self-attention部分代码注释

 多头注意力机制（Multi-Head Attention, MHA），是 Transformer 模型的核心组件之一。以下是对代码的逐行解析和详细说明：   

attention-is-all-you-need-pytorch-master\transformer\SubLayers.py

class MultiHeadAttention(nn.Module):
    ''' Multi-Head Attention module '''
    '''
        n_head: 多头注意力head数量 8
        d_model: 输入向量的维度 512
        d_k    : 单head中 Q, k 向量的维度 512 / 8 = 64
        d_v    : 单head中V向量的维度 d_k, 与d_v是独立的，可以相等也可以不等。在这里d_k = d_v 64
    '''

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

        self.n_head = n_head
        self.d_k = d_k
        self.d_v = d_v

        self.w_qs = nn.Linear(d_model, n_head * d_k, bias=False)
        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)

        self.attention = ScaledDotProductAttention(temperature=d_k ** 0.5)

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


    def forward(self, q, k, v, mask=None):
        '''
        q, k, v 的形状为 (32, 10, 512)（batch_size=32，seq_len=10，d_model=512）

        '''
        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)

        residual = q

        # Pass through the pre-attention projection: b x lq x (n*dv)
        # Separate different heads: b x lq x n x dv
        q = self.w_qs(q).view(sz_b, len_q, n_head, d_k) #view 为多头
        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, w的维度为(32, 10, 8, 64) 转换后维度为(32, 8, 10, 64)
        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.

        '''
        attn:  (32, 8, 10, 10)
        输出q:  (32, 8, 10, 64)
        '''
        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) #(32, 10, 512)
        q = self.dropout(self.fc(q))
        q += residual

        q = self.layer_norm(q)

        return q, attn

attention-is-all-you-need-pytorch-master\transformer\Modules.py

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):
        '''
        q: (32, 8, 10, 64)  -> 缩放 -> (32, 8, 10, 64)
        k: (32, 8, 10, 64)  -> 转置 -> (32, 8, 64, 10)
        attn = q @ k.transpose(2, 3): (32, 8, 10, 10)

        attn: (32, 8, 10, 10)  -> softmax -> (32, 8, 10, 10)
        v: (32, 8, 10, 64)
        output = attn @ v: (32, 8, 10, 64)
        '''
        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