@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 2d7b09b998
Fork 0

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

Deprecate `--disable-log-requests` and replace with `--enable-log-requests` (#21739)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

Harry Mellor committed Aug 1, 2025 at 17:16 UTC 2d7b09b998980b9ccbb3708632b47bc28de076aa
18 files changed +97 -79
.buildkite/nightly-benchmarks/README.md
-1
@@ -104,7 +104,6 @@ We test the throughput by using `vllm bench serve` with request rate = inf to co
104 "tensor_parallel_size": 1,
105 "swap_space": 16,
106 "disable_log_stats": "",
107 - "disable_log_requests": "",
107 "load_format": "dummy"
108 },
109 "client_parameters": {
.buildkite/nightly-benchmarks/tests/genai-perf-tests.json
-1
@@ -11,7 +11,6 @@
11 },
12 "vllm_server_parameters": {
13 "disable_log_stats": "",
14 - "disable_log_requests": "",
14 "gpu_memory_utilization": 0.9,
15 "num_scheduler_steps": 10,
16 "max_num_seqs": 512,
.buildkite/nightly-benchmarks/tests/nightly-tests.json
-6
@@ -35,7 +35,6 @@
35 },
36 "vllm_server_parameters": {
37 "disable_log_stats": "",
38 - "disable_log_requests": "",
38 "gpu_memory_utilization": 0.9,
39 "num_scheduler_steps": 10,
40 "max_num_seqs": 512,
@@ -90,7 +89,6 @@
89 },
90 "vllm_server_parameters": {
91 "disable_log_stats": "",
93 - "disable_log_requests": "",
92 "gpu_memory_utilization": 0.9,
93 "num_scheduler_steps": 10,
94 "max_num_seqs": 512,
@@ -145,7 +143,6 @@
143 },
144 "vllm_server_parameters": {
145 "disable_log_stats": "",
148 - "disable_log_requests": "",
146 "gpu_memory_utilization": 0.9,
147 "num_scheduler_steps": 10,
148 "max_num_seqs": 512,
@@ -197,7 +194,6 @@
194 },
195 "vllm_server_parameters": {
196 "disable_log_stats": "",
200 - "disable_log_requests": "",
197 "gpu_memory_utilization": 0.9,
198 "num_scheduler_steps": 10,
199 "max_num_seqs": 512,
@@ -251,7 +247,6 @@
247 },
248 "vllm_server_parameters": {
249 "disable_log_stats": "",
254 - "disable_log_requests": "",
250 "gpu_memory_utilization": 0.9,
251 "num_scheduler_steps": 10,
252 "max_num_seqs": 512,
@@ -305,7 +300,6 @@
300 },
301 "vllm_server_parameters": {
302 "disable_log_stats": "",
308 - "disable_log_requests": "",
303 "gpu_memory_utilization": 0.9,
304 "num_scheduler_steps": 10,
305 "max_num_seqs": 512,
.buildkite/nightly-benchmarks/tests/serving-tests-cpu-snc2.json
-6
@@ -17,7 +17,6 @@
17 "block_size": 128,
18 "trust_remote_code": "",
19 "disable_log_stats": "",
20 - "disable_log_requests": "",
20 "enforce_eager": "",
21 "max_num_batched_tokens": 2048,
22 "max_num_seqs": 256,
@@ -50,7 +49,6 @@
49 "block_size": 128,
50 "trust_remote_code": "",
51 "disable_log_stats": "",
53 - "disable_log_requests": "",
52 "enforce_eager": "",
53 "max_num_batched_tokens": 2048,
54 "max_num_seqs": 256,
@@ -83,7 +81,6 @@
81 "block_size": 128,
82 "trust_remote_code": "",
83 "disable_log_stats": "",
86 - "disable_log_requests": "",
84 "enforce_eager": "",
85 "max_num_batched_tokens": 2048,
86 "max_num_seqs": 256,
@@ -117,7 +114,6 @@
114 "trust_remote_code": "",
115 "enable_chunked_prefill": "",
116 "disable_log_stats": "",
120 - "disable_log_requests": "",
117 "enforce_eager": "",
118 "max_num_batched_tokens": 2048,
119 "max_num_seqs": 256,
@@ -153,7 +149,6 @@
149 "trust_remote_code": "",
