[GH-ISSUE #1512] Generation speed 3x slower after pulling newer docker image #47333

Closed
opened 2026-04-28 03:36:02 -05:00 by GiteaMirror · 1 comment
Owner

Originally created by @johnp-4dvanalytics on GitHub (Dec 14, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/1512

I updated the docker image recently and the inference was 3x slower (30s for new image vs. 10s for old image). I was able to get back to the previous inference speed by reverting to ollama/ollama:0.1.10. I tried ollama/ollama:0.1.15 and ollama/ollama:0.1.14 and both had the slower inference speed.

EDIT: model used was https://ollama.ai/library/openhermes2.5-mistral:latest

Originally created by @johnp-4dvanalytics on GitHub (Dec 14, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/1512 I updated the docker image recently and the inference was 3x slower (30s for new image vs. 10s for old image). I was able to get back to the previous inference speed by reverting to ollama/ollama:0.1.10. I tried ollama/ollama:0.1.15 and ollama/ollama:0.1.14 and both had the slower inference speed. EDIT: model used was https://ollama.ai/library/openhermes2.5-mistral:latest
Author
Owner

@jmorganca commented on GitHub (Feb 20, 2024):

Hi there, this should be fixed in newer versions of Ollama, but let me know if that isn't the case. Thanks for creating an issue!

<!-- gh-comment-id:1953345898 --> @jmorganca commented on GitHub (Feb 20, 2024): Hi there, this should be fixed in newer versions of Ollama, but let me know if that isn't the case. Thanks for creating an issue!
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#47333