Failed reading some audio files
Hello,
I am currently adding the support of GigaSpeech inside of SpeechBrain. However, I found that some audio files were not working properly for instance two of them seems to be corrupted :
/workspace/GigaSpeech/GigaSpeech/downloads/extracted/655045dc5f9c799e6605e87b9c6fb95df55525dc2f662183b7b2c73297368389/xl_chunks_0023/POD0000005964_S0000139.wav: Failed to open the input "/workspace/GigaSpeech/GigaSpeech/downloads/extracted/655045dc5f9c799e6605e87b9c6fb95df55525dc2f662183b7b2c73297368389/xl_chunks_0023/POD0000005964_S0000139.wav"
Failed reading /workspace/GigaSpeech/GigaSpeech/downloads/extracted/9ddfab35d1368ba99cbef6df90d81c2ff42be23d60c9fdc8d6b00a32ebee10b7/test_chunks_0002/YOU1000000126_S0000192.wav: Failed to decode audio.
It seems that some other colleagues had the same issue with this dataset as well. I did a custom modification of dataset.py to skip the failing audio files but maybe it would be worth investigating a bit more what is happening...
Thanks.
Did you ever find a resolution for this issue? I've just come across it myself with what appears to be the same example as you've had trouble with. The issue seems to be that value["bytes"] (https://github.com/huggingface/datasets/blob/main/src/datasets/features/audio.py#L136) is the empty string. This is what causes soundfile to fail.