Change adversary optim scheduler to sync with model optim scheduler
This commit is contained in:
parent
e7d2435a96
commit
2b7e559910
@ -102,7 +102,7 @@ def gpu_worker(local_rank, args):
|
|||||||
weight_decay=args.weight_decay,
|
weight_decay=args.weight_decay,
|
||||||
)
|
)
|
||||||
scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer,
|
scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer,
|
||||||
factor=0.5, patience=3, verbose=True)
|
factor=0.1, patience=10, verbose=True)
|
||||||
|
|
||||||
adv_model = adv_criterion = adv_optimizer = adv_scheduler = None
|
adv_model = adv_criterion = adv_optimizer = adv_scheduler = None
|
||||||
args.adv = args.adv_model is not None
|
args.adv = args.adv_model is not None
|
||||||
@ -130,7 +130,8 @@ def gpu_worker(local_rank, args):
|
|||||||
betas=(0.5, 0.999),
|
betas=(0.5, 0.999),
|
||||||
weight_decay=args.adv_weight_decay,
|
weight_decay=args.adv_weight_decay,
|
||||||
)
|
)
|
||||||
adv_scheduler = torch.optim.lr_scheduler.StepLR(adv_optimizer, 30, gamma=0.1)
|
adv_scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(adv_optimizer,
|
||||||
|
factor=0.1, patience=10, verbose=True)
|
||||||
|
|
||||||
if args.load_state:
|
if args.load_state:
|
||||||
state = torch.load(args.load_state, map_location=args.device)
|
state = torch.load(args.load_state, map_location=args.device)
|
||||||
@ -185,7 +186,7 @@ def gpu_worker(local_rank, args):
|
|||||||
|
|
||||||
scheduler.step(epoch_loss[0])
|
scheduler.step(epoch_loss[0])
|
||||||
if args.adv:
|
if args.adv:
|
||||||
adv_scheduler.step()
|
adv_scheduler.step(epoch_loss[0])
|
||||||
|
|
||||||
if args.rank == 0:
|
if args.rank == 0:
|
||||||
print(end='', flush=True)
|
print(end='', flush=True)
|
||||||
|
Loading…
Reference in New Issue
Block a user