Issues Setting Up for Training

[ForrestBakerIV]

Hi all,

I have looked over the discussions and walked through as close as I can everything.

So training via CPU is fine getting the GPU to engage has been problematic. Here are a few issues

1 ) Nvidia-docker2 - When I install this my CUDA is 11.4... I cannot seem to make this go to 10.2 or 10.0 WHICH from what I can tell is required for me to use the GPU. Any guidance on this would be great. a P.S. is that in the documents https://stt.readthedocs.io/en/latest/playbook/ENVIRONMENT.html it shows CUDA 11.0 - which from what I understand WILL NOT WORK...

Any and all suggestions to get this 'fixed' so I can utilize the GPU would be great!

[ForrestBakerIV]

Okay... So I have been banging my head against a wall...

So apparently the latest version of the docker image/container when using ghcr.io/coqui-ai/stt-train:latest is using CUDA 11.

This apparently is blowing everything up and not allowing things to train. I tried to install an earlier version of CUDA but ran into too many issues.

Not sure what is going on BUT I want to give people an update.