jax.experimental.pallas.BlockSpec#

class jax.experimental.pallas.BlockSpec(block_shape=None, index_map=None, indexing_mode=None, pipeline_mode=None, *, memory_space=None)[source]#

Specifies how an array should be sliced for each invocation of a kernel.

The block_shape is a sequence of int | None`s, or `BlockDim types (e.g. pl.Element, pl.Squeezed, pl.Blocked, pl.BoundedSlice). Each of these types specify the size of the block dimension. None is used to specify a dimension that is squeezed out of the kernel. The BlockDim types allow for more fine-grained control over the indexing of the dimension. The index_map needs to return a tuple of the same length as block_shape, which each entry depending on the type of BlockDim.

See BlockSpec, a.k.a. how to chunk up inputs and the individual BlockDim type docstrings for more details.

Parameters:

  • block_shape (Sequence[BlockDim | int | None] | None)

  • index_map (Callable[..., Any] | None)

  • indexing_mode (Any | None)

  • pipeline_mode (Buffered | None)

  • memory_space (Any | None)

__init__(block_shape=None, index_map=None, indexing_mode=None, pipeline_mode=None, *, memory_space=None)#
Parameters:

  • block_shape (Sequence[BlockDim | int | None] | None | None)

  • index_map (Callable[..., Any] | None | None)

  • indexing_mode (Any | None | None)

  • pipeline_mode (Buffered | None | None)

  • memory_space (Any | None | None)

Return type:

None

Methods

__init__([block_shape, index_map, ...])

replace(**changes)

Return a new object replacing specified fields with new values.

to_block_mapping(origin, array_aval, *, ...)

Attributes

block_shape

index_map

indexing_mode

memory_space

pipeline_mode