[GH-ISSUE #8683] Support release build without AVX #31386

Closed
opened 2026-04-22 11:47:34 -05:00 by GiteaMirror · 3 comments
Owner

Originally created by @yoonsio on GitHub (Jan 30, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/8683

Originally assigned to: @mxyng on GitHub.

Release image fails to detect the GPU when running on a CPU that does not support AVX.
Please add a non-AVX release build to the release pipeline.

msg="Dynamic LLM libraries" runners="[cpu_avx cpu cpu_avx2]"

Custom image can be built by overriding CUSTOM_CPU_FLAGS.

Example:

docker build --platform linux/amd64 --build-arg VERSION=noavx --build-arg CUSTOM_CPU_FLAGS= -t ollama/ollama:noavx -f Dockerfile .

Relevant issue:

Originally created by @yoonsio on GitHub (Jan 30, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/8683 Originally assigned to: @mxyng on GitHub. Release image fails to detect the GPU when running on a CPU that does not support AVX. Please add a non-AVX release build to the release pipeline. ``` msg="Dynamic LLM libraries" runners="[cpu_avx cpu cpu_avx2]" ``` Custom image can be built by overriding `CUSTOM_CPU_FLAGS`. #### Example: ``` docker build --platform linux/amd64 --build-arg VERSION=noavx --build-arg CUSTOM_CPU_FLAGS= -t ollama/ollama:noavx -f Dockerfile . ``` #### Relevant issue: * https://github.com/ollama/ollama/issues/2187
GiteaMirror added the feature request label 2026-04-22 11:47:34 -05:00
Author
Owner

@yoonsio commented on GitHub (Jan 30, 2025):

@mxyng I can create a PR if you don't have the bandwidth.

<!-- gh-comment-id:2625088153 --> @yoonsio commented on GitHub (Jan 30, 2025): @mxyng I can create a PR if you don't have the bandwidth.
Author
Owner

@mxyng commented on GitHub (Jan 31, 2025):

This should be fixed in the next release. The build has been updated to leverage ggmls dynamic backends which enables mixing cpu and gpu features

See #8539 for details

<!-- gh-comment-id:2626124432 --> @mxyng commented on GitHub (Jan 31, 2025): This should be fixed in the next release. The build has been updated to leverage ggmls dynamic backends which enables mixing cpu and gpu features See #8539 for details
Author
Owner

@diegocade1 commented on GitHub (Feb 2, 2025):

This should be fixed in the next release. The build has been updated to leverage ggmls dynamic backends which enables mixing cpu and gpu features

See #8539 for details

hello @mxyng , about what you say here, when will be this released?

<!-- gh-comment-id:2629567775 --> @diegocade1 commented on GitHub (Feb 2, 2025): > This should be fixed in the next release. The build has been updated to leverage ggmls dynamic backends which enables mixing cpu and gpu features > > See [#8539](https://github.com/ollama/ollama/pull/8539) for details hello @mxyng , about what you say here, when will be this released?
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#31386