[Multi Modal] Add an env var for message queue max chunk bytes (#19242)

Signed-off-by: yZhen <yZhen@fb.com>
Co-authored-by: yZhen <yZhen@fb.com>
This commit is contained in:
jennyyyyzhen
2025-06-08 21:39:12 +08:00
committed by GitHub
co-authored by yZhen
parent c123bc33f9
commit cda10fa3e2
2 changed files with 12 additions and 1 deletions
+7
View File
@@ -123,6 +123,7 @@ if TYPE_CHECKING:
VLLM_MAX_TOKENS_PER_EXPERT_FP4_MOE: int = 163840
VLLM_TOOL_PARSE_REGEX_TIMEOUT_SECONDS: int = 1
VLLM_SLEEP_WHEN_IDLE: bool = False
VLLM_MQ_MAX_CHUNK_BYTES_MB: int = 16
def get_default_cache_root():
@@ -847,6 +848,12 @@ environment_variables: dict[str, Callable[[], Any]] = {
# latency penalty when a request eventually comes.
"VLLM_SLEEP_WHEN_IDLE":
lambda: bool(int(os.getenv("VLLM_SLEEP_WHEN_IDLE", "0"))),
# Control the max chunk bytes (in MB) for the rpc message queue.
# Object larger than this threshold will be broadcast to worker
# processes via zmq.
"VLLM_MQ_MAX_CHUNK_BYTES_MB":
lambda: int(os.getenv("VLLM_MQ_MAX_CHUNK_BYTES_MB", "16")),
}
# --8<-- [end:env-vars-definition]