使用pytorch实现CNN和LSTM并列的文本分类代码

好的，这是一份使用 PyTorch 实现 CNN 和 LSTM 并列的文本分类代码的示例：

import torch
import torch.nn as nn

class TextClassifier(nn.Module):
    def __init__(self, vocab_size, embedding_dim, cnn_num_filters, cnn_filter_sizes, lstm_hidden_dim, output_dim, dropout):
        super().__init__()

        self.embedding = nn.Embedding(vocab_size, embedding_dim)
        self.cnn = nn.ModuleList([
            nn.Conv2d(in_channels=1, out_channels=cnn_num_filters, kernel_size=(fs, embedding_dim)) 
            for fs in cnn_filter_sizes
        ])
        self.lstm = nn.LSTM(embedding_dim, lstm_hidden_dim, bidirectional=True, batch_first=True)
        self.fc = nn.Linear(lstm_hidden_dim*2+len(cnn_filter_sizes)*cnn_num_filters, output_dim)
        self.dropout = nn.Dropout(dropout)
        
    def forward(self, x):
        # x.shape = (batch_size, seq_len)
        x = x.permute(1, 0)  # (seq_len, batch_size)
        embedded = self.dropout(self.embedding(x))  # (seq_len, batch_size, embedding_dim)
        embedded = embedded.unsqueeze(1)  # (seq_len, 1, batch_size, embedding_dim)
        cnn_outputs = [torch.relu(conv(embedded)).squeeze(3) for conv in self.cnn]  # [(batch_size, num_filters), ...]*len(filter_sizes)
        cnn_outputs = [torch.max_pool1d(i, i.size(2)).squeeze(2) for i in cnn_outputs]  # [(batch_size, num_filters), ...]*len(filter_sizes)
        lstm_output, (hidden, cell) = self.lstm(embedded)
        hidden = self.dropout(torch.cat((hidden[-2,:,:], hidden[-1,:,:]), dim=1))  # (batch_size, lstm_hidden_dim*2)
        concat = torch.cat((cnn_outputs, hidden), dim=1)  # (batch_size, num_filters*len(filter_sizes)+lstm_hidden_dim*2)
        logits = self.fc(concat)  # (batch_size, output_dim)
        return logits

在这段代码中，我们定义了一个名为 `TextClassifier`