exclude PNGs from download

Patho-Bench.py

@@ -30,12 +30,19 @@ class PathoBenchConfig(datasets.BuilderConfig):
 
 
 class PathoBenchDataset(datasets.GeneratorBasedBuilder):
+    """
+    Downloads only the .tsv and .yaml files needed to construct the dataset.
+    Excludes .png images so they don't break the builder.
+    """
     BUILDER_CONFIGS = [
         PathoBenchConfig(name="custom_config", version="1.0.0", description="PathoBench config")
     ]
     BUILDER_CONFIG_CLASS = PathoBenchConfig
 
-    
+    @property
+    def repo_id(self):
+        return "MahmoodLab/patho-bench"
+
     def _info(self):
         return datasets.DatasetInfo(
             description="PathoBench: collection of canonical computational pathology tasks",
@@ -56,36 +63,60 @@ class PathoBenchDataset(datasets.GeneratorBasedBuilder):
         # Ensure the base local directory exists
         os.makedirs(local_dir, exist_ok=True)
 
-        # download available_splits.yaml if not yet downloaded
+        # 1) Download the top-level available_splits.yaml
         snapshot_download(
             repo_id=repo_id,
-            allow_patterns=["available_splits.yaml"],
+            allow_patterns=["available_splits.yaml"],  # only this file
             repo_type="dataset",
             local_dir=local_dir,
             force_download=force_download,
         )
 
-        # open yaml and get a list of datasets implemented
+        # Read available splits
         with open(os.path.join(local_dir, "available_splits.yaml"), 'r') as file:
             available_splits = yaml.safe_load(file)
         
-        # ensure dataset_to_download is in implemented_datasets
+        # Basic validation
         if dataset_to_download != "*":
-            assert dataset_to_download in available_splits, f"{dataset_to_download} was not found. Available splits: ({available_splits})"
+            assert dataset_to_download in available_splits, (
+                f"{dataset_to_download} was not found. "
+                f"Available splits: {list(available_splits.keys())}"
+            )
             if task_in_dataset != "*":
-                assert task_in_dataset in available_splits[dataset_to_download], f"{task_in_dataset} was not found in {dataset_to_download}. Available tasks: ({available_splits[dataset_to_download]})"
+                assert task_in_dataset in available_splits[dataset_to_download], (
+                    f"{task_in_dataset} was not found in {dataset_to_download}. "
+                    f"Available tasks: {available_splits[dataset_to_download]}"
+                )
 
-        # Determine parent folder based on dataset naming
-        os.makedirs(local_dir, exist_ok=True)
-        
-        # Determine the download pattern
+        # 2) Decide what to allow based on dataset/task
+        #
+        # We only want .tsv and the relevant .yaml files (like about.yaml, config.yaml).
+        # That way, we skip .png images which can cause issues or be large in LFS.
         if dataset_to_download == "*":
-            allow_patterns = [f"*/*"]
+            # Download every dataset subfolder's .tsv and about.yaml/config.yaml
+            allow_patterns = [
+                "**/*.tsv",             # All tsv splits
+                "**/about.yaml",        # The about files
+                "**/config.yaml",       # The config files
+                "available_splits.yaml" # Already downloaded, but no harm
+            ]
         else:
-            task_path = "*" if task_in_dataset == '*' else f"{task_in_dataset}/*"
-            allow_patterns = [f"{dataset_to_download}/{task_path}"]
-    
-        # Download the required datasets
+            if task_in_dataset == "*":
+                allow_patterns = [
+                    f"{dataset_to_download}/**/*.tsv",
+                    f"{dataset_to_download}/**/about.yaml",
+                    f"{dataset_to_download}/**/config.yaml",
+                    "available_splits.yaml"
+                ]
+            else:
+                allow_patterns = [
+                    f"{dataset_to_download}/{task_in_dataset}/*.tsv",
+                    f"{dataset_to_download}/{task_in_dataset}/config.yaml",
+                    f"{dataset_to_download}/about.yaml",
+                    "available_splits.yaml"
+                ]
+
+        # 3) Download the requested patterns
         snapshot_download(
             repo_id=repo_id,
             allow_patterns=allow_patterns,
@@ -94,7 +125,7 @@ class PathoBenchDataset(datasets.GeneratorBasedBuilder):
             force_download=force_download,
         )
         
-        # Locate all .tsv files
+        # 4) Locate all .tsv files to pass to _generate_examples
         search_pattern = os.path.join(local_dir, '**', '*.tsv')
         all_tsv_splits = glob.glob(search_pattern, recursive=True)
         
@@ -105,7 +136,6 @@ class PathoBenchDataset(datasets.GeneratorBasedBuilder):
             )
         ]
 
-
     def _generate_examples(self, filepath):
         idx = 0
         for file in filepath: