Fix the way it reads the file name

On unix, you can indeed split the '/' to get the filename as the last item in the list.
On Windows, the separator may be a '\' so this doesn't work.
In streaming mode, URLs may look like 'zip://data/file.txt::https://foo.bar.data.zip' so this doesn't work either.

I fixed it by using `os.path.basename`, which works on Unix and Windows, and which `datasets` extends to also work with chained URLs.

NIH-Chest-X-ray-dataset.py

@@ -186,7 +186,7 @@ class ChestXray14(datasets.GeneratorBasedBuilder):
 		  logger.info(f"Batch for data_files: {batch}")
 		  path_files = dl_manager.iter_files(batch)
 		  for img in path_files:
-		    if img.split('/')[-1] in train_val_list:
+		    if os.path.basename(img) in train_val_list:
 		      train_files.append(img)
 		    else:
 		      test_files.append(img)
@@ -215,7 +215,7 @@ class ChestXray14(datasets.GeneratorBasedBuilder):
 		    for i, path in enumerate(files):
 		        file_name = os.path.basename(path)
 		        # Get image id to filter the respective row of the csv 	
-		        image_id = file_name.split('/')[-1]
+		        image_id = file_name
 		        image_labels = label_csv[label_csv["Image Index"] == image_id]["Finding Labels"].values[0].split("|")
 		        if file_name.endswith(".png"):
 		            yield i, {