Skip to content

Conversation

@HanClinto
Copy link
Contributor

As suggested by @slaren ( #6576 (comment) ), disabling Metal for test invocation of eval-callback to fix CI build on macOS-latest-cmake-arm64

Example failure:
https://github.com/ggerganov/llama.cpp/actions/runs/8651988935/job/23723903645#step:5:4773

@HanClinto
Copy link
Contributor Author

I would prefer it if I could add -ngl 0 only for the osx platform, but I'm not familiar enough with cmake to do that, and welcome suggestions / corrections.

@slaren
Copy link
Member

slaren commented Apr 11, 2024

I think -ngl 0 is fine, the github runners do not have a GPU anyway. And -ngl 0 is already the default for all backends except Metal. To test with a GPU it could be run in the ggml-ci (in ci/run.sh), but it's not important.

@HanClinto HanClinto merged commit f7001cc into ggml-org:master Apr 11, 2024
@phymbert
Copy link
Collaborator

Thanks, sorry I did not read the error log carefully in the PR as I was thinking the metal test randomly failed :/

@HanClinto
Copy link
Contributor Author

Thanks, sorry I did not read the error log carefully in the PR as I was thinking the metal test randomly failed :/

Not a problem at all! The eval-callback program looks awesome, and I'm glad it's added to the suite -- nice job on that!

tybalex pushed a commit to rubra-ai/tools.cpp that referenced this pull request Apr 17, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants