From 4a9ffc3ab6f1b3e3bdebb59c39c61b54639e14ac Mon Sep 17 00:00:00 2001 From: Vedant V Jhaveri Date: Mon, 30 Mar 2026 12:50:49 -0700 Subject: [PATCH] fix nemotron capture for non attention layers (#21436) --- python/sglang/srt/model_executor/model_runner.py | 2 ++ 1 file changed, 2 insertions(+) diff --git a/python/sglang/srt/model_executor/model_runner.py b/python/sglang/srt/model_executor/model_runner.py index 1a9579f96..6c1c7eceb 100644 --- a/python/sglang/srt/model_executor/model_runner.py +++ b/python/sglang/srt/model_executor/model_runner.py @@ -2424,6 +2424,8 @@ class ModelRunner(ModelRunnerKVCacheMixin): 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