pytorch transformer完整代码

PyTorch Transformer是一个基于Transformer模型的深度学习框架组件，它主要用于处理序列数据，如自然语言理解和机器翻译等任务。下面是一个简单的Transformer模型基础结构的代码示例，并非完整的训练流程，而是展示其核心部分：

import torch
import torch.nn as nn

class PositionalEncoding(nn.Module):
    def __init__(self, d_model, dropout=0.1, max_len=5000):
        super(PositionalEncoding, self).__init__()
        self.dropout = nn.Dropout(p=dropout)
        pe = torch.zeros(max_len, d_model)
        position = torch.arange(0, max_len, dtype=torch.float).unsqueeze(1)
        div_term = torch.exp(torch.arange(0, d_model, 2).float() * (-math.log(10000.0) / d_model))
        pe[:, 0::2] = torch.sin(position * div_term)
        pe[:, 1::2] = torch.cos(position * div_term)
        pe = pe.unsqueeze(0).transpose(0, 1)
        self.register_buffer('pe', pe)

    def forward(self, x):
        x = x + self.pe[:x.size(0), :]
        return self.dropout(x)

class TransformerBlock(nn.Module):
    def __init__(self, d_model, nhead, dim_feedforward=2048, dropout=0.1):
        super(TransformerBlock, self).__init__()
        self.self_attn = nn.MultiHeadAttention(d_model, nhead, dropout=dropout)
        self.norm1 = nn.LayerNorm(d_model)
        self.linear1 = nn.Linear(d_model, dim_feedforward)
        self.dropout = nn.Dropout(dropout)
        self.linear2 = nn.Linear(dim_feedforward, d_model)
        self.norm2 = nn.LayerNorm(d_model)

    def forward(self, src):
        src2 = self.norm1(src)
        attn_output = self.self_attn(src2, src2, src2)
        src = src + self.dropout(attn_output)
        src2 = self.norm2(src)
        linear_output = self.linear2(self.dropout(F.relu(self.linear1(src2))))
        src = src + self.dropout(linear_output)
        return src

# 完整模型示例
class TransformerModel(nn.Module):
    def __init__(self, input_dim, output_dim, num_layers, d_model, nhead, dim_feedforward, dropout):
        super(TransformerModel, self).__init__()
        self.embedding = nn.Embedding(input_dim, d_model)
        self.pos_encoder = PositionalEncoding(d_model, dropout)
        encoder_layer = TransformerBlock(d_model, nhead, dim_feedforward, dropout)
        self.transformer_encoder = nn.TransformerEncoder(encoder_layer, num_layers)
        self.fc_out = nn.Linear(d_model, output_dim)

    def forward(self, src):
        embedded = self.embedding(src) * math.sqrt(self.d_model)
        pos_encoded = self.pos_encoder(embedded)
        transformed = self.transformer_encoder(pos_encoded)
        output = self.fc_out(transformed.mean(dim=1))  # 对编码后的平均池化
        return output

# 使用示例
input_size = 100  # 假设输入词汇表大小
output_size = 50   # 假设输出类别数
model = TransformerModel(input_size, output_size, num_layers=6, d_model=512, nhead=8, dim_feedforward=2048, dropout=0.1)