nntoolbox.optim.utils module

nntoolbox.optim.utils.change_lr(optim: torch.optim.optimizer.Optimizer, lrs: Union[float, List[float]])

Change the learning rate of an optimizer

Parameters

• optim – optimizer

• lrs – target learning rate

nntoolbox.optim.utils.get_lr(optim: torch.optim.optimizer.Optimizer) → List[float]

nntoolbox.optim.utils.load_optimizer(optimizer: torch.optim.optimizer.Optimizer, path: str)

Load optimizer state for resuming training

Parameters

• optimizer –

• path –

nntoolbox.optim.utils.plot_schedule(schedule_fn: Callable[[int], float], iterations: int = 30)

Plot the learning rate schedule function

Parameters

• schedule_fn – a function that returns a learning rate given an iteration

• iterations – maximum number of iterations (or epochs)

Returns

nntoolbox.optim.utils.save_optimizer(optimizer: torch.optim.optimizer.Optimizer, path: str)

Save optimizer state for resuming training

Parameters

• optimizer –

• path –