Gradually warm-up(increasing) learning rate for pytorch's optimizer. Proposed in 'Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour'.
$ pip install git+https://github.com/ildoonet/pytorch-gradual-warmup-lr.git
from warmup_scheduler import GradualWarmupScheduler
scheduler_plateau = torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer, patience=3, verbose=True)
scheduler_warmup = GradualWarmupScheduler(optimizer, multiplier=8, total_epoch=10, after_scheduler=scheduler_plateau)
for epoch in range(train_epoch):
scheduler_warmup.step() # 10 epoch warmup, after that schedule as scheduler_plateau
...