When I wanted to run the repository for the first time I encountered this problem.
File "/home/nahue/dev-personal/FullSubNet-plus/speech_enhance/audio_zen/acoustics/feature.py", line 58, in istft
return torch.istft(
RuntimeError: istft requires a complex-valued input tensor matching the output from stft with return_complex=True.
It seems to be a bug or a major change in the way torch.istft works when moving from torch to version 2.x. Now, torch.istft only accepts complex input, so I was able to fix the problem and use the code with this modification.
Tests made on:
Processor: Intel® Core™ i5-10210U CPU @ 1.60GHz × 8
OS: Linux Ubuntu 22.04.4 LTS (64 bits)
Torch version: 2.2.2
Description:
When I wanted to run the repository for the first time I encountered this problem.
It seems to be a bug or a major change in the way
torch.istftworks when moving fromtorchto version2.x. Now,torch.istftonly accepts complex input, so I was able to fix the problem and use the code with this modification.Tests made on:
Processor: Intel® Core™ i5-10210U CPU @ 1.60GHz × 8 OS: Linux Ubuntu 22.04.4 LTS (64 bits) Torch version: 2.2.2