From 94216a9cc45d200af7bb4c946f2011e1f74a8e37 Mon Sep 17 00:00:00 2001
From: fzyzcjy <5236035+fzyzcjy@users.noreply.github.com>
Date: Tue, 25 Nov 2025 03:16:50 +0800
Subject: [PATCH] Fix quantized moe checker fail for Qwen3 dense fp8 model
 (#13853)

---
 python/sglang/srt/model_executor/model_runner.py | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/python/sglang/srt/model_executor/model_runner.py b/python/sglang/srt/model_executor/model_runner.py
index 11c69a309..a0272029a 100644
--- a/python/sglang/srt/model_executor/model_runner.py
+++ b/python/sglang/srt/model_executor/model_runner.py
@@ -597,9 +597,12 @@ class ModelRunner:
                 )
             moe_tp_size = self.tp_size // self.moe_ep_size
 
-            moe_intermediate_size = (
-                self.model_config.hf_text_config.moe_intermediate_size
+            moe_intermediate_size = getattr(
+                self.model_config.hf_text_config, "moe_intermediate_size", None
             )
+            if moe_intermediate_size is None:
+                return
+
             if moe_intermediate_size % moe_tp_size != 0:
                 raise ValueError(
                     f"moe_intermediate_size {moe_intermediate_size} must be divisible by moe_tp_size ({moe_tp_size}) which is tp_size ({self.tp_size}) divided by moe_ep_size ({self.moe_ep_size})."