Spaces:
Running
on
T4
Running
on
T4
Apply for community grant: Academic project
#3
by
hbredin
- opened
pyannote.audio speaker diarization is kind of slow on CPU.
Would be nice to have access to a GPU for this demo!
hbredin
changed discussion status to
closed
Than
hbredin
changed discussion status to
open
Thanks @reach-vb
However, I am having trouble using this GPU with pytorch.
I think I did what the documentation suggests (https://huggingface.co/docs/hub/spaces-gpus#pytorch) by updating the requirements.txt and adding those two lines:
--extra-index-url https://download.pytorch.org/whl/cu113
torch
However, False is still returned by torch.cuda.is_available(). Any idea?
Forcing torch==2.0.1 (instead of newly release 2.1) seems to do the trick!
hbredin
changed discussion status to
closed