update qwen&qwen2-moe and baichuan layout

vllm/model_executor/models/baichuan.py

@@ -489,7 +489,7 @@ class BaiChuanBaseForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
                 weight_loader = getattr(param, "weight_loader",
                                         default_weight_loader)
                 weight_loader(param, loaded_weight)
-                loaded_params.add(name)
+            loaded_params.add(name)
         
         if self.use_llama_nn and self.quant_method is None :
             lay_key_words = [

vllm/model_executor/models/qwen.py

@@ -1028,7 +1028,7 @@ class QWenBaseModel(nn.Module, SupportsPP, SupportsLoRA):
                 weight_loader = getattr(param, "weight_loader",
                                         default_weight_loader)
                 weight_loader(param, loaded_weight)
-                loaded_params.add(name)
+            loaded_params.add(name)
 
         if self.use_llama_nn and self.quant_method is None :
             lay_key_words = [

vllm/model_executor/models/qwen2_moe.py

@@ -553,7 +553,7 @@ class Qwen2MoeForCausalLM(nn.Module, SupportsPP):
                     weight_loader = getattr(param, "weight_loader",
                                             default_weight_loader)
                     weight_loader(param, loaded_weight)
-                    loaded_params.add(name)
+                loaded_params.add(name)
 
         if self.use_llama_nn and self.quant_method is None:
             lay_key_words = [