Pytorch简单实现线性回归_动手学深度学习_Note

1. 从0开始实现

1.1 导入相关的库

import torch
import random
import numpy as np

1.2 生成数据集与标签

# 特征数量：房价（面积、时间）
num_inputs = 2
# 样本数量
num_examples = 1000
# 真实权重
true_w = [2, -4.5]
# 真实偏差
true_b = 2.6

# 从numpy数组生成服从均值为0、标准差为1的正态分布数据，Size为（1000，2）
features = torch.from_numpy(np.random.normal(0, 1, (num_examples, num_inputs)))
labels = true_w[0] * features[:. 0] + true_w[1] * features[:, 1] + b
label += torch.from_numpy(np.random.normal(0, 0.01, size=label.szie()))

1.3 初始化权重参数

# 权重均值为0、标准差为0.01的正态随机数，偏差则初始化成0
w = torch.tensor(np.random.normal(0, 0.01, (num_inputs, 1)), dtype=torch.float64)
b = torch.zeros(1, dtype=torch.float64)

# 追踪梯度，利用链式法则进行梯度传播
w.requires_grad_(requires_grad=True)
b.requires_grad_(requires_grad=True)

1.4 抽取数据

def data_iter(batch_size, features, labels):
	num_examples = len(features) # 数据长度
	indices = list(range(num_examples)) # 数据
	random.shuffle(indices)
	for i in range(0, num_examples, batch_size):
		j = torch.longTensor(indices[i: min(i + batch_size, num_examples)])
		yield features.index_select(0, j), labels.index_select(0, j)

1.5 定义模型、损失函数、优化算法

# 线性模型
def linreg(X, w, b):
	return torch.mm(X, w) + b
# 损失函数
def squared_loss(y_hat, y):
	return (y_hat - y.view(y_hat.size())) ** 2 / 2
# 优化算法
def sgd(params, lr, batch_size):
	for param in params:
		param.data -= lr * param.grad / batch_size

1.6 模型训练

lr = 0.03
batch_size = 10
num_epochs = 3
net = linreg
loss = squared_loss

for epoch in range(num_epochs): 
	"""
	训练模型一共需要num_epochs个迭代周期
    每一个迭代周期中，会使用训练数据集中所有样本一次（假设样本数能够被批量大小整除）
    X和y分别是小批量样本的特征和标签
	"""
    for X, y in data_iter(batch_size, features, labels):
        l = loss(net(X, w, b), y).sum() # l是有关小批量X和y的损失
        # 变量l并不是一个标量，运行l.backward()将对l中元素求和得到新的变量，再求该变量有关模型参数的梯度
        l.backward() # 小批量的损失对模型参数求梯度
        sgd([w, b], lr, batch_size) # 使用小批量随机梯度下降迭代模型参数
        
        # 不要忘了梯度清零
        w.grad.data.zero_()
        b.grad.data.zero_()
    train_l = loss(net(features, w, b), labels)
    print('epoch %d, loss %f' % (epoch + 1, train_l.mean().item()))

out:
	epoch 1, loss 0.038098
	epoch 2, loss 0.000144
	epoch 3, loss 0.000047

1.7 结果

print(true_w, '\n', w)
print(true_b, '\n', b)

out:

[2, -4.5] 
 tensor([[ 2.0006],
        [-4.4990]], dtype=torch.float64, requires_grad=True)
2.6 
 tensor([2.5990], dtype=torch.float64, requires_grad=True)

2. 简洁实现

import torch.utils.data as Data
from torch.nn import init
import torch.optim as optim
from torch import nn

num_inputs = 2
num_examples = 1000
true_w = [2, -3.4]
true_b = 4.2
features = torch.tensor(np.random.normal(0, 1, (num_examples, num_inputs)), dtype=torch.float)
labels = true_w[0] * features[:, 0] + true_w[1] * features[:, 1] + true_b
labels += torch.tensor(np.random.normal(0, 0.01, size=labels.size()), dtype=torch.float)

batch_size = 10
dataset = Data.TensorDataset(features, labels)
data_iter = Data.DataLoader(dataset, batch_size, shuffle=True)

class LinearNet(nn.Module):
    def __init__(self, n_feature):
        super(LinearNet, self).__init__()
        self.linear = nn.Sequential(
            nn.Linear(n_feature, 1)
        )

    def forward(self, x):
        y = self.linear(x)
        return y

net = LinearNet(num_inputs)
init.normal_(net.linear[0].weight, mean=0, std=0.01)
init.constant_(net.linear[0].bias, val=0)

loss = nn.MSELoss()
optimizer = optim.SGD(net.parameters(), lr=0.03)

num_epochs = 3

for epoch in range(1, num_epochs + 1):
    for X, y in data_iter:
        output = net(X)
        l = loss(output, y.view(-1, 1))
        optimizer.zero_grad() # 梯度清零，等价于net.zero_grad()
        l.backward()
        optimizer.step()
    print('epoch %d, loss: %f' % (epoch, l.item()))
out:
	epoch 1, loss: 0.000319
	epoch 2, loss: 0.000089
	epoch 3, loss: 0.000087

dense = net.linear[0]
print(true_w, dense.weight)
print(true_b, dense.bias)
out:
	[2, -3.4] Parameter containing:
	tensor([[ 2.0006, -3.3997]], requires_grad=True)
	4.2 Parameter containing:
	tensor([4.1997], requires_grad=True)