[Intel GPU] import flash_attn functions from sgl_kernel only (#22438)
This commit is contained in:
@@ -20,11 +20,7 @@ if TYPE_CHECKING:
|
||||
from sglang.srt.model_executor.model_runner import ModelRunner
|
||||
|
||||
from sgl_kernel import merge_state_v2
|
||||
|
||||
from sglang.jit_kernel.flash_attention import (
|
||||
flash_attn_varlen_func,
|
||||
flash_attn_with_kvcache,
|
||||
)
|
||||
from sgl_kernel.flash_attn import flash_attn_varlen_func, flash_attn_with_kvcache
|
||||
|
||||
|
||||
class XPUAttentionBackend(AttentionBackend):
|
||||
|
||||
Reference in New Issue
Block a user