_steps_per_second": 7.242,
       "step": 1200
     },
     {
       "epoch": 9.84,
-      "learning_rate": 5.0887096774193554e-05,
-      "loss": 0.4715,
+      "learning_rate": 6.725806451612904e-05,
+      "loss": 0.4989,
       "step": 1220
     },
     {
       "epoch": 10.0,
-      "learning_rate": 5.008064516129033e-05,
-      "loss": 0.4828,
+      "learning_rate": 6.672043010752689e-05,
+      "loss": 0.4829,
       "step": 1240
     },
     {
       "epoch": 10.16,
-      "learning_rate": 4.92741935483871e-05,
-      "loss": 0.4171,
+      "learning_rate": 6.618279569892474e-05,
+      "loss": 0.458,
       "step": 1260
     },
     {
       "epoch": 10.32,
-      "learning_rate": 4.846774193548387e-05,
-      "loss": 0.4482,
+      "learning_rate": 6.564516129032258e-05,
+      "loss": 0.4135,
       "step": 1280
     },
     {
       "epoch": 10.48,
-      "learning_rate": 4.766129032258065e-05,
-      "loss": 0.3781,
+      "learning_rate": 6.510752688172043e-05,
+      "loss": 0.4365,
       "step": 1300
     },
     {
       "epoch": 10.48,
-      "eval_accuracy": 0.818705035971223,
-      "eval_loss": 0.6289911866188049,
-      "eval_runtime": 12.5799,
-      "eval_samples_per_second": 55.247,
-      "eval_steps_per_second": 6.916,
+      "eval_accuracy": 0.7971223021582734,
+      "eval_loss": 0.672641932964325,
+      "eval_runtime": 12.1208,
+      "eval_samples_per_second": 57.339,
+      "eval_steps_per_second": 7.178,
       "step": 1300
     },
     {
       "epoch": 10.65,
-      "learning_rate": 4.685483870967742e-05,
-      "loss": 0.4689,
+      "learning_rate": 6.456989247311829e-05,
+      "loss": 0.4852,
       "step": 1320
     },
     {
       "epoch": 10.81,
-      "learning_rate": 4.60483870967742e-05,
-      "loss": 0.379,
+      "learning_rate": 6.403225806451614e-05,
+      "loss": 0.3868,
       "step": 1340
     },
     {
       "epoch": 10.97,
-      "learning_rate": 4.5241935483870966e-05,
-      "loss": 0.4008,
+      "learning_rate": 6.349462365591398e-05,
+      "loss": 0.4663,
       "step": 1360
     },
     {
       "epoch": 11.13,
-      "learning_rate": 4.4435483870967745e-05,
-      "loss": 0.3681,
+      "learning_rate": 6.295698924731183e-05,
+      "loss": 0.396,
       "step": 1380
     },
     {
       "epoch": 11.29,
-      "learning_rate": 4.362903225806452e-05,
-      "loss": 0.457,
+      "learning_rate": 6.241935483870967e-05,
+      "loss": 0.4517,
       "step": 1400
     },
     {
       "epoch": 11.29,
-      "eval_accuracy": 0.8345323741007195,
-      "eval_loss": 0.6152310967445374,
-      "eval_runtime": 12.4473,
-      "eval_samples_per_second": 55.836,
-      "eval_steps_per_second": 6.989,
+      "eval_accuracy": 0.8244604316546763,
+      "eval_loss": 0.6046626567840576,
+      "eval_runtime": 12.0781,
+      "eval_samples_per_second": 57.542,
+      "eval_steps_per_second": 7.203,
       "step": 1400
     },
     {
       "epoch": 11.45,
-      "learning_rate": 4.282258064516129e-05,
-      "loss": 0.4215,
+      "learning_rate": 6.188172043010752e-05,
+      "loss": 0.4272,
       "step": 1420
     },
     {
       "epoch": 11.61,
-      "learning_rate": 4.201612903225807e-05,
-      "loss": 0.3423,
+      "learning_rate": 6.134408602150538e-05,
+      "loss": 0.3438,
       "step": 1440
     },
     {
       "epoch": 11.77,
-      "learning_rate": 4.120967741935484e-05,
-      "loss": 0.3994,
+      "learning_rate": 6.080645161290322e-05,
+      "loss": 0.3741,
       "step": 1460
     },
     {
       "epoch": 11.94,
-      "learning_rate": 4.0403225806451614e-05,
-      "loss": 0.3745,
+      "learning_rate": 6.0268817204301075e-05,
+      "loss": 0.3633,
       "step": 1480
