Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

feat(llama.cpp): enable ROCm/HIPBLAS support #1100

Merged
merged 3 commits into from
Sep 28, 2023
Merged

feat(llama.cpp): enable ROCm/HIPBLAS support #1100

merged 3 commits into from
Sep 28, 2023

Conversation

65a
Copy link
Contributor

@65a 65a commented Sep 24, 2023

Description

This PR fixes lack of HIPBLAS support in LocalAI.

Notes for Reviewers
This PR builds on go-skynet/go-llama.cpp#235 to enable ROCm/HIPBLAS support for gguf models running under llama.cpp backend (not the stable ggml one). It can be enabled by using BUILD_TYPE=hipblas. This was tested on a gfx1100 card, but should work for gfx900,gfx1030 and other cards. Card support can be set with AMDGPU_TARGETS environment variable.

Signed commits

  • Yes, I signed my commits.

Need to dance around the fact llama-stable doesn't support this (I think?) by using a plain CPU build type there. I guess clblas would be ideal, but it requires additional parameters.

Signed-off-by: 65a <65a@63bit.net>
Signed-off-by: 65a <65a@63bit.net>
Signed-off-by: 65a <65a@63bit.net>
@65a
Copy link
Contributor Author

65a commented Sep 24, 2023

Additional testing:

  • ✅ Clean checkout of PR builds correctly
  • ✅ Inference works on gfx900 and gfx1100

@65a 65a marked this pull request as ready for review September 24, 2023 19:28
@lunamidori5 lunamidori5 added the enhancement New feature or request label Sep 25, 2023
Copy link
Owner

@mudler mudler left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looking good here - not merging it yet as I'd like to cut a release of the current code first as it is well tested up to now, and I prefer to keep bumps separated.

ETA for merging this is 28-29 September.

@mudler mudler changed the title Enable ROCm/HIPBLAS support for LocalAI feat(llama.cpp): enable ROCm/HIPBLAS support Sep 27, 2023
@mudler
Copy link
Owner

mudler commented Sep 28, 2023

Thanks for the contribution @65a , merging it

@mudler mudler merged commit 55e38fe into mudler:master Sep 28, 2023
8 checks passed
mudler added a commit to go-skynet/localai-website that referenced this pull request Sep 30, 2023
mudler added a commit to go-skynet/localai-website that referenced this pull request Oct 4, 2023
* wip: new sections

Signed-off-by: mudler <mudler@localai.io>

* document hipblas mudler/LocalAI#1100

* add vllm, vall-e-x, minor updates

Signed-off-by: mudler <mudler@localai.io>

* Add development docs: wip

Signed-off-by: mudler <mudler@localai.io>

---------

Signed-off-by: mudler <mudler@localai.io>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement New feature or request
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants