Training process of searched architecture

D-X-Y / AutoDL-Projects

Automated deep learning algorithms implemented in PyTorch.

MIT License

1.56k stars 281 forks source link

Training process of searched architecture #17

Closed PluviophileYU closed 5 years ago

PluviophileYU commented 5 years ago

Hi, I read the codes from DARTS firstly and I didn't find the part of training the architecture in your codes, eg: the gumbel trick in your paper. Look forward to your replying!

D-X-Y commented 5 years ago

Please refer to #16 and #2 . Thanks for your patience and understanding.

D-X-Y commented 4 years ago

@PluviophileYU Sorry for the inconvenience. Please see the instructions at here https://github.com/D-X-Y/NAS-Projects#searching-on-the-nasnet-search-space You can use GDAS to search on the DARTS search space by:

CUDA_VISIBLE_DEVICES=0 bash ./scripts-search/GDAS-search-NASNet-space.sh cifar10 1 -1

or on the NAS-bench-102 search space by:

CUDA_VISIBLE_DEVICES=0 bash ./scripts-search/algos/GDAS.sh     cifar10 -1