From 1e5ad9b74f70f4690dff629598d414ee27116b85 Mon Sep 17 00:00:00 2001
From: Jee Jee Li <pandaleefree@gmail.com>
Date: Fri, 27 Feb 2026 11:46:30 +0800
Subject: [PATCH] [Bugfix] Fix Qwen3NextForCausalLM packed_modules_mapping
 (#35413)

Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
---
 vllm/model_executor/models/qwen3_next.py | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/vllm/model_executor/models/qwen3_next.py b/vllm/model_executor/models/qwen3_next.py
index 777d1d7bf..c57265cc7 100644
--- a/vllm/model_executor/models/qwen3_next.py
+++ b/vllm/model_executor/models/qwen3_next.py
@@ -412,6 +412,8 @@ class Qwen3NextGatedDeltaNet(nn.Module, MambaBase):
             prefix=f"{prefix}.in_proj_qkvz",
         )
         # ba_proj doesn't support blockwise fp8 quantization.
+        # # in_proj_ba is defined as MergedColumnParallelLinear for
+        # compatibility with Qwen3_5.
         self.in_proj_ba = MergedColumnParallelLinear(
             input_size=self.hidden_size,
             output_sizes=[self.num_v_heads] * 2,
@@ -1326,6 +1328,8 @@ class Qwen3NextForCausalLM(
             "v_proj",
         ],
         "gate_up_proj": ["gate_proj", "up_proj"],
+        "in_proj_qkvz": ["in_proj_qkvz"],
+        "in_proj_ba": ["in_proj_ba"],
     }
 
     def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):