Skip to content

Run CI on Modal, upgrade Bitsandbytes #641

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

Merged
merged 58 commits into from
Mar 15, 2025
Merged

Run CI on Modal, upgrade Bitsandbytes #641

merged 58 commits into from
Mar 15, 2025

Conversation

mryab
Copy link
Member

@mryab mryab commented Feb 10, 2025

This PR switches the execution of tests from GitHub Actions to Modal, unlocking the option to use GPUs in those tests in the future. Since Modal workers can have multiple CPUs, we can also run tests in parallel, which speeds them up by approximately 4x — from 8 minutes to under 2 minutes.

Also, as multiple tests seem to be unstable, this PR also disables them for the time being or marks them as flaky with pytest.mark.xfail. This will be fixed in future PRs.

Lastly, since the current version of Bitsandbytes is outdated, the PR upgrades it

Copy link

codecov bot commented Feb 23, 2025

Codecov Report

Attention: Patch coverage is 74.07407% with 7 lines in your changes missing coverage. Please review.

Project coverage is 85.20%. Comparing base (d20e810) to head (cfa51d2).
Report is 24 commits behind head on master.

Files with missing lines Patch % Lines
hivemind/moe/client/moe.py 50.00% 6 Missing ⚠️
hivemind/moe/server/runtime.py 0.00% 1 Missing ⚠️
Additional details and impacted files
@@            Coverage Diff             @@
##           master     #641      +/-   ##
==========================================
- Coverage   85.39%   85.20%   -0.20%     
==========================================
  Files          81       81              
  Lines        8006     8049      +43     
==========================================
+ Hits         6837     6858      +21     
- Misses       1169     1191      +22     
Files with missing lines Coverage Δ
hivemind/compression/base.py 94.36% <100.00%> (ø)
hivemind/compression/quantization.py 94.53% <100.00%> (+0.31%) ⬆️
hivemind/moe/server/connection_handler.py 90.72% <100.00%> (-0.95%) ⬇️
hivemind/moe/server/runtime.py 22.40% <0.00%> (-53.21%) ⬇️
hivemind/moe/client/moe.py 68.13% <50.00%> (-24.65%) ⬇️

... and 32 files with indirect coverage changes

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

Comment on lines +383 to +385
samples_with_tasks = {sample_idx for sample_idx, _ in task_to_indices.values()}
pending_samples = len(samples_with_tasks) # samples for which we have less than k_min results
assert pending_samples <= num_samples
Copy link
Member Author

Choose a reason for hiding this comment

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

It's better to recompute the number of pending_samples, as some samples might have 0 tasks (for example, in test_call_many)

@dvmazur dvmazur requested review from justheuristic and removed request for justheuristic March 15, 2025 10:16
Copy link
Member

@justheuristic justheuristic left a comment

Choose a reason for hiding this comment

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

Appreciate the version updates and a face lift to the tests. LGTM; pending @dvmazur

@mryab mryab merged commit 767afa5 into master Mar 15, 2025
14 checks passed
@mryab mryab deleted the modal-ci branch March 15, 2025 10:25
mryab added a commit that referenced this pull request Apr 20, 2025
* Run CI on Modal, upgrade Bitsandbytes

* Extract the blocksize for quantization into a constant

(cherry picked from commit 767afa5)
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