比较与tf.train.ProximalAdagradOptimizer的功能差异

查看源文件

tf.train.ProximalAdagradOptimizer

tf.train.ProximalAdagradOptimizer(
    learning_rate, initial_accumulator_value=0.1, l1_regularization_strength=0.0,
    l2_regularization_strength=0.0, use_locking=False, name='ProximalAdagrad'
)

更多内容详见tf.train.ProximalAdagradOptimizer

mindspore.nn.ProximalAdagrad

mindspore.nn.ProximalAdagrad(
    params, accum=0.1, learning_rate=0.001, l1=0.0, l2=0.0,
    use_locking=False, loss_scale=1.0, weight_decay=0.0)

更多内容详见mindspore.nn.ProximalAdagrad

使用方式

一般使用场景:

  • MindSpore:一般情况下,在实例化一个优化器子类之后,将其作为mindspore.model高阶API的入参参与训练,用法请参考代码示例;或使用mindspore.nn.TrainOneStepCell,通过传入优化器和一个mindspore.nn.WithLossCell的实例,自定义训练网络,具体实现方式可以参考官网教程

  • TensorFlow:一般情况下,在实例化一个优化器子类之后,将其作为tf.keras.models.Model高阶API的入参参与训练;或调用minimize()(包含compute_gradients()apply_gradients())方法单步执行。

其他功能差异:

  • 参数分组:MindSpore提供参数分组功能,且支持为不同参数组设置不同配置值,通过入参params传入参数组字典实现,mindspore.nn.ProximalAdagrad支持参数分组;TensorFlow没有此入参配置。

  • 动态学习率:MindSpore支持动态学习率,分别在nn.dynamic_lrnn.learning_rate_schedule模块中有不同的实现方法,mindspore.nn.ProximalAdagrad支持动态学习率;TensorFlow也支持此功能,学习率设置封装在tf.train模块中,tf.train.ProximalAdagradOptimizer支持动态学习率。

  • 权重衰减和混合精度:MindSpore的mindspore.nn.Optimizer基类支持通过配置入参weight_decayloss_scale来进行权重衰减及混合精度设置;TensorFlow的优化器没有相关入参配置,但提供了tf.keras.regularizerstf.keras.mixed_precision模块提供相似的功能,配合优化器使用。

代码示例

MindSpore:

# The following implements  ProximalAdagrad with MindSpore.
import mindspore.nn as nn
import mindspore as ms

class Net(nn.Cell):
    def __init__(self):
        super(Net, self).__init__()
        self.conv = nn.Conv2d(3, 64, 3)
        self.bn = nn.BatchNorm2d(64)

    def construct(self, x):
        x = self.conv(x)
        x = self.bn(x)
        return x

net = Net()

# 1) All parameters use the same learning rate and weight decay
optim = nn.ProximalAdagrad(params=net.trainable_params())

# 2) Use parameter groups and set different values
conv_params = list(filter(lambda x: 'conv' in x.name, net.trainable_params()))
no_conv_params = list(filter(lambda x: 'conv' not in x.name, net.trainable_params()))

group_params = [{'params': conv_params, 'weight_decay': 0.01, 'grad_centralization':True},
                {'params': no_conv_params, 'lr': 0.01},
                {'order_params': net.trainable_params()}]

optim = nn.ProximalAdagrad(group_params, learning_rate=0.1, weight_decay=0.0)
loss = nn.SoftmaxCrossEntropyWithLogits(sparse=True, reduction='mean')
model = ms.Model(net, loss_fn=loss, optimizer=optim, metrics={"accuracy"})

TensorFlow:

以下训练输出结果具有随机性。

# The following implements ProximalAdagrad with tensorflow.
import tensorflow as tf
from tensorflow.keras import layers
tf.enable_eager_execution()

# build model and instantiate ProximalAdagrad optimizer
model = tf.keras.Sequential()
model.add(layers.Dense(1, kernel_initializer='uniform', input_shape=(3,)))
model.add(layers.Activation('relu'))

inputs = tf.constant([[1., 2., 3.], [3., 4., 5.]], dtype=tf.float32)
outputs = tf.constant([[0.5], [0.6]], dtype=tf.float32)

optim = tf.train.ProximalAdagradOptimizer(learning_rate=0.01)
loss = tf.keras.losses.MeanSquaredError()

model.compile(optimizer=optim, loss=loss)
model.fit(inputs, outputs)

# out: Train on 2 samples
# 2/2 [==============================] - 0s 16ms/sample - loss: 0.0596