[plugin][OOT] make OOT MLA work with vLLM 0.17#304
Merged
XiaobingSuper merged 10 commits intoROCm:mainfrom Mar 11, 2026
Merged
Conversation
XiaobingSuper
changed the title
Contributor
There was a problem hiding this comment.
Pull request overview
This PR updates ATOM’s vLLM plugin-mode integration for vLLM 0.17 compatibility and adds MLA-related patches/buffer handling to support a persistent-kernel style decode path.
Changes:
- Update vLLM platform integration to match vLLM 0.17 attention-backend selection signature changes.
- Introduce a dedicated MLA patch module to hook vLLM’s
MLAAttentioninitialization / weight-loading / forward behavior in plugin mode. - Extend MLA plugin-mode attention/metadata flow to provide persistent decode metadata buffers and output-buffer writing for v_up projection.
Reviewed changes
Copilot reviewed 4 out of 4 changed files in this pull request and generated 4 comments.
Show a summary per file
| File | Description |
|---|---|
| atom/plugin/vllm/register.py | Wires in MLA patching during model registration and keeps attention weight-loading patching. |
| atom/plugin/vllm/platform.py | Updates get_attn_backend_cls override signature for vLLM 0.17. |
| atom/plugin/vllm/mla_patch.py | New module implementing vLLM MLAAttention monkey patches for plugin mode. |
| atom/plugin/attention_mla.py | Adjusts MLA plugin-mode v_up projection to write into a provided output buffer; adds do_kv_cache_update stub. |
| atom/plugin/attention.py | Adds persistent MLA metadata buffers and passes them through AttentionMetaData; updates env var access. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
XiaobingSuper
force-pushed
the
xiaobing/vllm_0.17
branch
from
916fffc to
ca4b8c4
Compare
XiaobingSuper
force-pushed
the
xiaobing/vllm_0.17
branch
from
ca4b8c4 to
aa278a2
Compare
XiaobingSuper
force-pushed
the
xiaobing/vllm_0.17
branch
from
aa278a2 to
39c7099
Compare
Contributor
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 4 out of 4 changed files in this pull request and generated 4 comments.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
wuhuikx
reviewed
Mar 11, 2026
wuhuikx
reviewed
Mar 11, 2026
Contributor
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
XiaobingSuper
force-pushed
the
xiaobing/vllm_0.17
branch
from
2470275 to
9de25e1
Compare
zejunchen-zejun
changed the title
wuhuikx
approved these changes
Mar 11, 2026
valarLip
approved these changes
Mar 11, 2026
Jasen2201
pushed a commit
to Jasen2201/ATOM
that referenced
this pull request
Apr 10, 2026
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
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.
This PR is to make plugin mode works for vllm 0.17 and also add the mla persistent kernel.
I test this PR for kimi-2 and deepseek, the accuracy are: