restore qwen-vl to default layout

vllm/model_executor/model_loader/utils.py

@@ -22,10 +22,14 @@ def set_default_torch_dtype(dtype: torch.dtype):
 def get_model_architecture(
         model_config: ModelConfig) -> Tuple[Type[nn.Module], str]:
     architectures = getattr(model_config.hf_config, "architectures", [])
+    visual = getattr(model_config.hf_config, "visual", [])
     support_nn_architectures = ['LlamaForCausalLM', 'QWenLMHeadModel', 'Qwen2ForCausalLM', 'ChatGLMModel', 'BaichuanForCausalLM', 'BloomForCausalLM']  
     if any(arch in architectures for arch in support_nn_architectures): 
         if os.getenv('LLAMA_NN') != '0': 
-            os.environ['LLAMA_NN'] = '1'
+             if architectures == ['QWenLMHeadModel'] and visual != []:
+                os.environ['LLAMA_NN'] = '0'
+             else:
+                os.environ['LLAMA_NN'] = '1'
         if os.getenv('GEMM_PAD') != '1': 
             os.environ['GEMM_PAD'] = '0'
         if os.getenv('FA_PAD') != '1':