mindspore.experimental

The experimental modules.

Experimental Optimizer

API Name

Description

Supported Platforms

mindspore.experimental.optim.Optimizer

Base class for all optimizers.

Ascend GPU CPU

mindspore.experimental.optim.Adadelta

Implements Adadelta algorithm.

Ascend GPU CPU

mindspore.experimental.optim.Adagrad

Implements Adagrad algorithm.

Ascend GPU CPU

mindspore.experimental.optim.Adam

Implements Adam algorithm.

Ascend GPU CPU

mindspore.experimental.optim.Adamax

Implements Adamax algorithm (a variant of Adam based on infinity norm).

Ascend GPU CPU

mindspore.experimental.optim.AdamW

Implements Adam Weight Decay algorithm.

Ascend GPU CPU

mindspore.experimental.optim.ASGD

Implements Averaged Stochastic Gradient Descent algorithm.

Ascend GPU CPU

mindspore.experimental.optim.NAdam

Implements NAdam algorithm.

Ascend GPU CPU

mindspore.experimental.optim.RAdam

Implements RAdam algorithm.

Ascend GPU CPU

mindspore.experimental.optim.RMSprop

Implements RMSprop algorithm.

Ascend GPU CPU

mindspore.experimental.optim.Rprop

Implements Rprop algorithm.

Ascend GPU CPU

mindspore.experimental.optim.SGD

Stochastic Gradient Descent optimizer.

Ascend GPU CPU

LRScheduler Class

The dynamic learning rates in this module are all subclasses of LRScheduler, this module should be used with optimizers in mindspore.experimental.optim, pass the optimizer instance to a LRScheduler when used. During the training process, the LRScheduler subclass dynamically changes the learning rate by calling the step method.

import mindspore
from mindspore import nn
from mindspore.experimental import optim

# Define the network structure of LeNet5. Refer to
# https://gitee.com/mindspore/docs/blob/r2.3.0/docs/mindspore/code/lenet.py

net = LeNet5()
loss_fn = nn.SoftmaxCrossEntropyWithLogits(sparse=True)
optimizer = optim.Adam(net.trainable_params(), lr=0.05)
scheduler = optim.lr_scheduler.StepLR(optimizer, step_size=2, gamma=0.1)
def forward_fn(data, label):
    logits = net(data)
    loss = loss_fn(logits, label)
    return loss, logits
grad_fn = mindspore.value_and_grad(forward_fn, None, optimizer.parameters, has_aux=True)
def train_step(data, label):
    (loss, _), grads = grad_fn(data, label)
    optimizer(grads)
    return loss
for epoch in range(6):
    # Create the dataset taking MNIST as an example. Refer to
    # https://gitee.com/mindspore/docs/blob/r2.3.0/docs/mindspore/code/mnist.py

    for data, label in create_dataset(need_download=False):
        train_step(data, label)
    scheduler.step()

API Name

Description

Supported Platforms

mindspore.experimental.optim.lr_scheduler.LRScheduler

Basic class of learning rate schedule.

Ascend GPU CPU

mindspore.experimental.optim.lr_scheduler.ConstantLR

Decays the learning rate of each parameter group by a small constant factor until the number of epoch reaches a pre-defined milestone: total_iters.

Ascend GPU CPU

mindspore.experimental.optim.lr_scheduler.CosineAnnealingLR

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

Ascend GPU CPU

mindspore.experimental.optim.lr_scheduler.CosineAnnealingWarmRestarts

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

Ascend GPU CPU

mindspore.experimental.optim.lr_scheduler.CyclicLR

Sets the learning rate of each parameter group according to cyclical learning rate policy (CLR).

Ascend GPU CPU

mindspore.experimental.optim.lr_scheduler.ExponentialLR

For each epoch, the learning rate decays exponentially, multiplied by gamma.

Ascend GPU CPU

mindspore.experimental.optim.lr_scheduler.LambdaLR

Sets the learning rate of each parameter group to the initial lr times a given function.

Ascend GPU CPU

mindspore.experimental.optim.lr_scheduler.LinearLR

Decays the learning rate of each parameter group by linearly changing small multiplicative factor until the number of epoch reaches a pre-defined milestone: total_iters.

Ascend GPU CPU

mindspore.experimental.optim.lr_scheduler.MultiplicativeLR

Multiply the learning rate of each parameter group by the factor given in the specified function.

Ascend GPU CPU

mindspore.experimental.optim.lr_scheduler.MultiStepLR

Multiply the learning rate of each parameter group by gamma once the number of epoch reaches one of the milestones.

Ascend GPU CPU

mindspore.experimental.optim.lr_scheduler.PolynomialLR

For each epoch, the learning rate is adjusted by polynomial fitting.

Ascend GPU CPU

mindspore.experimental.optim.lr_scheduler.ReduceLROnPlateau

Reduce learning rate when a metric has stopped improving.

Ascend GPU CPU

mindspore.experimental.optim.lr_scheduler.SequentialLR

Receives the list of schedulers that is expected to be called sequentially during optimization process and milestone points that provides exact intervals to reflect which scheduler is supposed to be called at a given epoch.

Ascend GPU CPU

mindspore.experimental.optim.lr_scheduler.StepLR

Decays the learning rate of each parameter group by gamma every step_size epochs.

Ascend GPU CPU