Skip to content

Official Repository for the VISAPP 2023 paper "Dynamically Modular and Sparse General Continual Learning".

License

Notifications You must be signed in to change notification settings

NeurAI-Lab/DynamicContinualLearning

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Dynamos

This is the official repository of VISAPP 2023 paper "Dynamically Modular and Sparse General Continual Learning" by Arnav Varma, Elahe Arani, Bahram Zonooz.
The repo is based on the Mammoth framework.

Schematic of Dynamos

image info

Setup

  • Use requirements.txt to set up environment.
  • Use main_policy.py to run experiments.
  • Datasets are assumed to be accessible from a parent directory in the format /path/to/datasets/{dataset name}
  • Sample commands:
python main_policy.py --model dynamos --dataset {name of dataset} --tensorboard --data-path {/path/to/dataset/} --save-path {/path/to/output/directory} --csv_log --exp-name {name of exp} --buffer_size 500 --seed 42 --load_best_args
python main_policy.py --model dynamos --dataset {name of dataset} --tensorboard --data-path {/path/to/dataset/} --save-path {/path/to/output/directory} --csv_log --exp-name {name of exp} --buffer_size 500 --seed 42 --n_epochs 1 --batch_size 10 --lr 0.07 --alpha 0.2 --beta 2.0 --minibatch_size 10 --nf 32 --policy-alpha 0.2 --prototype-loss 0.3 --policy-penalty -500 --reward-weight 0.5 --keep-ratio 0.7

Cite Our Work

If you find the code useful in your research, please consider citing our paper:

@conference{visapp23,
author={Arnav Varma. and Elahe Arani. and Bahram Zonooz.},
title={Dynamically Modular and Sparse General Continual Learning},
booktitle={Proceedings of the 18th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 5: VISAPP,},
year={2023},
pages={262-273},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0011790200003417},
isbn={978-989-758-634-7},
issn={2184-4321},
}

License

This project is licensed under the terms of the MIT license.

About

Official Repository for the VISAPP 2023 paper "Dynamically Modular and Sparse General Continual Learning".

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages