MeCo/README.md
HamsterMimi 1a57decf65 update
2023-12-23 11:08:58 +08:00

83 lines
2.4 KiB
Markdown

# MeCo: Zero-Cost Proxy for NAS Via Minimum Eigenvalue of Correlation on Feature Maps
## Installation
```
Python >= 3.6
PyTorch >= 2.0.0
nas-bench-201
```
## Preparation
1. Download three datasets (CIFAR-10, CIFAR-100, ImageNet16-120) from [Google Drive](https://drive.google.com/drive/folders/1T3UIyZXUhMmIuJLOBMIYKAsJknAtrrO4), place them into the directory `./data`
2. Download the [`data` directory](https://drive.google.com/drive/folders/18Eia6YuTE5tn5Lis_43h30HYpnF9Ynqf?usp=sharing) and save it to the root folder of this repo.
3. Download the benchmark files of NAS-Bench-201 from [Google Drive](https://drive.google.com/file/d/1SKW0Cu0u8-gb18zDpaAGi0f74UdXeGKs/view) , put them into the directory `./data`
4. Download the [NAS-Bench-101 dataset](https://storage.googleapis.com/nasbench/nasbench_only108.tfrecord), put it into the directory `./data`
5. Install `zero-cost-nas`
```bash
cd zero-cost-nas
pip install .
cd ..
```
## Usage/Examples
### Correlation Experiment
```bash
cd correlation
python NAS_Bench_101.py
python NAS_Bench_201.py
```
### Experiments on NAS-Bench-201
1. Run Zero-Cost-PT with appointed zero-cost proxy:
```bash
cd exp_scripts
bash zerocostpt_nb201_pipline.sh --metric [metric] --batch_size [batch_size] --seed [seed]
```
You can choice metric from `['snip', 'fisher', 'synflow', 'grad_norm', 'grasp', 'jacob_cov','tenas', 'zico', 'meco'] `
### Experiments on DARTS-CNN Space
#### 1. DARTS CNN Space
```bash
cd exp_scripts
bash zerocostpt_darts_pipline.sh --metric [metric] --batch_size [batch_size] --seed [seed]
```
#### 2. DARTS Subspaces S1-S4
````bash
cd exp_scripts
bash zerocostpt_darts_pipline.sh --metric [metric] --batch_size [batch_size] --seed [seed] --space [s1-s4]
````
## Reference
Our code is based on [Zero-Cost-PT](https://github.com/zerocostptnas/zerocost_operation_score) and [Zero-Cost-NAS](https://github.com/SamsungLabs/zero-cost-nas).
## Cite
If the code is found useful, we would appreciate it if our paper could be cited with the following Bibtex format
```
@inproceedings{
jiang2023meco,
title={MeCo: Zero-Shot {NAS} with One Data and Single Forward Pass via Minimum Eigenvalue of Correlation},
author={Tangyu Jiang and Haodi Wang and Rongfang Bie},
booktitle={Thirty-seventh Conference on Neural Information Processing Systems},
year={2023},
url={https://openreview.net/forum?id=KFm2lZiI7n}
}
```
For any inquiries, bugs, and assistance on building and running the code, please contact me at jty@mail.bnu.edu.cn