Skip to content

[enflame] fix stride mismatch int flash_attention_forward test #1100

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

Open
xingxing588 wants to merge 1 commit into
base: master
Choose a base branch
from
Open

[enflame] fix stride mismatch int flash_attention_forward test #1100

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
2 changes: 1 addition & 1 deletion src/flag_gems/ops/attention.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1065,7 +1065,7 @@ def flash_attention_forward(
else:
non_null_window_right = -1

out = torch.empty_like(query)
out = torch.empty(query.shape, device=query.device, dtype=query.dtype)
Copy link
Collaborator

@0x45f 0x45f Nov 26, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

why using torch.empty instead of torch.empty_like

if cumulative_sequence_length_q is not None:
out, q, k, v, lse, philox_seed, philox_offset, p = mha_varlan_fwd(
query,
Expand Down
3 changes: 2 additions & 1 deletion tests/test_attention_ops.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -192,6 +192,7 @@ def attention_ref(
else:
attention_drop = attention
output = torch.einsum("bhts,bshd->bthd", attention_drop, v * dropout_scaling)
output = output.contiguous()
if query_padding_mask is not None:
output.masked_fill_((~query_padding_mask)[:, :, None, None], 0.0)
return output.to(dtype=dtype_og), attention.to(dtype=dtype_og)
Expand Down Expand Up @@ -662,7 +663,7 @@ def test_flash_fwd_gqa_alibi_softcap(
@pytest.mark.flash_attention_forward
@pytest.mark.parametrize(
["batch", "num_head", "num_head_k", "q_seq_len", "kv_seq_len"],
[(1, 4, 1, 1, 1024), (4, 4, 4, 1, 519)],
[(1, 4, 1, 1, 1024), (4, 4, 4, 1, 519), (1, 4, 1, 2, 16)],
)
@pytest.mark.parametrize("head_size", [128, 192])
@pytest.mark.parametrize("is_causal", [False, True])
Expand Down
Loading