diff --git a/python/sglang/srt/model_executor/model_runner.py b/python/sglang/srt/model_executor/model_runner.py index 1a9579f96791..6c1c7eceb887 100644 --- a/python/sglang/srt/model_executor/model_runner.py +++ b/python/sglang/srt/model_executor/model_runner.py @@ -2424,6 +2424,8 @@ def init_piecewise_cuda_graphs(self): if attn_layer is not None: self.attention_layers.append(attn_layer) + elif hasattr(layer, "mixer"): + self.attention_layers.append(None) moe_block = None moe_fusion = None