     },
     {
       "epoch": 12.1,
-      "learning_rate": 3.959677419354839e-05,
-      "loss": 0.4174,
+      "learning_rate": 5.973118279569893e-05,
+      "loss": 0.4114,
       "step": 1500
     },
     {
       "epoch": 12.1,
-      "eval_accuracy": 0.814388489208633,
-      "eval_loss": 0.6348451375961304,
-      "eval_runtime": 12.3076,
-      "eval_samples_per_second": 56.469,
-      "eval_steps_per_second": 7.069,
+      "eval_accuracy": 0.823021582733813,
+      "eval_loss": 0.6088372468948364,
+      "eval_runtime": 12.1467,
+      "eval_samples_per_second": 57.217,
+      "eval_steps_per_second": 7.162,
       "step": 1500
     },
     {
       "epoch": 12.26,
-      "learning_rate": 3.879032258064516e-05,
-      "loss": 0.3009,
+      "learning_rate": 5.9193548387096774e-05,
+      "loss": 0.3284,
       "step": 1520
     },
     {
       "epoch": 12.42,
-      "learning_rate": 3.798387096774194e-05,
-      "loss": 0.3015,
+      "learning_rate": 5.8655913978494627e-05,
+      "loss": 0.3329,
       "step": 1540
     },
     {
       "epoch": 12.58,
-      "learning_rate": 3.717741935483871e-05,
-      "loss": 0.3543,
+      "learning_rate": 5.811827956989247e-05,
+      "loss": 0.3656,
       "step": 1560
     },
     {
       "epoch": 12.74,
-      "learning_rate": 3.637096774193549e-05,
-      "loss": 0.3493,
+      "learning_rate": 5.7580645161290325e-05,
+      "loss": 0.3581,
       "step": 1580
     },
     {
       "epoch": 12.9,
-      "learning_rate": 3.556451612903226e-05,
-      "loss": 0.3778,
+      "learning_rate": 5.704301075268818e-05,
+      "loss": 0.426,
       "step": 1600
     },
     {
       "epoch": 12.9,
-      "eval_accuracy": 0.8158273381294964,
-      "eval_loss": 0.6418473720550537,
-      "eval_runtime": 12.4779,
-      "eval_samples_per_second": 55.699,
-      "eval_steps_per_second": 6.972,
+      "eval_accuracy": 0.8201438848920863,
+      "eval_loss": 0.6164522171020508,
+      "eval_runtime": 12.2797,
+      "eval_samples_per_second": 56.597,
+      "eval_steps_per_second": 7.085,
       "step": 1600
     },
     {
       "epoch": 13.06,
-      "learning_rate": 3.475806451612903e-05,
-      "loss": 0.3284,
+      "learning_rate": 5.6505376344086024e-05,
+      "loss": 0.4441,
       "step": 1620
     },
     {
       "epoch": 13.23,
-      "learning_rate": 3.395161290322581e-05,
-      "loss": 0.3104,
+      "learning_rate": 5.599462365591398e-05,
+      "loss": 0.3141,
       "step": 1640
     },
     {
       "epoch": 13.39,
-      "learning_rate": 3.3145161290322585e-05,
-      "loss": 0.3097,
+      "learning_rate": 5.545698924731183e-05,
+      "loss": 0.3072,
       "step": 1660
     },
     {
       "epoch": 13.55,
-      "learning_rate": 3.233870967741936e-05,
-      "loss": 0.3491,
+      "learning_rate": 5.491935483870968e-05,
+      "loss": 0.3348,
       "step": 1680
     },
     {
       "epoch": 13.71,
-      "learning_rate": 3.153225806451613e-05,
-      "loss": 0.3391,
+      "learning_rate": 5.438172043010753e-05,
+      "loss": 0.3456,
       "step": 1700
     },
     {
       "epoch": 13.71,
-      "eval_accuracy": 0.8172661870503597,
-      "eval_loss": 0.6456794738769531,
-      "eval_runtime": 12.5081,
-      "eval_samples_per_second": 55.564,
-      "eval_steps_per_second": 6.955,
+      "eval_accuracy": 0.8258992805755395,
+      "eval_loss": 0.6133091449737549,
+      "eval_runtime": 12.0446,
+      "eval_samples_per_second": 57.702,
+      "eval_steps_per_second": 7.223,
       "step": 1700
     },
     {
       "epoch": 13.87,
-      "learning_rate": 3.076612903225807e-05,
-      "loss": 0.3287,
+      "learning_rate": 5.384408602150538e-05,
+      "loss": 0.3338,
       "step": 1720
     },
     {
       "epoch": 14.03,
-      "learning_rate": 2.9959677419354842e-05,
-      "loss": 0.3154,
+      "learning_rate": 5.330645161290323e-05,
+      "loss": 0.3473,
       "step": 1740
     },
     {
       "epoch": 14.19,
-      "learning_rate": 2.915322580645161e-05,
-      "loss": 0.3297,
+      "learning_rate": 5.276881720430108e-05,
+      "loss": 0.3563,
       "step": 1760
     },
     {
       "epoch": 14.35,
-      "learning_rate": 2.8346774193548387e-05,
-      "loss": 0.2665,
+      "learning_rate": 5.223118279569893e-05,
+      "loss": 0.3086,
       "step": 1780
     },
     {
       "epoch": 14.52,
-      "learning_rate": 2.7540322580645163e-05,
-      "loss": 0.309,
+      "learning_rate": 5.169354838709678e-05,
+      "loss": 0.332,
       "step": 1800
     },
     {
       "epoch": 14.52,
-      "eval_accuracy": 0.8100719424460432,
-      "eval_loss": 0.6428855657577515,
-      "eval_runtime": 12.5753,
-      "eval_samples_per_second": 55.267,
-      "eval_steps_per_second": 6.918,
+      "eval_accuracy": 0.8201438848920863,
+      "eval_loss": 0.6735997200012207,
+      "eval_runtime": 11.991,
+      "eval_samples_per_second": 57.96,
+      "eval_steps_per_second": 7.255,
       "step": 1800
     },
     {
       "epoch": 14.68,
-      "learning_rate": 2.6733870967741935e-05,
-      "loss": 0.3145,
+      "learning_rate": 5.115591397849463e-05,
+      "loss": 0.3146,
       "step": 1820
     },
     {
       "epoch": 14.84,
-      "learning_rate": 2.592741935483871e-05,
-      "loss": 0.3275,
+      "learning_rate": 5.061827956989248e-05,
+      "loss": 0.3475,
       "step": 1840
     },
     {
       "epoch": 15.0,
-      "learning_rate": 2.5120967741935486e-05,
-      "loss": 0.3119,
+      "learning_rate": 5.008064516129033e-05,
+      "loss": 0.3397,
       "step": 1860
     },
     {
       "epoch": 15.16,
-      "learning_rate": 2.431451612903226e-05,
-      "loss": 0.2758,
+      "learning_rate": 4.954301075268817e-05,
+      "loss": 0.2829,
       "step": 1880
     },
     {
       "epoch": 15.32,
-      "learning_rate": 2.3508064516129034e-05,
-      "loss": 0.3543,
+      "learning_rate": 4.9005376344086024e-05,
+      "loss": 0.3646,
       "step": 1900
     },
     {
       "epoch": 15.32,
-      "eval_accuracy": 0.8201438848920863,
-      "eval_loss": 0.6085776090621948,
-      "eval_runtime": 12.4853,
-      "eval_samples_per_second": 55.665,
-      "eval_steps_per_second": 6.968,
+      "eval_accuracy": 0.8172661870503597,
+      "eval_loss": 0.6405801177024841,
+      "eval_runtime": 12.0646,
+      "eval_samples_per_second": 57.607,
+      "eval_steps_per_second": 7.211,
       "step": 1900
     },
     {
       "epoch": 15.48,
-      "learning_rate": 2.2701612903225807e-05,
-      "loss": 0.2298,
+      "learning_rate": 4.846774193548387e-05,
+      "loss": 0.2571,
       "step": 1920
     },
     {
       "epoch": 15.65,
-      "learning_rate": 2.1895161290322582e-05,
-      "loss": 0.3242,
+      "learning_rate": 4.793010752688172e-05,
+      "loss": 0.3495,
       "step": 1940
     },
     {
       "epoch": 15.81,
-      "learning_rate": 2.1088709677419355e-05,
-      "loss": 0.2483,
+      "learning_rate": 4.7392473118279576e-05,
+      "loss": 0.2647,
       "step": 1960
     },
     {
       "epoch": 15.97,
-      "learning_rate": 2.028225806451613e-05,
-      "loss": 0.2685,
+      "learning_rate": 4.685483870967742e-05,
+      "loss": 0.3063,
       "step": 1980
     },
     {
       "epoch": 16.13,
-      "learning_rate": 1.9475806451612906e-05,
-      "loss": 0.3232,
+      "learning_rate": 4.6317204301075275e-05,
+      "loss": 0.3287,
       "step": 2000
     },
     {
       "epoch": 16.13,
-      "eval_accuracy": 0.8244604316546763,
-      "eval_loss": 0.6374786496162415,
-      "eval_runtime": 12.4055,
-      "eval_samples_per_second": 56.024,
-      "eval_steps_per_second": 7.013,
+      "eval_accuracy": 0.7971223021582734,
+      "eval_loss": 0.6978276968002319,
+      "eval_runtime": 12.1198,
+      "eval_samples_per_second": 57.344,
+      "eval_steps_per_second": 7.178,
       "step": 2000
     },
     {
       "epoch": 16.29,
-      "learning_rate": 1.8669354838709678e-05,
-      "loss": 0.2504,
+      "learning_rate": 4.577956989247312e-05,
+      "loss": 0.2456,
       "step": 2020
     },
     {
       "epoch": 16.45,
-      "learning_rate": 1.786290322580645e-05,
-      "loss": 0.3025,
+      "learning_rate": 4.5241935483870966e-05,
+      "loss": 0.3242,
       "step": 2040
     },
     {
       "epoch": 16.61,
-      "learning_rate": 1.7056451612903226e-05,
-      "loss": 0.3046,
+      "learning_rate": 4.470430107526882e-05,
+      "loss": 0.3186,
       "step": 2060
     },
     {
       "epoch": 16.77,
-      "learning_rate": 1.6250000000000002e-05,
-      "loss": 0.3086,
+      "learning_rate": 4.4166666666666665e-05,
+      "loss": 0.3462,
       "step": 2080
     },
     {
       "epoch": 16.94,
-      "learning_rate": 1.5443548387096778e-05,
-      "loss": 0.2687,
+      "learning_rate": 4.362903225806452e-05,
+      "loss": 0.2793,
       "step": 2100
     },
     {
       "epoch": 16.94,
-      "eval_accuracy": 0.8273381294964028,
-      "eval_loss": 0.6281846165657043,
-      "eval_runtime": 12.5909,
-      "eval_samples_per_second": 55.199,
-      "eval_steps_per_second": 6.91,
+      "eval_accuracy": 0.8172661870503597,
+      "eval_loss": 0.6432561278343201,
+      "eval_runtime": 12.0081,
+      "eval_samples_per_second": 57.878,
+      "eval_steps_per_second": 7.245,
       "step": 2100
     },
     {
       "epoch": 17.1,
-      "learning_rate": 1.4637096774193548e-05,
-      "loss": 0.3014,
+      "learning_rate": 4.309139784946237e-05,
+      "loss": 0.2849,
       "step": 2120
     },
     {
       "epoch": 17.26,
-      "learning_rate": 1.3830645161290324e-05,
-      "loss": 0.2911,
+      "learning_rate": 4.2553763440860216e-05,
+      "loss": 0.3371,
       "step": 2140
     },
     {
       "epoch": 17.42,
-      "learning_rate": 1.3024193548387098e-05,
-      "loss": 0.262,
+      "learning_rate": 4.201612903225807e-05,
+      "loss": 0.2561,
       "step": 2160
     },
     {
       "epoch": 17.58,
-      "learning_rate": 1.2217741935483872e-05,
-      "loss": 0.2739,
+      "learning_rate": 4.1478494623655915e-05,
+      "loss": 0.2628,
       "step": 2180
     },
     {
       "epoch": 17.74,
-      "learning_rate": 1.1411290322580646e-05,
-      "loss": 0.2551,
+      "learning_rate": 4.094086021505377e-05,
+      "loss": 0.2924,
       "step": 2200
     },
     {
       "epoch": 17.74,
-      "eval_accuracy": 0.8215827338129497,
-      "eval_loss": 0.6021825075149536,
-      "eval_runtime": 12.4226,
-      "eval_samples_per_second": 55.946,
-      "eval_steps_per_second": 7.003,
+      "eval_accuracy": 0.814388489208633,
+      "eval_loss": 0.6474089622497559,
+      "eval_runtime": 12.255,
+      "eval_samples_per_second": 56.711,
+      "eval_steps_per_second": 7.099,
       "step": 2200
     },
     {
       "epoch": 17.9,
-      "learning_rate": 1.060483870967742e-05,
-      "loss": 0.3001,
+      "learning_rate": 4.0403225806451614e-05,
+      "loss": 0.2935,
       "step": 2220
     },
     {
       "epoch": 18.06,
-      "learning_rate": 9.798387096774194e-06,
-      "loss": 0.5205,
+      "learning_rate": 3.9865591397849466e-05,
+      "loss": 0.5124,
