Support softcap and softmax_precision in Attention(CUDA) #27714
titaiwangms
merged
6 commits into
main
from
titaiwang/support_softcap_softmax_precision
Support softcap and softmax_precision in CUDA Attention operator
8dd5424e
Enable CUDA softcap tests and update backend filters
cb18f61e
Fix review findings: stale comments and softcap guard consistency
f76338f6
copilot-pull-request-reviewer
commented
on 2026-03-17
copilot-pull-request-reviewer
commented
on 2026-03-17
Fix inaccurate softmax_precision comment
9f297839
Add softmax_precision validation, fix error message, add test
2562d302
copilot-pull-request-reviewer
commented
on 2026-03-17
Fix CI filters and softmax_precision DOUBLE validation
0e95ff19
tianleiwu
approved these changes
on 2026-03-19
Assignees
No one assigned
Labels
ep:CUDA
Milestone
No milestone
Login to write a write a comment.
Login via GitHub