fix: cherry-pick combined projection fixes (#1324, #1357) into r0.2.1#1388
Open
fix: cherry-pick combined projection fixes (#1324, #1357) into r0.2.1#1388
Conversation
* fix: compbined projection Signed-off-by: Zhiyu Li <zhiyul@NVIDIA.com> * lint Signed-off-by: Zhiyu Li <zhiyul@NVIDIA.com> * address comment Signed-off-by: Zhiyu Li <zhiyul@NVIDIA.com> * add tp output parity tests Signed-off-by: Zhiyu Li <zhiyul@NVIDIA.com> --------- Signed-off-by: Zhiyu Li <zhiyul@NVIDIA.com> Signed-off-by: HuiyingLi <willwin.lee@gmail.com> Signed-off-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…ort (#1357) * fix: FSDP pre-shard combined projections on dim 1 for Qwen2.5-7B support Signed-off-by: Zhiyu Li <zhiyul@NVIDIA.com> * revert recipe change Signed-off-by: Zhiyu Li <zhiyul@NVIDIA.com> * lint Signed-off-by: Zhiyu Li <zhiyul@NVIDIA.com> --------- Signed-off-by: Zhiyu Li <zhiyul@NVIDIA.com> Signed-off-by: HuiyingLi <willwin.lee@gmail.com> Signed-off-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Contributor
Author
|
Hi @ZhiyuLi-Nvidia , I cherrypicked these two commits. However the TP2 cases still see high loss. Do you know anything else missing? Thanks! |
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
These fixes correct the QKV and gate_up weight layout from naive concatenation to KV-head-grouped interleaving (QKV) and row interleaving (gate_up), ensuring each TP rank receives complete head groups under ColwiseParallel sharding.
Test plan
Status: NOT YET VERIFIED — needs to be validated before merge.
Signed-off-by: HuiyingLi willwin.lee@gmail.com
Signed-off-by: Claude Opus 4.6 (1M context) noreply@anthropic.com