binary_cross_entropy_with_logits的PyTorch实现

参考链接

需要的包

import torch
import torch.nn as nn
import torch.nn.functional as F

模拟的输入x变量：4分类问题

batch_size, n_classes = 10, 4
x = torch.randn(batch_size, n_classes)
x.shape

x维度

torch.Size([10, 4])

运行：

x

Out:

tensor([[ 2.3611, -0.8813, -0.5006, -0.2178],
        [ 0.0419,  0.0763, -1.0457, -1.6692],
        [-1.0494,  0.8111,  1.5723,  1.2315],
        [ 1.3081,  0.6641,  1.1802, -0.2547],
        [ 0.5292,  0.7636,  0.3692, -0.8318],
        [ 0.5100,  0.9849, -1.2905,  0.2821],
        [ 1.4662,  0.4550,  0.9875,  0.3143],
        [-1.2121,  0.1262,  0.0598, -1.6363],
        [ 0.3214, -0.8689,  0.0689, -2.5094