[Minor] Improve logging and rename the health check endpoint name (#1180)

python/sglang/srt/managers/controller_multi.py

@@ -21,7 +21,6 @@ Each data parallel worker can manage multiple tensor parallel workers.
 import dataclasses
 import logging
 import multiprocessing
-import os
 from enum import Enum, auto
 
 import numpy as np

python/sglang/srt/managers/controller_single.py

@@ -17,7 +17,6 @@ limitations under the License.
 
 import logging
 import multiprocessing
-import os
 from typing import List
 
 import zmq

python/sglang/srt/managers/detokenizer_manager.py

@@ -39,6 +39,8 @@ asyncio.set_event_loop_policy(uvloop.EventLoopPolicy())
 
 @dataclasses.dataclass
 class DecodeStatus:
+    """Store the status of incremental decoding."""
+
     vid: int
     decoded_text: str
     decode_ids: List[int]
@@ -47,6 +49,8 @@ class DecodeStatus:
 
 
 class DetokenizerManager:
+    """DetokenizerManager is a process that detokenizes the token ids."""
+
     def __init__(
         self,
         server_args: ServerArgs,

python/sglang/srt/managers/tokenizer_manager.py

@@ -62,12 +62,16 @@ logger = logging.getLogger(__name__)
 
 @dataclasses.dataclass
 class ReqState:
+    """Store the state a request."""
+
     out_list: List
     finished: bool
     event: asyncio.Event
 
 
 class TokenizerManager:
+    """TokenizerManager is a process that tokenizes the text."""
+
     def __init__(
         self,
         server_args: ServerArgs,
@@ -481,11 +485,7 @@ class TokenizerManager:
 
             # Log requests
             if self.server_args.log_requests and state.finished:
-                if obj.text is None:
-                    in_obj = {"input_ids": obj.input_ids}
-                else:
-                    in_obj = {"text": obj.text}
-                logger.info(f"in={in_obj}, out={out}")
+                logger.info(f"in={obj}, out={out}")
 
             state.out_list = []
             if state.finished: