Upload folder using huggingface_hub

ZoomLDM-demo-dataset.py

@@ -86,41 +86,30 @@ class TCGADataset(datasets.GeneratorBasedBuilder):
     def _split_generators(self, dl_manager):
 
 
-        original_script_dir = Path(self.base_path).resolve()
-        mag_folder_name = self.config.data_dir
-
-        mag_data_abs_path = original_script_dir / "data" / mag_folder_name
-        
-        print(f"base path: {self.base_path}")
-        print(f"Original script directory: {original_script_dir}")
-        print(f"Using data directory: {mag_folder_name}")
-        print(f"Absolute path to data: {mag_data_abs_path}")
-
-
+        mag_folder =  f"data/{self.config.mag_level}"  
         
         return [
             datasets.SplitGenerator(
                 name=datasets.Split.TRAIN,
                 gen_kwargs={
-                    "mag_folder_abs_path": mag_data_abs_path, 
+                    "mag_folder": mag_folder, 
                     "mag_level": self.config.mag_level,
                 },
             ),
         ]
 
-    def _generate_examples(self, mag_folder_abs_path: Path, mag_level: str):
+    def _generate_examples(self, mag_folder: Path, mag_level: str):
         idx = 0
         for i in range(16):
             img_filename = f"{i}.jpg"
             feat_filename = f"{i}_ssl_feat.npy"
 
-            img_path = mag_folder_abs_path / img_filename
-            img = Image.open(img_filename).convert("RGB")
+            img_path = f"{self.base_path}/{mag_folder}/{img_filename}"
+            img = Image.open(img_path).convert("RGB")
             img_np = np.array(img)
-            feat_path = mag_folder_abs_path / feat_filename
-
 
-            ssl_feat_data = np.load(feat_filename)
+            feat_path = f"{self.base_path}/{mag_folder}/{feat_filename}"
+            ssl_feat_data = np.load(feat_path)
             processed_feature = preprocess_features(ssl_feat_data)
 
             h = np.sqrt(processed_feature.shape[1]).astype(int)