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 7, 11:30 UTC
[Doc] Improve UX of `--enable-log-requests` (#35723)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Cyrus Leung committed
Mar 3, 2026 at 00:24 UTC
792a74b9731f3ce27b6ac4c00064d6cacd86ef13
5 files changed
+21
-11
benchmarks/auto_tune/auto_tune.sh
-1
@@ -85,7 +85,6 @@ start_server() {
85
# Each argument and its value are separate elements.
86
local common_args_array=(
87
"$MODEL"
88
- "--disable-log-requests"
88
"--port" "8004"
89
"--host" "$HOSTNAME"
90
"--gpu-memory-utilization" "$gpu_memory_utilization"
benchmarks/multi_turn/README.md
+1
-1
@@ -7,7 +7,7 @@ First start serving your model
7
```bash
8
export MODEL_PATH=/models/meta-llama/Meta-Llama-3.1-8B-Instruct/
9
10
-vllm serve $MODEL_PATH --served-model-name Llama --disable-log-requests
10
+vllm serve $MODEL_PATH --served-model-name Llama
11
```
12
13
The variable `MODEL_PATH` should be a path to the model files (e.g. downloaded from huggingface).
vllm/engine/arg_utils.py
+4
-8
@@ -2187,14 +2187,10 @@ class AsyncEngineArgs(EngineArgs):
2187
"--enable-log-requests",
2188
action=argparse.BooleanOptionalAction,
2189
default=AsyncEngineArgs.enable_log_requests,
2190
- help="Enable logging requests.",
2191
- )
2192
- parser.add_argument(
2193
- "--disable-log-requests",
2194
- action=argparse.BooleanOptionalAction,
2195
- default=not AsyncEngineArgs.enable_log_requests,
2196
- help="[DEPRECATED] Disable logging requests.",
2197
- deprecated=True,
2190
+ help="Enable logging request information, dependant on log level:\n"
2191
+ "- INFO: Request ID, parameters and LoRA request.\n"
2192
+ "- DEBUG: Prompt inputs (e.g: text, token IDs).\n"
2193
+ "You can set the minimum log level via `VLLM_LOGGING_LEVEL`.",
2194
)
2195
current_platform.pre_register_and_update(parser)
2196
return parser
vllm/entrypoints/logger.py
+14
@@ -18,6 +18,20 @@ class RequestLogger:
18
def __init__(self, *, max_log_len: int | None) -> None:
19
self.max_log_len = max_log_len
20
21
+ if not logger.isEnabledFor(logging.INFO):
22
+ logger.warning_once(
23
+ "`--enable-log-requests` is set but "
24
+ "the minimum log level is higher than INFO. "
25
+ "No request information will be logged."
26
+ )
27
+ elif not logger.isEnabledFor(logging.DEBUG):
28
+ logger.info_once(
29
+ "`--enable-log-requests` is set but "
30
+ "the minimum log level is higher than DEBUG. "
31
+ "Only limited information will be logged to minimize overhead. "
32
+ "To view more details, set `VLLM_LOGGING_LEVEL=DEBUG`."
33
+ )
34
+
35
def log_inputs(
36
self,
37
request_id: str,
vllm/entrypoints/openai/cli_args.py
+2
-1
@@ -143,7 +143,8 @@ class BaseFrontendArgs:
143
templates and other tokenizer configuration."""
144
enable_log_outputs: bool = False
145
"""If set to True, log model outputs (generations).
146
- Requires --enable-log-requests."""
146
+ Requires `--enable-log-requests`. As with `--enable-log-requests`,
147
+ information is only logged at INFO level at maximum."""
148
enable_log_deltas: bool = True
149
"""If set to False, output deltas will not be logged. Relevant only if
150
--enable-log-outputs is set.