Fix biased_grouped_topk_cpu (#9420)

2025-08-21 10:18:48 +08:00
parent af1973b871
commit c674bf9c6b
1 changed files with 2 additions and 0 deletions
--- a/python/sglang/srt/layers/moe/topk.py
+++ b/python/sglang/srt/layers/moe/topk.py
@@ -709,8 +709,10 @@ def biased_grouped_topk_cpu(
    routed_scaling_factor: Optional[float] = None,
    num_token_non_padded: Optional[torch.Tensor] = None,
    expert_location_dispatch_info: Optional[ExpertLocationDispatchInfo] = None,
+    apply_routed_scaling_factor_on_output: Optional[bool] = False,
 ):
    assert expert_location_dispatch_info is None
+    assert not apply_routed_scaling_factor_on_output, "Not implemented"
    return torch.ops.sgl_kernel.biased_grouped_topk_cpu(
        hidden_states,
        gating_output,