End of training

all_results.json

@@ -1,13 +1,13 @@
 {
-    "epoch": 18.823529411764707,
-    "eval_accuracy": 0.9116022099447514,
-    "eval_loss": 0.28710222244262695,
-    "eval_runtime": 1.5215,
-    "eval_samples_per_second": 118.965,
-    "eval_steps_per_second": 3.944,
-    "total_flos": 7.600391915087462e+17,
-    "train_loss": 0.3423821290334066,
-    "train_runtime": 426.6472,
-    "train_samples_per_second": 76.128,
-    "train_steps_per_second": 0.563
+    "epoch": 3.0,
+    "eval_accuracy": 0.9911111111111112,
+    "eval_loss": 0.02681696228682995,
+    "eval_runtime": 1.8717,
+    "eval_samples_per_second": 120.214,
+    "eval_steps_per_second": 4.274,
+    "total_flos": 1.5048656676458496e+17,
+    "train_loss": 0.5111757349222898,
+    "train_runtime": 85.6924,
+    "train_samples_per_second": 470.987,
+    "train_steps_per_second": 3.734
 }

eval_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 18.823529411764707,
-    "eval_accuracy": 0.9116022099447514,
-    "eval_loss": 0.28710222244262695,
-    "eval_runtime": 1.5215,
-    "eval_samples_per_second": 118.965,
-    "eval_steps_per_second": 3.944
+    "epoch": 3.0,
+    "eval_accuracy": 0.9911111111111112,
+    "eval_loss": 0.02681696228682995,
+    "eval_runtime": 1.8717,
+    "eval_samples_per_second": 120.214,
+    "eval_steps_per_second": 4.274
 }

train_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 18.823529411764707,
-    "total_flos": 7.600391915087462e+17,
-    "train_loss": 0.3423821290334066,
-    "train_runtime": 426.6472,
-    "train_samples_per_second": 76.128,
-    "train_steps_per_second": 0.563
+    "epoch": 3.0,
+    "total_flos": 1.5048656676458496e+17,
+    "train_loss": 0.5111757349222898,
+    "train_runtime": 85.6924,
+    "train_samples_per_second": 470.987,
+    "train_steps_per_second": 3.734
 }

trainer_state.json

@@ -1,364 +1,80 @@
 {
-  "best_metric": 0.9116022099447514,
-  "best_model_checkpoint": "/kaggle/working/swin-brain-modality-classification/checkpoint-51",
-  "epoch": 18.823529411764707,
+  "best_metric": 0.9911111111111112,
+  "best_model_checkpoint": "/kaggle/working/swin-brain-modality-classification/checkpoint-48",
+  "epoch": 3.0,
   "eval_steps": 500,
-  "global_step": 240,
+  "global_step": 48,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.7843137254901961,
-      "grad_norm": 5.9126081466674805,
-      "learning_rate": 2.0833333333333336e-05,
-      "loss": 1.3069,
+      "epoch": 0.625,
+      "grad_norm": 9.655741691589355,
+      "learning_rate": 1.5625e-05,
+      "loss": 1.364,
       "step": 10
     },
     {
-      "epoch": 0.9411764705882353,
-      "eval_accuracy": 0.6077348066298343,
-      "eval_loss": 0.9999544024467468,
-      "eval_runtime": 1.4322,
-      "eval_samples_per_second": 126.379,
-      "eval_steps_per_second": 4.189,
-      "step": 12
+      "epoch": 1.0,
+      "eval_accuracy": 0.8266666666666667,
+      "eval_loss": 0.6507545113563538,
+      "eval_runtime": 1.8483,
+      "eval_samples_per_second": 121.736,
+      "eval_steps_per_second": 4.328,
+      "step": 16
     },
     {
-      "epoch": 1.5686274509803921,
-      "grad_norm": 7.789252758026123,
-      "learning_rate": 4.166666666666667e-05,
-      "loss": 0.8924,
+      "epoch": 1.25,
+      "grad_norm": 6.807246685028076,
+      "learning_rate": 3.125e-05,
+      "loss": 0.7521,
       "step": 20
     },
     {
