flashinfer.CakeFmhaRequestOrderedDecodePlan

class flashinfer.CakeFmhaRequestOrderedDecodePlan(module_name: str, batch_size: int, q_len: int, workspace_parts: int, grid: tuple[int, int, int], total_tiles: int, write_lse: bool)

Immutable host-selected launch plan for graph-safe request ordering.

__init__(module_name: str, batch_size: int, q_len: int, workspace_parts: int, grid: tuple[int, int, int], total_tiles: int, write_lse: bool) → None

Methods

__init__(module_name, batch_size, q_len, ...)

Attributes

module_name
batch_size
q_len
workspace_parts
grid
total_tiles
write_lse