# Neural Architecture Search Without Training 

> :warning: Note: this repository has been updated to reflect the second version of the paper to appear on arXiv 1 March. :warning:
> For the original version, refer to the [version v1.0](https://github.com/BayesWatch/nas-without-training/releases/tag/v1.0).

## Usage 

Create a conda environment using the env.yml file

```bash
conda env create -f env.yml
```

Activate the environment and follow the instructions to install

Install nasbench (see https://github.com/google-research/nasbench)

Download the NDS data from https://github.com/facebookresearch/nds and place the json files in naswot-codebase/nds_data/
Download the NASbench101 data (see https://github.com/google-research/nasbench)
Download the NASbench201 data (see https://github.com/D-X-Y/NAS-Bench-201)

Reproduce all of the results by running 

```bash
./scorehook.sh
```

The code is licensed under the MIT licence.

## Citing us

If you use or build on our work, please consider citing us:

```bibtex
@misc{mellor2020neural,
    title={Neural Architecture Search without Training},
    author={Joseph Mellor and Jack Turner and Amos Storkey and Elliot J. Crowley},
    year={2020},
    eprint={2006.04647},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}
```