Skip to content

Fix lr_schedule update for multi-gpu settings on train_vae #15

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
miquel-espinosa wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Fix lr_schedule update for multi-gpu settings on train_vae #15

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 5 additions & 5 deletions train_vae.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -153,6 +153,9 @@ def __init__(
dl = self.accelerator.prepare(data_loader)
self.dl = cycle(dl)

# step counter state
self.step = 0

# optimizer

self.opt_ae = torch.optim.AdamW(list(model.encoder.parameters())+
Expand All @@ -162,7 +165,7 @@ def __init__(
lr=train_lr)
self.opt_disc = torch.optim.AdamW(model.loss.discriminator.parameters(), lr=train_lr)
min_lr = cfg['trainer']['min_lr']
lr_lambda = lambda iter: max((1 - iter / train_num_steps) ** 0.95, min_lr/train_lr)
lr_lambda = lambda _: max((1 - self.step / train_num_steps) ** 0.95, min_lr/train_lr)
self.lr_scheduler_ae = torch.optim.lr_scheduler.LambdaLR(self.opt_ae, lr_lambda=lr_lambda)
self.lr_scheduler_disc = torch.optim.lr_scheduler.LambdaLR(self.opt_disc, lr_lambda=lr_lambda)
# for logging results in a folder periodically
Expand All @@ -174,9 +177,6 @@ def __init__(
self.results_folder = Path(results_folder)
self.results_folder.mkdir(exist_ok = True)

# step counter state

self.step = 0

# prepare model, dataloader, optimizer with accelerator

Expand Down Expand Up @@ -330,4 +330,4 @@ def train(self):
if __name__ == "__main__":
args = parse_args()
main(args)
pass
pass