Schedulers

class ezflow.functional.scheduler.CosineWarmupScheduler(optimizer, warmup=100, max_iters=200)[source]

Coaine learning rate warmup scheduler

Parameters

  • optimizer (torch.optim.Optimizer) – Optimizer to be used with the scheduler

  • warmup (int) – Number of epochs to warmup the learning rate

  • max_iters (int) – Maximum number of iterations to train the model

get_lr_factor(epoch)[source]
Parameters

epoch (int) – Current epoch

Returns

Learning rate factor

Return type

float