-
Notifications
You must be signed in to change notification settings - Fork 415
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
[LLM] Fix mixtral example for Azure #3017
Conversation
llm/mixtral/serve.yaml
Outdated
@@ -34,6 +34,9 @@ setup: | | |||
pip list | grep megablocks || pip install megablocks | |||
|
|||
run: | | |||
# Remove the default nccl.conf which causes failure on Azure | |||
sudo mv /etc/nccl.conf /etc/nccl.conf.bak || true |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
What is this file for? Will this have any side-effects?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Also is it a common issue for other workloads? Should we add it in azure-ray.yml.j2
instead?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is a good point! This file is wrongly configured on Azure with multiple GPUs, causing the issue with the distributed workload using nccl. We should include this in the azure-ray.yml.j2
.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Fixed. PTAL.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
c24169d
to
79ea123
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
Fixes #2905
Fix mixtral example for Azure by removing the problemetic nccl config file on Azure.
With this PR, after #2434 is merged, we should be able to change the examples to
disk_tier: best
, so that Azure is allowed to be part of the candidates.Tested (run the relevant ones):
bash format.sh
sky launch --disk-tier none -c test-mixtral --cloud azure llm/mixtral/serve.yaml
sky launch -c test-gcp-mixtral --cloud gcp --use-spot llm/mixtral/serve.yaml
pytest tests/test_smoke.py
pytest tests/test_smoke.py::test_fill_in_the_name
bash tests/backward_comaptibility_tests.sh