fix(fsdp): keep hybrid shard CI always enabled - #101
Merged
Conversation
Run this foundational regression on every PR and update its state-loading helper after the backend refactor moved that API. Authored with Claude. Co-authored-by: Cursor <cursoragent@cursor.com>
Match the repository isort configuration so the always-on regression can reach GPU validation. Authored with Claude. Co-authored-by: Cursor <cursoragent@cursor.com>
Keep this foundational four-GPU regression in the earlier five-GPU suite so failures surface before stage C end-to-end jobs. Authored with Claude. Co-authored-by: Cursor <cursoragent@cursor.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
stage-b-5-gpu-h200on every PR instead of gating it behindrun-ci-fsdp.broadcast_full_state_to_fsdpafter the state-loading helper moved out ofactor.py.Why
Hybrid Shard validates foundational FSDP topology, gradient scaling, weight synchronization, and checkpoint behavior, so it should not depend on an opt-in domain label. The previous label filtering hid a stale import introduced by the backend refactor.
Validation
PYTHONPATH=/tmp/miles_hybrid_fix_20260807 python tests/fast-gpu/backends/fsdp_utils/test_hybrid_shard.pyFiles
tests/fast-gpu/backends/fsdp_utils/test_hybrid_shard.py: move the suite to stage B and make it unconditional.tests/fast-gpu/backends/fsdp_utils/_hybrid_shard_worker.py: use the current state broadcast helper.Checklist
pre-commit run --all-filespasses — not run locally.pytest -xis green — full suite not run; the targeted four-case GPU suite passes.python3 train.py --helpstill parses — not applicable.This PR was authored with Claude.
Made with Cursor