import torch
from time import time
a = torch.ones(1000)
b = torch.ones(1000)
start = time()
c = torch.zeros(1000)
for i in range(1000):
c[i] = a[i] + b[i]
print(time() - start) # 0.02039504051208496
start = time()
d = a + b
print(time() - start) # 0.0008330345153808594
结论:后者比前者更省时,所以使用矢量计算可以提升计算效率
import torch
from IPython import display
from matplotlib import pyplot as plt
import numpy as np
import random
# 生成数据集
num_inputs = 2
num_examples = 1000
true_w = [2, -3.4]
true_b = 4.2
features = torch.randn(num_examples, num_inputs, dtype=torch.float32)
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.float32) # 添加噪音
def use_svg_display():
# 用矢量图显示
display.set_matplotlib_formats('svg')
def set_figsize(figsize=(3.5, 2.5)):
use_svg_display()
# 设置图的尺寸
plt.rcParams['figure.figsize'] = figsize
set_figsize()
plt.scatter(features[:, 1].numpy(), labels.numpy(), 1);
# 读取数据
def data_iter(batch_size, features, labels):
num_examples = len(features)
indices = list(range(num_examples))
random.shuffle(indices) # 样本的读取顺序是随机的
for i in range(0, num_examples, batch_size):
j = torch.LongTensor(indices[i: min(i + batch_size, num_examples)]) # 最后一次可能不足一个batch
yield features.index_select(0, j), labels.index_select(0, j)
# test
batch_size = 10
for X, y in data_iter(batch_size, features, labels):
print(X, y)
break
'''
tensor([[-1.4239, -1.3788],
[ 0.0275, 1.3550],
[ 0.7616, -1.1384],
[ 0.2967, -0.1162],
[ 0.0822, 2.0826],
[-0.6343, -0.7222],
[ 0.4282, 0.0235],
[ 1.4056, 0.3506],
[-0.6496, -0.5202],
[-0.3969, -0.9951]])
tensor([ 6.0394, -0.3365, 9.5882, 5.1810, -2.7355, 5.3873, 4.9827, 5.7962,
4.6727, 6.7921])
'''
# 初始化数据模型
w = torch.tensor(np.random.normal(0, 0.01, (num_inputs, 1)), dtype=torch.float32)
b = torch.zeros(1, dtype=torch.float32)
w.requires_grad_(requires_grad=True)
b.requires_grad_(requires_grad=True)
# 定义模型
def linreg(X, w, b):
return torch.mm(X, w) + b # torch.mm()
# 定义损失函数
def squared_loss(y_hat, y): # 本函数已保存在d2lzh_pytorch包中方便以后使用
# 注意这里返回的是向量, 另外, pytorch里的MSELoss并没有除以 2
return (y_hat - y.view(y_hat.size())) ** 2 / 2
# 定义优化函数
def sgd(params, lr, batch_size): # 本函数已保存在d2lzh_pytorch包中方便以后使用
for param in params:
param.data -= lr * param.grad / batch_size # 注意这里更改param时用的param.data
# 训练模型
lr = 0.03
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
# 在每一个迭代周期中,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
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()))
plt.show()
# 生成数据集
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)
# 读取数据
import torch.utils.data as Data
batch_size = 10
# 将训练数据的特征和标签组合
dataset = Data.TensorDataset(features, labels)
# 随机读取小批量
data_iter = Data.DataLoader(dataset, batch_size, shuffle=True)
# 定义模型
class LinearNet(nn.Module):
def __init__(self, n_feature):
super(LinearNet, self).__init__()
self.linear = nn.Linear(n_feature, 1) # 输入, 输出
# forward 定义前向传播
def forward(self, x):
y = self.linear(x)
return y
net = LinearNet(num_inputs)
print(net) # 使用print可以打印出网络的结构
'''
LinearNet(
(linear): Linear(in_features=2, out_features=1, bias=True)
)
'''
# 查看模型参数
for param in net.parameters(): # net.parameters()来查看模型所有的可学习参数,此函数将返回一个生成器。
print(param)
# 初始化模型参数
from torch.nn import init # 初始化模型参数
init.normal_(net[0].weight, mean=0, std=0.01)
init.constant_(net[0].bias, val=0) # 也可以直接修改bias的data: net[0].bias.data.fill_(0)
# 定义损失函数
loss = nn.MSELoss()
# 定义优化函数
import torch.optim as optim # 提供很多优化函数
optimizer = optim.SGD(net.parameters(), lr=0.03)
print(optimizer)
'''
SGD (
Parameter Group 0
dampening: 0
lr: 0.03
momentum: 0
nesterov: False
weight_decay: 0
)
'''
# 设置不同子网络的学习率
optimizer =optim.SGD([
# 如果对某个参数不指定学习率,就使用最外层的默认学习率
{'params': net.subnet1.parameters()}, # lr=0.03
{'params': net.subnet2.parameters(), 'lr': 0.01}
], lr=0.03)
# 训练模型
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()))
troch.utils.data 模板提供了有关数据处理的工具
troch,nn 模板定义了大量神经网络的层
torch.init 模板定义了各种初始化方法
torch.optim模块提供了很多常用的优化算法
https://tangshusen.me/Dive-into-DL-PyTorch/#/chapter03_DL-basics/3.3_linear-regression-pytorch