-      "epoch": 1.9607843137254903,
-      "eval_accuracy": 0.8784530386740331,
-      "eval_loss": 0.43337252736091614,
-      "eval_runtime": 1.4151,
-      "eval_samples_per_second": 127.902,
-      "eval_steps_per_second": 4.24,
-      "step": 25
-    },
-    {
-      "epoch": 2.3529411764705883,
-      "grad_norm": 6.7294697761535645,
-      "learning_rate": 4.8611111111111115e-05,
-      "loss": 0.5365,
+      "epoch": 1.875,
+      "grad_norm": 5.148420333862305,
+      "learning_rate": 4.6875e-05,
+      "loss": 0.2053,
       "step": 30
     },
     {
-      "epoch": 2.980392156862745,
-      "eval_accuracy": 0.9005524861878453,
-      "eval_loss": 0.3143160939216614,
-      "eval_runtime": 1.4092,
-      "eval_samples_per_second": 128.446,
-      "eval_steps_per_second": 4.258,
-      "step": 38
+      "epoch": 2.0,
+      "eval_accuracy": 0.9555555555555556,
+      "eval_loss": 0.08017182350158691,
+      "eval_runtime": 1.8504,
+      "eval_samples_per_second": 121.592,
+      "eval_steps_per_second": 4.323,
+      "step": 32
     },
     {
-      "epoch": 3.1372549019607843,
-      "grad_norm": 8.53775691986084,
-      "learning_rate": 4.62962962962963e-05,
-      "loss": 0.4119,
+      "epoch": 2.5,
+      "grad_norm": 9.427779197692871,
+      "learning_rate": 4.8611111111111115e-05,
+      "loss": 0.0841,
       "step": 40
     },
     {
-      "epoch": 3.9215686274509802,
-      "grad_norm": 8.774258613586426,
-      "learning_rate": 4.3981481481481486e-05,
-      "loss": 0.3814,
-      "step": 50
-    },
-    {
-      "epoch": 4.0,
-      "eval_accuracy": 0.9116022099447514,
-      "eval_loss": 0.28710222244262695,
-      "eval_runtime": 1.4257,
-      "eval_samples_per_second": 126.96,
-      "eval_steps_per_second": 4.209,
-      "step": 51
-    },
-    {
-      "epoch": 4.705882352941177,
-      "grad_norm": 5.285433769226074,
-      "learning_rate": 4.166666666666667e-05,
-      "loss": 0.3336,
-      "step": 60
-    },
-    {
-      "epoch": 4.9411764705882355,
-      "eval_accuracy": 0.9116022099447514,
-      "eval_loss": 0.2963091731071472,
-      "eval_runtime": 1.415,
-      "eval_samples_per_second": 127.916,
-      "eval_steps_per_second": 4.24,
-      "step": 63
-    },
-    {
-      "epoch": 5.490196078431373,
-      "grad_norm": 7.04965353012085,
-      "learning_rate": 3.935185185185186e-05,
-      "loss": 0.353,
-      "step": 70
-    },
-    {
-      "epoch": 5.96078431372549,
-      "eval_accuracy": 0.8729281767955801,
-      "eval_loss": 0.31954672932624817,
-      "eval_runtime": 1.4556,
-      "eval_samples_per_second": 124.344,
-      "eval_steps_per_second": 4.122,
-      "step": 76
-    },
-    {
-      "epoch": 6.2745098039215685,
-      "grad_norm": 5.833162307739258,
-      "learning_rate": 3.7037037037037037e-05,
-      "loss": 0.3069,
-      "step": 80
-    },
-    {
-      "epoch": 6.980392156862745,
-      "eval_accuracy": 0.9116022099447514,
-      "eval_loss": 0.29521241784095764,
-      "eval_runtime": 1.4158,
-      "eval_samples_per_second": 127.846,
-      "eval_steps_per_second": 4.238,
-      "step": 89
-    },
-    {
-      "epoch": 7.0588235294117645,
-      "grad_norm": 5.050061225891113,
-      "learning_rate": 3.472222222222222e-05,
-      "loss": 0.2789,
-      "step": 90
-    },
-    {
-      "epoch": 7.8431372549019605,
-      "grad_norm": 4.222379207611084,
-      "learning_rate": 3.240740740740741e-05,
