Hi, first of thank you for this repo which helps in a great way. I have question:
spec = torch.sqrt(spec.pow(2).sum(-1) + 1e-6)
This line of code is present in spectrogram calculation in mel_processing.py. The problem is that we lose the information about frame_length when we use sum(-1). That is why, output of spectrogram function becomes [8, 513] (8 is batch size).
Why did you use sum(-1)?
Hi, first of thank you for this repo which helps in a great way. I have question:
spec = torch.sqrt(spec.pow(2).sum(-1) + 1e-6)
This line of code is present in spectrogram calculation in mel_processing.py. The problem is that we lose the information about frame_length when we use sum(-1). That is why, output of spectrogram function becomes [8, 513] (8 is batch size). Why did you use sum(-1)?
Thanks in advance