Organize server_args (#277)

python/sglang/srt/server.py

@@ -1,6 +1,7 @@
 """SRT: SGLang Runtime"""
 
 import asyncio
+import dataclasses
 import json
 import multiprocessing as mp
 import os
@@ -86,6 +87,11 @@ async def get_model_info():
     return result
 
 
+@app.get("/get_server_args")
+async def get_server_args():
+    return dataclasses.asdict(tokenizer_manager.server_args)
+
+
 @app.get("/flush_cache")
 async def flush_cache():
     await tokenizer_manager.flush_cache()
@@ -548,7 +554,6 @@ class Runtime:
         max_prefill_num_token: int = ServerArgs.max_prefill_num_token,
         context_length: int = ServerArgs.context_length,
         tp_size: int = 1,
-        model_mode: List[str] = (),
         schedule_heuristic: str = "lpm",
         attention_reduce_in_fp32: bool = False,
         random_seed: int = 42,
@@ -571,7 +576,6 @@ class Runtime:
             max_prefill_num_token=max_prefill_num_token,
             context_length=context_length,
             tp_size=tp_size,
-            model_mode=model_mode,
             schedule_heuristic=schedule_heuristic,
             attention_reduce_in_fp32=attention_reduce_in_fp32,
             random_seed=random_seed,