Open TheSunWillRise opened 4 years ago
I believe you have got the correct understanding of it.
The released train codes are not for training super-net, right? I find it is hard to train super-net without detail setting. Could you please release the codes or configs for training super-net?
right, btw, how you split the validation and training set during architecture search? @gmh14
@815961618 @gmh14 Is there a way to obtain the trained super-net on any dataset?
Hi, in your paper, you say "After training the supernet, we randomly sample 1,000 architectures from the supernet and finetune each of them for 3 epochs", how to do this? Constructing a network according to a randomly generated genotype_list and evaluate this network?