2020-01-15 15:43:07 +01:00
# DARTS: Differentiable Architecture Search
DARTS: Differentiable Architecture Search is accepted by ICLR 2019.
In this paper, Hanxiao proposed a differentiable neural architecture search method, named as DARTS.
Recently, DARTS becomes very popular due to its simplicity and performance.
2020-01-17 12:28:33 +01:00
## Run DARTS on the NAS-Bench-201 search space
2020-01-15 15:43:07 +01:00
```
CUDA_VISIBLE_DEVICES=0 bash ./scripts-search/algos/DARTS-V2.sh cifar10 1 -1
CUDA_VISIBLE_DEVICES=0 bash ./scripts-search/algos/GDAS.sh cifar10 1 -1
```
2020-01-17 12:28:33 +01:00
## Run the first-order DARTS on the NASNet/DARTS search space
This command will start to use the first-order DARTS to search architectures on the DARTS search space.
2020-01-17 12:14:47 +01:00
```
CUDA_VISIBLE_DEVICES=0 bash ./scripts-search/DARTS1V-search-NASNet-space.sh cifar10 -1
```
2021-03-01 14:02:29 +01:00
After searching, if you want to train the searched architecture found by the above scripts, you need to add the config of that architecture (will be printed in log) in [genotypes.py ](https://github.com/D-X-Y/AutoDL-Projects/blob/main/lib/nas_infer_model/DXYs/genotypes.py ).
2020-01-17 12:28:33 +01:00
In future, I will add a more eligent way to train the searched architecture from the DARTS search space.
2020-01-15 15:43:07 +01:00
# Citation
```
@inproceedings {liu2019darts,
title = {{DARTS}: Differentiable architecture search},
author = {Liu, Hanxiao and Simonyan, Karen and Yang, Yiming},
booktitle = {International Conference on Learning Representations (ICLR)},
year = {2019}
}
```