8.复习保存和提取

'''
Author: 365JHWZGo
Description: 8.复习保存和提取
Date: 2021-10-25 16:34:53
FilePath: pytorchpytorchday07-1.py
LastEditTime: 2021-10-25 17:00:28
LastEditors: 365JHWZGo
'''

# 导包
import torch
import matplotlib.pyplot as plt
from torch.autograd import Variable
# 创建数据
x = torch.unsqueeze(torch.linspace(-1, 1, 100), dim=1)		#注意此处不宜过大范围，否则偏移不明显
y = x.pow(2) + 0.2*torch.rand(x.size())

x, y = Variable(x, requires_grad=True), Variable(y, requires_grad=True)

# 搭建神经层


def save():
    net1 = torch.nn.Sequential(  # care:   torch.nn.Sequential
        torch.nn.Linear(1, 10),
        torch.nn.ReLU(),
        torch.nn.Linear(10, 1)
    )

    # 创建优化器
    optimizer = torch.optim.SGD(net1.parameters(), lr=0.5)

    # 损失函数
    loss_func = torch.nn.MSELoss()  # Mean-Squared Loss均方差

    # 训练
    for i in range(100):
        prediction = net1(x)
        loss = loss_func(prediction, y)

        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

    plt.figure(1, figsize=(10, 3))
    plt.subplot(131)
    plt.scatter(x.data.numpy(), y.data.numpy())
    plt.title('net1')
    plt.plot(x.data.numpy(), prediction.data.numpy(), 'r-', lw=5)

    torch.save(net1, 'net1.pkl')
    torch.save(net1.state_dict(), 'net1_params.pkl')


def restore_net():
    net2 = torch.load('net1.pkl')

    prediction = net2(x)
    plt.subplot(132)
    plt.scatter(x.data.numpy(), y.data.numpy())
    plt.title('net2')
    plt.plot(x.data.numpy(), prediction.data.numpy(), 'r-', lw=5)


def restore_params():
    net3 = torch.nn.Sequential(
        torch.nn.Linear(1, 10),
        torch.nn.ReLU(),
        torch.nn.Linear(10, 1)
    )
    net3.load_state_dict(torch.load('net1_params.pkl'))
    prediction = net3(x)
    plt.subplot(133)
    plt.scatter(x.data.numpy(), y.data.numpy())
    plt.title('net3')
    plt.plot(x.data.numpy(), prediction.data.numpy(), 'r-', lw=5)
    plt.show()


save()
restore_net()
restore_params()