This is a PyTorch implementation of the paper Parameter-Efficient Fine-Tuning with Controls. The goal of our work is to provide a pure control view for the well-known LoRA algorithm.
- CUDA 11.2 + PyTorch 2.1.0 + torchvision 0.16.0
- timm 1.0.7
- easydict
The mae_pretrain_vit_b model is available here.
Start
# image
python main.py \
--batch_size 128 --cls_token \
--drop_path 0.0 --lr_decay 0.97 \
--dataset cifar100 --ffn_adaptThe project is based on MAE and AdaptFormer. Thank all the authors for their awesome works.
@inproceedings{zhangparameter,
title={Parameter-Efficient Fine-Tuning with Controls},
author={Zhang, Chi and Jingpu, Cheng and Xu, Yanyu and Li, Qianxiao},
booktitle={Forty-first International Conference on Machine Learning}
}
This project is under the MIT license. See LICENSE for details.