-
Notifications
You must be signed in to change notification settings - Fork 0
/
multi_head_attention.py
47 lines (39 loc) · 1.84 KB
/
multi_head_attention.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
import torch
import torch.nn as nn
import math
#implement multihead attention mechanism
class MultiHeadAttention(nn.Module):
def __init__(self, d_model, num_heads):
super(MultiHeadAttention, self).__init__()
assert d_model % num_heads == 0, "d_model must be divisible by num_heads"
self.d_model = d_model
self.num_heads = num_heads
self.d_k = d_model // num_heads
#create queries, keys, values and output
self.W_q = nn.Linear(d_model, d_model)
self.W_k = nn.Linear(d_model, d_model)
self.W_v = nn.Linear(d_model, d_model)
self.W_o = nn.Linear(d_model, d_model)
#impelement attention formula
def scaled_dot_product_attention(self, Q, K, V, mask=None):
attn_scores = torch.matmul(Q, K.transpose(-2, -1)) / math.sqrt(self.d_k)
if mask is not None:
attn_scores = attn_scores.masked_fill(mask == 0, -1e9)
attn_probs = torch.softmax(attn_scores, dim=-1)
output = torch.matmul(attn_probs, V)
return output
#function to split input values into multi-head format
def split_heads(self, x):
batch_size, seq_length, d_model = x.size()
return x.view(batch_size, seq_length, self.num_heads, self.d_k).transpose(1, 2)
#function to concatenate multihead attention values together
def combine_heads(self, x):
batch_size, _, seq_length, d_k = x.size()
return x.transpose(1, 2).contiguous().view(batch_size, seq_length, self.d_model)
def forward(self, Q, K, V, mask=None):
Q = self.split_heads(self.W_q(Q))
K = self.split_heads(self.W_k(K))
V = self.split_heads(self.W_v(V))
attn_output = self.scaled_dot_product_attention(Q, K, V, mask)
output = self.W_o(self.combine_heads(attn_output))
return output