通過繼承nn.Module 類來自定義深度學(xué)習(xí)模型是 PyTorch 中常見的做法秉版。nn.Module 是所有神經(jīng)網(wǎng)絡(luò)模塊的基類，提供了許多有用的方法和屬性。自定義模型主要涉及以下幾個步驟：

1. 定義模型類达椰，繼承 nn.Module
2. 在 init 方法中定義模型的層
3. 在 forward 方法中定義前向傳播的邏輯
4. 實例化模型并使用

示例代碼

• MLP

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

class MLP(nn.Module):
    def __init__(self, input_size, hidden_size, output_size):
        super(MLP, self).__init__()
        self.fc1 = nn.Linear(input_size, hidden_size)
        self.fc2 = nn.Linear(hidden_size, hidden_size)
        self.fc3 = nn.Linear(hidden_size, output_size)
        self.relu = nn.ReLU()

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

# 示例用法
input_size = 784  # 例如28x28的圖像
hidden_size = 128
output_size = 10  # 10類分類
model = MLP(input_size, hidden_size, output_size)
print(model)

• CNN

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

class CNN(nn.Module):
    def __init__(self):
        super(CNN, self).__init__()
        self.conv1 = nn.Conv2d(in_channels=1, out_channels=32, kernel_size=3, padding=1)
        self.conv2 = nn.Conv2d(in_channels=32, out_channels=64, kernel_size=3, padding=1)
        self.pool = nn.MaxPool2d(kernel_size=2, stride=2, padding=0)
        self.fc1 = nn.Linear(64 * 7 * 7, 128)
        self.fc2 = nn.Linear(128, 10)

    def forward(self, x):
        x = self.pool(F.relu(self.conv1(x)))
        x = self.pool(F.relu(self.conv2(x)))
        x = x.view(-1, 64 * 7 * 7)
        x = F.relu(self.fc1(x))
        x = self.fc2(x)
        return x

# 示例用法
model = CNN()
print(model)

• Transformer

import torch
import torch.nn as nn

class TransformerModel(nn.Module):
    def __init__(self, input_dim, model_dim, num_heads, num_layers, output_dim):
        super(TransformerModel, self).__init__()
        self.embedding = nn.Embedding(input_dim, model_dim)
        encoder_layer = nn.TransformerEncoderLayer(d_model=model_dim, nhead=num_heads)
        self.transformer_encoder = nn.TransformerEncoder(encoder_layer, num_layers=num_layers)
        self.fc = nn.Linear(model_dim, output_dim)

    def forward(self, src):
        src = self.embedding(src)  # [seq_len, batch_size, model_dim]
        src = src.permute(1, 0, 2)  # Transformer expects [batch_size, seq_len, model_dim]
        output = self.transformer_encoder(src)
        output = self.fc(output.mean(dim=1))
        return output

# 示例用法
input_dim = 10000  # 詞匯表大小
model_dim = 512
num_heads = 8
num_layers = 6
output_dim = 10  # 10類分類
model = TransformerModel(input_dim, model_dim, num_heads, num_layers, output_dim)

print(model)