You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I am using docker cpu-latest image from wav2letter. I got two test fails in docker - Seq2seq test and w2lModule test. When I ran the Test binary file to get transcripts of some audio files, there were only the references but no predictions were generated. And it showed 100% WER. Previously, I used the docker Cuda version on another computer with GPU and didn't face any trouble. How to get around this problem?
We had problem with cpu backend before and it was not consistent with cuda (so cuda trained model didn't give the same output with cpu). This is fixed in current fl master with switching to onednn and improving the operations itself. So please if you can use latest master for cpu train/inference. cc @jacobkahn on more detail on this.
I am using docker cpu-latest image from wav2letter. I got two test fails in docker - Seq2seq test and w2lModule test. When I ran the Test binary file to get transcripts of some audio files, there were only the references but no predictions were generated. And it showed 100% WER. Previously, I used the docker Cuda version on another computer with GPU and didn't face any trouble. How to get around this problem?
*** I am using Ubuntu 20.04
The text was updated successfully, but these errors were encountered: