training time and number of GPUs

XPixelGroup / RankSRGAN

ICCV 2019 (oral) RankSRGAN: Generative Adversarial Networks with Ranker for Image Super-Resolution. PyTorch implementation

269 stars 60 forks source link

training time and number of GPUs #2

Open Marshall-yao opened 5 years ago

Marshall-yao commented 5 years ago

Hi, thanks for your wonderful work and opening source. Could you please tell me how long did you train the model , the kind of GPU and number of GPUs?

Best regards

WenlongZhang0517 commented 5 years ago

Thanks for your interest in our work! It takes about 55 hours (total 500 K iterations) to train the RankSRGAN using 4 Titan Xp GPUs. The training time is slower than standard SRGAN. The reason is that the patch size of standard SRGAN is set to 128, due to Ranker needs a larger size to predict the ranking score, we set the patch size to 296.

Best regards

Marshall-yao commented 5 years ago

@WenlongZhang0724 Thanks very much for your patient reply. Yes, due to larger patch size, training time is longer.