[Misc] Log vLLM logo when starting server (#32796)

Signed-off-by: Nick Hill <nickhill123@gmail.com>

vllm/entrypoints/grpc_server.py

@@ -30,6 +30,7 @@ from grpc_reflection.v1alpha import reflection
 
 from vllm import SamplingParams, TextPrompt, TokensPrompt
 from vllm.engine.arg_utils import AsyncEngineArgs
+from vllm.entrypoints.utils import log_version_and_model
 from vllm.grpc import vllm_engine_pb2, vllm_engine_pb2_grpc
 from vllm.logger import init_logger
 from vllm.outputs import RequestOutput
@@ -408,8 +409,8 @@ async def serve_grpc(args: argparse.Namespace):
     Args:
         args: Parsed command line arguments
     """
-    logger.info("vLLM gRPC server version %s", VLLM_VERSION)
-    logger.info("args: %s", args)
+    log_version_and_model(logger, VLLM_VERSION, args.model)
+    logger.info("vLLM gRPC server args: %s", args)
 
     start_time = time.time()
 

vllm/entrypoints/openai/api_server.py

@@ -63,6 +63,7 @@ from vllm.entrypoints.serve.tokenize.serving import OpenAIServingTokenization
 from vllm.entrypoints.utils import (
     cli_env_setup,
     log_non_default_args,
+    log_version_and_model,
     process_lora_modules,
     sanitize_message,
 )
@@ -867,7 +868,7 @@ def setup_server(args):
     """Validate API server args, set up signal handler, create socket
     ready to serve."""
 
-    logger.info("vLLM API server version %s", VLLM_VERSION)
+    log_version_and_model(logger, VLLM_VERSION, args.model)
     log_non_default_args(args)
 
     if args.tool_parser_plugin and len(args.tool_parser_plugin) > 3:

vllm/entrypoints/utils.py

@@ -6,6 +6,8 @@ import dataclasses
 import functools
 import os
 from argparse import Namespace
+from logging import Logger
+from string import Template
 from typing import TYPE_CHECKING, Any
 
 import regex as re
@@ -13,8 +15,9 @@ from fastapi import Request
 from fastapi.responses import JSONResponse, StreamingResponse
 from starlette.background import BackgroundTask, BackgroundTasks
 
+from vllm import envs
 from vllm.engine.arg_utils import EngineArgs
-from vllm.logger import init_logger
+from vllm.logger import current_formatter_type, init_logger
 from vllm.platforms import current_platform
 from vllm.utils.argparse_utils import FlexibleArgumentParser
 
@@ -295,3 +298,28 @@ def process_lora_modules(
 def sanitize_message(message: str) -> str:
     # Avoid leaking memory address from object reprs
     return re.sub(r" at 0x[0-9a-f]+>", ">", message)
+
+
+def log_version_and_model(lgr: Logger, version: str, model_name: str) -> None:
+    if envs.VLLM_DISABLE_LOG_LOGO or (formatter := current_formatter_type(lgr)) is None:
+        message = "vLLM server version %s, serving model %s"
+    else:
+        logo_template = Template(
+            "\n       ${w}█     █     █▄   ▄█${r}\n"
+            " ${o}▄▄${r} ${b}▄█${r} ${w}█     █     █ ▀▄▀ █${r}  version ${w}%s${r}\n"
+            "  ${o}█${r}${b}▄█▀${r} ${w}█     █     █     █${r}  model   ${w}%s${r}\n"
+            "   ${b}▀▀${r}  ${w}▀▀▀▀▀ ▀▀▀▀▀ ▀     ▀${r}\n"
+        )
+        colors = {
+            "w": "\033[97;1m",  # white
+            "o": "\033[93m",  # orange
+            "b": "\033[94m",  # blue
+            "r": "\033[0m",  # reset
+        }
+        if formatter != "color":
+            # monochrome logo (no ansi escape codes)
+            colors = dict.fromkeys(colors, "")
+
+        message = logo_template.substitute(colors)
+
+    lgr.info(message, version, model_name)

vllm/envs.py

@@ -249,6 +249,7 @@ if TYPE_CHECKING:
     VLLM_USE_V2_MODEL_RUNNER: bool = False
     VLLM_LOG_MODEL_INSPECTION: bool = False
     VLLM_DEBUG_MFU_METRICS: bool = False
+    VLLM_DISABLE_LOG_LOGO: bool = False
 
 
 def get_default_cache_root():
@@ -1610,6 +1611,8 @@ environment_variables: dict[str, Callable[[], Any]] = {
     "VLLM_DEBUG_MFU_METRICS": lambda: bool(
         int(os.getenv("VLLM_DEBUG_MFU_METRICS", "0"))
     ),
+    # Disable logging of vLLM logo at server startup time.
+    "VLLM_DISABLE_LOG_LOGO": lambda: bool(int(os.getenv("VLLM_DISABLE_LOG_LOGO", "0"))),
 }
 
 # --8<-- [end:env-vars-definition]

vllm/logger.py

@@ -17,6 +17,7 @@ from types import MethodType
 from typing import Any, Literal, cast
 
 import vllm.envs as envs
+from vllm.logging_utils import ColoredFormatter, NewLineFormatter
 
 _FORMAT = (
     f"{envs.VLLM_LOGGING_PREFIX}%(levelname)s %(asctime)s "
@@ -224,6 +225,18 @@ def suppress_logging(level: int = logging.INFO) -> Generator[None, Any, None]:
     logging.disable(current_level)
 
 
+def current_formatter_type(lgr: Logger) -> Literal["color", "newline", None]:
+    while lgr is not None:
+        if lgr.handlers and len(lgr.handlers) == 1 and lgr.handlers[0].name == "vllm":
+            formatter = lgr.handlers[0].formatter
+            if isinstance(formatter, ColoredFormatter):
+                return "color"
+            if isinstance(formatter, NewLineFormatter):
+                return "newline"
+        lgr = lgr.parent
+    return None
+
+
 # The root logger is initialized when the module is imported.
 # This is thread-safe as the module is only imported once,
 # guaranteed by the Python GIL.