       "step": 2240
     },
     {
       "epoch": 18.23,
-      "learning_rate": 8.991935483870968e-06,
-      "loss": 0.285,
+      "learning_rate": 3.932795698924731e-05,
+      "loss": 0.2954,
       "step": 2260
     },
     {
       "epoch": 18.39,
-      "learning_rate": 8.185483870967742e-06,
-      "loss": 0.2684,
+      "learning_rate": 3.879032258064516e-05,
+      "loss": 0.2584,
       "step": 2280
     },
     {
       "epoch": 18.55,
-      "learning_rate": 7.379032258064517e-06,
-      "loss": 0.2711,
+      "learning_rate": 3.825268817204301e-05,
+      "loss": 0.2605,
       "step": 2300
     },
     {
       "epoch": 18.55,
-      "eval_accuracy": 0.818705035971223,
-      "eval_loss": 0.6055982708930969,
-      "eval_runtime": 12.4967,
-      "eval_samples_per_second": 55.615,
-      "eval_steps_per_second": 6.962,
+      "eval_accuracy": 0.8287769784172662,
+      "eval_loss": 0.627909243106842,
+      "eval_runtime": 12.1775,
+      "eval_samples_per_second": 57.072,
+      "eval_steps_per_second": 7.144,
       "step": 2300
     },
     {
       "epoch": 18.71,
-      "learning_rate": 6.57258064516129e-06,
-      "loss": 0.2776,
+      "learning_rate": 3.7715053763440864e-05,
+      "loss": 0.312,
       "step": 2320
     },
     {
       "epoch": 18.87,
-      "learning_rate": 5.766129032258065e-06,
-      "loss": 0.2575,
+      "learning_rate": 3.717741935483871e-05,
+      "loss": 0.2679,
       "step": 2340
     },
     {
       "epoch": 19.03,
-      "learning_rate": 4.9596774193548395e-06,
-      "loss": 0.2816,
+      "learning_rate": 3.663978494623656e-05,
+      "loss": 0.2654,
       "step": 2360
     },
     {
       "epoch": 19.19,
-      "learning_rate": 4.1532258064516135e-06,
-      "loss": 0.2369,
+      "learning_rate": 3.610215053763441e-05,
+      "loss": 0.2524,
       "step": 2380
     },
     {
       "epoch": 19.35,
-      "learning_rate": 3.3467741935483875e-06,
-      "loss": 0.2371,
+      "learning_rate": 3.556451612903226e-05,
+      "loss": 0.2016,
       "step": 2400
     },
     {
       "epoch": 19.35,
-      "eval_accuracy": 0.823021582733813,
-      "eval_loss": 0.6057971715927124,
-      "eval_runtime": 12.6132,
-      "eval_samples_per_second": 55.101,
-      "eval_steps_per_second": 6.898,
+      "eval_accuracy": 0.8215827338129497,
+      "eval_loss": 0.6360746026039124,
+      "eval_runtime": 12.0929,
+      "eval_samples_per_second": 57.472,
+      "eval_steps_per_second": 7.194,
       "step": 2400
     },
     {
       "epoch": 19.52,
-      "learning_rate": 2.5403225806451615e-06,
-      "loss": 0.2299,
+      "learning_rate": 3.502688172043011e-05,
+      "loss": 0.2691,
       "step": 2420
     },
     {
       "epoch": 19.68,
-      "learning_rate": 1.7338709677419355e-06,
-      "loss": 0.203,
+      "learning_rate": 3.448924731182796e-05,
+      "loss": 0.2068,
       "step": 2440
     },
     {
       "epoch": 19.84,
-      "learning_rate": 9.274193548387097e-07,
-      "loss": 0.301,
+      "learning_rate": 3.395161290322581e-05,
+      "loss": 0.3017,
       "step": 2460
     },
     {
       "epoch": 20.0,
-      "learning_rate": 1.2096774193548387e-07,
-      "loss": 0.2444,
+      "learning_rate": 3.341397849462366e-05,
+      "loss": 0.2318,
       "step": 2480
     },
     {
-      "epoch": 20.0,
-      "step": 2480,
-      "total_flos": 6.103469271205233e+18,
-      "train_loss": 0.6948060337574251,
-      "train_runtime": 2796.6359,
-      "train_samples_per_second": 28.155,
-      "train_steps_per_second": 0.887
+      "epoch": 20.16,
+      "learning_rate": 3.2876344086021504e-05,
+      "loss": 0.2524,
+      "step": 2500
+    },
+    {
+      "epoch": 20.16,
