[Feature]: Enable directly use OmniLLM init AR model (#1821)

docs/configuration/stage_configs.md

@@ -18,7 +18,7 @@ If users want to modify some part of it. The custom stage_configs file can be in
 For offline (Assume necessary dependencies have ben imported):
 ```python
 model_name = "Qwen/Qwen2.5-Omni-7B"
-omni_llm = OmniLLM(model=model_name, stage_configs_path="/path/to/custom_stage_configs.yaml")
+omni_llm = Omni(model=model_name, stage_configs_path="/path/to/custom_stage_configs.yaml")
 ```
 
 For online serving:

vllm_omni/config/model.py

@@ -48,7 +48,7 @@ class OmniModelConfig(ModelConfig):
     stage_id: int = 0
     async_chunk: bool = False
     model_stage: str = "thinker"
-    model_arch: str = "Qwen2_5OmniForConditionalGeneration"
+    model_arch: str | None = None
     worker_type: str | None = None
     engine_output_type: str | None = None
     hf_config_name: str | None = None
@@ -69,7 +69,9 @@ class OmniModelConfig(ModelConfig):
 
     @property
     def architectures(self) -> list[str]:
-        return [self.model_arch]
+        if self.model_arch is not None:
+            return [self.model_arch]
+        return super().architectures
 
     @property
     def embedding_size(self):

vllm_omni/engine/arg_utils.py

@@ -71,7 +71,7 @@ class OmniEngineArgs(EngineArgs):
 
     stage_id: int = 0
     model_stage: str = "thinker"
-    model_arch: str = "Qwen2_5OmniForConditionalGeneration"
+    model_arch: str | None = None
     engine_output_type: str | None = None
     hf_config_name: str | None = None
     custom_process_next_stage_input_func: str | None = None
@@ -233,7 +233,7 @@ class AsyncOmniEngineArgs(AsyncEngineArgs):
 
     stage_id: int = 0
     model_stage: str = "thinker"
-    model_arch: str = "Qwen2_5OmniForConditionalGeneration"
+    model_arch: str | None = None
     engine_output_type: str | None = None
     hf_config_name: str | None = None
     custom_process_next_stage_input_func: str | None = None

vllm_omni/entrypoints/omni_llm.py

@@ -25,8 +25,6 @@ from vllm_omni.engine.input_processor import OmniInputProcessor, reinject_omni_f
 from vllm_omni.engine.output_processor import MultimodalOutputProcessor
 from vllm_omni.entrypoints.utils import (
     filter_dataclass_kwargs,
-    load_stage_configs_from_model,
-    load_stage_configs_from_yaml,
     resolve_model_config_path,
 )
 
@@ -43,8 +41,6 @@ class OmniLLM(LLM):
 
     Args:
         model: Model name or path to load
-        stage_configs_path: Optional path to YAML file containing stage
-            configurations. If None, configurations are loaded from the model.
         log_stats: Whether to enable statistics logging
         compilation_config: Optional compilation configuration. Can be an
             integer (compilation level), dict, or CompilationConfig instance.
@@ -68,7 +64,6 @@ class OmniLLM(LLM):
     def __init__(
         self,
         model: str,
-        stage_configs_path: str | None = None,
         log_stats: bool = False,
         compilation_config: int | dict[str, Any] | CompilationConfig | None = None,
         hf_overrides: dict[str, Any] | None = None,
@@ -86,13 +81,8 @@ class OmniLLM(LLM):
         self.batch_timeout = batch_timeout
         self.log_stats: bool = bool(log_stats)
 
-        # Load stage configurations
-        if stage_configs_path is None:
-            self.config_path = resolve_model_config_path(model)
-            self.stage_configs = load_stage_configs_from_model(model)
-        else:
-            self.config_path = stage_configs_path
-            self.stage_configs = load_stage_configs_from_yaml(stage_configs_path)
+        # Resolve model config path for connectors
+        self.config_path = resolve_model_config_path(model)
 
         # Initialize connectors
         self.omni_transfer_config, self.connectors = initialize_orchestrator_connectors(
@@ -154,6 +144,7 @@ class OmniLLM(LLM):
             compilation_config=compilation_config_instance,
             structured_outputs_config=structured_outputs_instance,
             omni_kv_config=omni_kv_config,
+            hf_overrides=hf_overrides or {},
             **filter_dataclass_kwargs(OmniEngineArgs, kwargs),
         )