Request
deadline_ms (property)
Soft deadline in milliseconds, or None.
inputs (property)
The named inputs, as a dict of tensors.
priority (property)
Higher runs sooner.
streaming (property)
True when the session emits per-step chunks.
__init__
__init____init__(self, inputs: dict, *, priority: int = 0, deadline_ms: int | None = None, streaming: bool = False) -> None
init(self, inputs: dict, *, priority: int = 0, deadline_ms: int | None = None, streaming: bool = False) -> None
A request: named input tensors (a dict keyed by the pipeline's input names) plus scheduling hints. streaming=True asks for per-step chunks; Pipeline.stream() sets it for you.