alr.training.pl_mixup_cyclic

Classses

CyclicPLMixupTrainer

class alr.training.pl_mixup_cyclic.CyclicPLMixupTrainer(model: torch.nn.modules.module.Module, optimiser: str, train_transform: Callable, test_transform: Callable, optimiser_kwargs: dict, loader_kwargs: dict, rfls_length: int, log_dir: Optional[str] = None, alpha: Optional[float] = 1.0, min_labelled: Union[float, int, None] = 16, num_classes: Optional[int] = 10, data_augmentation: Optional[Callable] = None, batch_size: Optional[int] = 100, patience: Union[Tuple[int, int], int, None] = (5, 25), lr_patience: Optional[int] = 10, device: Union[str, torch.device, None] = None)[source]

Bases: alr.training.pl_mixup.PLMixupTrainer

fit(train: torch.utils.data.dataset.Dataset, val: torch.utils.data.dataset.Dataset, pool: torch.utils.data.dataset.Dataset, epochs: Optional[Tuple[int, int, int]] = (50, 400, 60))[source]

Functions

cyclic_annealer

alr.training.pl_mixup_cyclic.cyclic_annealer(t, T, M, init_lr=0.2)[source]