Building a Neural Network

A neural network model consists of multiple data operation layers. mindspore.nn provides various basic network modules.

The following uses LeNet as an example to describe how MindSpore builds a neural network model.

Import the required modules and APIs:

import numpy as np
import mindspore
import mindspore.nn as nn
from mindspore import Tensor

Defining a Model Class

The Cell class of MindSpore is the base class for building all networks and the basic unit of a network. When a neural network is required, you need to inherit the Cell class and overwrite the __init__ and construct methods.

class LeNet5(nn.Cell):
    """
    Lenet network structure
    """
    def __init__(self, num_class=10, num_channel=1):
        super(LeNet5, self).__init__()
        # Define the required operation.
        self.conv1 = nn.Conv2d(num_channel, 6, 5, pad_mode='valid')
        self.conv2 = nn.Conv2d(6, 16, 5, pad_mode='valid')
        self.fc1 = nn.Dense(16 * 5 * 5, 120)
        self.fc2 = nn.Dense(120, 84)
        self.fc3 = nn.Dense(84, num_class)
        self.relu = nn.ReLU()
        self.max_pool2d = nn.MaxPool2d(kernel_size=2, stride=2)
        self.flatten = nn.Flatten()

    def construct(self, x):
        # Use the defined operation to build a forward network.
        x = self.conv1(x)
        x = self.relu(x)
        x = self.max_pool2d(x)
        x = self.conv2(x)
        x = self.relu(x)
        x = self.max_pool2d(x)
        x = self.flatten(x)
        x = self.fc1(x)
        x = self.relu(x)
        x = self.fc2(x)
        x = self.relu(x)
        x = self.fc3(x)
        return x

Model Layers

The following describes the key member functions of the Cell class used in LeNet, and then describes how to use the Cell class to access model parameters through the instantiation network.

nn.Conv2d

Add the nn.Conv2d layer and add a convolution function to the network to help the neural network extract features.

conv2d = nn.Conv2d(1, 6, 5, has_bias=False, weight_init='normal', pad_mode='valid')
input_x = Tensor(np.ones([1, 1, 32, 32]), mindspore.float32)

print(conv2d(input_x).shape)

    (1, 6, 28, 28)

nn.ReLU

Add the nn.ReLU layer and add a non-linear activation function to the network to help the neural network learn various complex features.

relu = nn.ReLU()
input_x = Tensor(np.array([-1, 2, -3, 2, -1]), mindspore.float16)
output = relu(input_x)

print(output)

    [0. 2. 0. 2. 0.]

nn.MaxPool2d

Initialize the nn.MaxPool2d layer and down-sample the 6 x 28 x 28 array to a 6 x 14 x 14 array.

max_pool2d = nn.MaxPool2d(kernel_size=2, stride=2)
input_x = Tensor(np.ones([1, 6, 28, 28]), mindspore.float32)

print(max_pool2d(input_x).shape)

    (1, 6, 14, 14)

nn.Flatten

Initialize the nn.Flatten layer and convert the 16 x 5 x 5 array into 400 consecutive arrays.

flatten = nn.Flatten()
input_x = Tensor(np.ones([1, 16, 5, 5]), mindspore.float32)
output = flatten(input_x)

print(output.shape)

    (1, 400)

nn.Dense

Initialize the nn.Dense layer and perform linear transformation on the input matrix.

dense = nn.Dense(400, 120, weight_init='normal')
input_x = Tensor(np.ones([1, 400]), mindspore.float32)
output = dense(input_x)

print(output.shape)

    (1, 120)

Model Parameters

The convolutional layer and fully-connected layer in the network will have weights and offsets after being instantiated, and these weight and offset parameters are optimized in subsequent training. In nn.Cell, the parameters_and_names() method is used to access all parameters.

In the example, we traverse each parameter and display the name and attribute of each layer in the network.

model = LeNet5()
for m in model.parameters_and_names():
    print(m)

    ('conv1.weight', Parameter (name=conv1.weight))
    ('conv2.weight', Parameter (name=conv2.weight))
    ('fc1.weight', Parameter (name=fc1.weight))
    ('fc1.bias', Parameter (name=fc1.bias))
    ('fc2.weight', Parameter (name=fc2.weight))
    ('fc2.bias', Parameter (name=fc2.bias))
    ('fc3.weight', Parameter (name=fc3.weight))
    ('fc3.bias', Parameter (name=fc3.bias))