150 "enable_chunked_prefill": "",
151 "disable_log_stats": "",
156 - "disable_log_requests": "",
152 "enforce_eager": "",
153 "max_num_batched_tokens": 2048,
154 "max_num_seqs": 256,
@@ -189,7 +184,6 @@
184 "trust_remote_code": "",
185 "enable_chunked_prefill": "",
186 "disable_log_stats": "",
192 - "disable_log_requests": "",
187 "enforce_eager": "",
188 "max_num_batched_tokens": 2048,
189 "max_num_seqs": 256,
.buildkite/nightly-benchmarks/tests/serving-tests-cpu-snc3.json
-6
@@ -17,7 +17,6 @@
17 "block_size": 128,
18 "trust_remote_code": "",
19 "disable_log_stats": "",
20 - "disable_log_requests": "",
20 "enforce_eager": "",
21 "max_num_batched_tokens": 2048,
22 "max_num_seqs": 256,
@@ -50,7 +49,6 @@
49 "block_size": 128,
50 "trust_remote_code": "",
51 "disable_log_stats": "",
53 - "disable_log_requests": "",
52 "enforce_eager": "",
53 "max_num_batched_tokens": 2048,
54 "max_num_seqs": 256,
@@ -84,7 +82,6 @@
82 "block_size": 128,
83 "trust_remote_code": "",
84 "disable_log_stats": "",
87 - "disable_log_requests": "",
85 "enforce_eager": "",
86 "max_num_batched_tokens": 2048,
87 "max_num_seqs": 256,
@@ -118,7 +115,6 @@
115 "trust_remote_code": "",
116 "enable_chunked_prefill": "",
117 "disable_log_stats": "",
121 - "disable_log_requests": "",
118 "enforce_eager": "",
119 "max_num_batched_tokens": 2048,
120 "max_num_seqs": 256,
@@ -154,7 +150,6 @@
150 "trust_remote_code": "",
151 "enable_chunked_prefill": "",
152 "disable_log_stats": "",
157 - "disable_log_requests": "",
153 "enforce_eager": "",
154 "max_num_batched_tokens": 2048,
155 "max_num_seqs": 256,
@@ -191,7 +186,6 @@
186 "trust_remote_code": "",
187 "enable_chunked_prefill": "",
188 "disable_log_stats": "",
194 - "disable_log_requests": "",
189 "enforce_eager": "",
190 "max_num_batched_tokens": 2048,
191 "max_num_seqs": 256,
.buildkite/nightly-benchmarks/tests/serving-tests-cpu.json
-5
@@ -17,7 +17,6 @@
17 "block_size": 128,
18 "trust_remote_code": "",
19 "disable_log_stats": "",
20 - "disable_log_requests": "",
20 "enforce_eager": "",
21 "max_num_batched_tokens": 2048,
22 "max_num_seqs": 256,
@@ -50,7 +49,6 @@
49 "block_size": 128,
50 "trust_remote_code": "",
51 "disable_log_stats": "",
53 - "disable_log_requests": "",
52 "enforce_eager": "",
53 "max_num_batched_tokens": 2048,
54 "max_num_seqs": 256,
@@ -83,7 +81,6 @@
81 "block_size": 128,
82 "trust_remote_code": "",
83 "disable_log_stats": "",
86 - "disable_log_requests": "",
84 "enforce_eager": "",
85 "max_num_batched_tokens": 2048,
86 "max_num_seqs": 256,
@@ -117,7 +114,6 @@
114 "trust_remote_code": "",
115 "enable_chunked_prefill": "",
116 "disable_log_stats": "",
120 - "disable_log_requests": "",
117 "enforce_eager": "",
118 "max_num_batched_tokens": 2048,
119 "max_num_seqs": 256,
@@ -153,7 +149,6 @@
149 "trust_remote_code": "",
150 "enable_chunked_prefill": "",
151 "disable_log_stats": "",
156 - "disable_log_requests": "",
152 "enforce_eager": "",
153 "max_num_batched_tokens": 2048,
154 "max_num_seqs": 256,
.buildkite/nightly-benchmarks/tests/serving-tests.json
+1 -5
@@ -7,7 +7,6 @@
7 "tensor_parallel_size": 1,
8 "swap_space": 16,
9 "disable_log_stats": "",
10 - "disable_log_requests": "",
10 "load_format": "dummy"
11 },
12 "client_parameters": {
@@ -26,7 +25,6 @@
