[plugin] Custom model_runner/model support (#3186)

* support custom model&&model_runner

* fix merge

* add test && update doc

* fix codestyle

* fix unittest

* load model in rl

fastdeploy/entrypoints/llm.py

@@ -28,6 +28,7 @@ from tqdm import tqdm
 from fastdeploy.engine.args_utils import EngineArgs
 from fastdeploy.engine.engine import LLMEngine
 from fastdeploy.engine.sampling_params import SamplingParams
+from fastdeploy.plugins.model_register import load_model_register_plugins
 from fastdeploy.utils import (
     deprecated_kwargs_warning,
     llm_logger,
@@ -76,6 +77,7 @@ class LLM:
     ):
         deprecated_kwargs_warning(**kwargs)
 
+        load_model_register_plugins()
         model = retrive_model_from_server(model, revision)
         engine_args = EngineArgs(
             model=model,

fastdeploy/entrypoints/openai/api_server.py

@@ -46,6 +46,7 @@ from fastdeploy.metrics.metrics import (
     main_process_metrics,
 )
 from fastdeploy.metrics.trace_util import inject_to_metadata, instrument
+from fastdeploy.plugins.model_register import load_model_register_plugins
 from fastdeploy.utils import (
     FlexibleArgumentParser,
     api_server_logger,
@@ -393,6 +394,7 @@ def launch_controller_server():
 def main():
     """main函数"""
 
+    load_model_register_plugins()
     if load_engine() is None:
         return