Core

Core#

Optimizer#

class Optimizer(params, defaults: 'dict'):

Base class for all optimizers, inspired by PyTorch’s optim.Optimizer.

Parameters

  • params : iterable – An iterable of parameters to optimize or dicts defining parameter groups.

  • defaults : dict – A dict containing default values of optimization options (e.g. learning rate, momentum).

Methods#

step#

def step(self) -> 'None':

Performs a single optimization step.

zero_grad#

def zero_grad(self) -> 'None':

Clears the gradients of all optimized parameters.

Contents