uniform dataset ids

spgispeech.py

@@ -102,7 +102,7 @@ class Spgispeech(datasets.GeneratorBasedBuilder):
             description=_DESCRIPTION,
             features=datasets.Features(
                 {
-                    "wav_filename": datasets.Value("string"),
+                    "id": datasets.Value("string"),
                     "audio": datasets.Audio(sampling_rate=16_000),
                     "wav_filesize": datasets.Value("int32"),
                     "text": datasets.Value("string"),
@@ -230,6 +230,7 @@ class Spgispeech(datasets.GeneratorBasedBuilder):
                 # get the .wav filename by removing the directory path from the audio filename
                 wav_filename = "/".join(audio_filename.split("/")[-2:])
                 example = dict(metadata[wav_filename])
+                example["id"] = example.pop("wav_filename")
                 example["text"] = example.pop("transcript")
                 example["audio"] = {"path": path, "bytes": audio_file.read()}
                 example["whisper_transcript"] = whisper_transcripts[idx]