ConstrainShardingByName

class penzai.toolshed.sharding_util.ConstrainShardingByName

Bases: Layer

A layer that constrains the sharding of a tree of NamedArrays by name.

Note: Defined using the experimental v2 API, but compatible with both v1 and v2 APIs (due to not having any parameters or state variables).

Variables:

• mesh (jax.sharding.Mesh) – The Mesh to shard the tree to.

• axis_name_to_mesh_name (dict[str, str | tuple[str, ...]] | None) – A mapping from array axis names to mesh axis names. If an axis name is not present, that axis will not be sharded. If a mesh axis name is a tuple, the corresponding axis will be sharded to multiple mesh axes. If this dictionary is not provided, it will be inferred as an “identity” mapping, where each axis is sharded to a mesh axis with the same name (if present).

Methods

__init__(mesh[, axis_name_to_mesh_name])
__call__(tree, **_unused_side_inputs)

Attributes

axis_name_to_mesh_name
mesh

Inherited Methods

(expand to view inherited methods)

attributes_dict()	Constructs a dictionary with all of the fields in the class.
bind_variables(variables[, allow_unused])	Convenience function to bind variables to a layer.
from_attributes(**field_values)	Directly instantiates a struct given all of its fields.
key_for_field(field_name)	Generates a JAX PyTree key for a given field name.
select()	Wraps this struct in a selection, enabling functional-style mutations.
stateless_call(variable_values, argument, /, ...)	Calls a layer with temporary variables, without modifying its state.
tree_flatten()	Flattens this tree node.
tree_flatten_with_keys()	Flattens this tree node with keys.
tree_unflatten(aux_data, children)	Unflattens this tree node.
treescope_color()	Computes a CSS color to display for this object in treescope.