-      "loss": 0.293,
-      "step": 100
-    },
-    {
-      "epoch": 8.0,
-      "eval_accuracy": 0.8895027624309392,
-      "eval_loss": 0.3174145817756653,
-      "eval_runtime": 1.4186,
-      "eval_samples_per_second": 127.591,
-      "eval_steps_per_second": 4.23,
-      "step": 102
-    },
-    {
-      "epoch": 8.627450980392156,
-      "grad_norm": 7.039156436920166,
-      "learning_rate": 3.0092592592592593e-05,
-      "loss": 0.2667,
-      "step": 110
-    },
-    {
-      "epoch": 8.941176470588236,
-      "eval_accuracy": 0.8950276243093923,
-      "eval_loss": 0.3225868344306946,
-      "eval_runtime": 1.4137,
-      "eval_samples_per_second": 128.03,
-      "eval_steps_per_second": 4.244,
-      "step": 114
-    },
-    {
-      "epoch": 9.411764705882353,
-      "grad_norm": 5.598822593688965,
-      "learning_rate": 2.777777777777778e-05,
-      "loss": 0.2424,
-      "step": 120
-    },
-    {
-      "epoch": 9.96078431372549,
-      "eval_accuracy": 0.8895027624309392,
-      "eval_loss": 0.3213161826133728,
-      "eval_runtime": 1.4313,
-      "eval_samples_per_second": 126.455,
-      "eval_steps_per_second": 4.192,
-      "step": 127
-    },
-    {
-      "epoch": 10.196078431372548,
-      "grad_norm": 4.785697937011719,
-      "learning_rate": 2.5462962962962965e-05,
-      "loss": 0.2544,
-      "step": 130
-    },
-    {
-      "epoch": 10.980392156862745,
-      "grad_norm": 5.349719047546387,
-      "learning_rate": 2.314814814814815e-05,
-      "loss": 0.2605,
-      "step": 140
-    },
-    {
-      "epoch": 10.980392156862745,
-      "eval_accuracy": 0.8895027624309392,
-      "eval_loss": 0.31716108322143555,
-      "eval_runtime": 1.4269,
-      "eval_samples_per_second": 126.846,
-      "eval_steps_per_second": 4.205,
-      "step": 140
-    },
-    {
-      "epoch": 11.764705882352942,
-      "grad_norm": 6.121713161468506,
-      "learning_rate": 2.0833333333333336e-05,
-      "loss": 0.232,
-      "step": 150
-    },
-    {
-      "epoch": 12.0,
-      "eval_accuracy": 0.8895027624309392,
-      "eval_loss": 0.33846884965896606,
-      "eval_runtime": 1.406,
-      "eval_samples_per_second": 128.737,
-      "eval_steps_per_second": 4.268,
-      "step": 153
-    },
-    {
-      "epoch": 12.549019607843137,
-      "grad_norm": 7.647618770599365,
-      "learning_rate": 1.8518518518518518e-05,
-      "loss": 0.242,
-      "step": 160
-    },
-    {
-      "epoch": 12.941176470588236,
-      "eval_accuracy": 0.8950276243093923,
-      "eval_loss": 0.32744264602661133,
-      "eval_runtime": 1.4273,
-      "eval_samples_per_second": 126.813,
-      "eval_steps_per_second": 4.204,
-      "step": 165
-    },
-    {
-      "epoch": 13.333333333333334,
-      "grad_norm": 6.248785972595215,
-      "learning_rate": 1.6203703703703704e-05,
-      "loss": 0.215,
-      "step": 170
-    },
-    {
-      "epoch": 13.96078431372549,
-      "eval_accuracy": 0.8950276243093923,
-      "eval_loss": 0.33850720524787903,
-      "eval_runtime": 1.4433,
-      "eval_samples_per_second": 125.407,
-      "eval_steps_per_second": 4.157,
-      "step": 178
-    },
-    {
-      "epoch": 14.117647058823529,
-      "grad_norm": 5.345800876617432,
-      "learning_rate": 1.388888888888889e-05,
-      "loss": 0.2123,
-      "step": 180
-    },
-    {
-      "epoch": 14.901960784313726,
-      "grad_norm": 5.421293258666992,
