forked from Karylab-cklius/vllm
[XPU] WA of topk_softplus_sqrt arg mismatch on XPU (#49408)
Signed-off-by: xiaolong <xiaolong.guo@intel.com>
This commit is contained in:
@@ -2447,6 +2447,21 @@ def topk_hash_softplus_sqrt(
|
||||
hash_indices_table: torch.Tensor | None = None,
|
||||
is_padding: torch.Tensor | None = None,
|
||||
) -> None:
|
||||
if current_platform.is_xpu():
|
||||
# TODO: Remove after vllm-xpu-kernels supports is_padding.
|
||||
torch.ops._moe_C.topk_softplus_sqrt(
|
||||
topk_weights,
|
||||
topk_indices,
|
||||
token_expert_indices,
|
||||
gating_output,
|
||||
renormalize,
|
||||
routed_scaling_factor,
|
||||
e_score_correction_bias,
|
||||
input_tokens,
|
||||
hash_indices_table,
|
||||
)
|
||||
|
||||
return
|
||||
torch.ops._moe_C.topk_softplus_sqrt(
|
||||
topk_weights,
|
||||
topk_indices,
|
||||
|
||||
Reference in New Issue
Block a user