Skip to content

[Flex][SM100] Replay expand fix on sm100 #2209

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

Merged
drisspg merged 1 commit into from
Jan 26, 2026
Merged

[Flex][SM100] Replay expand fix on sm100 #2209

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
6 changes: 5 additions & 1 deletion flash_attn/cute/flash_bwd_sm100.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -412,8 +412,12 @@ def __call__(
assert self.dv_dtype.width == 32, "Must accumulate dV in float precision for GQA"

# Assume all strides are divisible by 128 bits except the last stride
# Skip assume for Python ints (e.g., stride=0 from GQA expand)
new_stride = lambda t: (
*(cute.assume(s, divby=128 // t.element_type.width) for s in t.stride[:-1]),
*(
s if isinstance(s, int) else cute.assume(s, divby=128 // t.element_type.width)
for s in t.stride[:-1]
),
t.stride[-1],
)
(
Expand Down
3 changes: 2 additions & 1 deletion flash_attn/cute/flash_fwd_sm100.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -290,8 +290,9 @@ def __call__(
self.v_dtype = mV.element_type
self.o_dtype = mO.element_type
# Assume all strides are divisible by 128 bits except the last stride
# Skip assume for Python ints (e.g., stride=0 from GQA expand)
new_stride = lambda t: (
*(cute.assume(s, divby=128 // t.element_type.width) for s in t.stride[:-1]),
*(s if isinstance(s, int) else cute.assume(s, divby=128 // t.element_type.width) for s in t.stride[:-1]),
t.stride[-1],
)
mQ, mK, mV, mO = [
Expand Down