-      "learning_rate": 1.1574074074074075e-05,
-      "loss": 0.2131,
-      "step": 190
-    },
-    {
-      "epoch": 14.980392156862745,
-      "eval_accuracy": 0.8950276243093923,
-      "eval_loss": 0.34223416447639465,
-      "eval_runtime": 1.409,
-      "eval_samples_per_second": 128.462,
-      "eval_steps_per_second": 4.258,
-      "step": 191
-    },
-    {
-      "epoch": 15.686274509803921,
-      "grad_norm": 4.188720703125,
-      "learning_rate": 9.259259259259259e-06,
-      "loss": 0.201,
-      "step": 200
-    },
-    {
-      "epoch": 16.0,
-      "eval_accuracy": 0.8784530386740331,
-      "eval_loss": 0.341948539018631,
-      "eval_runtime": 1.4191,
-      "eval_samples_per_second": 127.549,
-      "eval_steps_per_second": 4.228,
-      "step": 204
-    },
-    {
-      "epoch": 16.470588235294116,
-      "grad_norm": 4.887516498565674,
-      "learning_rate": 6.944444444444445e-06,
-      "loss": 0.1976,
-      "step": 210
-    },
-    {
-      "epoch": 16.941176470588236,
-      "eval_accuracy": 0.9005524861878453,
-      "eval_loss": 0.3447644114494324,
-      "eval_runtime": 1.4043,
-      "eval_samples_per_second": 128.89,
-      "eval_steps_per_second": 4.273,
-      "step": 216
-    },
-    {
-      "epoch": 17.254901960784313,
-      "grad_norm": 5.946260452270508,
-      "learning_rate": 4.6296296296296296e-06,
-      "loss": 0.1886,
-      "step": 220
-    },
-    {
-      "epoch": 17.96078431372549,
-      "eval_accuracy": 0.8895027624309392,
-      "eval_loss": 0.3459985554218292,
-      "eval_runtime": 1.4593,
-      "eval_samples_per_second": 124.035,
-      "eval_steps_per_second": 4.112,
-      "step": 229
-    },
-    {
-      "epoch": 18.03921568627451,
-      "grad_norm": 6.021714210510254,
-      "learning_rate": 2.3148148148148148e-06,
-      "loss": 0.2,
-      "step": 230
-    },
-    {
-      "epoch": 18.823529411764707,
-      "grad_norm": 3.8110241889953613,
-      "learning_rate": 0.0,
-      "loss": 0.1972,
-      "step": 240
-    },
-    {
-      "epoch": 18.823529411764707,
-      "eval_accuracy": 0.8895027624309392,
-      "eval_loss": 0.34530630707740784,
-      "eval_runtime": 1.454,
-      "eval_samples_per_second": 124.486,
-      "eval_steps_per_second": 4.127,
-      "step": 240
+      "epoch": 3.0,
+      "eval_accuracy": 0.9911111111111112,
+      "eval_loss": 0.02681696228682995,
+      "eval_runtime": 1.886,
+      "eval_samples_per_second": 119.3,
+      "eval_steps_per_second": 4.242,
+      "step": 48
     },
     {
-      "epoch": 18.823529411764707,
-      "step": 240,
-      "total_flos": 7.600391915087462e+17,
-      "train_loss": 0.3423821290334066,
-      "train_runtime": 426.6472,
-      "train_samples_per_second": 76.128,
-      "train_steps_per_second": 0.563
+      "epoch": 3.0,
+      "step": 48,
+      "total_flos": 1.5048656676458496e+17,
+      "train_loss": 0.5111757349222898,
+      "train_runtime": 85.6924,
+      "train_samples_per_second": 470.987,
+      "train_steps_per_second": 3.734
     }
   ],
   "logging_steps": 10,
-  "max_steps": 240,
+  "max_steps": 320,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 20,
   "save_steps": 500,
@@ -383,7 +99,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 7.600391915087462e+17,
+  "total_flos": 1.5048656676458496e+17,
   "train_batch_size": 32,
   "trial_name": null,
   "trial_params": null