channels

all_results.json

@@ -1,12 +1,12 @@
 {
     "epoch": 6.0,
-    "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_accuracy": 0.6390977443609023,
+    "eval_loss": 0.7639745473861694,
+    "eval_runtime": 0.7192,
+    "eval_samples_per_second": 184.925,
+    "eval_steps_per_second": 23.637,
+    "train_loss": 0.8484223491013653,
+    "train_runtime": 32.6308,
+    "train_samples_per_second": 190.127,
+    "train_steps_per_second": 6.068
 }

config.json

@@ -26,7 +26,7 @@
   },
   "layer_type": "basic",
   "model_type": "resnet",
-  "num_channels": 1,
+  "num_channels": 3,
   "problem_type": "single_label_classification",
   "torch_dtype": "float32",
   "transformers_version": "4.21.0.dev0"

eval_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 6.0,
-    "eval_accuracy": 0.518796992481203,
-    "eval_loss": 0.9825071096420288,
-    "eval_runtime": 0.5938,
-    "eval_samples_per_second": 223.97,
-    "eval_steps_per_second": 28.628
+    "eval_accuracy": 0.6390977443609023,
+    "eval_loss": 0.7639745473861694,
+    "eval_runtime": 0.7192,
+    "eval_samples_per_second": 184.925,
+    "eval_steps_per_second": 23.637
 }

preprocessor_config.json

@@ -4,9 +4,13 @@
   "do_resize": true,
   "feature_extractor_type": "ConvNextFeatureExtractor",
   "image_mean": [
+    0.45,
+    0.45,
     0.45
   ],
   "image_std": [
+    0.22,
+    0.22,
     0.22
   ],
   "resample": 3,

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f75ed21005e99f438f2ed571a7b8db6b3dd13e53225752be4b204a6efb71388d
-size 761689
+oid sha256:64d031cd790aedfb724b3db0bccf4be88ab3b13655bab905ec329dfa39a8c0d6
+size 786777

train.py

@@ -7,7 +7,7 @@ import datasets
 import torch
 import transformers
 from torchinfo import summary
-from torchvision.transforms import Compose, Normalize, ToTensor
+from torchvision.transforms import Compose, Normalize, ToTensor, Resize, CenterCrop
 from transformers import (
     ConvNextFeatureExtractor,
     HfArgumentParser,
@@ -103,7 +103,7 @@ def main():
         dataset["validation"] = split["test"]
 
     feature_extractor = ConvNextFeatureExtractor(
-        do_resize=True, do_normalize=True, image_mean=[0.45], image_std=[0.22]
+        do_resize=True, do_normalize=True, image_mean=[0.45, 0.45, 0.45], image_std=[0.22, 0.22, 0.22]
     )
 
     # Prepare label mappings.
@@ -115,7 +115,7 @@ def main():
         id2label[str(i)] = label
 
     config = ResNetConfig(
-        num_channels=1,
+        num_channels=3,
         layer_type="basic",
         depths=[2, 2],
         hidden_sizes=[32, 64],
@@ -129,12 +129,17 @@ def main():
 
     # Define torchvision transforms to be applied to each image.
     normalize = Normalize(mean=feature_extractor.image_mean, std=feature_extractor.image_std)
-    _transforms = Compose([ToTensor(), normalize])
+    _transforms = Compose([
+        Resize(feature_extractor.size),
+        CenterCrop(feature_extractor.size),
+        ToTensor(),
+        normalize]
+    )
 
     def transforms(example_batch):
         """Apply _train_transforms across a batch."""
         # black and white
-        example_batch["pixel_values"] = [_transforms(pil_img.convert("L")) for pil_img in example_batch["image"]]
+        example_batch["pixel_values"] = [_transforms(pil_img.convert("RGB")) for pil_img in example_batch["image"]]
         return example_batch
 
     # Load the accuracy metric from the datasets package

train_results.json

@@ -1,7 +1,7 @@
 {
     "epoch": 6.0,
-    "train_loss": 0.9973401926984691,
-    "train_runtime": 45.8453,
-    "train_samples_per_second": 135.325,
-    "train_steps_per_second": 4.319
+    "train_loss": 0.8484223491013653,
+    "train_runtime": 32.6308,
+    "train_samples_per_second": 190.127,
+    "train_steps_per_second": 6.068
 }

trainer_state.json

@@ -10,16 +10,16 @@
     {
       "epoch": 6.0,
       "step": 198,
-      "total_flos": 1708758414000000.0,
-      "train_loss": 0.9973401926984691,
-      "train_runtime": 45.8453,
-      "train_samples_per_second": 135.325,
-      "train_steps_per_second": 4.319
+      "total_flos": 1064007556964352.0,
+      "train_loss": 0.8484223491013653,
+      "train_runtime": 32.6308,
+      "train_samples_per_second": 190.127,
+      "train_steps_per_second": 6.068
     }
   ],
   "max_steps": 198,
   "num_train_epochs": 6,
-  "total_flos": 1708758414000000.0,
+  "total_flos": 1064007556964352.0,
   "trial_name": null,
   "trial_params": null
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a971e6637bccdbec3fdbb73349771760dd99e3c2a66a65ad63f1ab48570fa252
+oid sha256:b5c30f97a34f52c729b4381ca5704c7c50d318011f10167ec3bfaa9db9be0777
 size 3247