[BugFix] Fix vllm_flash_attn install issues (#17267)

Signed-off-by: Lucas Wilkinson <lwilkinson@neuralmagic.com>
Co-authored-by: Jee Jee Li <pandaleefree@gmail.com>
Co-authored-by: Aaron Pham <contact@aarnphm.xyz>
This commit is contained in:
Lucas Wilkinson
2025-04-27 17:27:56 -07:00
committed by GitHub
co-authored by Jee Jee Li Aaron Pham
parent 20e489eaa1
commit d8bccde686
11 changed files with 28 additions and 284 deletions
+2 -2
View File
@@ -11,11 +11,11 @@ from vllm.attention.backends.abstract import (AttentionBackend, AttentionImpl,
AttentionMetadata, AttentionType,
is_quantized_kv_cache)
from vllm.attention.ops.merge_attn_states import merge_attn_states
from vllm.attention.utils.fa_utils import (flash_attn_supports_fp8,
get_flash_attn_version)
from vllm.logger import init_logger
from vllm.platforms import current_platform
from vllm.utils import cdiv
from vllm.vllm_flash_attn.fa_utils import (flash_attn_supports_fp8,
get_flash_attn_version)
if TYPE_CHECKING:
from vllm.v1.core.sched.output import SchedulerOutput
+1 -1
View File
@@ -197,6 +197,7 @@ from vllm.attention.backends.abstract import (AttentionBackend, AttentionLayer,
MLAAttentionImpl)
from vllm.attention.backends.utils import get_mla_dims
from vllm.attention.ops.merge_attn_states import merge_attn_states
from vllm.attention.utils.fa_utils import get_flash_attn_version
from vllm.logger import init_logger
from vllm.model_executor.layers.linear import (ColumnParallelLinear,
LinearBase, RowParallelLinear,
@@ -204,7 +205,6 @@ from vllm.model_executor.layers.linear import (ColumnParallelLinear,
from vllm.model_executor.layers.rotary_embedding import RotaryEmbedding
from vllm.platforms import current_platform
from vllm.utils import cdiv, round_down
from vllm.vllm_flash_attn.fa_utils import get_flash_attn_version
try:
from vllm.vllm_flash_attn import flash_attn_varlen_func