[CPU] Support GQA for flash attention#157893
Closed
Valentine233 wants to merge 4 commits intomainfrom
Closed
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/157893
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 289b47c with merge base b146ca7 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
mingfeima
approved these changes
Jul 10, 2025
Collaborator
mingfeima
left a comment
There was a problem hiding this comment.
genernally OK, just simplify the test cases a little bit to remove the duplicated code.
| @parametrize("dtype", [torch.float64, torch.float32, torch.bfloat16, torch.float16]) | ||
| @parametrize("n_heads", [[65, 5], [16, 4], [27, 1], [5, 1]]) | ||
| @parametrize("train", [False, True]) | ||
| def test_scaled_dot_product_fused_attention_gqa_vs_math_cpu( |
Collaborator
There was a problem hiding this comment.
combine this one with test_scaled_dot_product_fused_attention_mask_vs_math_cpu to remove duplicated code.
### impls
def test_sdpa_vs_math_cpu_helper(...)
def test_scaled_dot_product_fused_attention_mask_vs_math_cpu()
test_sdpa_vs_math_cpu_helper(...)
def test_scaled_dot_product_fused_attention_gqa_vs_math_cpu()
test_sdpa_vs_math_cpu_helper(...)
Collaborator
Author
There was a problem hiding this comment.
Thanks, UT updated.
b7aa830 to
289b47c
Compare
jansel
approved these changes
Jul 12, 2025
Collaborator
Author
|
@pytorchbot merge |
Collaborator
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
fduwjj
added a commit
that referenced
this pull request
Jul 14, 2025
As many models require GQA, we support it in flash attention for CPU path. Approved by: https://github.com/mingfeima, https://github.com/jansel [ghstack-poisoned]
facebook-github-bot
pushed a commit
that referenced
this pull request
Jul 18, 2025
Summary: For `scaled_dot_product_attention(..., enable_gqa=True)`: - the Math backend passes the flag through, performing the extra [KV broadcast](https://github.com/pytorch/pytorch/blob/6e07d6a0ff386d99d8c2f1d25978b0683988a4cb/aten/src/ATen/native/transformers/attention.cpp#L902) if set to True - the Flash backend has no flag, and relies on correct indexing in the C++ kernel - Export used to default to Math for `enable_gqa=True`, but #157893 landed and enabled Flash. At the same time, there's an export-only [decomp](https://github.com/pytorch/pytorch/blob/6e07d6a0ff386d99d8c2f1d25978b0683988a4cb/torch/_decomp/decompositions.py#L4968) redirecting flash -> math, calling with `enable_gqa` unset, because that info isn't available. This led to https://fb.workplace.com/groups/1028545332188949/posts/1264609398582540 crashing, calling the Math non-GQA variant, with GQA inputs. This assumes GQA for seqlen mismatches in the export decomp, setting `enable_gqa = <q seqlen> != <kv seqlen>`, relying on prior backend checks to raise on invalid input shapes. Test Plan: test_export Rollback Plan: Differential Revision: D78524147
facebook-github-bot
pushed a commit
that referenced
this pull request
Jul 23, 2025
Summary: For `scaled_dot_product_attention(..., enable_gqa=True)`: - the Math backend passes the flag through, performing the extra [KV broadcast](https://github.com/pytorch/pytorch/blob/6e07d6a0ff386d99d8c2f1d25978b0683988a4cb/aten/src/ATen/native/transformers/attention.cpp#L902) if set to True - the Flash backend has no flag, and relies on correct indexing in the C++ kernel - Export used to default to Math for `enable_gqa=True`, but #157893 landed and enabled Flash. At the same time, there's an export-only [decomp](https://github.com/pytorch/pytorch/blob/6e07d6a0ff386d99d8c2f1d25978b0683988a4cb/torch/_decomp/decompositions.py#L4968) redirecting flash -> math, calling with `enable_gqa` unset, because that info isn't available. This led to https://fb.workplace.com/groups/1028545332188949/posts/1264609398582540 crashing, calling the Math non-GQA variant, with GQA inputs. This assumes GQA for seqlen mismatches in the export decomp, setting `enable_gqa = <q seqlen> != <kv seqlen>`, relying on prior backend checks to raise on invalid input shapes. Test Plan: test_export Rollback Plan: Reviewed By: angelayi Differential Revision: D78524147
pytorchmergebot
pushed a commit
that referenced
this pull request
Jul 24, 2025
Differential Revision: D78524147 For `scaled_dot_product_attention(..., enable_gqa=True)`: - the Math backend passes the flag through, performing the extra [KV broadcast](https://github.com/pytorch/pytorch/blob/6e07d6a0ff386d99d8c2f1d25978b0683988a4cb/aten/src/ATen/native/transformers/attention.cpp#L902) if set to True - the Flash backend has no flag, and relies on correct indexing in the C++ kernel - Export used to default to Math for `enable_gqa=True`, but #157893 landed and enabled Flash. At the same time, there's an export-only [decomp](https://github.com/pytorch/pytorch/blob/6e07d6a0ff386d99d8c2f1d25978b0683988a4cb/torch/_decomp/decompositions.py#L4968) redirecting flash -> math, calling with `enable_gqa` unset, because that info isn't available. This led to https://fb.workplace.com/groups/1028545332188949/posts/1264609398582540 crashing, calling the Math non-GQA variant, with GQA inputs. This assumes GQA for seqlen mismatches in the export decomp, setting `enable_gqa = <q seqlen> != <kv seqlen>`, relying on prior backend checks to raise on invalid input shapes. Pull Request resolved: #158604 Approved by: https://github.com/angelayi, https://github.com/drisspg
yangw-dev
pushed a commit
that referenced
this pull request
Aug 1, 2025
Differential Revision: D78524147 For `scaled_dot_product_attention(..., enable_gqa=True)`: - the Math backend passes the flag through, performing the extra [KV broadcast](https://github.com/pytorch/pytorch/blob/6e07d6a0ff386d99d8c2f1d25978b0683988a4cb/aten/src/ATen/native/transformers/attention.cpp#L902) if set to True - the Flash backend has no flag, and relies on correct indexing in the C++ kernel - Export used to default to Math for `enable_gqa=True`, but #157893 landed and enabled Flash. At the same time, there's an export-only [decomp](https://github.com/pytorch/pytorch/blob/6e07d6a0ff386d99d8c2f1d25978b0683988a4cb/torch/_decomp/decompositions.py#L4968) redirecting flash -> math, calling with `enable_gqa` unset, because that info isn't available. This led to https://fb.workplace.com/groups/1028545332188949/posts/1264609398582540 crashing, calling the Math non-GQA variant, with GQA inputs. This assumes GQA for seqlen mismatches in the export decomp, setting `enable_gqa = <q seqlen> != <kv seqlen>`, relying on prior backend checks to raise on invalid input shapes. Pull Request resolved: #158604 Approved by: https://github.com/angelayi, https://github.com/drisspg
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
5 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.
As many models require GQA, we support it in flash attention for CPU path.
cc @jgong5 @mingfeima @XiaobingSuper @sanchitintel @ashokei @jingxu10 @jerryzh168