def get_lr(self, runner, base_lr):
if self.by_epoch:
progress = runner.epoch
max_progress = runner.max_epochs
else:
progress = runner.iter
max_progress = runner.max_iters
target_lr = self.min_lr
if self.warmup is not None:
progress = progress - self.warmup_iters
max_progress = max_progress - self.warmup_iters
factor = progress / max_progress
return annealing_cos(base_lr, target_lr, factor)
def get_lr(self, runner, base_lr):
if self.last_epoch == -1:
self.last_epoch = runner.max_epochs
if self.by_epoch:
progress = runner.epoch
max_progress = self.last_epoch
else:
progress = runner.iter
max_progress = runner.max_iters
if self.min_lr_ratio is not None:
target_lr = base_lr * self.min_lr_ratio
else:
target_lr = self.min_lr
if self.by_epoch and self.last_epoch < progress:
return target_lr
return annealing_cos(base_lr, target_lr, progress / max_progress)
def get_lr(self, runner, base_lr):
last_iter = len(runner.data_loader) * self.num_last_epochs
if self.by_epoch:
progress = runner.epoch
max_progress = runner.max_epochs
else:
progress = runner.iter
max_progress = runner.max_iters
progress += 1
if self.min_lr_ratio is not None:
target_lr = base_lr * self.min_lr_ratio
else:
target_lr = self.min_lr
if progress >= max_progress - last_iter:
# fixed learning rate
return target_lr
else:
return annealing_cos(
base_lr, target_lr, (progress - self.warmup_iters) /
(max_progress - self.warmup_iters - last_iter))