Update audioset.py

audioset.py

@@ -58,7 +58,7 @@ class AudioSet(datasets.GeneratorBasedBuilder):
                     "label": datasets.Sequence(datasets.features.ClassLabel(names=CLASSES)), 
                 }
             ),
-            name="balanced", 
+            name="20k", 
             description="",
         ), 
         AudioSetConfig(
@@ -70,7 +70,19 @@ class AudioSet(datasets.GeneratorBasedBuilder):
                     "label": datasets.Sequence(datasets.features.ClassLabel(names=CLASSES)), 
                 }
             ),
-            name="unbalanced", 
+            name="500k", 
+            description="",
+        ), 
+        AudioSetConfig(
+            features=datasets.Features(
+                {
+                    "file": datasets.Value("string"),
+                    "audio": datasets.Audio(sampling_rate=SAMPLE_RATE),
+                    "sound": datasets.Sequence(datasets.Value("string")), 
+                    "label": datasets.Sequence(datasets.features.ClassLabel(names=CLASSES)), 
+                }
+            ),
+            name="2m", 
             description="",
         ), 
     ]
@@ -118,10 +130,13 @@ class AudioSet(datasets.GeneratorBasedBuilder):
         extensions = ['.wav']
 
         if split == 'train':
-            if self.config.name == 'balanced':
+            if self.config.name == '20k':
                 archive_path = os.path.join(data_dir, 'audios', 'balanced_train_segments')
                 metadata_url = 'https://huggingface.co/datasets/confit/audioset/resolve/main/metadata/audioset-20k.jsonl'
-            elif self.config.name == 'unbalanced':
+            elif self.config.name == '500k':
+                archive_path = os.path.join(data_dir, 'audios', 'unbalanced_train_segments')
+                metadata_url = 'https://huggingface.co/datasets/confit/audioset/resolve/main/metadata/audioset-500k.jsonl'
+            elif self.config.name == '2m':
                 archive_path = os.path.join(data_dir, 'audios', 'unbalanced_train_segments')
                 metadata_url = 'https://huggingface.co/datasets/confit/audioset/resolve/main/metadata/audioset-2m.jsonl'
         elif split == 'test':
@@ -133,7 +148,7 @@ class AudioSet(datasets.GeneratorBasedBuilder):
             # Split the content by lines and parse each line as JSON
             # Each line is like {"filename":"YN6UbMsh-q1c.wav","label":["Vehicle","Car"]}
             data_list = [json.loads(line) for line in response.text.splitlines()]
-            fileid2labels = {item['filename']:item['label'] for item in data_list}
+            fileid2labels = {item['filename']:item['labels'] for item in data_list}
         else:
             logger.info(f"Failed to retrieve data: Status code {response.status_code}")