pytorch代码解析：loss = y_hat - y.view(y_hat.size())

pytorch代码解析：pytorch中loss = y_hat - y.view(y_hat.size())

import torch

y_hat = torch.tensor([[-0.0044],
        [-0.0015],
        [-0.0010],
        [-0.0053],
        [-0.0107],
        [ 0.0055],
        [-0.0073],
        [ 0.0070],
        [ 0.0041],
        [-0.0031]])  # （10，1）

y = torch.tensor([ 6.1894,  7.2714,  4.4549,  5.2858, 10.8752,  2.6536,  6.6904, -1.6226,
         0.6733,  5.6745])  # （10，）

loss = y_hat - y.view(y_hat.size())

print('y_hat.size()为', y_hat.size())
print('y.view(y_hat.size())为：',y.view(y_hat.size()))

print(y.view([5,2])

y_hat.size() 获取 y_hat 的形状 例子为[10,1] 的二维张量

y.view() 是按要求改变y 的形状 ,例如 y.view([5,2]) 把 y 变成[5,2]的二维张量。

注：y 元素个数一定要刚好符合 10 才可以。

运行结果

y_hat.size()为 torch.Size([10, 1])
y.view(y_hat.size())为： tensor([[ 6.1894],
        [ 7.2714],
        [ 4.4549],
        [ 5.2858],
        [10.8752],
        [ 2.6536],
        [ 6.6904],
        [-1.6226],
        [ 0.6733],
        [ 5.6745]])
tensor([[ 6.1894,  7.2714],
        [ 4.4549,  5.2858],
        [10.8752,  2.6536],
        [ 6.6904, -1.6226],
        [ 0.6733,  5.6745]])