vllm.tasks ¶

GENERATION_TASKS: tuple[GenerationTask, ...] = get_args(
    GenerationTask
)

GenerationTask = Literal[
    "generate", "transcription", "realtime"
]

POOLING_TASKS: tuple[PoolingTask, ...] = get_args(
    PoolingTask
)

PoolingTask = Literal[
    "embed",
    "classify",
    "score",
    "token_embed",
    "token_classify",
    "plugin",
]

SupportedTask = Literal[GenerationTask, PoolingTask]