[Dev] Fix single grouped weight when enabling MXFP8 primary weight #4621
Draft
zhongbozhu wants to merge 1 commit into
Draft
[Dev] Fix single grouped weight when enabling MXFP8 primary weight #4621zhongbozhu wants to merge 1 commit into
zhongbozhu wants to merge 1 commit into
Conversation
zhongbozhu
commented
May 5, 2026
| return quantized_members | ||
|
|
||
|
|
||
| def modify_grouped_tensor_underlying_storage( |
Contributor
Author
There was a problem hiding this comment.
Note: this is to support the following combinations
- single grouped weight + mxfp8 recipe + turned off fp8 param gather & reuse grad buffer
- single grouped weight + bf16 recipe
9623fbd to
9f6354a
Compare
Signed-off-by: Zhongbo Zhu <zhongboz@nvidia.com>
Connor-XY
added a commit
to Connor-XY/Megatron-LM
that referenced
this pull request
May 13, 2026
Upstream dev PR NVIDIA#4621 ("Fix single grouped weight when enabling MXFP8 primary weight") fixes a numerical bug in the dist-optimizer's quantized-param shard path on the single-grouped- weight storage. That fix is still on dev and overlaps with code this PR already touches. The buggy combo isn't exercised by this PR's tests (bf16, no fp8), so this PR can ship without the fix — but users who set moe_single_grouped_weight=True outside the mxfp8 recipe would silently train on a broken numerical path. Add a __post_init__ guard that raises early with a clear pointer to NVIDIA#4621. Scoped to moe_single_grouped_weight only, not moe_single_grouped_bias: the fix in NVIDIA#4621 touches only the quantized-tensor code paths in distrib_optimizer.py and param_and_grad_buffer.py, and biases don't enter those paths (they aren't quantized in mxfp8 training). Upstream NVIDIA#4621's validation gates both for safety; ours is tighter because the actual bug is weight-only. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
3 tasks
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 does this PR do ?
Issue tracking
For PRs from open-source community contributors:
Linked issue:
Contribution process
Pre-checks
Code review
Feel free to message or comment the @mcore-oncall to help accelerate your merge into main. The less complex your PR is, the faster it will be approved and merged!
All PRs start as draft. If you open a non-draft PR, it will be automatically converted to draft.
Step 1: Mark PR as "Ready for Review"
.github/CODEOWNERS.Final Review might get declined if these requirements are not fulfilled.
Step 2: Final Review
For PRs that change
megatron/core, once all expert reviewers have approved, theFinal Reviewlabel is applied automatically and final reviewers are assigned.For PRs outside
megatron/core, this step is skipped.Step 3: Approved
Once all required reviewers have approved, the
Approvedlabel is applied automatically.Merge
Any member of mcore-engineers will be able to merge your PR.
For MRs into `dev` branch
The proposed review process for `dev` branch is under active discussion.MRs are mergable after one approval by either
eharper@nvidia.comorzijiey@nvidia.com.