25 "tensor_parallel_size": 4,
26 "swap_space": 16,
27 "disable_log_stats": "",
29 - "disable_log_requests": "",
28 "load_format": "dummy"
29 },
30 "client_parameters": {
@@ -45,7 +43,6 @@
43 "tensor_parallel_size": 2,
44 "swap_space": 16,
45 "disable_log_stats": "",
48 - "disable_log_requests": "",
46 "load_format": "dummy"
47 },
48 "client_parameters": {
@@ -60,8 +57,7 @@
57 "test_name": "serving_llama70B_tp4_sharegpt_specdecode",
58 "qps_list": [2],
59 "server_parameters": {
63 - "model": "meta-llama/Meta-Llama-3.1-70B-Instruct",
64 - "disable_log_requests": "",
60 + "model": "meta-llama/Meta-Llama-3.1-70B-Instruct",
61 "tensor_parallel_size": 4,
62 "swap_space": 16,
63 "speculative_config": {
tests/config/test_mp_reducer.py
-1
@@ -28,7 +28,6 @@ def test_mp_reducer(monkeypatch):
28 max_model_len=32,
29 gpu_memory_utilization=0.1,
30 disable_log_stats=True,
31 - disable_log_requests=True,
31 )
32
33 async_llm = AsyncLLM.from_engine_args(
tests/mq_llm_engine/test_load.py
+1 -1
@@ -16,7 +16,7 @@ NUM_EXPECTED_TOKENS = 10
16 NUM_REQUESTS = 10000
17
18 # Scenarios to test for num generated token.
19 -ENGINE_ARGS = AsyncEngineArgs(model=MODEL, disable_log_requests=True)
19 +ENGINE_ARGS = AsyncEngineArgs(model=MODEL)
20
21
22 @pytest.fixture(scope="function")
tests/v1/engine/test_async_llm.py
+1 -3
@@ -26,12 +26,10 @@ if not current_platform.is_cuda():
26 TEXT_ENGINE_ARGS = AsyncEngineArgs(
27 model="meta-llama/Llama-3.2-1B-Instruct",
28 enforce_eager=True,
29 - disable_log_requests=True,
29 )
30
31 VISION_ENGINE_ARGS = AsyncEngineArgs(model="Qwen/Qwen2-VL-2B-Instruct",
33 - enforce_eager=True,
34 - disable_log_requests=True)
32 + enforce_eager=True)
33
34 TEXT_PROMPT = "Hello my name is Robert and"
35
tests/v1/test_async_llm_dp.py
-1
@@ -25,7 +25,6 @@ DP_SIZE = int(os.getenv("DP_SIZE", 2))
25 engine_args = AsyncEngineArgs(
26 model="ibm-research/PowerMoE-3b",
27 enforce_eager=True,
28 - disable_log_requests=True,
28 tensor_parallel_size=int(os.getenv("TP_SIZE", 1)),
29 data_parallel_size=DP_SIZE,
30 )
vllm/engine/arg_utils.py
+26 -4
@@ -18,7 +18,7 @@ from typing import (TYPE_CHECKING, Annotated, Any, Callable, Dict, List,
18 import regex as re
19 import torch
20 from pydantic import TypeAdapter, ValidationError
21 -from typing_extensions import TypeIs
21 +from typing_extensions import TypeIs, deprecated
22
23 import vllm.envs as envs
24 from vllm.config import (BlockSize, CacheConfig, CacheDType, CompilationConfig,
@@ -1704,7 +1704,23 @@ class EngineArgs:
1704 @dataclass
1705 class AsyncEngineArgs(EngineArgs):
1706 """Arguments for asynchronous vLLM engine."""
1707 - disable_log_requests: bool = False
1707 + enable_log_requests: bool = False
1708 +
1709 + @property
1710 + @deprecated(
1711 + "`disable_log_requests` is deprecated and has been replaced with "
1712 + "`enable_log_requests`. This will be removed in v0.12.0. Please use "
1713 + "`enable_log_requests` instead.")
1714 + def disable_log_requests(self) -> bool:
1715 + return not self.enable_log_requests
1716 +
1717 + @disable_log_requests.setter
1718 + @deprecated(
1719 + "`disable_log_requests` is deprecated and has been replaced with "
1720 + "`enable_log_requests`. This will be removed in v0.12.0. Please use "
1721 + "`enable_log_requests` instead.")
