Neural Architecture Search on ImageNet in Four GPU Hours: A Theoretically Inspired Perspective [PDF]
Wuyang Chen, Xinyu Gong, Zhangyang Wang
In ICLR 2021.
We present TE-NAS, the first published training-free neural architecture search method with extremely fast search speed (no gradient descent at all!) and high-quality performance.
Highlights:
- Trainig-free and label-free NAS: we achieved extreme fast neural architecture search without a single gradient descent.
- Bridging the theory-application gap: We identified two training-free indicators to rank the quality of deep networks: the condition number of their NTKs, and the number of linear regions in their input space.
- SOTA: TE-NAS achieved extremely fast search speed (one 1080Ti, 20 minutes on NAS-Bench-201 space / four hours on DARTS space on ImageNet) and maintains competitive accuracy.
- Ubuntu 16.04
- Python 3.6.9
- CUDA 10.1 (lower versions may work but were not tested)
- NVIDIA GPU + CuDNN v7.3
This repository has been tested on GTX 1080Ti. Configurations may need to be changed on different platforms.
- Clone this repo:
git clone https://github.com/chenwydj/TENAS.git
cd TENAS
- Install dependencies:
pip install -r requirements.txt
- Please follow the guideline here to prepare the CIFAR-10/100 and ImageNet dataset, and also the NAS-Bench-201 database.
- Remember to properly set the
TORCH_HOME
anddata_paths
in theprune_launch.py
.
python prune_launch.py --space nas-bench-201 --dataset cifar10 --gpu 0
python prune_launch.py --space nas-bench-201 --dataset cifar100 --gpu 0
python prune_launch.py --space nas-bench-201 --dataset ImageNet16-120 --gpu 0
python prune_launch.py --space darts --dataset cifar10 --gpu 0
python prune_launch.py --space darts --dataset imagenet-1k --gpu 0
- For architectures searched on
nas-bench-201
, the accuracies are immediately available at the end of search (from the console output). - For architectures searched on
darts
, please use DARTS_evaluation for training the searched architecture from scratch and evaluation.
@inproceedings{chen2020tenas,
title={Neural Architecture Search on ImageNet in Four GPU Hours: A Theoretically Inspired Perspective},
author={Chen, Wuyang and Gong, Xinyu and Wang, Zhangyang},
booktitle={International Conference on Learning Representations},
year={2021}
}
- Code base from NAS-Bench-201.