Replies: 1 comment
-
Does this support that one speaker can speake multilingual language? |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
I thought I'd share this in case anyone is finding the notebook for training no longer works. Here's a fix I found worked for me. For anyone not familiar with collab notebooks, click 'show code' on the Install Software step and do this:
Change this line:
!pip install -q cython>=0.29.0 piper-phonemize==1.1.0 librosa>=0.9.2 numpy==1.24 onnxruntime>=1.11.0 pytorch-lightning==1.7.7 torch==1.13.0+cu117 --extra-index-url https://download.pytorch.org/whl/cu117
To this:
!pip install -q cython>=0.29.0 piper-phonemize==1.1.0 librosa>=0.9.2 numpy==1.26.4 onnxruntime>=1.11.0 pytorch-lightning==1.9.0 torch==2.0.0 --extra-index-url https://download.pytorch.org/whl/cu117
Change this line:
!pip install -q torchtext==0.14.0 torchvision==0.14.0
To this:
!pip install -q torchtext==0.15.1 torchvision==0.15.1
Change this line:
!pip install -q torchaudio==0.13.0 torchmetrics==0.11.4 faster_whisper
To this:
!pip install -q torchaudio==2.0.0 torchmetrics==0.11.4 faster_whisper
What we've done:
Updated numpy from 1.24 to 1.26.4
Updated pytorch-lightning 1.7.7 to 1.9.0
Updated torch 1.13.0+cu117 to 2.0.0
Updated torchtext 0.14.0 to 0.15.1
Updated torchvision from 0.14.0 to 0.15.1
Updated torch audio from 0.13.0 to 2.0.0
One additional note I found useful: if selecting a model to use as a starting point for training, make sure it matches the quality you've selected in the settings. So if you selected 'high' you must use a high quality model. I was using Amy medium so I went and selected 'medium' in settings.
I hope this helps someone.
Beta Was this translation helpful? Give feedback.
All reactions