【论文阅读】Transformer-Attention is all you need（简介+代码）

class Encoder(nn.Module):
    ''' A encoder model with self attention mechanism. '''

    def __init__(
            self, n_src_vocab, d_word_vec, n_layers, n_head, d_k, d_v,
            d_model, d_inner, pad_idx, dropout=0.1, n_position=200, scale_emb=False):

        super().__init__()
	# 编码层，使用torch.nn中的Embedding层，前两个参数分别为词表长度、嵌入维度
        self.src_word_emb = nn.Embedding(n_src_vocab, d_word_vec, padding_idx=pad_idx)
	# 位置编码
        self.position_enc = PositionalEncoding(d_word_vec, n_position=n_position)
        self.dropout = nn.Dropout(p=dropout)
	# 多个编码层堆叠
        self.layer_stack = nn.ModuleList([
            EncoderLayer(d_model, d_inner, n_head, d_k, d_v, dropout=dropout)
            for _ in range(n_layers)])
	# 层归一化
        self.layer_norm = nn.LayerNorm(d_model, eps=1e-6)
        self.scale_emb = scale_emb
        self.d_model = d_model

    def forward(self, src_seq, src_mask, return_attns=False):

        enc_slf_attn_list = []
        # 预处理：embedding + 层归一化
        enc_output = self.src_word_emb(src_seq)
        if self.scale_emb:
            enc_output *= self.d_model ** 0.5
        enc_output = self.dropout(self.position_enc(enc_output))
        enc_output = self.layer_norm(enc_output)
	# 通过多个编码层
        for enc_layer in self.layer_stack:
            enc_output, enc_slf_attn = enc_layer(enc_output, slf_attn_mask=src_mask)
            enc_slf_attn_list += [enc_slf_attn] if return_attns else []

        if return_attns:
            return enc_output, enc_slf_attn_list
        return enc_output,

self.src_word_emb = nn.Embedding(n_src_vocab, d_word_vec, padding_idx=pad_idx)

class EncoderLayer(nn.Module):
    ''' Compose with two layers '''
    def __init__(self, d_model, d_inner, n_head, d_k, d_v, dropout=0.1):
        super(EncoderLayer, self).__init__()
	# 多头注意力层
        self.slf_attn = MultiHeadAttention(n_head, d_model, d_k, d_v, dropout=dropout)
	# 前馈网络
        self.pos_ffn = PositionwiseFeedForward(d_model, d_inner, dropout=dropout)

    def forward(self, enc_input, slf_attn_mask=None):
	# 多头注意力计算得到注意力输出和注意力分数
        enc_output, enc_slf_attn = self.slf_attn(enc_input, enc_input, enc_input, mask=slf_attn_mask)
      	# 输入线性层前向计算
	enc_output = self.pos_ffn(enc_output)
	# 返回计算结果、注意力分数
        return enc_output, enc_slf_attn

class MultiHeadAttention(nn.Module):
    ''' Multi-Head Attention module '''

    def __init__(self, n_head, d_model, d_k, d_v, dropout=0.1):
        super().__init__()

        self.n_head = n_head		# 头的个数
        self.d_k = d_k
        self.d_v = d_v

        self.w_qs = nn.Linear(d_model, n_head * d_k, bias=False)	# 线性层处理QKV的输入
        self.w_ks = nn.Linear(d_model, n_head * d_k, bias=False)
        self.w_vs = nn.Linear(d_model, n_head * d_v, bias=False)
        self.fc = nn.Linear(n_head * d_v, d_model, bias=False)		# 线性输出层

        self.attention = ScaledDotProductAttention(temperature=d_k ** 0.5)	# 缩放点积注意力

        self.dropout = nn.Dropout(dropout)
        self.layer_norm = nn.LayerNorm(d_model, eps=1e-6)


    def forward(self, q, k, v, mask=None):

        d_k, d_v, n_head = self.d_k, self.d_v, self.n_head
        sz_b, len_q, len_k, len_v = q.size(0), q.size(1), k.size(1), v.size(1)	# batch_size, 查询的个数, 键的个数, 值的个数

        residual = q	# 残差计算

        # Pass through the pre-attention projection: b x lq x (n*dv)
        # Separate different heads: 对QKV输入进行线性层计算，并且处理为多个头的输入
        q = self.w_qs(q).view(sz_b, len_q, n_head, d_k)
        k = self.w_ks(k).view(sz_b, len_k, n_head, d_k)
        v = self.w_vs(v).view(sz_b, len_v, n_head, d_v)

        # Transpose for attention dot product: b x n x lq x dv
        q, k, v = q.transpose(1, 2), k.transpose(1, 2), v.transpose(1, 2)

        if mask is not None:
            mask = mask.unsqueeze(1)   # For head axis broadcasting.
	# 进行缩放点积注意力计算，得到计算结果和注意力分数
        q, attn = self.attention(q, k, v, mask=mask)

        # Transpose to move the head dimension back: b x lq x n x dv
        # Combine the last two dimensions to concatenate all the heads together: b x lq x (n*dv)
        q = q.transpose(1, 2).contiguous().view(sz_b, len_q, -1)
        q = self.dropout(self.fc(q))
        q += residual		# 残差连接

        q = self.layer_norm(q)	# 层归一化

        return q, attn

class ScaledDotProductAttention(nn.Module):
    ''' Scaled Dot-Product Attention '''

    def __init__(self, temperature, attn_dropout=0.1):
        super().__init__()
        self.temperature = temperature
        self.dropout = nn.Dropout(attn_dropout)

    def forward(self, q, k, v, mask=None):

        attn = torch.matmul(q / self.temperature, k.transpose(2, 3))

        if mask is not None:
            attn = attn.masked_fill(mask == 0, -1e9)

        attn = self.dropout(F.softmax(attn, dim=-1))
        output = torch.matmul(attn, v)

        return output, attn

class PositionwiseFeedForward(nn.Module):
    ''' A two-feed-forward-layer module '''

    def __init__(self, d_in, d_hid, dropout=0.1):
        super().__init__()
        self.w_1 = nn.Linear(d_in, d_hid) # position-wise
        self.w_2 = nn.Linear(d_hid, d_in) # position-wise
        self.layer_norm = nn.LayerNorm(d_in, eps=1e-6)
        self.dropout = nn.Dropout(dropout)

    def forward(self, x):

        residual = x

        x = self.w_2(F.relu(self.w_1(x)))
        x = self.dropout(x)
        x += residual

        x = self.layer_norm(x)

        return x