[人工智能]使用pytorch高级api实现线性回归

import torch
from matplotlib import pyplot as plt
import numpy as np
from torch.utils import data
import torch
from torch import nn
from collections.abc import Iterator,Iterable
# y = 7*x1+8*x2+5
true_w = torch.tensor([7.0,8.0])
true_b = torch.tensor([5.0])

def generate_data(num_examples):
    x = torch.normal(0,1,[num_examples,2])
    y = torch.matmul(x,true_w)+true_b+torch.normal(0,0.01,[num_examples])
    y = y.reshape([-1, 1]) #reshape 不是原地操作
    print(y.shape)
    return x,y



def load_data(data_array,batch_size,is_train=True):#返回的是可迭代的,一次返回batch_size个
    train_data = data.TensorDataset(*data_array)
    return data.DataLoader(train_data,batch_size,shuffle=is_train)


# 定义损失函数 两个位数应该一样,默认是取均值了
loss = nn.MSELoss()


# 定义层,上一层输入是下一层输出输入为?*2
net = nn.Sequential(nn.Linear(2,1))

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

data_iter = load_data(generate_data(1000),5)

for x,y in data_iter:
    l = loss(net(x),y)
    l.backward()
    trainer.step() #执行一次梯度下降
    trainer.zero_grad() # 必须清零梯度,否则会累加

print(net.state_dict())