opto.optimizers.optimizer
#
Module Contents#
Classes#
An optimizer is responsible for updating the parameters based on the feedback. |
|
API#
- class AbstractOptimizer(parameters: List[opto.trace.nodes.ParameterNode], *args, **kwargs)[source]#
An optimizer is responsible for updating the parameters based on the feedback.
Initialization
- abstract property propagator#
Return a Propagator object that can be used to propagate feedback in backward.
- class Optimizer(parameters: List[opto.trace.nodes.ParameterNode], *args, propagator: opto.trace.propagators.propagators.Propagator = None, **kwargs)[source]#
Bases:
opto.optimizers.optimizer.AbstractOptimizer
- property propagator#
- property trace_graph#
Aggregate the graphs of all the parameters.
- update(update_dict: Dict[opto.trace.nodes.ParameterNode, Any])[source]#
Update the trainable parameters given a dictionary of new data.
- backward(node: opto.trace.nodes.Node, *args, **kwargs)[source]#
Propagate the feedback backward.