Review - Transformer

回顾《 attention is all your need 》，transformer 的结构如下图所示，Inputs 包括 embedding 和 positional encodeing，将词嵌入结合位置信息；Encoder 包括 N 个堆叠的层，每个层中的多头注意力机制和前馈神经网络后都进行了残差和归一化连接；Decoder 包括 N 个堆叠的层，与 Encoder 不同的是，它还多了一层 masked multi-head attention；Output 包括简单的线性层和 softmax 。

Inputs

embedding 将文本处理为向量，如word embedding。

class Embeddings(nn.Module):
    def __init__(self, d_model, vocab):
        super(Embeddings, self).__init__()
        self.lut = nn.Embedding(vocab, d_model)
        self.d_model = d_model

    def forward(self, x):
        embedds = self.lut(x)
        return embedds * math.sqrt(self.d_model)
        # 这里在给词向量添加位置编码之前，扩大词向量的数值目的是让位置编码相对较小。
        # 这意味着向词向量添加位置编码时，词向量的原始含义不会丢失。

positional ecoding 添加位置信息，采用正余弦可以避免句子长短不一时对位置带来的影响。

class PositionalEncoding(nn.Module):
    def __init__(self, d_model, dropout, max_len=5000):
        super(PositionalEncoding, self).__init__()
        self.dropout = nn.Dropout(p=dropout)
        # 为防止当1000的幂作为分母导致的float溢出，对公式进行转换
        pe = torch.zeros(max_len, d_model)
        position = torch.arange(0, max_len).unsqueeze(1)
        div_term = torch.exp(torch.arange(0, d_model, 2) *
                             -(math.log(10000.0) / d_model))
        pe[:, 0::2] = torch.sin(position * div_term) # 奇数
        pe[:, 1::2] = torch.cos(position * div_term) # 偶数
        pe = pe.unsqueeze(0)
        self.register_buffer('pe', pe)
        
    def forward(self, x):
        x = x + Variable(self.pe[:, :x.size(1)], requires_grad=False)
        # self.pe[:, :x.size(1)]取到x的实际长度
        return self.dropout(x)

可视化位置信息

Encoder

encoder由 N 层堆叠，将层复制 N 次。

def clones(module, N):
    return nn.ModuleList([copy.deepcopy(module) for _ in range(N)])

class Encoder(nn.Module):
    def __init__(self, layer, N):
        super(Encoder, self).__init__()
        self.layers = clones(layer, N)
        self.norm = LayerNorm(layer.size)
        
    def forward(self, x, mask):
        for layer in self.layers:
            x = layer(x, mask) 
        return self.norm(x)

构造掩码，这里掩码的作用是屏蔽空白区域，decoder中掩码还有屏蔽未来信息的作用。

def subsequent_mask(size):
    attn_shape = (1, size, size)
    subsequent_mask = np.triu(np.ones(attn_shape), k=1).astype('uint8')
    return torch.from_numpy(subsequent_mask) == 0

encoder层包括多头注意力层和前馈全连接层这两个子层，每个子层后面都用归一和残差连接。

class SublayerConnection(nn.Module):
    def __init__(self, size, dropout):
        super(SublayerConnection, self).__init__()
        self.norm = LayerNorm(size)
        self.dropout = nn.Dropout(dropout)

    def forward(self, x, sublayer):
        x_norm = self.norm(x + self.dropout(sublayer(x)))
        # 有的把x提出来加速收敛 x_norm = x + self.norm(self.dropout(sublayer(x))) 
        return x_norm

规范化层

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

class EncoderLayer(nn.Module):
    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):
        # 多注意力层
        x = self.sublayer[0](x, lambda x: self.self_attn(x, x, x, mask))
        # 前馈传播层
        z = self.sublayer[1](x, self.feed_forward)
        return z

attention层，采用多头注意力机制。

单头注意力中，QK矩阵内积求出相关性系数scores，判断是否使用掩码，对scores进行softmax，乘上V得到输出。

def attention(query, key, value, mask=None, dropout=None):
    d_k = query.size(-1)
    scores = torch.matmul(query, key.transpose(-2, -1)) / math.sqrt(d_k)
    if mask is not None:
        scores = scores.masked_fill(mask == 0, -1e9)
    p_attn = F.softmax(scores, dim = -1)
    if dropout is not None:
        p_attn = dropout(p_attn)
    return torch.matmul(p_attn, value), p_attn

多头注意力，设计多种Q均衡偏差，让词义有多种表达。给每个头分配等量的词特征，四个线性层中有三个分别对应QKV，最后一个是对应拼接后的。

class MultiHeadedAttention(nn.Module):
    def __init__(self, h, d_model, dropout=0.1):
        super(MultiHeadedAttention, self).__init__()
        assert d_model % h == 0
        self.d_k = d_model // h
        self.h = h
        self.linears = clones(nn.Linear(d_model, d_model), 4)
        self.attn = None
        self.dropout = nn.Dropout(p=dropout)
        
    def forward(self, query, key, value, mask=None):
        if mask is not None:
            mask = mask.unsqueeze(1)
        nbatches = query.size(0)

        query, key, value = \
            [l(x).view(nbatches, -1, self.h, self.d_k).transpose(1, 2)
             for l, x in zip(self.linears, (query, key, value))]
         
        x, self.attn = attention(query, key, value, mask=mask, 
                                 dropout=self.dropout)
  
        x = x.transpose(1, 2).contiguous() \
             .view(nbatches, -1, self.h * self.d_k)
        return self.linears[-1](x)

feed forward 层包括两个线性层和一个relu层。

class PositionwiseFeedForward(nn.Module):
    def __init__(self, d_model, d_ff, dropout=0.1):
        super(PositionwiseFeedForward, self).__init__()
        self.w_1 = nn.Linear(d_model, d_ff)
        self.w_2 = nn.Linear(d_ff, d_model)
        self.dropout = nn.Dropout(dropout)

    def forward(self, x):
        return self.w_2(self.dropout(F.relu(self.w_1(x))))

Decoder

decoder 根据 encoder 的输出和上一次的预测结果，预测序列的下一个输出，由 N 个相同的层堆叠。

class Decoder(nn.Module):
    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 = layer(x, memory, src_mask, tgt_mask)
        return self.norm(x)

class DecoderLayer(nn.Module):
    def __init__(self, size, self_attn, src_attn, feed_forward, dropout):
        super(DecoderLayer, self).__init__()
        self.size = size
        self.self_attn = self_attn
        self.src_attn = src_attn
        self.feed_forward = feed_forward
        self.sublayer = clones(SublayerConnection(size, dropout), 3)
 
    def forward(self, x, memory, src_mask, tgt_mask):
        m = memory
        x = self.sublayer[0](x, lambda x: self.self_attn(x, x, x, tgt_mask))
        x = self.sublayer[1](x, lambda x: self.src_attn(x, m, m, src_mask))
        return self.sublayer[2](x, self.feed_forward)

Output

class Generator(nn.Module):
    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)

Overview

class EncoderDecoder(nn.Module):

    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):
        memory = self.encode(src, src_mask)
        res = self.decode(memory, src_mask, tgt, tgt_mask)
        return res
    
    def encode(self, src, src_mask):
        src_embedds = self.src_embed(src)
        return self.encoder(src_embedds, src_mask)
    
    def decode(self, memory, src_mask, tgt, tgt_mask):
        target_embedds = self.tgt_embed(tgt)
        return self.decoder(target_embedds, memory, src_mask, tgt_mask)

def make_model(src_vocab, tgt_vocab, N=6, d_model=512, d_ff=2048, h=8, dropout=0.1):
    c = copy.deepcopy
    attn = MultiHeadedAttention(h, d_model)
    ff = PositionwiseFeedForward(d_model, d_ff, dropout)
    position = PositionalEncoding(d_model, dropout)
    model = EncoderDecoder(
        Encoder(EncoderLayer(d_model, c(attn), c(ff), dropout), N),
        Decoder(DecoderLayer(d_model, c(attn), c(attn), c(ff), dropout), N),
        nn.Sequential(Embeddings(d_model, src_vocab), c(position)),
        nn.Sequential(Embeddings(d_model, tgt_vocab), c(position)),
        Generator(d_model, tgt_vocab))

    for p in model.parameters():
        if p.dim() > 1:
            nn.init.xavier_uniform_(p)
    return model