January 9, 20251 yr I've noticed my Ollama docker container is not using the GPU even though it is available when you exec into it so looking for help. I have the nVidia plugin and driver installed. I have tried restarting, removing the container and pulling it back down but it's still not working. Config below. I have deliberately pinned it to 4 cores to make sure it does not consumer all my CPU's whilst it's running and also limited the memory (bad experience with llama3.3). Any help would be appreciated. Docker Template: SMI output from within the container: Top showing it's using CPU executed inside the container:
January 10, 20251 yr Author 8 hours ago, tjb_altf4 said: You're using the AMD GPU specific image (rocm) That did the trick. Thank you!
March 15, 20251 yr I think I might be having the same issue you were however mine was working fine and then all of a sudden ollama started pegging the cpu when running a command in open webui. My nvidia-smi looks like yours except I'm running v570.124.04. I'm running unraid 7.0.1. Any suggestions? Thanks in advance.
March 15, 20251 yr Well I've got something else going on. n8n calls to Ollama are working fine but if I use Open Webui the requests just use the cpu. So I created a new ID in Open Webui and the new ID works fine. Both account settings look the same so I'll probably just nuke the bad account if I can't find the issue. I could also nuke the webui.db file and start from scratch.
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.