BranchAndMultiplyTogether

class penzai.nn.combinators.BranchAndMultiplyTogether

Bases: Layer

A data-flow branch with multiplicative interactions between branches.

BranchAndMultiplyTogether can be used to compose multiple operations that all produce values of the same shape (or broadcast-compatible shapes). Each branch receives the same input, and the outputs of all branches are multiplied together.

This is a common pattern in “gated” neural networks, where one branch computes a gate using a nonlinear activation function, and the other branch computes a value that is multiplied by the gate.

Variables:

branches (list[layer_base.Layer]) – A list of layers that will be run on the same input, and have their outputs multiplied together.

Methods

__init__(branches)
__call__(arg, **side_inputs)

Attributes

branches

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.