1722 + def disable_log_requests(self, value: bool):
1723 + self.enable_log_requests = not value
1724
1725 @staticmethod
1726 def add_cli_args(parser: FlexibleArgumentParser,
@@ -1715,9 +1731,15 @@ class AsyncEngineArgs(EngineArgs):
1731 load_general_plugins()
1732 if not async_args_only:
1733 parser = EngineArgs.add_cli_args(parser)
1734 + parser.add_argument('--enable-log-requests',
1735 + action=argparse.BooleanOptionalAction,
1736 + default=AsyncEngineArgs.enable_log_requests,
1737 + help='Enable logging requests.')
1738 parser.add_argument('--disable-log-requests',
1719 - action='store_true',
1720 - help='Disable logging requests.')
1739 + action=argparse.BooleanOptionalAction,
1740 + default=not AsyncEngineArgs.enable_log_requests,
1741 + help='[DEPRECATED] Disable logging requests.',
1742 + deprecated=True)
1743 current_platform.pre_register_and_update(parser)
1744 return parser
1745
vllm/engine/async_llm_engine.py
+16 -10
@@ -30,7 +30,7 @@ from vllm.sampling_params import SamplingParams
30 from vllm.sequence import ExecuteModelRequest
31 from vllm.transformers_utils.tokenizer import AnyTokenizer
32 from vllm.usage.usage_lib import UsageContext
33 -from vllm.utils import Device, weak_bind
33 +from vllm.utils import Device, deprecate_kwargs, weak_bind
34
35 logger = init_logger(__name__)
36 ENGINE_ITERATION_TIMEOUT_S = envs.VLLM_ENGINE_ITERATION_TIMEOUT_S
@@ -554,14 +554,20 @@ class AsyncLLMEngine(EngineClient):
554 return LLMEngine._get_executor_cls(engine_config)
555
556 @classmethod
557 + @deprecate_kwargs(
558 + "disable_log_requests",
559 + additional_message=("This argument will have no effect. "
560 + "Use `enable_log_requests` instead."),
561 + )
562 def from_vllm_config(
558 - cls,
559 - vllm_config: VllmConfig,
560 - start_engine_loop: bool = True,
561 - usage_context: UsageContext = UsageContext.ENGINE_CONTEXT,
562 - stat_loggers: Optional[dict[str, StatLoggerBase]] = None,
563 - disable_log_requests: bool = False,
564 - disable_log_stats: bool = False,
563 + cls,
564 + vllm_config: VllmConfig,
565 + start_engine_loop: bool = True,
566 + usage_context: UsageContext = UsageContext.ENGINE_CONTEXT,
567 + stat_loggers: Optional[dict[str, StatLoggerBase]] = None,
568 + enable_log_requests: bool = False,
569 + disable_log_stats: bool = False,
570 + disable_log_requests: bool = True, # Deprecated, will be removed
571 ) -> "AsyncLLMEngine":
572 """Create an AsyncLLMEngine from the EngineArgs."""
573
@@ -569,7 +575,7 @@ class AsyncLLMEngine(EngineClient):
575 vllm_config=vllm_config,
576 executor_class=cls._get_executor_cls(vllm_config),
577 start_engine_loop=start_engine_loop,
572 - log_requests=not disable_log_requests,
578 + log_requests=enable_log_requests,
579 log_stats=not disable_log_stats,
580 usage_context=usage_context,
581 stat_loggers=stat_loggers,
@@ -598,7 +604,7 @@ class AsyncLLMEngine(EngineClient):
604 usage_context=usage_context,
605 stat_loggers=stat_loggers,
606 disable_log_stats=engine_args.disable_log_stats,
601 - disable_log_requests=engine_args.disable_log_requests,
607 + enable_log_requests=engine_args.enable_log_requests,
608 )
609
610 @property
vllm/engine/multiprocessing/engine.py
+19 -8
@@ -34,6 +34,7 @@ from vllm.outputs import RequestOutput
34 from vllm.transformers_utils.config import (
35 maybe_register_config_serialize_by_value)
36 from vllm.usage.usage_lib import UsageContext
37 +from vllm.utils import deprecate_kwargs
38 from vllm.worker.model_runner_base import InputProcessingError
39
40 logger = init_logger(__name__)
