if i need train it first and than evalutate?

chenxin061 / pdarts

Codes for our paper "Progressive Differentiable Architecture Search:Bridging the Depth Gap between Search and Evaluation"

Other

359 stars 83 forks source link

if i need train it first and than evalutate? #15

Closed johsnows closed 5 years ago

johsnows commented 5 years ago

or evalutate code train_cifar include the search process?

Catosine commented 5 years ago

Hi @johsnows !

Sorry I can't catch you. To evaluate a searched model generated by PDARTS, you would always need to retrain it by provided codes such as train_cifar.py, or implement your own version of retrain code.

GL,

johsnows commented 5 years ago

就是先用train_search.py 再train_cifar.py这样是嘛

chenxin061 commented 5 years ago

Use train_search.py to do architecture search and train_cifar.py/train_imagenet.py to do architecture evaluation.

johsnows commented 4 years ago

after search on the train_search.py, we get many genotype, no restrict skipconnetct genotype and genotypes with number of skip-connect:7,6,5,4,3,2,1,0, which one is best? which one is recommended to evaluate?