MultiHeadAttention的代码实现

转载自：MultiHeadAttention实现详解

import torch
import torch.nn as nn
import torch.nn.functional as F
import math, copy

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

def attention(query, key, value, mask=None, dropout=None):
    "Compute 'Scaled Dot Product Attention'"
    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

class MultiHeadAttention(nn.Module):
    def __init__(self, h, d_model, dropout=0.1):
        "Take in model size and number of heads."
        super(MultiHeadAttention, self).__init__()
        assert d_model % h == 0
        # We assume d_v always equals d_k
        self.d_k = d_model // h
        self.h = h
        self.linears = clones(nn.Linear(d_model, d_model), 4) # create 4 linear layers
        self.attn = None
        self.dropout = nn.Dropout(p=dropout)
        
    def forward(self, query, key, value, mask=None):
        if mask is not None:
            # Same mask applied to all h heads.
            mask = mask.unsqueeze(1)
        batch_size = query.size(0)
        
        print ('Before transform query: ' + str(query.size())) # (batch_size, seq_length, d_model)        

        query, key, value = [l(x) for l, x in zip(self.linears, (query, key, value))] # (batch_size, seq_length, d_model), use first 3 self.linears
        query, key, value = [x.view(batch_size, -1, self.h, self.d_k).transpose(1, 2)
                             for x in (query, key, value)] # (batch_size, h, seq_length, d_k)
                
        print ('After transform query: ' + str(query.size()))
        
        # 2) Apply attention on all the projected vectors in batch. 
        x, self.attn = attention(query, key, value, mask=mask, dropout=self.dropout)
        
        # 3) "Concat" using a view and apply a final linear. 
        x = x.transpose(1, 2).contiguous().view(batch_size, -1, self.h * self.d_k)
        return self.linears[-1](x)

    
h = 8
d_model = 512
batch_size = 1
seq_length = 10
model = MultiHeadAttention(h, d_model)

query = torch.randn([batch_size, seq_length, d_model])
key = query
value = query

print ('Input size: ' + str(query.size()))

m = model(query, key, value)

print ('Output size: ' + str(m.size()))

也可以用使用pytorch自带的类nn.MultiheadAttention，可以查看官方文档
使用方法如下：

multihead_attn = nn.MultiheadAttention(embed_dim, num_heads)
attn_output, attn_output_weights = multihead_attn(query, key, value)