【人工智能作业】1-使用Pytorch实现反向传播

代码如下

import torch

x_data = [1.0, 2.0, 3.0]
y_data = [2.0, 4.0, 6.0]

w = torch.tensor([1.0]) 
w.requires_grad = True 

def forward(x):
    return x * w 


def loss(x, y):
    y_pred = forward(x)
    return (y_pred - y) ** 2


print("predict (before training)", 4, forward(4).item())

for epoch in range(100):
    for x, y in zip(x_data, y_data):
        l = loss(x, y) 
        l.backward()  
        print('\tgrad:', x, y, w.grad.item()) 
        w.data = w.data - 0.01 * w.grad.data 
        w.grad.data.zero_()  
    print('progress:', epoch, l.item())
print("predict (after training)", 4, forward(4).item())

训练前参数为

第一轮运行结果如下