search

D-X-Y / AutoDL-Projects

Automated deep learning algorithms implemented in PyTorch.
MIT License
1.57k stars 282 forks source link

Where can I get the code for regularized_ea train and evaluation #118

Closed xielinzhen closed 2 years ago

xielinzhen commented 2 years ago

Which Algorithm regularized_ea

Describe the Question python ./exps/NATS-algos/regularized_ea.py --dataset cifar10 --search_space tss --ea_cycles 200 --ea_population 10 --ea_sample_size 3 --rand_seed 1 I would like to execute the above command to get the time and accuracy of the actual training evaluation on my own device, but there is only simulation training and evaluation, https://github.com/D-X-Y/AutoDL-Projects/blob/main/exps/NATS-algos/regularized_ea.py#L172

where can I get the code for the actual training evaluation

D-X-Y commented 2 years ago

Good question!

The actual training and evaluation codes, which are used when I built this benchmark, is at: https://github.com/D-X-Y/AutoDL-Projects/blob/main/exps/NATS-Bench/main-tss.py#L37. However, if you want to integrate it with the REA algorithm, you need to modify the function in main-tss.py and replace the simulate_train_eval call with the modified function.

xielinzhen commented 2 years ago

Thanks for your reply