vllm_gaudi.envs
¶
VLLM_USE_HPU_CONTIGUOUS_CACHE_FETCH
module-attribute
¶
VLLM_USE_HPU_CONTIGUOUS_CACHE_FETCH: bool = True
environment_variables
module-attribute
¶
environment_variables: dict[str, Callable[[], Any]] = {
"VLLM_USE_HPU_CONTIGUOUS_CACHE_FETCH": lambda: (
lower() in ("1", "true")
),
"VLLM_HPU_FORCE_CHANNEL_FP8": lambda: (
lower() in ("1", "true")
and get("QUANT_CONFIG", None) is None
),
"VLLM_HPU_HETERO_KV_LAYOUT": lambda: (
lower() in ("0", "false")
),
"VLLM_HPU_MULTI_MODEL_CONFIG": lambda: get(
"VLLM_HPU_MULTI_MODEL_CONFIG", None
),
}