线性回归的简洁实现

上文我们自己编写了，模型，损失函数，优化算法；如果经常进行模型训练，就需要经常编写这些函数，效率低，pytorch有一些现成的资源可供调用，我们就使用现成的资源，重写一下上文的训练过程

import torch
from torch.utils import data
from torch import nn

def synthetic_data(w, b, num_examples):
    X = torch.normal(0, 1, (num_examples, len(w)))
    y = torch.matmul(X, w) + b
    y += torch.normal(0, 0.01, y.shape)
    return X, y.reshape((-1, 1))

def load_array(data_arrays, batch_size, is_shuffle=True):
    dataset = data.TensorDataset(*data_arrays)
    return data.DataLoader(dataset, batch_size, is_shuffle)

net = nn.Sequential(nn.Linear(2, 1))

loss = nn.MSELoss()

trainer = torch.optim.SGD(net.parameters(), lr=0.03)

net[0].weight.data.normal_(0, 0.01)
net[0].bias.data.fill_(0)

num_epochs = 3
true_w = torch.tensor([2, -3.4])
true_b = 4.3
batch_size = 10

features, lables = synthetic_data(true_w, true_b, 1000)

data_iter = load_array((features, lables), batch_size)

for epoch in range(num_epochs):
    for X, y in data_iter:
        l = loss(net(X), y)
        trainer.zero_grad()
        l.backward()
        trainer.step()
    with torch.no_grad():
        l = loss(net(features), lables)
        print('loss=', l ,'weight=', net[0].weight.data, 'bias=', net[0].bias.data)

运行结果：

loss= tensor(0.0002) weight= tensor([[ 1.9987, -3.3966]]) bias= tensor([4.2902])
loss= tensor(0.0001) weight= tensor([[ 1.9994, -3.4013]]) bias= tensor([4.2997])
loss= tensor(0.0001) weight= tensor([[ 1.9999, -3.3992]]) bias= tensor([4.2988])