Skip to content
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

[Kernel] Pipe attn_logits_soft_cap through paged attention TPU kernels #12482

Merged
merged 1 commit into from
Jan 28, 2025
Merged

[Kernel] Pipe attn_logits_soft_cap through paged attention TPU kernels #12482

merged 1 commit into from
Jan 28, 2025

Conversation

fenghuizhang
Copy link
Contributor

@fenghuizhang fenghuizhang commented Jan 27, 2025 •
edited by github-actions bot
Loading

Pipe attn_logits_soft_cap through paged_attention, this will unblock some of our models' adoption.

Note that the changed code currently doesn't have unit tests. I will add one later.

This is the same as #12294. We created a new PR as I messed up the sign off and previous commit chain.

Thanks,

@github-actions GitHub Actions
Copy link

👋 Hi! Thank you for contributing to the vLLM project.
Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can do one of these:

  • Add ready label to the PR
  • Enable auto-merge.

🚀

@mergify mergify bot added the ci/build label Jan 27, 2025
@miladm
Copy link
Collaborator

miladm commented Jan 27, 2025

cc @vanbasten23

alexm-redhat
alexm-redhat approved these changes Jan 28, 2025
View reviewed changes
Copy link
Collaborator

@alexm-redhat alexm-redhat left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@fenghuizhang thanks for doing this! LGTM

vllm/attention/backends/pallas.py
context_lens,
block_tables,
pages_per_compute_block,
megacore_mode=megacore_mode,
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

megacore_mode is checked internally for None?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yep, we fixed it on the ptxla side. No workaround needed here.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

cool

@mgoin mgoin added tpu Related to Google TPUs ready ONLY add when PR is ready to merge/full CI is needed labels Jan 28, 2025
@alexm-redhat alexm-redhat enabled auto-merge (squash) January 28, 2025 22:35
@alexm-redhat alexm-redhat merged commit 80fcc3e into vllm-project:main Jan 28, 2025
63 checks passed
rasmith pushed a commit to rasmith/vllm that referenced this pull request Jan 30, 2025
@fenghuizhang @rasmith
[Kernel] Pipe attn_logits_soft_cap through paged attention TPU kernels (
d76a704 
vllm-project#12482)

Signed-off-by: Fenghui Zhang <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@alexm-redhat alexm-redhat alexm-redhat approved these changes

@mgoin mgoin mgoin approved these changes

Assignees
No one assigned
Labels
ci/build ready ONLY add when PR is ready to merge/full CI is needed tpu Related to Google TPUs
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@fenghuizhang @miladm @mgoin @alexm-redhat