1、使用了PyTorch的nn.Module类来定义神经网络模型;使用nn.Linear来创建全连接层。(CPU)

import torch.nn as nn
import torch.nn.functional as F
from torchsummary import summary

# 定义神经网络模型
class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.fc1 = nn.Linear(in_features=250, out_features=100, bias=True)  # 输入层到隐藏层1，具有250个输入特征和100个神经元
        self.fc2 = nn.Linear(100, 50)  # 隐藏层2，具有100到50个神经元
        self.fc3 = nn.Linear(50, 25)   # 隐藏层3，具有50到25个神经元
        self.fc4 = nn.Linear(25, 10)   # 隐藏层4，具有25到10个神经元
        self.fc5 = nn.Linear(10, 2)    # 输出层，具有10到2个神经元，用于二分类任务

    # 前向传播函数
    def forward(self, x):
        x = x.view(-1, 250)  # 将输入数据展平成一维张量
        x = F.relu(self.fc1(x))  # 使用ReLU激活函数传递到隐藏层1
        x = F.relu(self.fc2(x))  # 使用ReLU激活函数传递到隐藏层2
        x = F.relu(self.fc3(x))  # 使用ReLU激活函数传递到隐藏层3
        x = F.relu(self.fc4(x))  # 使用ReLU激活函数传递到隐藏层4
        x = self.fc5(x)         # 输出层，没有显式激活函数
        return x

if __name__ == '__main__':
    print(Net())
    model = Net()
    summary(model, (250,))  # 打印模型摘要信息，输入大小为(250,)

 

2、GPU版本

import torch
import torch.nn as nn
import torch.nn.functional as F
from torchsummary import summary

class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.fc1 = nn.Linear(784, 100).to(device='cuda:0')
        self.fc2 = nn.Linear(100, 50).to(device='cuda:0')
        self.fc3 = nn.Linear(50, 25).to(device='cuda:0')
        self.fc4 = nn.Linear(25, 10).to(device='cuda:0')

    def forward(self, x):
        x = F.relu(self.fc1(x))
        x = F.relu(self.fc2(x))
        x = F.relu(self.fc3(x))
        x = F.relu(self.fc4(x))
        return x

device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
model = Net().to(device)
input_data = torch.randn(784, 100).to(device)

summary(model, (784, ))