fix: trying to pair metadata and tarfile audios

asr_dataset.py

@@ -99,18 +99,14 @@ class ASRDataset(datasets.GeneratorBasedBuilder):
 
     def _generate_examples(self, prompts_path, path_to_clips, audio_files):
         """Yields examples as (key, example) tuples."""
-        logger.info('hey there!')
-        examples = {}
+
+        sentence_map  = {}
         with open(prompts_path, encoding="utf-8") as f:
             for row in f:
                 data = row.strip().split("\t", 1)
                 file_name = data[0].strip()
-                audio_path = file_name # "/".join([path_to_clips, file_name])
-                examples[audio_path] = {
-                    "path": audio_path,
-                    "sentence": data[1],
-                }
-        inside_clips_dir = False
+                sentence_map[audio_path] = sentence
+
         id_ = 0
 
         tar_files = [
@@ -139,8 +135,15 @@ class ASRDataset(datasets.GeneratorBasedBuilder):
         for tar_file in tar_files:
             with tarfile.open(tar_file, 'r') as tar:
                 for member in tar.getmembers():
-                    path = member.name #"/".join([path_to_clips, member.name])
+                    file_name = member.name #"/".join([path_to_clips, member.name])
+
+                    sentence = sentence_map.get(file_name, "")
+
                     audio_file =tar.extractfile(member)
-                    audio = {"path": path, "bytes": audio_file.read()}
-                    yield id_, {**examples[path], "audio": audio}
+                    audio = {"path": file_name, "bytes": audio_file.read()}
+                    yield id_, {
+                        "file_name": file_name,
+                        "audio": audio,
+                        "sentence": sentence
+                    }
                     id_ += 1