+      "eval_accuracy": 0.8258992805755395,
+      "eval_loss": 0.6393768191337585,
+      "eval_runtime": 12.2058,
+      "eval_samples_per_second": 56.94,
+      "eval_steps_per_second": 7.128,
+      "step": 2500
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 3.233870967741936e-05,
+      "loss": 0.2603,
+      "step": 2520
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 3.18010752688172e-05,
+      "loss": 0.207,
+      "step": 2540
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 3.1263440860215056e-05,
+      "loss": 0.1927,
+      "step": 2560
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 3.07258064516129e-05,
+      "loss": 0.1947,
+      "step": 2580
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 3.0188172043010754e-05,
+      "loss": 0.2017,
+      "step": 2600
+    },
+    {
+      "epoch": 20.97,
+      "eval_accuracy": 0.8158273381294964,
+      "eval_loss": 0.6682714223861694,
+      "eval_runtime": 12.0188,
+      "eval_samples_per_second": 57.826,
+      "eval_steps_per_second": 7.239,
+      "step": 2600
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 2.9650537634408604e-05,
+      "loss": 0.2351,
+      "step": 2620
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 2.9112903225806453e-05,
+      "loss": 0.1842,
+      "step": 2640
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 2.8575268817204302e-05,
+      "loss": 0.2131,
+      "step": 2660
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 2.8037634408602155e-05,
+      "loss": 0.2509,
+      "step": 2680
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 2.7500000000000004e-05,
+      "loss": 0.2082,
+      "step": 2700
+    },
+    {
+      "epoch": 21.77,
+      "eval_accuracy": 0.8345323741007195,
+      "eval_loss": 0.6389289498329163,
+      "eval_runtime": 12.0715,
+      "eval_samples_per_second": 57.574,
+      "eval_steps_per_second": 7.207,
+      "step": 2700
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 2.6962365591397854e-05,
+      "loss": 0.2671,
+      "step": 2720
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 2.6424731182795696e-05,
+      "loss": 0.2534,
+      "step": 2740
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 2.588709677419355e-05,
+      "loss": 0.205,
+      "step": 2760
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 2.5349462365591398e-05,
+      "loss": 0.1904,
+      "step": 2780
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 2.4811827956989248e-05,
+      "loss": 0.2751,
+      "step": 2800
+    },
+    {
+      "epoch": 22.58,
+      "eval_accuracy": 0.837410071942446,
+      "eval_loss": 0.6141177415847778,
+      "eval_runtime": 12.0253,
+      "eval_samples_per_second": 57.795,
+      "eval_steps_per_second": 7.235,
+      "step": 2800
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 2.4274193548387097e-05,
+      "loss": 0.1863,
+      "step": 2820
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 2.3736559139784946e-05,
+      "loss": 0.2385,
+      "step": 2840
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 2.31989247311828e-05,
+      "loss": 0.1666,
+      "step": 2860
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 2.266129032258065e-05,
+      "loss": 0.2087,
+      "step": 2880
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 2.2123655913978494e-05,
+      "loss": 0.207,
+      "step": 2900
+    },
+    {
+      "epoch": 23.39,
+      "eval_accuracy": 0.8258992805755395,
+      "eval_loss": 0.6051694750785828,
+      "eval_runtime": 12.011,
+      "eval_samples_per_second": 57.864,
+      "eval_steps_per_second": 7.243,
+      "step": 2900
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 2.1586021505376344e-05,
