Skip to content

Disable SWA related decode kernel tests since they are not supported #4911

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
wants to merge 1 commit into from
Closed

Disable SWA related decode kernel tests since they are not supported #4911

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
12 changes: 4 additions & 8 deletions fbgemm_gpu/experimental/gen_ai/test/attention/blackwell_fmha_test.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -439,23 +439,17 @@ def _execute_cutlass_blackwell_attn_varlen(
seqlen_k,
batch_size,
is_mqa,
window_size,
sm_scale,
)
for seqlen_k in [64, 128, 256, 1024]
for batch_size in [1, 2]
for is_mqa in [True]
for window_size in [(-1, -1), (0, 0), (0, 128), (128, 0), (1024, 0)]
for sm_scale in [None, 1.0 / 128]
]
)
def test_decode(
self,
seqlen_k: int,
batch_size: int,
is_mqa: bool,
window_size: tuple[int, int],
sm_scale: Optional[float],
q_heads: int = 8,
dtype: torch.dtype = torch.float8_e4m3fn,
) -> None:
Expand All @@ -473,10 +467,12 @@ def test_decode(
head_dim=128,
dtype=dtype,
causal=causal,
window_size=window_size,
# Decode kernel does not support sliding window attention yet
window_size=(-1, -1),
fwd_only=True,
deterministic=False,
sm_scale=sm_scale,
# Decode kernel does not support sm_scale
sm_scale=None,
)

@skip_cuda_lt_sm100
Expand Down
Loading