[Feature] Flash Attention to Support Attention Mask (#760)

Signed-off-by: Didan Deng <33117903+wtomin@users.noreply.github.com>
This commit is contained in:
Didan Deng authored and GitHub committed 2026-01-16 11:01:11 +08:00
1 parent 35f994e255
commit 43ec7f701b
6 files changed
+566 -12

No files matched your search

@@ -23,7 +23,7 @@ from vllm_omni.diffusion.attention.backends.abstract import (
AttentionMetadata,
)
from vllm_omni.diffusion.attention.layer import Attention
from vllm_omni.diffusion.attention.selector import get_attn_backend
from vllm_omni.diffusion.attention.selector import _BACKENDS_SUPPORT_ATTENTION_MASK, get_attn_backend
from vllm_omni.diffusion.cache.base import CachedTransformer
from vllm_omni.diffusion.data import OmniDiffusionConfig
from vllm_omni.diffusion.distributed.parallel_state import (
@@ -790,8 +790,8 @@ class QwenImageTransformer2DModel(CachedTransformer):
sp_size = get_sequence_parallel_world_size()
if seq_len % sp_size != 0:
# flash_attn, ring_attn, sage_attn do not support attention_mask
if get_attn_backend(-1).get_name() != "SDPA" and get_attn_backend(-1).get_name() != "ASCEND":
# ring_attn, sage_attn do not support attention_mask
if get_attn_backend(-1).get_name() not in _BACKENDS_SUPPORT_ATTENTION_MASK:
raise ValueError(
f"When generating image shape that the sequence length is NOT divisible by sp_size={sp_size},"
f"cannot use {get_attn_backend(-1).get_name()} which does not support attention_mask."