diff --git a/python/sglang/srt/layers/moe/topk.py b/python/sglang/srt/layers/moe/topk.py index 1b23554f7..a355707be 100644 --- a/python/sglang/srt/layers/moe/topk.py +++ b/python/sglang/srt/layers/moe/topk.py @@ -652,6 +652,8 @@ def fused_topk_torch_native( return gating_output.softmax(dim=-1) elif scoring_func == "sigmoid": return gating_output.sigmoid() + elif scoring_func == "sqrtsoftplus": + return F.softplus(gating_output).sqrt() else: raise ValueError(f"Invalid scoring function: {scoring_func}")