线性回归的简洁实现—调用pytorch中封装好的函数
#线性回归的简洁实现
import numpy as np
import torch
from torch.utils import data
from d2l import torch as d2l
from torch import nn # nn是神经网络的缩写
true_w = torch.tensor([2, -3.4])
true_b = 4.2
features, labels = d2l.synthetic_data(true_w, true_b, 1000) # 生成features和 labels
def load_array(data_arrays, batch_size, is_train=True):
dataset = data.TensorDataset(*data_arrays) # 将features, labels传入TensorDataset 得到pytorch的dataset
return data.DataLoader(dataset, batch_size, shuffle=is_train) # DataLoader每次从dataset中挑选batch_size个样本 shuffle随机打乱
batch_size = 10
data_iter = load_array((features, labels), batch_size)
next(iter(data_iter))
net = nn.Sequential(nn.Linear(2, 1)) # 线性层 2是输入维度 1是输出维度 Sequential是将层按顺序放在sequential容器里
net[0].weight.data.normal_(0, 0.01) # net[0]是第一层 data.normal_(0, 0.01) 使用均值为0 方差为0.01的正态分布替换掉data
net[0].bias.data.fill_(0) # 将偏差设成0
loss = nn.MSELoss()
trainer = torch.optim.SGD(net.parameters(), lr=0.03) # net.parameters()传入所有参数
num_epochs = 3
for epoch in range(num_epochs):
for X, y in data_iter:
l = loss(net(X), y)
trainer.zero_grad() # 梯度清零
l.backward()
trainer.step() # 进行一次模型更新
l = loss(net(features), labels)
print(f'epoch {epoch + 1}, loss {l:f}') # 打印l 格式为浮点型f