Closed e-osipov closed 1 year ago
Ok, I think torch is somehow not installed properly with cuda support, this is weird....
I will update the Dockerfile to explicitly specify torch version with cuda=11.3.
Sorry about that!
Oh, I guess cuda=11.3.1 is too old for torch=2.1.1, will just do
pip install torch==1.12.1+cu113 --extra-index-url https://download.pytorch.org/whl/cu113
ok, resolved with d5f5787!! thanks for reporting this issue!
Hi everyone,
I wonder how to properly create docker image and run container in order to utilize GPU? For now, chroma is running using CPU only.
What I did:
Dockerfile
usingdocker build -t chroma .
commanddocker run -i -t --runtime=nvidia --gpus all chroma
nvidia-smi
within container will show that I have GPU available However, any example will run on CPU, using all available cores.