04反向传播

反向传播

import numpy as np
import matplotlib.pyplot as plt
import torch

x_data = np.random.randint(1, 6, size=(3))
y_data = x_data * 3

w = torch.Tensor([1.0])
w.requires_grad = True   # 需要计算梯度

def forward(x):
    return x * w

def loss(x, y):
    y_pred = forward(x)
    return (y_pred - y) ** 2    # 构建计算图


for epoch in range(100):
    for x, y in zip(x_data, y_data):
        l = loss(x, y)    # 构建计算图
        l.backward()      # 反向计算梯度并data，释放计算图
        print("\tgrad:", x, y, w.grad.item())    # item，防止又构建计算图
        w.data = w.data - 0.01 * w.grad.data   # 用data，防止构建新的计算图

        w.grad.data.zero_()   # 梯度清零，不然会累加
    print("epoch: {}\tl: {}\t".format(epoch, l.item()))
    print("w: ", w.item())