A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 22:35 UTC
[BugFix] Don't change title of top-level process (#22032)
Signed-off-by: Nick Hill <nhill@redhat.com>
Nick Hill committed
Aug 1, 2025 at 15:37 UTC
3146519add735bc51a6a983af9e9c4a8b8d3373e
2 files changed
+8
-7
vllm/entrypoints/cli/serve.py
+6
-5
@@ -18,7 +18,8 @@ from vllm.entrypoints.utils import (VLLM_SUBCMD_PARSER_EPILOG,
18
show_filtered_argument_or_group_from_help)
19
from vllm.logger import init_logger
20
from vllm.usage.usage_lib import UsageContext
21
-from vllm.utils import FlexibleArgumentParser, decorate_logs, get_tcp_uri
21
+from vllm.utils import (FlexibleArgumentParser, decorate_logs, get_tcp_uri,
22
+ set_process_title)
23
from vllm.v1.engine.core import EngineCoreProc
24
from vllm.v1.engine.utils import CoreEngineProcManager, launch_core_engines
25
from vllm.v1.executor.abstract import Executor
@@ -74,7 +75,7 @@ def run_headless(args: argparse.Namespace):
75
76
if args.api_server_count > 1:
77
raise ValueError("api_server_count can't be set in headless mode")
77
- # set_process_title("Headless_ProcManager")
78
+
79
# Create the EngineConfig.
80
engine_args = vllm.AsyncEngineArgs.from_cli_args(args)
81
usage_context = UsageContext.OPENAI_API_SERVER
@@ -139,8 +140,6 @@ def run_multi_api_server(args: argparse.Namespace):
140
141
orig_disable_mm_preprocessor_cache = args.disable_mm_preprocessor_cache
142
142
- # set_process_title("ProcManager")
143
-
143
if num_api_servers > 1:
144
setup_multiprocess_prometheus()
145
@@ -225,7 +224,9 @@ def run_api_server_worker_proc(listen_address,
224
**uvicorn_kwargs) -> None:
225
"""Entrypoint for individual API server worker processes."""
226
228
- # Add process-specific prefix to stdout and stderr.
227
+ # Set process title and add process-specific prefix to stdout and stderr.
228
+ server_index = client_config.get("client_index", 0) if client_config else 0
229
+ set_process_title("APIServer", str(server_index))
230
decorate_logs()
231
232
uvloop.run(
vllm/entrypoints/openai/api_server.py
+2
-2
@@ -102,7 +102,7 @@ from vllm.transformers_utils.tokenizer import MistralTokenizer
102
from vllm.usage.usage_lib import UsageContext
103
from vllm.utils import (Device, FlexibleArgumentParser, decorate_logs,
104
get_open_zmq_ipc_path, is_valid_ipv6_address,
105
- set_process_title, set_ulimit)
105
+ set_ulimit)
106
from vllm.v1.metrics.prometheus import get_prometheus_registry
107
from vllm.version import __version__ as VLLM_VERSION
108
@@ -1824,7 +1824,7 @@ async def run_server_worker(listen_address,
1824
ToolParserManager.import_tool_parser(args.tool_parser_plugin)
1825
1826
server_index = client_config.get("client_index", 0) if client_config else 0
1827
- set_process_title("APIServer", str(server_index))
1827
+
1828
# Load logging config for uvicorn if specified
1829
log_config = load_log_config(args.log_config_file)
1830
if log_config is not None: