Open fumiama opened 2 weeks ago
我的环境如下: fairseq==0.12.2 torchaudio==2.3.0 torchfcpe=0.0.4 python版本3.10.7 但是我无法复现此错误,在这个环境里可以正常工作
This error can be solved by the commit https://github.com/fumiama/Retrieval-based-Voice-Conversion-WebUI/commit/d9a116f4f7369474dd55547b646045f4ceab250d
You can compare the training behavior at this commit and one commit before it.
If import this lib first like
then use
The torchaudio in fairseq will got crash, here's the log
If avoid to import this lib, the fairseq can be imported successfully.