![](figures/motivation.png)
The idea is simple: we view existing parameter-efficient tuning modules, including Adapter, LoRA and VPT, as prompt modules and propose to search the optimal configuration via neural architecture search. Our approach is named NOAH (Neural prOmpt seArcH).
[05/2022] arXiv paper has been released.
conda create -n NOAH python=3.8
conda activate NOAH
pip install -r requirements.txt
cd data/vtab-source
python get_vtab1k.py
-
Images
Please refer to DATASETS.md to download the datasets.
-
Train/Val/Test splits
Please refer to files under
data/XXX/XXX/annotations
for the detail information.
We use the VTAB experiments as examples.
Model | Link |
---|---|
ViT B/16 | link |
sh configs/NOAH/VTAB/supernet/slurm_train_vtab.sh PATH-TO-YOUR-PRETRAINED-MODEL
sh configs/NOAH/VTAB/search/slurm_search_vtab.sh PARAMETERS-LIMITES
sh configs/NOAH/VTAB/subnet/slurm_retrain_vtab.sh PATH-TO-YOUR-PRETRAINED-MODEL
We add the optimal subnet architecture of each dataset in the experiments/NOAH/subnet/VTAB
.
If you use this code in your research, please kindly cite this work.
@inproceedings{zhang2022NOAH,
title={Neural Prompt Search},
author={Yuanhan Zhang and Kaiyang Zhou and Ziwei Liu},
year={2022},
archivePrefix={arXiv},
}
Part of the code is borrowed from CoOp, AutoFormer, timm and mmcv.
Thanks to Chong Zhou (https://chongzhou96.github.io/) for the code of downloading the VTAB-1k.