+      "loss": 0.2026,
+      "step": 2920
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 2.1048387096774193e-05,
+      "loss": 0.2073,
+      "step": 2940
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 2.0510752688172046e-05,
+      "loss": 0.1829,
+      "step": 2960
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 1.9973118279569895e-05,
+      "loss": 0.2048,
+      "step": 2980
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 1.9435483870967744e-05,
+      "loss": 0.1791,
+      "step": 3000
+    },
+    {
+      "epoch": 24.19,
+      "eval_accuracy": 0.823021582733813,
+      "eval_loss": 0.6331909894943237,
+      "eval_runtime": 12.2518,
+      "eval_samples_per_second": 56.726,
+      "eval_steps_per_second": 7.101,
+      "step": 3000
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 1.889784946236559e-05,
+      "loss": 0.2243,
+      "step": 3020
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 1.836021505376344e-05,
+      "loss": 0.2148,
+      "step": 3040
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 1.7822580645161292e-05,
+      "loss": 0.2568,
+      "step": 3060
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 1.728494623655914e-05,
+      "loss": 0.1782,
+      "step": 3080
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 1.674731182795699e-05,
+      "loss": 0.1719,
+      "step": 3100
+    },
+    {
+      "epoch": 25.0,
+      "eval_accuracy": 0.8402877697841726,
+      "eval_loss": 0.5942012071609497,
+      "eval_runtime": 12.1124,
+      "eval_samples_per_second": 57.379,
+      "eval_steps_per_second": 7.183,
+      "step": 3100
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 1.620967741935484e-05,
+      "loss": 0.1881,
+      "step": 3120
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 1.5672043010752686e-05,
+      "loss": 0.2375,
+      "step": 3140
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 1.5134408602150537e-05,
+      "loss": 0.1779,
+      "step": 3160
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 1.4596774193548388e-05,
+      "loss": 0.1753,
+      "step": 3180
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 1.4059139784946238e-05,
+      "loss": 0.1685,
+      "step": 3200
+    },
+    {
+      "epoch": 25.81,
+      "eval_accuracy": 0.8359712230215828,
+      "eval_loss": 0.612082839012146,
+      "eval_runtime": 12.1399,
+      "eval_samples_per_second": 57.249,
+      "eval_steps_per_second": 7.166,
+      "step": 3200
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 1.3521505376344087e-05,
+      "loss": 0.1407,
+      "step": 3220
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 1.2983870967741938e-05,
+      "loss": 0.1854,
+      "step": 3240
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 1.2446236559139786e-05,
+      "loss": 0.1956,
+      "step": 3260
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 1.1908602150537635e-05,
+      "loss": 0.2,
+      "step": 3280
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 1.1370967741935484e-05,
+      "loss": 0.1557,
+      "step": 3300
+    },
+    {
+      "epoch": 26.61,
+      "eval_accuracy": 0.8345323741007195,
+      "eval_loss": 0.6236761212348938,
+      "eval_runtime": 11.9687,
+      "eval_samples_per_second": 58.068,
+      "eval_steps_per_second": 7.269,
+      "step": 3300
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 1.0833333333333334e-05,
+      "loss": 0.212,
+      "step": 3320
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 1.0295698924731183e-05,
+      "loss": 0.1893,
+      "step": 3340
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 9.758064516129032e-06,
+      "loss": 0.159,
+      "step": 3360
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 9.220430107526883e-06,
+      "loss": 0.1755,
+      "step": 3380
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 8.68279569892473e-06,
+      "loss": 0.1694,
+      "step": 3400
+    },
+    {
+      "epoch": 27.42,
+      "eval_accuracy": 0.8316546762589928,
+      "eval_loss": 0.6371967792510986,
+      "eval_runtime": 12.2075,
+      "eval_samples_per_second": 56.932,
+      "eval_steps_per_second": 7.127,
+      "step": 3400
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 8.145161290322582e-06,
+      "loss": 0.1552,
+      "step": 3420
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 7.607526881720431e-06,
+      "loss": 0.1848,
+      "step": 3440
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 7.06989247311828e-06,
+      "loss": 0.1467,
+      "step": 3460
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 6.532258064516129e-06,
+      "loss": 0.1292,
+      "step": 3480
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 5.994623655913978e-06,
+      "loss": 0.1927,
+      "step": 3500
+    },
+    {
+      "epoch": 28.23,
+      "eval_accuracy": 0.8273381294964028,
+      "eval_loss": 0.6377986073493958,
+      "eval_runtime": 12.0256,
+      "eval_samples_per_second": 57.793,
+      "eval_steps_per_second": 7.235,
+      "step": 3500
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 5.4569892473118285e-06,
+      "loss": 0.2078,
+      "step": 3520
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 4.919354838709678e-06,
+      "loss": 0.1992,
+      "step": 3540
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 4.381720430107527e-06,
+      "loss": 0.1773,
+      "step": 3560
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 3.8440860215053765e-06,
+      "loss": 0.1334,
+      "step": 3580
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 3.3064516129032262e-06,
+      "loss": 0.1375,
+      "step": 3600
+    },
+    {
+      "epoch": 29.03,
+      "eval_accuracy": 0.8330935251798561,
+      "eval_loss": 0.6257502436637878,
+      "eval_runtime": 12.1079,
+      "eval_samples_per_second": 57.4,
+      "eval_steps_per_second": 7.185,
+      "step": 3600
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 2.768817204301075e-06,
+      "loss": 0.1842,
+      "step": 3620
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 2.231182795698925e-06,
+      "loss": 0.169,
+      "step": 3640
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 1.6935483870967744e-06,
+      "loss": 0.1623,
+      "step": 3660
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 1.1559139784946237e-06,
+      "loss": 0.1658,
+      "step": 3680
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 6.182795698924732e-07,
+      "loss": 0.1653,
+      "step": 3700
+    },
+    {
+      "epoch": 29.84,
+      "eval_accuracy": 0.8330935251798561,
+      "eval_loss": 0.626188337802887,
+      "eval_runtime": 12.1517,
+      "eval_samples_per_second": 57.193,
+      "eval_steps_per_second": 7.159,
+      "step": 3700
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 8.064516129032259e-08,
+      "loss": 0.1634,
+      "step": 3720
+    },
+    {
+      "epoch": 30.0,
+      "step": 3720,
+      "total_flos": 9.155203906807849e+18,
+      "train_loss": 0.5316014153983003,
+      "train_runtime": 4018.8325,
+      "train_samples_per_second": 29.389,
+      "train_steps_per_second": 0.926
     }
   ],
   "logging_steps": 20,
-  "max_steps": 2480,
-  "num_train_epochs": 20,
+  "max_steps": 3720,
+  "num_train_epochs": 30,
   "save_steps": 100,
-  "total_flos": 6.103469271205233e+18,
+  "total_flos": 9.155203906807849e+18,
   "trial_name": null,
   "trial_params": null
 }