[GH-ISSUE #1116] vicuna33b not executing on GPU #26322

Closed
opened 2026-04-22 02:31:25 -05:00 by GiteaMirror · 2 comments
Owner

Originally created by @nshern on GitHub (Nov 13, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/1116

I am using Debian with an A4000. Vicuna13b will execute on GPU just fine but Vicuna33b will not.

Is this a bug or am I misunderstanding something in the documentation or configuration?

Originally created by @nshern on GitHub (Nov 13, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/1116 I am using Debian with an A4000. Vicuna13b will execute on GPU just fine but Vicuna33b will not. Is this a bug or am I misunderstanding something in the documentation or configuration?
Author
Owner

@BruceMacD commented on GitHub (Nov 13, 2023):

Vicuna-33b is too big to entirely offload to an A4000 GPU (quick Google search says that its got 16GB of VRAM) so the load will be shared between CPU and GPU. If you are only seeing CPU usage try checking the ollama service logs:
sudo journalctl -u ollama.service > ollama_logs.txt

Happy to help with any errors you may see in there.

<!-- gh-comment-id:1809153803 --> @BruceMacD commented on GitHub (Nov 13, 2023): Vicuna-33b is too big to entirely offload to an A4000 GPU (quick Google search says that its got 16GB of VRAM) so the load will be shared between CPU and GPU. If you are only seeing CPU usage try checking the ollama service logs: `sudo journalctl -u ollama.service > ollama_logs.txt` Happy to help with any errors you may see in there.
Author
Owner

@jmorganca commented on GitHub (Nov 14, 2023):

Hi @nshern thanks for the issue. I'll close this one for now if it's answered, but feel free to re-open it otherwise!

<!-- gh-comment-id:1809465898 --> @jmorganca commented on GitHub (Nov 14, 2023): Hi @nshern thanks for the issue. I'll close this one for now if it's answered, but feel free to re-open it otherwise!
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#26322