Skip to content

Update THD sink attention logic for cudnn >=9.18.0 #2561

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

Closed
cuichenx wants to merge 3 commits into from
Closed

Update THD sink attention logic for cudnn >=9.18.0 #2561

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
fbb0702
Update THD sink attention logic for newer cudnn versions
cuichenx Jan 5, 2026
01848c0
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Jan 5, 2026
442699c
Merge branch 'main' into patch-1
cyanguwa Jan 6, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 6 additions & 4 deletions transformer_engine/pytorch/attention/dot_product_attention/utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -716,10 +716,12 @@ def _is_fa3_supported(num_heads, num_gqa_groups, head_dim_qk, head_dim_v, qkv_dt
)
use_unfused_attention = False
if qkv_format == "thd":
logger.debug(
"Disabling FusedAttention for softmax_type = %s and qkv_format = thd", softmax_type
)
use_fused_attention = False
if cudnn_version < (9, 18, 0):
logger.debug(
"Disabling FusedAttention for softmax_type = %s and qkv_format = thd",
softmax_type,
)
use_fused_attention = False
logger.debug(
"Disabling UnfusedDotProductAttention for softmax_type = %s and qkv_format = thd",
softmax_type,
Expand Down
Loading