resize and crop

all_results.json

@@ -1,12 +1,12 @@
 {
     "epoch": 6.0,
-    "eval_accuracy": 0.5263157894736842,
-    "eval_loss": 0.9798622727394104,
-    "eval_runtime": 0.5254,
-    "eval_samples_per_second": 253.138,
-    "eval_steps_per_second": 32.356,
-    "train_loss": 0.9850856896602747,
-    "train_runtime": 45.3029,
-    "train_samples_per_second": 136.945,
-    "train_steps_per_second": 4.371
+    "eval_accuracy": 0.518796992481203,
+    "eval_loss": 0.9825071096420288,
+    "eval_runtime": 0.5938,
+    "eval_samples_per_second": 223.97,
+    "eval_steps_per_second": 28.628,
+    "train_loss": 0.9973401926984691,
+    "train_runtime": 45.8453,
+    "train_samples_per_second": 135.325,
+    "train_steps_per_second": 4.319
 }

eval_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 6.0,
-    "eval_accuracy": 0.5263157894736842,
-    "eval_loss": 0.9798622727394104,
-    "eval_runtime": 0.5254,
-    "eval_samples_per_second": 253.138,
-    "eval_steps_per_second": 32.356
+    "eval_accuracy": 0.518796992481203,
+    "eval_loss": 0.9825071096420288,
+    "eval_runtime": 0.5938,
+    "eval_samples_per_second": 223.97,
+    "eval_steps_per_second": 28.628
 }

preprocessor_config.json

@@ -1,7 +1,7 @@
 {
   "crop_pct": null,
-  "do_normalize": false,
-  "do_resize": false,
+  "do_normalize": true,
+  "do_resize": true,
   "feature_extractor_type": "ConvNextFeatureExtractor",
   "image_mean": [
     0.45

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e97a7d6e40e590c8dc7ffdafc47eb810e50b51a33a98732caa5bd854c31fadc0
+oid sha256:f75ed21005e99f438f2ed571a7b8db6b3dd13e53225752be4b204a6efb71388d
 size 761689

train.py

@@ -103,7 +103,7 @@ def main():
         dataset["validation"] = split["test"]
 
     feature_extractor = ConvNextFeatureExtractor(
-        do_resize=False, do_normalize=False, image_mean=[0.45], image_std=[0.22]
+        do_resize=True, do_normalize=True, image_mean=[0.45], image_std=[0.22]
     )
 
     # Prepare label mappings.

train_results.json

@@ -1,7 +1,7 @@
 {
     "epoch": 6.0,
-    "train_loss": 0.9850856896602747,
-    "train_runtime": 45.3029,
-    "train_samples_per_second": 136.945,
-    "train_steps_per_second": 4.371
+    "train_loss": 0.9973401926984691,
+    "train_runtime": 45.8453,
+    "train_samples_per_second": 135.325,
+    "train_steps_per_second": 4.319
 }

trainer_state.json

@@ -11,10 +11,10 @@
       "epoch": 6.0,
       "step": 198,
       "total_flos": 1708758414000000.0,
-      "train_loss": 0.9850856896602747,
-      "train_runtime": 45.3029,
-      "train_samples_per_second": 136.945,
-      "train_steps_per_second": 4.371
+      "train_loss": 0.9973401926984691,
+      "train_runtime": 45.8453,
+      "train_samples_per_second": 135.325,
+      "train_steps_per_second": 4.319
     }
   ],
   "max_steps": 198,

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:70d8482fcb9abee83a681eed3ba240f26ef087833064a1271f60d549c4fa686a
+oid sha256:a971e6637bccdbec3fdbb73349771760dd99e3c2a66a65ad63f1ab48570fa252
 size 3247