paraqeet.optimizable.Optimizable

class paraqeet.optimizable.Optimizable

Bases: ABC

Optimizable parameter model.

This interface must be implemented by any class that provides optimizable parameters. The optimizer will collect all parameters (by reference) and update their values.

Parameters:

• _name (str | None) – The name of the object.

• _optimizable_parameters (list[Quantity] = []) – The optimizable parameters associated with the object.

• _all_optimizable_parameters (list[Quantity] = []) – All the optimizable parameters considered in the optimization. This is needed to set correctly the size of the gradient and its ordering.

__init__()

Methods

__init__()
get_parameters()	Return all parameters of this class that can be optimized.
set_all_optimizable_parameters(all_params)	Set all optimizable parameters in the optimization.
set_optimizable_parameters(params)	Set which parameters associated with the object shall be considered during optimization.

Attributes

all_optimizable_parameters	Get the optimizable parameters
name	Get the name of the parameter.
optimizable_parameters	Get the optimizable parameters

property all_optimizable_parameters: list[Quantity]

Get the optimizable parameters

Returns:

The list of all the optimizable parameters considered in the optimization

Return type:

list[Quantity]

abstract get_parameters()

Return all parameters of this class that can be optimized.

Raises:

NotImplementedError – Subclasses derived from this class must implement this method.

Return type:

list[Quantity]

property name: str | None

Get the name of the parameter.

Returns:

Name of the parameter.

Return type:

str | None

property optimizable_parameters: list[Quantity]

Get the optimizable parameters

Returns:

The list of optimizable parameters associated with the object.

Return type:

list[Quantity]

set_all_optimizable_parameters(all_params)

Set all optimizable parameters in the optimization.

Parameters:

• params (List[Quantity]) – List of optimizable parameters to be set.

• all_params (list[Quantity])

Return type:

None

set_optimizable_parameters(params)

Set which parameters associated with the object shall be considered during optimization.

All quantities that are not in the response of get_parameters will be filtered out. This function is called by the optimizer before gradient based optimization to tell the layers which gradients to compute.

Parameters:

params (list[Quantity]) – List of optimizable parameters to be set.

Return type:

None