use resnet50 to train a cifar10 classifier

utils/dataset.py

@@ -0,0 +1,26 @@
+# prepare.py
+#
+# author: deng
+# date  : 20231229
+
+from pathlib import PosixPath
+
+import torch
+import numpy as np
+from torch.utils.data import Dataset
+
+
+class ProcessedDataset(Dataset):
+    """"Load processed data"""
+    def __init__(self, dataset_dir: PosixPath):
+        self.dataset_dir = dataset_dir
+        self.file_paths = list(self.dataset_dir.glob('*.npz'))
+
+    def __len__(self):
+        return len(self.file_paths)
+
+    def __getitem__(self, idx):
+        npz = np.load(self.file_paths[idx])
+        x = torch.from_numpy(npz['x'])
+        y = torch.from_numpy(npz['y'])
+        return x, y