Fix fsdp prob diff computation when tis is disabled - #82
Merged
Conversation
fzyzcjy
added a commit
that referenced
this pull request
Mar 19, 2026
JD-ETH
pushed a commit
to JensenFire/miles
that referenced
this pull request
Apr 11, 2026
GuanxingLu
pushed a commit
to GuanxingLu/miles
that referenced
this pull request
Apr 21, 2026
This was referenced Aug 13, 2026
yushengsu-thu
added a commit
that referenced
this pull request
Aug 13, 2026
…per-slot grad norm" This reverts commit 1eb5036. The sqrt(TP) grad-norm inflation it patched is now fixed at the root by Megatron-LM #82 (the backend's ci-megatron-pr directive; exercised by tests/fast-gpu/test_layerwise_expert_dependencies.py): LayerWise shards expert optimizer parameters over expt_dp and filters allreduce=False parameters with the expert TP group, so each logical expert-adapter parameter is owned by exactly one rank and counted once in the world-reduced per-slot norm — with tensor_model_parallel=True intact. Keeping the pre-wrap flag-clearing hook on top of that would be wrong twice over: the dedup was topology-specific (correct at TP2/ETP1/DP1, under-counting DP2/EP2 ownership), and clearing the flag now contradicts the dependency contract the new image-level test pins (grouped-expert adapter weights keep tensor_model_parallel=True; ownership, not attribute filtering, provides the dedup). The backend PR history dropped the hook for the same reason; this removes it from the frontend stack so the final tree never carries it.
This was referenced Aug 13, 2026
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
No description provided.