[BUGFIX] KeyError 'layers.14.mlp.gate.g_idx' for Qwen3-MoE with GPTQ on ROCm (vllm-project#22017)

JartX · jingyu · commit 33ee436801c0 · 2025-08-11T20:23:28.000Z
Signed-off-by: jingyu &lt;jingyu@omniml.ai&gt;
diff --git a/vllm/model_executor/models/qwen3_moe.py b/vllm/model_executor/models/qwen3_moe.py
@@ -149,7 +149,7 @@ def __init__(
         self.gate = ReplicatedLinear(config.hidden_size,
                                      config.num_experts,
                                      bias=False,
-                                     quant_config=None,
+                                     quant_config=quant_config,
                                      prefix=f"{prefix}.gate")
 
     def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
