Transformer：Pytorch版本的源码解析

1. 首先加载包：

import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
import math, copy, time
from torch.autograd import Variable
import matplotlib.pyplot as plt
import seaborn
seaborn.set_context(context="talk")

2. 常用的编码器 - 解码器结构、文本生成结构如下：

encode－函数，encoder－网络结构

src－输入文本，src_mask－掩码的输入文本，src_embed－嵌入后的输入文本

tgt－目标文本，tgt_mask－掩码的目标文本，tgt_embed－嵌入后的目标文本

memory－记忆

编码　encode：encoder(self.src_embed(src), src_mask)

解码　decode：decoder(self.tgt_embed(tgt), memory, src_mask, tgt_mask)

向前传播　forward：decode(self.encode(src, src_mask), src_mask,tgt, tgt_mask)

class EncoderDecoder(nn.Module):
    """
    A standard Encoder-Decoder architecture. Base for this and many 
    other models.
    """
    def __init__(self, encoder, decoder, src_embed, tgt_embed, generator):
        super(EncoderDecoder, self).__init__()
        self.encoder = encoder
        self.decoder = decoder
        self.src_embed = src_embed
        self.tgt_embed = tgt_embed
        self.generator = generator
        
    def forward(self, src, tgt, src_mask, tgt_mask):
        "Take in and process masked src and target sequences."
        return self.decode(self.encode(src, src_mask), src_mask,
                            tgt, tgt_mask)
    
    def encode(self, src, src_mask):
        return self.encoder(self.src_embed(src), src_mask)
    
    def decode(self, memory, src_mask, tgt, tgt_mask):
        return self.decoder(self.tgt_embed(tgt), memory, src_mask, tgt_mask)

文本生成结构如下：

线性层：nn.Linear(d_model, vocab)

Softmax层：F.log_softmax(self.proj(x), dim=-1)

class Generator(nn.Module):
    "Define standard linear + softmax generation step."
    def __init__(self, d_model, vocab):
        super(Generator, self).__init__()
        self.proj = nn.Linear(d_model, vocab)

    def forward(self, x):
        return F.log_softmax(self.proj(x), dim=-1)

3. 编码器

首先定义编码器，六个相同的层采用深拷贝搭建：

def clones(module, N):
    # 产生N个相同的层，N=6
    # ModuleList 可以像常规Python列表一样编制索引，包含的模块已正确注册
    # copy.copy 浅拷贝 只拷贝父对象，不会拷贝对象的内部的子对象
    # copy.deepcopy 深拷贝 拷贝对象及其子对象
    return nn.ModuleList([copy.deepcopy(module) for _ in range(N)])

class Encoder(nn.Module):
    # "Core encoder is a stack of N layers"
    def __init__(self, layer, N):
        super(Encoder, self).__init__()
        self.layers = clones(layer, N)
        # 归一化层 LayerNorm(normalized_shape, eps=1e-05, elementwise_affine=True)
        # normalized_shape 输入尺寸  [∗×normalized_shape[0]×normalized_shape[1]×…×normalized_shape[−1]]
        # eps-为保证数值稳定性（分母不能趋近或取0）,给分母加上的值。默认为1e-5
        # elementwise_affine 布尔值，当设为true，给该层添加可学习的仿射变换参数
        self.norm = LayerNorm(layer.size)
        
    def forward(self, x, mask):
        # "Pass the input (and mask) through each layer in turn."
        for layer in self.layers:
            x = layer(x, mask)
        return self.norm(x)

然后构建LayerNorm，在两个子层中分别使用残余连接，然后是层标准化

class LayerNorm(nn.Module):
    def __init__(self, features, eps=1e-6):
        super(LayerNorm, self).__init__()
        self.a_2 = nn.Parameter(torch.ones(features))
        self.b_2 = nn.Parameter(torch.zeros(features))
        self.eps = eps

    def forward(self, x):
        mean = x.mean(-1, keepdim=True)
        std = x.std(-1, keepdim=True)
        return self.a_2 * (x - mean) / (std + self.eps) + self.b_2

每个子层的输出是，其中是由子层本身实现的函数。将dropout应用于每个子层的输出，然后将其添加到子层输入并进行规范化。为了促进这些残余连接，模型中的所有子层以及嵌入层都产生维度为　的输出。

class SublayerConnection(nn.Module):
    """
    A residual connection followed by a layer norm.
    Note for code simplicity the norm is first as opposed to last.
    """
    def __init__(self, size, dropout):
        super(SublayerConnection, self).__init__()
        self.norm = LayerNorm(size)
        self.dropout = nn.Dropout(dropout)

    def forward(self, x, sublayer):
        "Apply residual connection to any sublayer with the same size."
        return x + self.dropout(sublayer(self.norm(x)))

每层有两个子层组成，一是多头自我关注机制multi-head self-attention mechanism，即实现了“多头”的 Self-attention，二是位置完全连接的前馈网络position-wise fully connected feed- forward network。

class EncoderLayer(nn.Module):
    "Encoder is made up of self-attn and feed forward (defined below)"
    def __init__(self, size, self_attn, feed_forward, dropout):
        super(EncoderLayer, self).__init__()
        self.self_attn = self_attn
        self.feed_forward = feed_forward
        self.sublayer = clones(SublayerConnection(size, dropout), 2)
        self.size = size

    def forward(self, x, mask):
        "Follow Figure 1 (left) for connections."
        x = self.sublayer[0](x, lambda x: self.self_attn(x, x, x, mask))
        return self.sublayer[1](x, self.feed_forward)

4. 解码器

新增加了一个编码－解码多头注意力层，其他与编码器相同，也是由6个相同层组成：

class Decoder(nn.Module):
    "Generic N layer decoder with masking."
    # N=6
    def __init__(self, layer, N):
        super(Decoder, self).__init__()
        self.layers = clones(layer, N)
        self.norm = LayerNorm(layer.size)
        
    def forward(self, x, memory, src_mask, tgt_mask):
        for layer in self.layers:
            x = la