bark icon indicating copy to clipboard operation
bark copied to clipboard

No GPU being used

Open mike2003 opened this issue 1 year ago • 10 comments

I have this message. No GPU being used. Careful

But my is Geforce 1660 Super What's wrong?

Driver Version: 472.12 CUDA Version: 11.4 Win10x64

mike2003 avatar Apr 20 '23 10:04 mike2003

Hm, what do you get for: torch.cuda.is_available()?

gkucsko avatar Apr 20 '23 12:04 gkucsko

false

mike2003 avatar Apr 20 '23 12:04 mike2003

Hm then this is a PyTorch issue. Sorry, hard to debug remotely. Have you used torch on gpu successfully in the past?

gkucsko avatar Apr 20 '23 12:04 gkucsko

My first time with python.

mike2003 avatar Apr 20 '23 13:04 mike2003

ah, yeah then i would look at torch tutorials (https://pytorch.org/get-started/locally/) unfortunately cuda is still a bit finicky, especially on older cards. in the meantime for bark you can just use the huggingface spaces demo which has a UI or a google colab notebook which has a built-in gpu. that should get you started faster

gkucsko avatar Apr 20 '23 13:04 gkucsko

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117 worked for me

Fictiverse avatar Apr 21 '23 00:04 Fictiverse

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117 worked for me

Did not work for me. I have CUDA 11.8 installed. I don't know what is missing.

hippopotamus1000 avatar Apr 21 '23 09:04 hippopotamus1000

So the problem you have to solve is when running this in python: import torch torch.cuda.device_count() it returns 0. There are lots of google results for debugging that issue, on it myself atm.

The pip3 install did not help for me either and I have stable diffusion running fine

jjonj avatar Apr 21 '23 14:04 jjonj

So the problem you have to solve is when running this in python: import torch torch.cuda.device_count() it returns 0. There are lots of google results for debugging that issue, on it myself atm.

The pip3 install did not help for me either and I have stable diffusion running fine

Try to run pip uninstall torch torchvision and then pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117

This worked for me. (ChatGPT recommended this)

pyth0on avatar Apr 21 '23 15:04 pyth0on

So the problem you have to solve is when running this in python: import torch torch.cuda.device_count() it returns 0. There are lots of google results for debugging that issue, on it myself atm. The pip3 install did not help for me either and I have stable diffusion running fine

Try to run pip uninstall torch torchvision and then pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117

This worked for me. (ChatGPT recommended this)

Worked for me as well.

najeebl avatar Apr 22 '23 00:04 najeebl

So the problem you have to solve is when running this in python: import torch torch.cuda.device_count() it returns 0. There are lots of google results for debugging that issue, on it myself atm. The pip3 install did not help for me either and I have stable diffusion running fine

Try to run pip uninstall torch torchvision and then pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117

This worked for me. (ChatGPT recommended this)

thanks its works for me too

shlomitgueta avatar Jun 07 '23 13:06 shlomitgueta

pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu121

jxyk2007 avatar Aug 21 '23 09:08 jxyk2007