@@ -120,10 +121,20 @@ class MQLLMEngine:
121 return ENGINE_DEAD_ERROR()
122
123 @classmethod
123 - def from_vllm_config(cls, vllm_config: VllmConfig,
124 - usage_context: UsageContext,
125 - disable_log_requests: bool, disable_log_stats: bool,
126 - ipc_path: str) -> "MQLLMEngine":
124 + @deprecate_kwargs(
125 + "disable_log_requests",
126 + additional_message=("This argument will have no effect. "
127 + "Use `enable_log_requests` instead."),
128 + )
129 + def from_vllm_config(
130 + cls,
131 + vllm_config: VllmConfig,
132 + usage_context: UsageContext,
133 + enable_log_requests: bool,
134 + disable_log_stats: bool,
135 + ipc_path: str,
136 + disable_log_requests: bool = True, # Deprecated, will be removed
137 + ) -> "MQLLMEngine":
138 # Setup plugins for each process
139 from vllm.plugins import load_general_plugins
140 load_general_plugins()
@@ -136,7 +147,7 @@ class MQLLMEngine:
147 ipc_path=ipc_path,
148 usage_context=usage_context,
149 use_async_sockets=use_async_sockets,
139 - log_requests=(not disable_log_requests),
150 + log_requests=enable_log_requests,
151 log_stats=(not disable_log_stats),
152 )
153
@@ -150,7 +161,7 @@ class MQLLMEngine:
161 ipc_path=ipc_path,
162 vllm_config=vllm_config,
163 usage_context=usage_context,
153 - disable_log_requests=engine_args.disable_log_requests,
164 + enable_log_requests=engine_args.enable_log_requests,
165 disable_log_stats=engine_args.disable_log_stats,
166 )
167
@@ -436,7 +447,7 @@ def signal_handler(*_) -> None:
447
448 def run_mp_engine(vllm_config: VllmConfig, usage_context: UsageContext,
449 ipc_path: str, disable_log_stats: bool,
439 - disable_log_requests: bool, engine_alive):
450 + enable_log_requests: bool, engine_alive):
451 try:
452 # Ensure we can serialize transformer config before spawning
453 maybe_register_config_serialize_by_value()
@@ -445,7 +456,7 @@ def run_mp_engine(vllm_config: VllmConfig, usage_context: UsageContext,
456 vllm_config=vllm_config,
457 usage_context=usage_context,
458 disable_log_stats=disable_log_stats,
448 - disable_log_requests=disable_log_requests,
459 + enable_log_requests=enable_log_requests,
460 ipc_path=ipc_path)
461
462 signal.signal(signal.SIGTERM, signal_handler)
vllm/entrypoints/openai/api_server.py
+6 -6
@@ -205,7 +205,7 @@ async def build_async_engine_client_from_engine_args(
205 async_llm = AsyncLLM.from_vllm_config(
206 vllm_config=vllm_config,
207 usage_context=usage_context,
208 - disable_log_requests=engine_args.disable_log_requests,
208 + enable_log_requests=engine_args.enable_log_requests,
209 disable_log_stats=engine_args.disable_log_stats,
210 client_addresses=client_config,
211 client_index=client_index)
@@ -227,7 +227,7 @@ async def build_async_engine_client_from_engine_args(
227 engine_client = AsyncLLMEngine.from_vllm_config(
228 vllm_config=vllm_config,
229 usage_context=usage_context,
230 - disable_log_requests=engine_args.disable_log_requests,
230 + enable_log_requests=engine_args.enable_log_requests,
231 disable_log_stats=engine_args.disable_log_stats)
232 yield engine_client
233 finally:
@@ -272,7 +272,7 @@ async def build_async_engine_client_from_engine_args(
272 target=run_mp_engine,
273 args=(vllm_config, UsageContext.OPENAI_API_SERVER, ipc_path,
274 engine_args.disable_log_stats,
275 - engine_args.disable_log_requests, engine_alive))
275 + engine_args.enable_log_requests, engine_alive))
276 engine_process.start()
277 engine_pid = engine_process.pid
278 assert engine_pid is not None, "Engine process failed to start."
@@ -1570,10 +1570,10 @@ async def init_app_state(
1570 else:
1571 served_model_names = [args.model]
1572
1573 - if args.disable_log_requests:
1574 - request_logger = None
1575 - else:
1573 + if args.enable_log_requests:
1574 request_logger = RequestLogger(max_log_len=args.max_log_len)
1575 + else:
1576 + request_logger = None
1577
1578 base_model_paths = [
1579 BaseModelPath(name=name, model_path=args.model)
vllm/entrypoints/openai/run_batch.py
+3 -3
@@ -324,10 +324,10 @@ async def run_batch(
324 else:
325 served_model_names = [args.model]
326
327 - if args.disable_log_requests:
328 - request_logger = None
329 - else:
327 + if args.enable_log_requests:
328 request_logger = RequestLogger(max_log_len=args.max_log_len)
329 + else:
330 + request_logger = None
331
332 base_model_paths = [
333 BaseModelPath(name=name, model_path=args.model)
vllm/utils/__init__.py
+6
@@ -1668,6 +1668,12 @@ class FlexibleArgumentParser(ArgumentParser):
1668 # Enable the deprecated kwarg for Python 3.12 and below
1669
1670 def parse_known_args(self, args=None, namespace=None):
1671 + if args is not None and "--disable-log-requests" in args:
1672 + # Special case warning because the warning below won't trigger
1673 + # if –-disable-log-requests because its value is default.
1674 + logger.warning_once(
1675 + "argument '--disable-log-requests' is deprecated. This "
1676 + "will be removed in v0.12.0.")
1677 namespace, args = super().parse_known_args(args, namespace)
1678 for action in FlexibleArgumentParser._deprecated:
1679 if (hasattr(namespace, dest := action.dest)
vllm/v1/engine/async_llm.py
+18 -12
@@ -27,7 +27,7 @@ from vllm.transformers_utils.config import (
27 from vllm.transformers_utils.tokenizer import AnyTokenizer
28 from vllm.transformers_utils.tokenizer_group import init_tokenizer_from_configs
29 from vllm.usage.usage_lib import UsageContext
30 -from vllm.utils import Device, cdiv
30 +from vllm.utils import Device, cdiv, deprecate_kwargs
31 from vllm.v1.engine import EngineCoreRequest
32 from vllm.v1.engine.core_client import EngineCoreClient
33 from vllm.v1.engine.exceptions import EngineDeadError, EngineGenerateError
@@ -142,16 +142,22 @@ class AsyncLLM(EngineClient):
142 pass
143
144 @classmethod
145 + @deprecate_kwargs(
146 + "disable_log_requests",
147 + additional_message=("This argument will have no effect. "
148 + "Use `enable_log_requests` instead."),
149 + )
150 def from_vllm_config(
146 - cls,
147 - vllm_config: VllmConfig,
148 - start_engine_loop: bool = True,
149 - usage_context: UsageContext = UsageContext.ENGINE_CONTEXT,
150 - stat_loggers: Optional[list[StatLoggerFactory]] = None,
151 - disable_log_requests: bool = False,
152 - disable_log_stats: bool = False,
153 - client_addresses: Optional[dict[str, str]] = None,
154 - client_index: int = 0,
151 + cls,
152 + vllm_config: VllmConfig,
153 + start_engine_loop: bool = True,
154 + usage_context: UsageContext = UsageContext.ENGINE_CONTEXT,
155 + stat_loggers: Optional[list[StatLoggerFactory]] = None,
156 + enable_log_requests: bool = False,
157 + disable_log_stats: bool = False,
158 + client_addresses: Optional[dict[str, str]] = None,
159 + client_index: int = 0,
160 + disable_log_requests: bool = True, # Deprecated, will be removed
161 ) -> "AsyncLLM":
162 if not envs.VLLM_USE_V1:
163 raise ValueError(
@@ -166,7 +172,7 @@ class AsyncLLM(EngineClient):
172 executor_class=Executor.get_class(vllm_config),
173 start_engine_loop=start_engine_loop,
174 stat_loggers=stat_loggers,
169 - log_requests=not disable_log_requests,
175 + log_requests=enable_log_requests,
176 log_stats=not disable_log_stats,
177 usage_context=usage_context,
178 client_addresses=client_addresses,
@@ -191,7 +197,7 @@ class AsyncLLM(EngineClient):
197 return cls(
198 vllm_config=vllm_config,
199 executor_class=executor_class,
194 - log_requests=not engine_args.disable_log_requests,
200 + log_requests=engine_args.enable_log_requests,
201 log_stats=not engine_args.disable_log_stats,
202 start_engine_loop=start_engine_loop,
203 usage_context=usage_context,