Skip to content
/ darts Public
forked from quark0/darts

Differentiable architecture search for convolutional and recurrent networks

License

Notifications You must be signed in to change notification settings

chicm-ms/darts

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Differentiable Architecture Search

Code accompanying the paper

DARTS: Differentiable Architecture Search
Hanxiao Liu, Karen Simonyan, Yiming Yang.
arXiv:1806.09055.

darts

The algorithm is based on continuous relaxation and gradient descent in the architecture space. It is able to efficiently design high-performance convolutional architectures for image classification (on CIFAR-10 and ImageNet) and recurrent architectures for language modeling (on Penn Treebank and WikiText-2). Only a single GPU is required.

Requirements

Python >= 3.5.5, PyTorch == 0.3.1, torchvision >= 0.2.1

PyTorch 0.4 will be supported soon.

Datasets

Instructions for acquiring PTB and WT2 can be found here. While CIFAR-10 can be automatically downloaded by torchvision, ImageNet needs to be manually downloaded (preferably to a SSD) following the instructions here.

Architecture Search

To carry out architecture search, run

cd cnn && python train_search.py --unrolled     # for conv cells on CIFAR-10
cd rnn && python train_search.py --unrolled     # for recurrent cells on PTB

Snapshots of the most likely convolutional & recurrent cells over time:

progress_convolutional progress_recurrent

Architecture Evaluation

To reproduce our results using the best cells, run

cd cnn && python train.py --auxiliary --cutout            # CIFAR-10
cd rnn && python train.py                                 # PTB
cd rnn && python train.py --data ../data/wikitext-2 \     # WT2
            --dropouth 0.15 --emsize 700 --nhidlast 700 --nhid 700 --wdecay 5e-7
cd cnn && python train_imagenet.py --auxiliary            # ImageNet

Customized architectures are supported through the --arch flag once specified in genotypes.py.

About

Differentiable architecture search for convolutional and recurrent networks

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%