[BugFix] bugfix for Flash Attention MLA with full cuda graph IMA following pr-25490#27128
Merged
LucasWilkinson merged 5 commits intovllm-project:mainfrom Oct 22, 2025
Merged
Conversation
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Code Review
This pull request aims to fix an illegal memory access error in Flash Attention MLA with full CUDA graph support by ensuring get_scheduler_metadata and FlashAttnMLAMetadata receive the same max_num_splits value. The changes correctly refactor the logic to calculate max_num_splits before it's used. However, I've identified a remaining logic issue where a similar discrepancy can occur when vllm_is_batch_invariant() is true, which could lead to the same bug under different conditions. I've provided a suggestion to fully resolve this.
…25490 Signed-off-by: qqma <qqma@amazon.com>
Daisy-Ma-coder
force-pushed
the
flash_attn_mla_ima_fix
branch
from
257c4e8 to
a8cdaba
Compare
added 2 commits
October 17, 2025 15:36
Signed-off-by: qqma <qqma@amazon.com>
Signed-off-by: qqma <qqma@amazon.com>
github-actions
bot
added
the
ready
Contributor
Author
|
seems like the failed tests are unrelated, is it fine to still merge it? |
usberkeley
pushed a commit
to usberkeley/vllm
that referenced
this pull request
Oct 23, 2025
…owing pr-25490 (vllm-project#27128) Signed-off-by: qqma <qqma@amazon.com> Co-authored-by: qqma <qqma@amazon.com>
845473182
pushed a commit
to raindaywhu/vllm
that referenced
this pull request
Oct 24, 2025
…o step_forward * 'step_forward' of https://github.com/raindaywhu/vllm: (148 commits) [Model] Add MoE support for NemotronH (vllm-project#25863) [Metrics] [KVConnector] Add connector prefix cache hit rate stats (vllm-project#26245) [CI] Reorganize entrypoints tests (vllm-project#27403) add SLA information into comparison graph for vLLM Benchmark Suite (vllm-project#25525) [CI/Build] Fix AMD CI: test_cpu_gpu.py (vllm-project#27388) [Bugfix] Fix args settings for guided decoding args (vllm-project#27375) [CI/Build] Fix Prithvi plugin test (vllm-project#27393) [Chore] Remove duplicate `has_` functions in vllm.utils (vllm-project#27372) [Model] Add num_cached_tokens for PoolingRequestOutput (vllm-project#27378) [V1][spec decode] return logprobs for spec decoding (vllm-project#26060) [CORE] Support Prefix Caching with Prompt Embeds (vllm-project#27219) [Bugfix][Core] running queue index leakage exception (vllm-project#26754) [Bugfix] Fix incorrect kv cache metrics in grafana.json (vllm-project#27133) [Bugfix] Fix SLA tuner initialization (vllm-project#27355) [Bugfix] Fix deepseek-ocr multi-image inference and add `merge_by_field_config=True` with tensor schema support (vllm-project#27361) [MLA] Bump FlashMLA (vllm-project#27354) [Chore] Separate out system utilities from vllm.utils (vllm-project#27201) [BugFix] bugfix for Flash Attention MLA with full cuda graph IMA following pr-25490 (vllm-project#27128) [Feature] publisher default set zmq in kv_event config (vllm-project#26915) [Prefix Cache] Use LoRA name for consistent KV-cache block hashing (vllm-project#27211) ...
0xrushi
pushed a commit
to 0xrushi/vllm
that referenced
this pull request
Oct 26, 2025
…owing pr-25490 (vllm-project#27128) Signed-off-by: qqma <qqma@amazon.com> Co-authored-by: qqma <qqma@amazon.com> Signed-off-by: 0xrushi <6279035+0xrushi@users.noreply.github.com>
0xrushi
pushed a commit
to 0xrushi/vllm
that referenced
this pull request
Oct 26, 2025
…owing pr-25490 (vllm-project#27128) Signed-off-by: qqma <qqma@amazon.com> Co-authored-by: qqma <qqma@amazon.com> Signed-off-by: 0xrushi <6279035+0xrushi@users.noreply.github.com>
ilmarkov
pushed a commit
to neuralmagic/vllm
that referenced
this pull request
Nov 7, 2025
…owing pr-25490 (vllm-project#27128) Signed-off-by: qqma <qqma@amazon.com> Co-authored-by: qqma <qqma@amazon.com>
rtourgeman
pushed a commit
to rtourgeman/vllm
that referenced
this pull request
Nov 10, 2025
…owing pr-25490 (vllm-project#27128) Signed-off-by: qqma <qqma@amazon.com> Co-authored-by: qqma <qqma@amazon.com>
devpatelio
pushed a commit
to SumanthRH/vllm
that referenced
this pull request
Nov 29, 2025
…owing pr-25490 (vllm-project#27128) Signed-off-by: qqma <qqma@amazon.com> Co-authored-by: qqma <qqma@amazon.com>
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
2 participants
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.
Bugfix for Flash Attention MLA with full cuda graph IMA following pr-25490
Run into illegal memory access error when testing some prompts with prefix caching enabled on Flash Attention MLA backend
Log below is generated with CUDA_LAUNCH_BLOCKING=1 which indicating it's flash attn mla.
And realized it's the same root cause as #25490 where
get_scheduler_metadatawas being called with a differentmax_num_splitsthan what was being passed toFlashAttnMLAMetadata.