jaryp / continualai

Continual learning framework

This is a Continual Learning library based on Pytorch, mainly born for personal use, which can be used for fast prototyping, training and to compare different build-in methods over a various numbers of scenarios and benchmarks.

Installation

Type

pip install continual-learning

Continual learning framework

The library is organized in four main modules:

• Benchmarks: This module contains the most used dataset in CL, reimplemented to give more flexibility.
• Extras: It contains many Cl methods, that can be easily used and evalauted.
• Training: This module contains many popular networks used to extract the features from the input samples.
• Evaluation: This modules provides a unified way to evaluate a method over a flexible numbers of metrics/
• Models: In this module you will find different solvers, used to classify the features extracted by a backbone network.

Disclaimer

This is an under development framework which is born to improve coding, and the reproducibility, of the papers in which I have worked during the years. It may be unstable.