mindspore.nn.cosine_decay_lr

mindspore.nn.cosine_decay_lr(min_lr, max_lr, total_step, step_per_epoch, decay_epoch)[source]

Calculates learning rate base on cosine decay function. The learning rate for each step will be stored in a list.

For the i-th step, the formula of computing decayed_learning_rate[i] is:

\[decayed\_learning\_rate[i] = min\_lr + 0.5 * (max\_lr - min\_lr) * (1 + cos(\frac{current\_epoch}{decay\_epoch}\pi))\]

Where \(current\_epoch=floor(\frac{i}{step\_per\_epoch})\).

Parameters
  • min_lr (float) – The minimum value of learning rate.

  • max_lr (float) – The maximum value of learning rate.

  • total_step (int) – The total number of steps.

  • step_per_epoch (int) – The number of steps in per epoch.

  • decay_epoch (int) – Number of epochs to decay over.

Returns

list[float]. The size of list is total_step.

Raises
  • TypeError – If min_lr or max_lr is not a float.

  • TypeError – If total_step or step_per_epoch or decay_epoch is not an int.

  • ValueError – If max_lr is not greater than 0 or min_lr is less than 0.

  • ValueError – If total_step or step_per_epoch or decay_epoch is less than 0.

  • ValueError – If min_lr is greater than or equal to max_lr.

Supported Platforms:

Ascend GPU CPU

Examples

>>> import mindspore.nn as nn
>>>
>>> min_lr = 0.01
>>> max_lr = 0.1
>>> total_step = 6
>>> step_per_epoch = 2
>>> decay_epoch = 2
>>> lr = nn.cosine_decay_lr(min_lr, max_lr, total_step, step_per_epoch, decay_epoch)
>>> net = nn.Dense(2, 3)
>>> optim = nn.SGD(net.trainable_params, learning_rate=lr)