A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 05:30 UTC
[XPU]Fix `flash_attn_varlen_func` interface on xpu (#22350)
Signed-off-by: Kunshang Ji <kunshang.ji@intel.com>
Kunshang Ji committed
Aug 7, 2025 at 10:28 UTC
6b47ef24de3d3b4f551aca0bc21b9f16f3d21b6a
1 file changed
+1
vllm/_ipex_ops.py
+1
@@ -271,6 +271,7 @@ class ipex_ops:
271
k_descale=None,
272
v_descale=None,
273
num_splits=0,
274
+ s_aux: Optional[torch.Tensor] = None,
275
):
276
if cu_seqlens_k is None:
277
# cu_seqlens_k is not used in ipex kernel.