[SDPA] Fix alignment check for efficient_attention #90413
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/90413
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 FailuresAs of commit 1c7bc96: The following jobs have failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
drisspg
changed the title
aten/src/ATen/native/transformers/cuda/mem_eff_attention/gemm_kernel_utils.h
Show resolved
Hide resolved
drisspg
force-pushed
the
update_sdp_constraints
branch
from
fca3a48
to
114953f
Compare
|
@pytorchbot merge |
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 |
Merge failedReason: The following mandatory check(s) failed (Rule Dig deeper by viewing the failures on hud Details for Dev Infra teamRaised by workflow job |
drisspg
force-pushed
the
update_sdp_constraints
branch
from
114953f
to
6615758
Compare
drisspg
force-pushed
the
update_sdp_constraints
branch
2 times, most recently
from
c186cf1
to
6f3e76d
Compare
drisspg
force-pushed
the
update_sdp_constraints
branch
from
5978173
to
cf15ace
Compare
|
@pytorchbot merge |
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 |
Merge failedReason: The following mandatory check(s) failed (Rule Dig deeper by viewing the failures on hud Details for Dev Infra teamRaised by workflow job |
|
@pytorchbot rebase |
|
@pytorchbot successfully started a rebase job. Check the current status here |
|
Tried to rebase and push PR #90413, but it was already up to date |
|
@pytorchbot rebase |
|
@pytorchbot successfully started a rebase job. Check the current status here |
|
Tried to rebase and push PR #90413, but it was already up to date |
drisspg
force-pushed
the
update_sdp_constraints
branch
from
cf15ace
to
1c7bc96
Compare
|
@pytorchbot merge -f "unrelated to my changes" |
Merge startedYour change will be merged immediately since you used the force (-f) flag, bypassing any CI checks (ETA: 1-5 minutes). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
drisspg
changed the title
Fixes a bug found using head_dim_size==100 on an a100 gpu. This PR contains stricter guards on the input shape. These constraints are taken from xformers: https://github.com/facebookresearch/xformers/blob/gh/danthe3rd/60/orig/xformers/ops/fmha/cutlass.py#L23