Code for Contrastive Continual Learning with Importance Sampling and Prototype-Instance Relation Distillation(AAAI 2024) paper. Our code is based on the implementation of Co2L.
Similar to previous work Co2L, CCLIS needs to train a model to learn contrastive representations and then test it with linear evaluation method. To run the code, you can run the script cifar10_200.sh or follow below two commands:
python CCLIS/main.py --batch_size 512 --dataset cifar10 --model resnet18 --mem_size 200 --epochs 100 --start_epoch 500 --learning_rate 1.0 --temp 0.5 --current_temp 0.2 --past_temp 0.1 --distill_type PRD --distill_power 0.6 --cosine --syncBN --seed 0
python CCLIS/main_linear_buffer.py --dataset cifar10 --learning_rate 0.05 --target_task 4 --ckpt ./save_weight_200/cifar10_models/cifar10_32_resnet18_lr_1.0_0.01_decay_0.0001_bsz_512_temp_0.5_trial_0_500_100_0.2_0.1_0.6_distill_type_PRD_freeze_prototypes_niters_5_seed_0_cosine_warm/ --logpt ./save_weight_200/logs/cifar10_32_resnet18_lr_1.0_0.01_decay_0.0001_bsz_512_temp_0.5_trial_0_500_100_0.2_0.1_0.6_distill_type_PRD_freeze_prototypes_niters_5_seed_0_cosine_warm/
If you find our code useful in your research, please consider citing our paper:
@inproceedings{li2024contrastive,
title={Contrastive Continual Learning with Importance Sampling and Prototype-Instance Relation Distillation},
author={Li, Jiyong and Azizov, Dilshod and Yang, LI and Liang, Shangsong},
booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
volume={38},
number={12},
pages={13554--13562},
year={2024}
}