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, 05:30 UTC
[Minor] Remove unused env variable (#27161)
Woosuk Kwon committed
Oct 18, 2025 at 18:48 UTC
fb860670da41eed17ac651d985c8098daadb91b3
1 file changed
-18
vllm/envs.py
-18
@@ -42,7 +42,6 @@ if TYPE_CHECKING:
42
VLLM_LOGGING_PREFIX: str = ""
43
VLLM_LOGGING_STREAM: str = "ext://sys.stdout"
44
VLLM_LOGGING_CONFIG_PATH: str | None = None
45
- VLLM_LOGITS_PROCESSOR_THREADS: int | None = None
45
VLLM_LOG_STATS_INTERVAL: float = 10.0
46
VLLM_TRACE_FUNCTION: int = 0
47
VLLM_ATTENTION_BACKEND: str | None = None
@@ -142,7 +141,6 @@ if TYPE_CHECKING:
141
VLLM_RAY_DP_PACK_STRATEGY: Literal["strict", "fill", "span"] = "strict"
142
VLLM_MARLIN_USE_ATOMIC_ADD: bool = False
143
VLLM_MXFP4_USE_MARLIN: bool | None = None
145
- VLLM_V0_USE_OUTLINES_CACHE: bool = False
144
VLLM_V1_USE_OUTLINES_CACHE: bool = False
145
VLLM_TPU_BUCKET_PADDING_GAP: int = 0
146
VLLM_TPU_MOST_MODEL_LEN: int | None = None
@@ -567,15 +565,6 @@ environment_variables: dict[str, Callable[[], Any]] = {
565
"VLLM_LOGGING_STREAM": lambda: os.getenv("VLLM_LOGGING_STREAM", "ext://sys.stdout"),
566
# if set, VLLM_LOGGING_PREFIX will be prepended to all log messages
567
"VLLM_LOGGING_PREFIX": lambda: os.getenv("VLLM_LOGGING_PREFIX", ""),
570
- # if set, vllm will call logits processors in a thread pool with this many
571
- # threads. This is useful when using custom logits processors that either
572
- # (a) launch additional CUDA kernels or (b) do significant CPU-bound work
573
- # while not holding the python GIL, or both.
574
- "VLLM_LOGITS_PROCESSOR_THREADS": lambda: int(
575
- os.getenv("VLLM_LOGITS_PROCESSOR_THREADS", "0")
576
- )
577
- if "VLLM_LOGITS_PROCESSOR_THREADS" in os.environ
578
- else None,
568
# If set, vllm will log stats at this interval in seconds
569
# If not set, vllm will log stats every 10 seconds.
570
"VLLM_LOG_STATS_INTERVAL": lambda: val
@@ -1066,13 +1055,6 @@ environment_variables: dict[str, Callable[[], Any]] = {
1055
"VLLM_MXFP4_USE_MARLIN": lambda: maybe_convert_bool(
1056
os.environ.get("VLLM_MXFP4_USE_MARLIN", None)
1057
),
1069
- # Whether to turn on the outlines cache for V0
1070
- # This cache is unbounded and on disk, so it's not safe to use in
1071
- # an environment with potentially malicious users.
1072
- "VLLM_V0_USE_OUTLINES_CACHE": lambda: os.environ.get(
1073
- "VLLM_V0_USE_OUTLINES_CACHE", "0"
1074
- )
1075
- == "1",
1058
# Whether to turn on the outlines cache for V1
1059
# This cache is unbounded and on disk, so it's not safe to use in
1060
# an environment with potentially malicious users.