File size: 1,450 Bytes
b98cec2 |
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 |
import torch
import torch.nn as nn
from torch.nn import functional as F
from torch.nn import MultiheadAttention
class attentionLayer(nn.Module):
def __init__(self, d_model, nhead, dropout=0.1):
super(attentionLayer, self).__init__()
self.self_attn = MultiheadAttention(d_model, nhead, dropout=dropout)
self.linear1 = nn.Linear(d_model, d_model * 4)
self.dropout = nn.Dropout(dropout)
self.linear2 = nn.Linear(d_model * 4, d_model)
self.norm1 = nn.LayerNorm(d_model)
self.norm2 = nn.LayerNorm(d_model)
self.dropout1 = nn.Dropout(dropout)
self.dropout2 = nn.Dropout(dropout)
self.activation = F.relu
def forward(self, src, tar, adjust=False, attn_mask=None):
# type: (Tensor, Optional[Tensor], Optional[Tensor]) -> Tensor
src = src.transpose(0, 1) # B, T, C -> T, B, C
tar = tar.transpose(0, 1) # B, T, C -> T, B, C
if adjust:
src2 = self.self_attn(src, tar, tar, attn_mask=None, key_padding_mask=None)[0]
else:
src2 = self.self_attn(tar, src, src, attn_mask=None, key_padding_mask=None)[0]
src = src + self.dropout1(src2)
src = self.norm1(src)
src2 = self.linear2(self.dropout(self.activation(self.linear1(src))))
src = src + self.dropout2(src2)
src = self.norm2(src)
src = src.transpose(0, 1) # T, B, C -> B, T, C
return src
|