A high-throughput and memory-efficient inference and serving engine for LLMs
[BUGFIX][XPU] Update fa interface for compatibility (#45394)
Signed-off-by: zhenwei-intel <zhenwei.liu@intel.com> Signed-off-by: Kunshang Ji <kunshang.ji@intel.com> Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>
liuzhenwei committed
Jun 12, 2026 at 19:45 UTC
8af550b39997d15808802cf8527a9cf6182c406b
1 file changed
+2
vllm/_xpu_ops.py
+2
@@ -784,8 +784,10 @@ class xpu_ops:
784
return_softmax_lse: bool | None = False,
785
s_aux: torch.Tensor | None = None,
786
return_attn_probs: bool | None = False,
787
+ dynamic_causal: torch.Tensor | None = None,
788
mask_mod: Callable | None = None,
789
aux_tensors: list | None = None,
790
+ **kwargs,
791
):
792
assert cu_seqlens_k is not None or seqused_k is not None, (
793
"cu_seqlens_k or seqused_k must be provided"