JOINT

Arguments

This module contains the implementation of the Joint CL model.

The Joint model is the upper bound of the CL scenario, as it has access to all the data from all the tasks. This model is required for the domain-il scenario, while class-il and task-il scenarios can use the –joint=1 flag.

Classes

class models.joint.Joint(backbone, loss, args, transform, dataset=None)

Bases: ContinualModel

COMPATIBILITY: List[str] = ['class-il', 'domain-il', 'task-il']

NAME: str = 'joint'

end_task(dataset)

This version of joint training simply saves all data from previous tasks and then trains on all data at the end of the last one.

observe(*args, **kwargs)