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.
Type
pip install continual-learning
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.
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.