Skip to content

fix: Fix FMHA-based MLA in the generation phase and add MLA unit test #3863

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.

Sign up for GitHub

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

Jump to bottom
Merged
merged 7 commits into from
Apr 29, 2025
Merged

fix: Fix FMHA-based MLA in the generation phase and add MLA unit test #3863

merged 7 commits into from
Apr 29, 2025

Conversation

@jinyangyuan-nvidia
Copy link
Collaborator

@jinyangyuan-nvidia jinyangyuan-nvidia commented Apr 25, 2025 •
edited
Loading

@PerkzZheng fixes a bug of FMHA-based MLA in the generation phase (cubins are generated using commit 56491367f0a36a4f7bba68d79b4c04f369dd5a91).
@peaceh-nv modifies the code for SM120.
@jinyangyuan-nvidia adds MLA unit test.
This PR also cherry-picks the modifications in #3675.

@jinyangyuan-nvidia jinyangyuan-nvidia self-assigned this Apr 25, 2025
@jinyangyuan-nvidia
Copy link
Collaborator Author

/bot run

@tensorrt-cicd
Copy link
Collaborator

PR_Github #3382 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #3382 [ run ] completed with state SUCCESS
/LLM/main/L0_MergeRequest_PR pipeline #2367 completed with status: 'FAILURE'

@jinyangyuan-nvidia jinyangyuan-nvidia force-pushed the dev/chore_mla_ut branch 2 times, most recently from 8befdc7 to bd144b7 Compare April 25, 2025 14:46
@jinyangyuan-nvidia
Copy link
Collaborator Author

/bot run

@tensorrt-cicd
Copy link
Collaborator

PR_Github #3399 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #3399 [ run ] completed with state FAILURE
/LLM/main/L0_MergeRequest_PR pipeline #2383 completed with status: 'FAILURE'

@jinyangyuan-nvidia
Copy link
Collaborator Author

/bot run

@tensorrt-cicd
Copy link
Collaborator

PR_Github #3510 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #3510 [ run ] completed with state SUCCESS
/LLM/main/L0_MergeRequest_PR pipeline #2472 completed with status: 'FAILURE'

@jinyangyuan-nvidia
Copy link
Collaborator Author

/bot run

@tensorrt-cicd
Copy link
Collaborator

PR_Github #3532 [ run ] triggered by Bot

@jinyangyuan-nvidia
Copy link
Collaborator Author

/bot kill

@jinyangyuan-nvidia
Copy link
Collaborator Author

/bot run

@tensorrt-cicd
Copy link
Collaborator

PR_Github #3548 [ kill ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #3549 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #3548 [ kill ] completed with state ABORTED

@tensorrt-cicd
Copy link
Collaborator

PR_Github #3532 [ run ] completed with state ABORTED

@jinyangyuan-nvidia
Copy link
Collaborator Author

/bot kill

@tensorrt-cicd
Copy link
Collaborator

PR_Github #3560 [ kill ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #3549 [ run ] completed with state ABORTED

@tensorrt-cicd
Copy link
Collaborator

PR_Github #3560 [ kill ] completed with state SUCCESS
Successfully killed previous jobs for commit 34c549d

bobboli
bobboli reviewed Apr 28, 2025
View reviewed changes
@jinyangyuan-nvidia
Copy link
Collaborator Author

/bot run

@tensorrt-cicd
Copy link
Collaborator

PR_Github #3600 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #3600 [ run ] completed with state SUCCESS
/LLM/main/L0_MergeRequest_PR pipeline #2539 completed with status: 'SUCCESS'

@jinyangyuan-nvidia jinyangyuan-nvidia enabled auto-merge (squash) April 28, 2025 15:36
PerkzZheng and others added 7 commits April 29, 2025 09:06
@PerkzZheng @jinyangyuan-nvidia
update cubins
b0df5c4 
Signed-off-by: Perkz Zheng <[email protected]>
@PerkzZheng @jinyangyuan-nvidia
fix kernel selection
fa4d61f
@peaceh-nv @jinyangyuan-nvidia
Add SM120 fix
534a2d4 
Signed-off-by: peaceh <[email protected]>
@DylanChen-NV @jinyangyuan-nvidia
fix FP8 kv accuracy
28eaba7 
Signed-off-by: Dylan Chen <[email protected]>
@DylanChen-NV @jinyangyuan-nvidia
update doc
489a77e 
Signed-off-by: Dylan Chen <[email protected]>
@jinyangyuan-nvidia
add MLA unit test
169b15f 
Signed-off-by: Jinyang Yuan <[email protected]>
@jinyangyuan-nvidia
update fmha_v2 cubin files
a9cec22 
Signed-off-by: Jinyang Yuan <[email protected]>
@niukuo niukuo merged commit dafc28f into NVIDIA:main Apr 29, 2025
1 check passed
@jinyangyuan-nvidia jinyangyuan-nvidia deleted the dev/chore_mla_ut branch April 29, 2025 01:27
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@bobboli bobboli bobboli approved these changes

@DylanChen-NV DylanChen-NV DylanChen-NV approved these changes

@EmmaQiaoCh EmmaQiaoCh Awaiting requested review from EmmaQiaoCh

@PerkzZheng PerkzZheng Awaiting requested review from PerkzZheng

@peaceh-nv peaceh-nv Awaiting requested review from peaceh-nv

Assignees

@jinyangyuan-nvidia jinyangyuan-nvidia

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

7 participants

@jinyangyuan-nvidia @tensorrt-cicd @bobboli @DylanChen-NV @niukuo @PerkzZheng @peaceh-nv