What is the reason to store file in device?

[zhangshaohu]

In interfaces.py:
# Storing waveform in the specified device
wavs, wav_lens = wavs.to(self.device), wav_lens.to(self.device)
Why do we need to save the audio? It takes so many resources! I commented on all the lines but it still stores the speech file. How can we don't save the audio file?

[Gastron]

This does not save the audio in any new hard drive or anything like that. Here basically the .to() methods just move data from CPU-side RAM to GPU-side VRAM. This is of course absolutely necessary - the data needs to be where the computation happens.

[zhangshaohu]

Thank you for your response. I used the simple demo below
classifier = foreign_class(source="speechbrain/emotion-recognition-wav2vec2-IEMOCAP", pymodule_file="custom_interface.py", classname="CustomEncoderWav2vec2Classifier")
but every time it copied the original speech file to the local directory. Anyone experienced this?

[Gastron]

Ah. I replied a little hastily last time. The code you quoted indeed does what I say, but I understand what you were really after. The fetch functionality does this type of copying/symlinking, and that has turned out to be a very annoying feature for many users. However we're going to implement changes soon. See discussion e.g. in PR #1268

[zhangshaohu]

This commit solve this problem
stachu86@887ed86

[Gastron]

Yes, it's just that we're trying to support all these different sources.