Brain MRI 🧠 fine tune - 5 more epochs.

README.md

@@ -3,6 +3,7 @@ library_name: transformers
 license: apache-2.0
 base_model: google/vit-base-patch16-224
 tags:
+- image-classification
 - generated_from_trainer
 metrics:
 - accuracy
@@ -16,10 +17,10 @@ should probably proofread and complete it, then remove this comment. -->
 
 # vit-base-brain-mri
 
-This model is a fine-tuned version of [google/vit-base-patch16-224](https://huggingface.co/google/vit-base-patch16-224) on an unknown dataset.
+This model is a fine-tuned version of [google/vit-base-patch16-224](https://huggingface.co/google/vit-base-patch16-224) on the BrainMRI dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.6249
-- Accuracy: 0.7770
+- Loss: 1.1441
+- Accuracy: 0.5457
 
 ## Model description
 

all_results.json

@@ -1,13 +1,13 @@
 {
-    "epoch": 5.0,
-    "eval_accuracy": 0.5304568527918782,
-    "eval_loss": 1.1719971895217896,
-    "eval_runtime": 7.6226,
-    "eval_samples_per_second": 51.688,
-    "eval_steps_per_second": 6.559,
-    "total_flos": 8.896239873264845e+17,
-    "train_loss": 0.0,
-    "train_runtime": 0.201,
-    "train_samples_per_second": 57114.909,
-    "train_steps_per_second": 1791.06
+    "epoch": 10.0,
+    "eval_accuracy": 0.5456852791878173,
+    "eval_loss": 1.1441423892974854,
+    "eval_runtime": 3.3533,
+    "eval_samples_per_second": 117.494,
+    "eval_steps_per_second": 14.91,
+    "total_flos": 1.779247974652969e+18,
+    "train_loss": 0.3026825189590454,
+    "train_runtime": 130.7944,
+    "train_samples_per_second": 175.543,
+    "train_steps_per_second": 5.505
 }

eval_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 5.0,
-    "eval_accuracy": 0.5304568527918782,
-    "eval_loss": 1.1719971895217896,
-    "eval_runtime": 7.6226,
-    "eval_samples_per_second": 51.688,
-    "eval_steps_per_second": 6.559
+    "epoch": 10.0,
+    "eval_accuracy": 0.5456852791878173,
+    "eval_loss": 1.1441423892974854,
+    "eval_runtime": 3.3533,
+    "eval_samples_per_second": 117.494,
+    "eval_steps_per_second": 14.91
 }

runs/Aug25_15-08-58_DESKTOP-4SGMSGR/events.out.tfevents.1724587887.DESKTOP-4SGMSGR.35484.4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4595eed64c344239b3d3c66fab8db26b02ca685a6cc37e2e543104046d2e4c74
+size 363

train_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 5.0,
-    "total_flos": 8.896239873264845e+17,
-    "train_loss": 0.0,
-    "train_runtime": 0.201,
-    "train_samples_per_second": 57114.909,
-    "train_steps_per_second": 1791.06
+    "epoch": 10.0,
+    "total_flos": 1.779247974652969e+18,
+    "train_loss": 0.3026825189590454,
+    "train_runtime": 130.7944,
+    "train_samples_per_second": 175.543,
+    "train_steps_per_second": 5.505
 }

trainer_state.json

@@ -1,9 +1,9 @@
 {
-  "best_metric": 0.7015281915664673,
-  "best_model_checkpoint": "./vit-base-brain-mri\\checkpoint-360",
-  "epoch": 5.0,
+  "best_metric": 0.6089015603065491,
+  "best_model_checkpoint": "./vit-base-brain-mri\\checkpoint-648",
+  "epoch": 10.0,
   "eval_steps": 500,
-  "global_step": 360,
+  "global_step": 720,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -75,19 +75,92 @@
       "step": 360
     },
     {
-      "epoch": 5.0,
-      "step": 360,
-      "total_flos": 8.896239873264845e+17,
-      "train_loss": 0.0,
-      "train_runtime": 0.201,
-      "train_samples_per_second": 57114.909,
-      "train_steps_per_second": 1791.06
+      "epoch": 5.555555555555555,
+      "grad_norm": 1.9345481395721436,
+      "learning_rate": 0.0001333333333333333,
+      "loss": 0.6651,
+      "step": 400
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.7630662020905923,
+      "eval_loss": 0.68391352891922,
+      "eval_runtime": 4.8866,
+      "eval_samples_per_second": 117.465,
+      "eval_steps_per_second": 14.734,
+      "step": 432
+    },
+    {
+      "epoch": 6.944444444444445,
+      "grad_norm": 2.201732635498047,
+      "learning_rate": 9.166666666666667e-05,
+      "loss": 0.6307,
+      "step": 500
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": 0.7700348432055749,
+      "eval_loss": 0.6623685359954834,
+      "eval_runtime": 4.9432,
+      "eval_samples_per_second": 116.118,
+      "eval_steps_per_second": 14.565,
+      "step": 504
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.7822299651567944,
+      "eval_loss": 0.6363218426704407,
+      "eval_runtime": 4.9241,
+      "eval_samples_per_second": 116.57,
+      "eval_steps_per_second": 14.622,
+      "step": 576
+    },
+    {
+      "epoch": 8.333333333333334,
+      "grad_norm": 1.8554637432098389,
+      "learning_rate": 4.9999999999999996e-05,
+      "loss": 0.5857,
+      "step": 600
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": 0.7822299651567944,
+      "eval_loss": 0.6089015603065491,
+      "eval_runtime": 4.9259,
+      "eval_samples_per_second": 116.527,
+      "eval_steps_per_second": 14.617,
+      "step": 648
+    },
+    {
+      "epoch": 9.722222222222221,
+      "grad_norm": 3.4061734676361084,
+      "learning_rate": 8.333333333333332e-06,
+      "loss": 0.576,
+      "step": 700
+    },
+    {
+      "epoch": 10.0,
+      "eval_accuracy": 0.7770034843205574,
+      "eval_loss": 0.6249111294746399,
+      "eval_runtime": 5.1184,
+      "eval_samples_per_second": 112.144,
+      "eval_steps_per_second": 14.067,
+      "step": 720
+    },
+    {
+      "epoch": 10.0,
+      "step": 720,
+      "total_flos": 1.779247974652969e+18,
+      "train_loss": 0.3026825189590454,
+      "train_runtime": 130.7944,
+      "train_samples_per_second": 175.543,
+      "train_steps_per_second": 5.505
     }
   ],
   "logging_steps": 100,
-  "max_steps": 360,
+  "max_steps": 720,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 5,
+  "num_train_epochs": 10,
   "save_steps": 500,
   "stateful_callbacks": {
     "TrainerControl": {
@@ -101,7 +174,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 8.896239873264845e+17,
+  "total_flos": 1.779247974652969e+18,
   "train_batch_size": 32,
   "trial_name": null,
   "trial_params": null