PyTorch基本用法(四)——回归

文章作者：Tyan
博客：noahsnail.com  |  CSDN  |  简书

本文主要是关于PyTorch的一些用法。

import torch
import torch.nn.functional as F
import matplotlib.pyplot as plt
from torch.autograd import Variable

# 生成数据
x = torch.unsqueeze(torch.linspace(-1, 1, 100), dim = 1)
y = x.pow(2) + 0.2 * torch.rand(x.size())

# 变为Variable
x, y = Variable(x), Variable(y)

# 绘制数据图像
plt.scatter(x.data.numpy(), y.data.numpy())
plt.show()

# 定义pytorch网络
class Net(torch.nn.Module):
    
    def __init__(self, n_features, n_hidden, n_output):
        super(Net, self).__init__()
        self.hidden = torch.nn.Linear(n_features, n_hidden)
        self.predict = torch.nn.Linear(n_hidden, n_output)
    
    def forward(self, x):
        x = F.relu(self.hidden(x))
        y = self.predict(x)
        return y

# 构建网络
net = Net(1, 10, 1)
print net

Net (
  (hidden): Linear (1 -> 10)
  (predict): Linear (10 -> 1)
)

# 选择优化方法
optimizer = torch.optim.SGD(net.parameters(), lr = 0.5)

# 选择损失函数
loss_func = torch.nn.MSELoss()

plt.ion()
# 训练
for i in xrange(100):
    # 对x进行预测
    prediction = net(x)
    # 计算损失
    loss = loss_func(prediction, y)
    # 每次迭代清空上一次的梯度
    optimizer.zero_grad()
    # 反向传播
    loss.backward()
    # 更新梯度
    optimizer.step()
    
    if i % 5 == 0:
        plt.cla()
        plt.scatter(x.data.numpy(), y.data.numpy())
        plt.plot(x.data.numpy(), prediction.data.numpy(), 'r-', lw = 5)
        plt.text(0.5, 0, 'Loss=%.4f' % loss.data[0], fontdict={'size': 10, 'color':  'red'})
        plt.pause(0.1)
plt.ioff()
plt.show()

# unsqueeze用法, 一维变二维
x = torch.Tensor([1, 2, 3, 4])
print x
print torch.unsqueeze(x, 0)
print torch.unsqueeze(x, 1)

 1
 2
 3
 4
[torch.FloatTensor of size 4]


 1  2  3  4
[torch.FloatTensor of size 1x4]


 1
 2
 3
 4
[torch.FloatTensor of size 4x1]

# rand用法, rand返回的是[0,1)之间的均匀分布
print torch.rand(4)
print torch.rand(2, 3)

 0.8473
 0.2252
 0.0599
 0.0777
[torch.FloatTensor of size 4]


 0.2864  0.1693  0.1261
 0.9013  0.2009  0.9854
[torch.FloatTensor of size 2x3]

参考资料

1. https://www.youtube.com/user/MorvanZhou