[GH-ISSUE #5428] bump llama.cpp for gemma2 fixes #49908

Closed
opened 2026-04-28 13:22:55 -05:00 by GiteaMirror · 7 comments
Owner

Originally created by @ki-manufaktur on GitHub (Jul 2, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/5428

What is the issue?

Hi,

there are at least 3 merged PRs in lamma.cpp that fix performance issues for gemma2 models

https://github.com/ggerganov/llama.cpp/pull/8197
https://github.com/ggerganov/llama.cpp/pull/8227
https://github.com/ggerganov/llama.cpp/pull/8244

OS

No response

GPU

No response

CPU

No response

Ollama version

0.1.48

Originally created by @ki-manufaktur on GitHub (Jul 2, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/5428 ### What is the issue? Hi, there are at least 3 merged PRs in lamma.cpp that fix performance issues for gemma2 models https://github.com/ggerganov/llama.cpp/pull/8197 https://github.com/ggerganov/llama.cpp/pull/8227 https://github.com/ggerganov/llama.cpp/pull/8244 ### OS _No response_ ### GPU _No response_ ### CPU _No response_ ### Ollama version 0.1.48
GiteaMirror added the bug label 2026-04-28 13:22:55 -05:00
Author
Owner

@Qualzz commented on GitHub (Jul 2, 2024):

amazing

<!-- gh-comment-id:2202718326 --> @Qualzz commented on GitHub (Jul 2, 2024): amazing
Author
Owner

@Bearsaerker commented on GitHub (Jul 3, 2024):

would be very important I think, gemma remains broken as it is right now.

<!-- gh-comment-id:2205609193 --> @Bearsaerker commented on GitHub (Jul 3, 2024): would be very important I think, gemma remains broken as it is right now.
Author
Owner

@ai-and-i commented on GitHub (Jul 3, 2024):

yes, especially the 27b version is quite unusable at the moment

<!-- gh-comment-id:2206295922 --> @ai-and-i commented on GitHub (Jul 3, 2024): yes, especially the 27b version is quite unusable at the moment
Author
Owner

@ki-manufaktur commented on GitHub (Jul 4, 2024):

most probably solved by https://github.com/ollama/ollama/pull/5475

<!-- gh-comment-id:2209445976 --> @ki-manufaktur commented on GitHub (Jul 4, 2024): most probably solved by https://github.com/ollama/ollama/pull/5475
Author
Owner

@ai-and-i commented on GitHub (Jul 5, 2024):

That PR is merged. Do the model files need to be regenerated as well?

<!-- gh-comment-id:2211448465 --> @ai-and-i commented on GitHub (Jul 5, 2024): That PR is merged. Do the model files need to be regenerated as well?
Author
Owner

@jmorganca commented on GitHub (Jul 6, 2024):

Hi folks this should be fixed in main and now in the pre-release which should be going out soon: https://github.com/ollama/ollama/releases/tag/v0.1.49-rc7. Will look into re-pushing Gemma 2 models on ollama.com

<!-- gh-comment-id:2211620756 --> @jmorganca commented on GitHub (Jul 6, 2024): Hi folks this should be fixed in main and now in the pre-release which should be going out soon: https://github.com/ollama/ollama/releases/tag/v0.1.49-rc7. Will look into re-pushing Gemma 2 models on ollama.com
Author
Owner

@Bearsaerker commented on GitHub (Jul 6, 2024):

Thank you very much for all you work! @jmorganca

<!-- gh-comment-id:2211703456 --> @Bearsaerker commented on GitHub (Jul 6, 2024): Thank you very much for all you work! @jmorganca
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#49908