search

chaofengc / PSFRGAN

PyTorch codes for "Progressive Semantic-Aware Style Transformation for Blind Face Restoration", CVPR2021
Other
370 stars 68 forks source link

lr_policy question #35

Closed zhangyunming closed 3 years ago

zhangyunming commented 3 years ago

企业微信截图_16218448218073

total_epochs is 50 , and lr_decay_iters is 50 , it means the lr is not change in trainning process? all be 0.0002?

if i do not use 'step', --lr_policy', type=str, default='step', help='learning rate policy. [linear | step | plateau | cosine]' which one is better?

chaofengc commented 3 years ago

Thank you for interest in our work. The learning rate fixes at g_lr = 0.0001, d_lr = 0.0004. We did not experiment other policy because this simple strategy is already good enough. You may try other options to see if the performance can be improved further.