Skip to content

fix: only enable flash_attn test on sm80 sm90 #7289

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
merged 1 commit into from
Jun 17, 2025
Merged

fix: only enable flash_attn test on sm80 sm90 #7289

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
8 changes: 6 additions & 2 deletions sgl-kernel/tests/test_flash_attention.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,7 +13,7 @@

def is_hopper():
# Only Hopper supports different V headdim
return torch.cuda.get_device_properties(0).major >= 9
return torch.cuda.get_device_properties(0).major == 9


def is_fa3_supported(device=None) -> bool:
Expand Down Expand Up @@ -451,7 +451,7 @@ def generate_qkv(

@pytest.mark.skipif(
not is_fa3_supported(),
reason="flash_attn at sgl-kernel is only supported on sm90 and above",
reason="flash_attn at sgl-kernel is only supported on sm90 or sm80",
)
# @pytest.mark.parametrize("dtype", [torch.float16, torch.bfloat16, torch.float8_e4m3fn])
@pytest.mark.parametrize(
Expand Down Expand Up @@ -1009,6 +1009,10 @@ def _generate_block_kvcache(
return k_cache, v_cache, page_table, k_cache_paged, v_cache_paged, num_blocks


@pytest.mark.skipif(
not is_fa3_supported(),
reason="flash_attn at sgl-kernel is only supported on sm90 or sm80",
)
# @pytest.mark.parametrize("dtype", [torch.float16, torch.bfloat16, torch.float8_e4m3fn])
@pytest.mark.parametrize(
"dtype", [torch.bfloat16] + ([torch.float8_e4m3fn] if not DISABLE_FP8 else [])
Expand Down
15 changes: 13 additions & 2 deletions sgl-kernel/tests/test_sparse_flash_attn.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,9 +8,8 @@
convert_vertical_slash_indexes,
convert_vertical_slash_indexes_mergehead,
sparse_attn_func,
sparse_attn_varlen_func,
)
from test_flash_attention import construct_local_mask
from test_flash_attention import construct_local_mask, is_fa3_supported


def ref_attn(
Expand Down Expand Up @@ -172,6 +171,10 @@ def ref_paged_attn(
return torch.cat(outputs, dim=0)


@pytest.mark.skipif(
not is_fa3_supported(),
reason="flash_attn at sgl-kernel is only supported on sm90 or sm80",
)
@pytest.mark.parametrize("batch_size", [1, 2])
@pytest.mark.parametrize(
"seq_lens",
Expand Down Expand Up @@ -257,6 +260,10 @@ def test_sparse_attention(

# sparse attention utils
# origin
@pytest.mark.skipif(
not is_fa3_supported(),
reason="flash_attn at sgl-kernel is only supported on sm90 or sm80",
)
@pytest.mark.parametrize("causal", [True, False])
def test_convert_vertical_slash_indexes(causal):
# Prepare small, hand-checkable inputs
Expand Down Expand Up @@ -311,6 +318,10 @@ def test_convert_vertical_slash_indexes(causal):


# mergehead
@pytest.mark.skipif(
not is_fa3_supported(),
reason="flash_attn at sgl-kernel is only supported on sm90 or sm80",
)
@pytest.mark.parametrize("causal", [True, False])
def test_convert_vertical_slash_indexes_mergehead(causal):
# Prepare small, hand-checkable inputs for mergehead version
Expand Down
Loading