About training with 8 gpus #67

New Issue

claunia · 2026-01-29T21:40:41Z

claunia commented

2026-01-29 21:40:41 +00:00

Originally created by @NNNNAI on GitHub (Sep 15, 2021).

Hi xintao, thanks for sharing your great work.
I currently trying to train GFPGAN with 8 gpus, which means the total batchsize will be double. Should I modified some hyperparameter in the train_gfpgan_v1.yml? Such as the learning rate and the totoal step, etc. Thanks again, have a nice day~.

Originally created by @NNNNAI on GitHub (Sep 15, 2021). Hi xintao, thanks for sharing your great work. I currently trying to train GFPGAN with 8 gpus, which means the total batchsize will be double. Should I modified some hyperparameter in the train_gfpgan_v1.yml? Such as the learning rate and the totoal step, etc. Thanks again, have a nice day~.

claunia closed this issue

2026-01-29 21:40:42 +00:00

claunia commented

2026-01-29 21:40:44 +00:00

@xinntao commented on GitHub (Sep 17, 2021):

I have also trained with 8 GPUs without modifying the hyper-parameters. The results are Ok with fewer iterations.
But it may be not the optimal~

@xinntao commented on GitHub (Sep 17, 2021): 1. I have also trained with 8 GPUs without modifying the hyper-parameters. The results are Ok with fewer iterations. But it may be not the optimal~

claunia commented

2026-01-29 21:40:44 +00:00

@NNNNAI commented on GitHub (Sep 17, 2021):

Thanks for your help! Have a great day ~

@NNNNAI commented on GitHub (Sep 17, 2021): Thanks for your help! Have a great day ~

claunia referenced this issue

2026-01-29 21:48:51 +00:00

[PR #67] Add Docker environment & web demo #562

Sign in to join this conversation.