ER ACE#

Arguments#

Options

--task_free0|1|True|False -> bool

Help: Enable task-free training (replay starts from second task)?.

  • Default: False

--use_custom_classifier0|1|True|False -> bool

Help: Use the custom classifier used in the original work.

  • Default: True

Rehearsal arguments

Arguments shared by all rehearsal-based methods.

--buffer_sizeint

Help: The size of the memory buffer.

  • Default: None

--minibatch_sizeint

Help: The batch size of the memory buffer.

  • Default: None

Classes#

class models.er_ace.CustomLinear(indim, outdim, weight=None)[source]#

Bases: Module

forward(x)[source]#
class models.er_ace.ErACE(backbone, loss, args, transform, dataset=None)[source]#

Bases: ContinualModel

Continual learning via Experience Replay with asymmetric cross-entropy.

COMPATIBILITY: List[str] = ['class-il', 'task-il']#
NAME: str = 'er_ace'#
static get_parser(parser)[source]#
Return type:

ArgumentParser

observe(inputs, labels, not_aug_inputs, epoch=None)[source]#