From 96fcd3c267a0b16e3f94abd59e05cb3708a67db8 Mon Sep 17 00:00:00 2001
From: BingjiaWang <wangbingjia.wbj@ant-intl.com>
Date: Thu, 8 Jan 2026 17:27:50 +0800
Subject: [PATCH] [Misc] Support qwen3-next lora (#31719)

---
 vllm/model_executor/models/qwen3_next.py | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/vllm/model_executor/models/qwen3_next.py b/vllm/model_executor/models/qwen3_next.py
index 3822aae00..9fded8e6b 100644
--- a/vllm/model_executor/models/qwen3_next.py
+++ b/vllm/model_executor/models/qwen3_next.py
@@ -145,7 +145,13 @@ class Qwen3NextSparseMoeBlock(nn.Module):
             prefix=f"{prefix}.gate",
         )
 
-        self.shared_expert_gate = torch.nn.Linear(config.hidden_size, 1, bias=False)
+        self.shared_expert_gate = ReplicatedLinear(
+            config.hidden_size,
+            1,
+            bias=False,
+            quant_config=None,
+            prefix=f"{prefix}.shared_expert_gate",
+        )
 
         if config.shared_expert_intermediate_size > 0:
             self.shared_expert = Qwen3NextMLP(