allow user chose log level by --log-level instead of fixed 'info'. (#3109)

Co-authored-by: zixiao <shunli.dsl@alibaba-inc.com>
Co-authored-by: Simon Mo <simon.mo@hey.com>

vllm/config.py

@@ -317,7 +317,7 @@ class CacheConfig:
         self.num_cpu_blocks = None
 
     def metrics_info(self):
-        # convert cache_config to dict(key: str, value:str) for prometheus metrics info
+        # convert cache_config to dict(key: str, value: str) for prometheus metrics info
         return {key: str(value) for key, value in self.__dict__.items()}
 
     def _verify_args(self) -> None:

vllm/engine/metrics.py

@@ -23,6 +23,7 @@ class Metrics:
             if hasattr(collector, "_name") and "vllm" in collector._name:
                 REGISTRY.unregister(collector)
 
+        # Config Information
         self.info_cache_config = Info(
             name='vllm:cache_config',
             documentation='information of cache_config')

vllm/entrypoints/openai/api_server.py

@@ -62,6 +62,12 @@ def parse_args():
         description="vLLM OpenAI-Compatible RESTful API server.")
     parser.add_argument("--host", type=str, default=None, help="host name")
     parser.add_argument("--port", type=int, default=8000, help="port number")
+    parser.add_argument(
+        "--uvicorn-log-level",
+        type=str,
+        default="info",
+        choices=['debug', 'info', 'warning', 'error', 'critical', 'trace'],
+        help="log level for uvicorn")
     parser.add_argument("--allow-credentials",
                         action="store_true",
                         help="allow credentials")
@@ -245,7 +251,7 @@ if __name__ == "__main__":
     uvicorn.run(app,
                 host=args.host,
                 port=args.port,
-                log_level="info",
+                log_level=args.uvicorn_log_level,
                 timeout_keep_alive=TIMEOUT_KEEP_ALIVE,
                 ssl_keyfile=args.ssl_keyfile,
                 ssl_certfile=args.ssl_certfile)