# CosineAnnealingLR

Source: https://www.tensorplay.cn/docs/generated/tensorplay.optim.lr_scheduler.CosineAnnealingLR.html

# CosineAnnealingLR

class tensorplay.optim.lr_scheduler.CosineAnnealingLR(optimizer: [Optimizer](tensorplay.optim.optimizer.Optimizer.html#tensorplay.optim.optimizer.Optimizer), T_max: [int](https://docs.python.org/3/library/functions.html#int), eta_min: [float](https://docs.python.org/3/library/functions.html#float) = 0.0, last_epoch: [int](https://docs.python.org/3/library/functions.html#int) = -1)[[source]](../_modules/tensorplay/optim/lr_scheduler.html#CosineAnnealingLR)

Set the learning rate of each parameter group using a cosine annealing schedule.

The learning rate is updated recursively using:

\[\eta_{t+1} = \eta_{\min} + (\eta_t - \eta_{\min}) \cdot
\frac{1 + \cos\left(\frac{(T_{cur}+1) \pi}{T_{max}}\right)}
    {1 + \cos\left(\frac{T_{cur} \pi}{T_{max}}\right)}\]

This implements a recursive approximation of the closed-form schedule proposed in
[SGDR: Stochastic Gradient Descent with Warm Restarts](https://arxiv.org/abs/1608.03983):

\[\eta_t = \eta_{\min} + \frac{1}{2}(\eta_{\max} - \eta_{\min}) \left(
    1 + \cos\left(\frac{T_{cur} \pi}{T_{max}}\right) \right)\]

where:

- \(\eta_t\) is the learning rate at step \(t\)

- \(T_{cur}\) is the number of epochs since the last restart

- \(T_{max}\) is the maximum number of epochs in a cycle

Note

Although SGDR includes periodic restarts, this implementation performs cosine annealing
without restarts, so \(T_{cur} = t\) and increases monotonically with each call
to [step()](#tensorplay.optim.lr_scheduler.CosineAnnealingLR.step).

Parameters:

- optimizer ([Optimizer](tensorplay.optim.optimizer.Optimizer.html#tensorplay.optim.optimizer.Optimizer)) – Wrapped optimizer.

- T_max ([int](https://docs.python.org/3/library/functions.html#int)) – Maximum number of iterations.

- eta_min ([float](https://docs.python.org/3/library/functions.html#float)) – Minimum learning rate. Default: 0.

- last_epoch ([int](https://docs.python.org/3/library/functions.html#int)) – The index of the last epoch. Default: -1.

Example

```
>>> # xdoctest: +SKIP
>>> num_epochs = 100
>>> scheduler = CosineAnnealingLR(optimizer, T_max=num_epochs)
>>> for epoch in range(num_epochs):
>>>     train(...)
>>>     validate(...)
>>>     scheduler.step()
```

get_last_lr() → [list](https://docs.python.org/3/library/stdtypes.html#list)[[float](https://docs.python.org/3/library/functions.html#float) | TensorBase]

Get the most recent learning rates computed by this scheduler.

Returns:

A [list](https://docs.python.org/3/library/stdtypes.html#list) of learning rates with entries
for each of the optimizer’s
param_groups, with the same types as
their group["lr"]s.

Return type:

[list](https://docs.python.org/3/library/stdtypes.html#list)[[float](https://docs.python.org/3/library/functions.html#float) | Tensor]

Note

The returned Tensors are copies, and never alias
the optimizer’s group["lr"]s.

get_lr() → [list](https://docs.python.org/3/library/stdtypes.html#list)[[float](https://docs.python.org/3/library/functions.html#float) | TensorBase][[source]](../_modules/tensorplay/optim/lr_scheduler.html#CosineAnnealingLR.get_lr)

Compute the next learning rate for each of the optimizer’s
param_groups.

Scales the group["lr"]s in the optimizer’s
param_groups such that their learning
rates approximate

\[\texttt{eta\_min} + \frac{1}{2} (\texttt{base\_lr} -
\texttt{eta\_min}) \left(1 + \cos\left(\pi \cdot
\frac{\texttt{last\_epoch}}{\texttt{T\_max}}\right) \right)\]

Returns:

A [list](https://docs.python.org/3/library/stdtypes.html#list) of learning rates for each of
the optimizer’s param_groups with the
same types as their current group["lr"]s.

Return type:

[list](https://docs.python.org/3/library/stdtypes.html#list)[[float](https://docs.python.org/3/library/functions.html#float) | Tensor]

Note

If you’re trying to inspect the most recent learning rate, use
[get_last_lr()](#tensorplay.optim.lr_scheduler.CosineAnnealingLR.get_last_lr) instead.

Note

The returned Tensors are copies, and never alias
the optimizer’s group["lr"]s.

load_state_dict(state_dict: [dict](https://docs.python.org/3/library/stdtypes.html#dict)[[str](https://docs.python.org/3/library/stdtypes.html#str), [Any](https://docs.python.org/3/library/typing.html#typing.Any)]) → [None](https://docs.python.org/3/library/constants.html#None)

Load the scheduler’s state.

Parameters:

state_dict ([dict](https://docs.python.org/3/library/stdtypes.html#dict)) – scheduler state. Should be an object returned
from a call to [state_dict()](#tensorplay.optim.lr_scheduler.CosineAnnealingLR.state_dict).

state_dict() → [dict](https://docs.python.org/3/library/stdtypes.html#dict)[[str](https://docs.python.org/3/library/stdtypes.html#str), [Any](https://docs.python.org/3/library/typing.html#typing.Any)]

Return the state of the scheduler as a [dict](https://docs.python.org/3/library/stdtypes.html#dict).

It contains an entry for every variable in self.__dict__ which
is not the optimizer.

step(epoch: [int](https://docs.python.org/3/library/functions.html#int) | [None](https://docs.python.org/3/library/constants.html#None) = None) → [None](https://docs.python.org/3/library/constants.html#None)

Step the scheduler.

Parameters:

epoch ([int](https://docs.python.org/3/library/functions.html#int), optional) – 
Deprecated since version 1.4: If provided, sets last_epoch to epoch and uses
_get_closed_form_lr() if it is available. This is not
universally supported. Use [step()](#tensorplay.optim.lr_scheduler.CosineAnnealingLR.step) without arguments
instead.

Note

Call this method after calling the optimizer’s
step().