chore(build): Use Purego with stablediffusion backend #6067
Conversation
✅ Deploy Preview for localai ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
|
I'm trying to figure out a sensible way to run the backend build/tests for the particular backend that was changed which possibly means using a dynamic matrix. In the meantime I'm just rebuilding all of the backends on my fork's CI: https://github.com/richiejp/LocalAI/actions/runs/16999488324 |
|
Let's try that again: https://github.com/richiejp/LocalAI/actions/runs/17005231171/job/48213847468 |
yup, as a one-off I usually edit LocalAI/.github/workflows/backend.yml Line 5 in 1149fb6 pull_request event
|
|
OK, I was a bit worried about doing that on the main repo in case it counted towards some usage limits, but it's probably fine in that case. Seem to be having some success: https://github.com/richiejp/LocalAI/actions/runs/17005231171/job/48213847641 |
|
and https://github.com/richiejp/LocalAI/actions/runs/17005231171/job/48213847442 There's another bug I noticed which could effect OSX, so I'll submit a fix for that as well. |
Signed-off-by: Richard Palethorpe <[email protected]>
richiejp
force-pushed
the
chore/sd-purego2
branch
from
dba407f to
361e1a6
Compare
Signed-off-by: Richard Palethorpe [email protected]
Description
Possibly the CI CPU or RAM is being over subscribed and either the runner process is not able to report progress or is killed by OOM.
The original Purego PR accidentally set CMake to use all the cores, now it defaults to nproc - 1
Compiling locally it doesn't appear that JOBS=4 uses more than 16GB of RAM (with JOBS=20 it certainly does), so I would guess that CPU is the issue. An alternative is to use nice (possibly from Docker https://github.com/docker/build-push-action?tab=readme-ov-file#environment-variables).
Notes for Reviewers
Signed commits