學習率預熱(transformers.get_linear_schedule_with_warmup)

學習率預熱

• 在預熱期間，學習率從0線性增加到優化器中的初始lr。

• 在預熱階段之後創建一個schedule，使其學習率從優化器中的初始lr線性降低到0

Parameters

• optimizer (Optimizer) – The optimizer for which to schedule the learning rate.

• num_warmup_steps (int) – The number of steps for the warmup phase.

• num_training_steps (int) – The total number of training steps.

• last_epoch (int, optional, defaults to -1) – The index of the last epoch when resuming training.

Returns

• torch.optim.lr_scheduler.LambdaLR with the appropriate schedule.

# training steps 的數量: [number of batches] x [number of epochs].
total_steps = len(train_dataloader) * epochs

# 設計 learning rate scheduler
scheduler = get_linear_schedule_with_warmup(optimizer, num_warmup_steps = 50, 
                                            num_training_steps = total_steps)