[Misc] Support qwen3-next lora (#31719)

2026-01-08 17:27:50 +08:00
parent 1f214290d6
commit 96fcd3c267
1 changed files with 7 additions and 1 deletions
--- 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(