Awesome

Learning to Continually Learn with the Bayesian Principle

This repository contains the code for our ICML 2024 paper titled Learning to Continually Learn with the Bayesian Principle.

pip install -r requirements.txt

The basic usage of the training script is as follows:

python train.py -c [config] -o [override options] -l [log directory]

In cfg/, we provide the configuration files for all the experiments in the paper.

After training, we evaluate the models using the following command:

python evaluate.py -l [log directory]

The SB-MCL (MAP) scores can be attained by turning on the map option.

python evaluate.py -l [SB-MCL log directory] -o "map=True"

All datasets except MS-Celeb-1M are downloaded automatically by the code. Note that downloading the CASIA dataset may take days.

Use BitTorrent to download the dataset from Academic Torrents.

transmission-cli https://academictorrents.com/download/9e67eb7cc23c9417f39778a8e06cca5e26196a97.torrent -w data