test with multiple folders

unit-test_PDFfolder.py

@@ -8,6 +8,7 @@ logger = datasets.logging.get_logger(__name__)
 
 _DESCRIPTION = "A generic pdf folder"
 
+_CLASSES = ["categoryA", "categoryB"]
 
 folder = "https://huggingface.co/datasets/jordyvl/unit-test_PDFfolder/resolve/main/data"
 
@@ -26,12 +27,8 @@ class PdfFolder(datasets.GeneratorBasedBuilder):
         if folder is None:
             raise RuntimeError()
         """
-        self.config.data_files = folder
-        import pdb; pdb.set_trace()  # breakpoint f8426718 //
-        
-
 
-        classes = sorted([x.name.lower() for x in Path(folder).glob("*/**")])
+        #classes = sorted([x.name.lower() for x in Path(folder).glob("*/**")])
 
         return datasets.DatasetInfo(
             description=_DESCRIPTION,
@@ -48,17 +45,17 @@ class PdfFolder(datasets.GeneratorBasedBuilder):
         self, dl_manager: datasets.DownloadManager
     ) -> List[datasets.SplitGenerator]:
 
-        data_files = self.config.data_files
+        self.config.data_files = folder
 
-        if isinstance(data_files, str):
+        if isinstance(self.config.data_files, str):
             return [
                 datasets.SplitGenerator(
-                    name=datasets.Split.TRAIN, gen_kwargs={"archive_path": data_files}
+                    name=datasets.Split.TRAIN, gen_kwargs={"archive_path": self.config.data_files}
                 )
             ]
-
+            
         splits = []
-        for split_name, folder in data_files.items():
+        for split_name, folder in self.config.data_files.items():
             splits.append(
                 datasets.SplitGenerator(name=split_name, gen_kwargs={"archive_path": folder})
             )
@@ -72,7 +69,7 @@ class PdfFolder(datasets.GeneratorBasedBuilder):
         for i, path in enumerate(Path(archive_path).glob("**/*")):
             if path.suffix in extensions:
 
-                images = pdf2image.convert_from_bytes(path.posix())
+                images = pdf2image.convert_from_bytes(path.posix()) #alternatively https://huggingface.co/docs/datasets/v2.8.0/en/package_reference/main_classes#datasets.Dataset.set_transform
 
                 # convert PDF to list of images
                 yield i, {"file": images, "labels": labels.encode_example(path.parent.name.lower())}