This repository contains PyTorch implementation code for awesome continual learning method L2P,
Wang, Zifeng, et al. "Learning to prompt for continual learning." CVPR. 2022.
The official Jax implementation is here.
The system I used and tested in
- Ubuntu 20.04.4 LTS
- Slurm 21.08.1
- NVIDIA GeForce RTX 3090
- Python 3.8
First, clone the repository locally:
git clone https://github.com/JH-LEE-KR/l2p-pytorch
cd l2p-pytorch
Then, install the packages below:
pytorch==1.12.1
torchvision==0.13.1
timm==0.6.7
pillow==9.2.0
matplotlib==3.5.3
These packages can be installed easily by
pip install -r requirements.txt
If you already have CIFAR-100 or 5-Datasets (MNIST, Fashion-MNIST, NotMNIST, CIFAR10, SVHN), pass your dataset path to --data-path.
The datasets aren't ready, change the download argument in datasets.py as follows
CIFAR-100
datasets.CIFAR100(download=True)
5-Datasets
datasets.CIFAR10(download=True)
MNIST_RGB(download=True)
FashionMNIST(download=True)
NotMNIST(download=True)
SVHN(download=True)
To train a model via command line:
Single node with single gpu
python -m torch.distributed.launch \
--nproc_per_node=1 \
--use_env main.py \
<cifar100_l2p or five_datasets_l2p> \
--model vit_base_patch16_224 \
--batch-size 16 \
--data-path /local_datasets/ \
--output_dir ./output
Single node with multi gpus
python -m torch.distributed.launch \
--nproc_per_node=<Num GPUs> \
--use_env main.py \
<cifar100_l2p or five_datasets_l2p> \
--model vit_base_patch16_224 \
--batch-size 16 \
--data-path /local_datasets/ \
--output_dir ./output
Also available in Slurm system by changing options on train_cifar100_l2p.sh or train_five_datasets.sh properly.
Distributed training is available via Slurm and submitit:
pip install submitit
To train a model on 2 nodes with 4 gpus each:
python run_with_submitit.py <cifar100_l2p or five_datasets_l2p> --shared_folder <Absolute Path of shared folder for all nodes>
Absolute Path of shared folder must be accessible from all nodes.
According to your environment, you can use NCLL_SOCKET_IFNAME=<Your own IP interface to use for communication> optionally.
To evaluate a trained model:
python -m torch.distributed.launch --nproc_per_node=1 --use_env main.py <cifar100_l2p or five_datasets_l2p> --eval
Test results on a single gpu.
| Name | Acc@1 | Forgetting |
|---|---|---|
| Pytorch-Implementation | 83.77 | 6.63 |
| Reproduce Official-Implementation | 82.59 | 7.88 |
| Paper Results | 83.83 | 7.63 |
| Name | Acc@1 | Forgetting |
|---|---|---|
| Pytorch-Implementation | 80.22 | 3.81 |
| Reproduce Official-Implementation | 79.68 | 3.71 |
| Paper Results | 81.14 | 4.64 |
Here are the metrics used in the test, and their corresponding meanings:
| Metric | Description |
|---|---|
| Acc@1 | Average evaluation accuracy up until the last task |
| Forgetting | Average forgetting up until the last task |
This repository is released under the Apache 2.0 license as found in the LICENSE file.
@inproceedings{wang2022learning,
title={Learning to prompt for continual learning},
author={Wang, Zifeng and Zhang, Zizhao and Lee, Chen-Yu and Zhang, Han and Sun, Ruoxi and Ren, Xiaoqi and Su, Guolong and Perot, Vincent and Dy, Jennifer and Pfister, Tomas},
booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
pages={139--149},
year={2022}
}