CosineAnnealingWarmRestarts

class paddle.optimizer.lr. CosineAnnealingWarmRestarts ( learning_rate, T_0, T_mult=1, eta_min=0, last_epoch=- 1, verbose=False ) [source]

Set the learning rate of each parameter group using a cosine annealing schedule, where \(\eta_{max}\) is set to the initial lr, \(T_{cur}\) is the number of epochs since the last restart and \(T_{i}\) is the number of epochs between two warm restarts in SGDR:

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

When \(T_{cur}=T_{i}\), set \(\eta_t = \eta_{min}\). When \(T_{cur}=0\) after restart, set \(\eta_t=\eta_{max}\).

It has been proposed in SGDR: Stochastic Gradient Descent with Warm Restarts.

Parameters

  • learning_rate (float) – Initial learning rate.

  • T_0 (int) – Number of iterations for the first restart.

  • T_mult (int, optional) – A factor increases \(T_{i}\) after a restart. Default: 1.

  • eta_min (float, optional) – Minimum learning rate. Default: 0.

  • last_epoch (int, optional) – The index of last epoch. Default: -1, means initial learning rate.

  • verbose (bool, optional) – If True, prints a message to stdout for each update. Default: False.

Returns

CosineAnnealingWarmRestarts instance to schedule learning rate.

Examples

>>> import paddle
>>> import numpy as np
>>> # train on default dynamic graph mode
>>> linear = paddle.nn.Linear(10, 10)
>>> scheduler = paddle.optimizer.lr.CosineAnnealingWarmRestarts(learning_rate=0.5, T_0=1, T_mult=2, verbose=True)
>>> adam = paddle.optimizer.Adam(learning_rate=scheduler, parameters=linear.parameters())
>>> for epoch in range(10):
...    for batch_id in range(10):
...        x = paddle.uniform([10, 10])
...        out = linear(x)
...        loss = paddle.mean(out)
...        loss.backward()
...        adam.step()
...        adam.clear_grad()
...    scheduler.step(epoch)        # You should update learning rate each step

>>> import paddle
>>> import numpy as np
>>> paddle.enable_static()
>>> main_prog = paddle.static.Program()
>>> start_prog = paddle.static.Program()
>>> with paddle.static.program_guard(main_prog, start_prog):
...    x = paddle.static.data(name='x', shape=[None, 4, 5])
...    y = paddle.static.data(name='y', shape=[None, 4, 5])
...    z = paddle.static.nn.fc(x, 100)
...    loss = paddle.mean(z)
...    scheduler = paddle.optimizer.lr.CosineAnnealingWarmRestarts(learning_rate=0.5, T_0=1, T_mult=2,verbose=True)
...    sgd = paddle.optimizer.SGD(learning_rate=scheduler)
...    sgd.minimize(loss)
>>> exe = paddle.static.Executor()
>>> exe.run(start_prog)
>>> for epoch in range(10):
...    for batch_id in range(10):
...        out = exe.run(
...            main_prog,
...            feed={
...                'x': np.random.randn(3, 4, 5).astype('float32'),
...                'y': np.random.randn(3, 4, 5).astype('float32')
...            },
...            fetch_list=loss.name)
...    scheduler.step(epoch)    # You should update learning rate each step
get_lr ( )

get_lr

For those subclass who overload LRScheduler (Base Class), User should have a custom implementation of get_lr() .

Otherwise, an NotImplementedError exception will be thrown.

set_dict ( state_dict )

set_dict

Loads the schedulers state.

set_state_dict ( state_dict )

set_state_dict

Loads the schedulers state.

state_dict ( )

state_dict

Returns the state of the scheduler as a dict.

It is a subset of self.__dict__ .

state_keys ( )

state_keys

For those subclass who overload LRScheduler (Base Class). Acquiescently, “last_epoch, last_lr” will be saved by self.keys = ['last_epoch', 'last_lr'] .

last_epoch is the current epoch num, and last_lr is the current learning rate.

If you want to change the default behavior, you should have a custom implementation of _state_keys() to redefine self.keys .

step ( epoch=None )

step

step should be called after optimizer.step() . It will update the learning rate in optimizer. The new learning rate will take effect on next epoch.

Parameters

epoch (int, None) – specify current epoch. Default: None. Auto-increment from last_epoch=-1.

Returns

None

Examples

Please refer to the example of current LRScheduler.