search

alterzero / DBPN-Pytorch

The project is an official implement of our CVPR2018 paper "Deep Back-Projection Networks for Super-Resolution" (Winner of NTIRE2018 and PIRM2018)
https://alterzero.github.io/projects/DBPN.html
MIT License
566 stars 164 forks source link

About training on GAN #56

Open yangyingni opened 4 years ago

yangyingni commented 4 years ago

Thank you for your work very much.I was training main_gan,and now i have three questions 1.Do i need to pretrain Generator and Discriminator firstly? 2.How can i continue before epoches i have ever trained? 3.'--pretrained_iter', type=int, default=100;
'--pretrained_sr', default='dl00DBPNLLPIRM_pretrained_50.pth'; '--load_pretrained', type=bool, default=True; '--pretrained_D', default='dnnDBPNLLPIRM_RESNET_epoch_Discriminator_499.pth' '--load_pretrained_D', type=bool, default=True What's the relationship between them? Hope to hear from you,thank you.

shazib-summar commented 4 years ago

Same here, the GAN takes a lot of time to train (I'm talking days on google colab). It would be helpful if it was explained explicitly how to continue training from a checkpoint saved previously. All the arguments are confusing as well.