update version

setup.py

@@ -488,11 +488,11 @@ def get_version_add(sha: Optional[str] = None) -> str:
             if sha is None:
                 sha = get_sha(vllm_root)
             if (major, minor) == ('2', '4'):
-                version = 'das.opt1.beta.' + sha[:7]
+                version = 'das.opt2.alpha.' + sha[:7]
                 # version = 'das.opt1.' + sha[:7]
     else:
         if (major, minor) == ('2', '4'):
-            version = 'das.opt1.beta'
+            version = 'das.opt2.alpha'
             # version = 'das.opt1'
 
 

vllm/model_executor/model_loader/utils.py

@@ -79,10 +79,9 @@ def get_model_architecture(
         model_config: ModelConfig) -> Tuple[Type[nn.Module], str]:
     architectures = getattr(model_config.hf_config, "architectures", [])
     visions = getattr(model_config.hf_config, "visual", []) or getattr(model_config.hf_config, "vision_config", [])
-    # TODO: support deepseek distillation series models ( 'LlamaForCausalLM', 'Qwen2ForCausalLM' )
-    # 'Qwen2VLForConditionalGeneration', 'Qwen2_5_VLForConditionalGeneration'
-    support_nn_architectures = ['QWenLMHeadModel', 
-                                'Qwen2MoeForCausalLM', 'ChatGLMModel', 'ChatGLMForConditionalGeneration', 
+    # 'Qwen2_5_VLForConditionalGeneration'
+    support_nn_architectures = ['LlamaForCausalLM', 'QWenLMHeadModel', 'Qwen2ForCausalLM', 'Qwen2MoeForCausalLM',
+                                'Qwen2VLForConditionalGeneration','ChatGLMModel', 'ChatGLMForConditionalGeneration', 
                                 'BaichuanForCausalLM', 'BloomForCausalLM', 'MedusaModel', 'MixtralForCausalLM', 
                                 'MLPSpeculatorPreTrainedModel', 'FalconForCausalLM', 'DeepseekV2ForCausalLM', 
                                 'DeepseekV3ForCausalLM', 'DeepSeekMTPModel']