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, 10:15 UTC
Fix TYPE_CHECKING stub defaults in envs.py to match actual runtime defaults (#35645)
lin-shh committed
Mar 3, 2026 at 00:59 UTC
8fa68a8ce45641420a080920fccd9139aba80613
1 file changed
+3
-3
vllm/envs.py
+3
-3
@@ -35,7 +35,7 @@ if TYPE_CHECKING:
35
VLLM_USAGE_STATS_SERVER: str = "https://stats.vllm.ai"
36
VLLM_NO_USAGE_STATS: bool = False
37
VLLM_DO_NOT_TRACK: bool = False
38
- VLLM_USAGE_SOURCE: str = ""
38
+ VLLM_USAGE_SOURCE: str = "production"
39
VLLM_CONFIGURE_LOGGING: bool = True
40
VLLM_LOGGING_LEVEL: str = "INFO"
41
VLLM_LOGGING_PREFIX: str = ""
@@ -48,7 +48,7 @@ if TYPE_CHECKING:
48
VLLM_USE_FLASHINFER_SAMPLER: bool | None = None
49
VLLM_PP_LAYER_PARTITION: str | None = None
50
VLLM_CPU_KVCACHE_SPACE: int | None = 0
51
- VLLM_CPU_OMP_THREADS_BIND: str = ""
51
+ VLLM_CPU_OMP_THREADS_BIND: str = "auto"
52
VLLM_CPU_NUM_OF_RESERVED_CPU: int | None = None
53
VLLM_CPU_SGL_KERNEL: bool = False
54
VLLM_XLA_CACHE_PATH: str = os.path.join(VLLM_CACHE_ROOT, "xla_cache")
@@ -89,7 +89,7 @@ if TYPE_CHECKING:
89
VLLM_LORA_RESOLVER_CACHE_DIR: str | None = None
90
VLLM_LORA_RESOLVER_HF_REPO_LIST: str | None = None
91
VLLM_USE_AOT_COMPILE: bool = False
92
- VLLM_USE_BYTECODE_HOOK: bool = False
92
+ VLLM_USE_BYTECODE_HOOK: bool = True
93
VLLM_FORCE_AOT_LOAD: bool = False
94
VLLM_USE_MEGA_AOT_ARTIFACT: bool = False
95
VLLM_USE_TRITON_AWQ: bool = False