-
Notifications
You must be signed in to change notification settings - Fork 0
/
model.py
204 lines (171 loc) · 8.44 KB
/
model.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
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
import torch
import torch.nn as nn
import math
class LayerNormalization(nn.Module):
def __init__(self, features: int, eps: float = 1e-6) -> None:
super().__init__()
self.eps = eps
self.alpha = nn.Parameter(torch.ones(features))
self.bias = nn.Parameter(torch.zeros(features))
def forward(self, x):
mean = x.mean(dim=-1, keepdim=True)
std = x.std(dim=-1, keepdim=True)
return self.alpha * (x - mean) / (std + self.eps) + self.bias
class FeedForwardBlock(nn.Module):
def __init__(self, d_model: int, d_ff: int, dropout: float) -> None:
super().__init__()
self.fc1 = nn.Linear(d_model, d_ff)
self.dropout = nn.Dropout(dropout)
self.fc2 = nn.Linear(d_ff, d_model)
def forward(self, x):
return self.fc2(self.dropout(torch.relu(self.fc1(x))))
class InputEmbeddings(nn.Module):
def __init__(self, d_model: int, vocab_size: int) -> None:
super().__init__()
self.d_model = d_model
self.embedding = nn.Embedding(vocab_size, d_model)
def forward(self, x):
return self.embedding(x) * math.sqrt(self.d_model)
class PositionalEncoding(nn.Module):
def __init__(self, d_model: int, seq_len: int, dropout: float) -> None:
super().__init__()
self.dropout = nn.Dropout(dropout)
pe = torch.zeros(seq_len, d_model)
position = torch.arange(0, seq_len, dtype=torch.float).unsqueeze(1)
div_term = torch.exp(torch.arange(0, d_model, 2).float() * (-math.log(10000.0) / d_model))
pe[:, 0::2] = torch.sin(position * div_term)
pe[:, 1::2] = torch.cos(position * div_term)
pe = pe.unsqueeze(0)
self.register_buffer('pe', pe)
def forward(self, x):
x = x + self.pe[:, :x.shape[1], :].requires_grad_(False)
return self.dropout(x)
class ResidualConnection(nn.Module):
def __init__(self, features: int, dropout: float) -> None:
super().__init__()
self.dropout = nn.Dropout(dropout)
self.norm = LayerNormalization(features)
def forward(self, x, sublayer):
return x + self.dropout(sublayer(self.norm(x)))
class MultiHeadAttentionBlock(nn.Module):
def __init__(self, d_model: int, num_heads: int, dropout: float) -> None:
super().__init__()
self.num_heads = num_heads
self.d_k = d_model // num_heads
self.w_q = nn.Linear(d_model, d_model, bias=False)
self.w_k = nn.Linear(d_model, d_model, bias=False)
self.w_v = nn.Linear(d_model, d_model, bias=False)
self.w_o = nn.Linear(d_model, d_model, bias=False)
self.dropout = nn.Dropout(dropout)
@staticmethod
def attention(query, key, value, mask, dropout: nn.Dropout):
d_k = query.shape[-1]
scores = (query @ key.transpose(-2, -1)) / math.sqrt(d_k)
if mask is not None:
scores.masked_fill_(mask == 0, -1e9)
scores = scores.softmax(dim=-1)
if dropout is not None:
scores = dropout(scores)
return scores @ value, scores
def forward(self, q, k, v, mask):
query = self.w_q(q)
key = self.w_k(k)
value = self.w_v(v)
query = query.view(query.shape[0], query.shape[1], self.num_heads, self.d_k).transpose(1, 2)
key = key.view(key.shape[0], key.shape[1], self.num_heads, self.d_k).transpose(1, 2)
value = value.view(value.shape[0], value.shape[1], self.num_heads, self.d_k).transpose(1, 2)
x, self.attention_scores = MultiHeadAttentionBlock.attention(query, key, value, mask, self.dropout)
x = x.transpose(1, 2).contiguous().view(x.shape[0], -1, self.num_heads * self.d_k)
return self.w_o(x)
class EncoderBlock(nn.Module):
def __init__(self, features: int, self_attention: MultiHeadAttentionBlock, feed_forward: FeedForwardBlock, dropout: float) -> None:
super().__init__()
self.self_attention = self_attention
self.feed_forward = feed_forward
self.residuals = nn.ModuleList([ResidualConnection(features, dropout) for _ in range(2)])
def forward(self, x, src_mask):
x = self.residuals[0](x, lambda x: self.self_attention(x, x, x, src_mask))
x = self.residuals[1](x, self.feed_forward)
return x
class Encoder(nn.Module):
def __init__(self, features: int, layers: nn.ModuleList) -> None:
super().__init__()
self.layers = layers
self.norm = LayerNormalization(features)
def forward(self, x, mask):
for layer in self.layers:
x = layer(x, mask)
return self.norm(x)
class DecoderBlock(nn.Module):
def __init__(self, features: int, self_attention: MultiHeadAttentionBlock, cross_attention: MultiHeadAttentionBlock, feed_forward: FeedForwardBlock, dropout: float) -> None:
super().__init__()
self.self_attention = self_attention
self.cross_attention = cross_attention
self.feed_forward = feed_forward
self.residuals = nn.ModuleList([ResidualConnection(features, dropout) for _ in range(3)])
def forward(self, x, encoder_output, src_mask, tgt_mask):
x = self.residuals[0](x, lambda x: self.self_attention(x, x, x, tgt_mask))
x = self.residuals[1](x, lambda x: self.cross_attention(x, encoder_output, encoder_output, src_mask))
x = self.residuals[2](x, self.feed_forward)
return x
class Decoder(nn.Module):
def __init__(self, features: int, layers: nn.ModuleList) -> None:
super().__init__()
self.layers = layers
self.norm = LayerNormalization(features)
def forward(self, x, encoder_output, src_mask, tgt_mask):
for layer in self.layers:
x = layer(x, encoder_output, src_mask, tgt_mask)
return self.norm(x)
class ProjectionLayer(nn.Module):
def __init__(self, d_model, vocab_size) -> None:
super().__init__()
self.proj = nn.Linear(d_model, vocab_size)
def forward(self, x) -> None:
return self.proj(x)
class Transformer(nn.Module):
def __init__(self, encoder: Encoder, decoder: Decoder, src_embed: InputEmbeddings, tgt_embed: InputEmbeddings, src_pos: PositionalEncoding, tgt_pos: PositionalEncoding, projection_layer: ProjectionLayer) -> None:
super().__init__()
self.encoder = encoder
self.decoder = decoder
self.src_embed = src_embed
self.tgt_embed = tgt_embed
self.src_pos = src_pos
self.tgt_pos = tgt_pos
self.projection_layer = projection_layer
def encode(self, src, src_mask):
src = self.src_embed(src)
src = self.src_pos(src)
return self.encoder(src, src_mask)
def decode(self, encoder_output: torch.Tensor, src_mask: torch.Tensor, tgt: torch.Tensor, tgt_mask: torch.Tensor):
tgt = self.tgt_embed(tgt)
tgt = self.tgt_pos(tgt)
return self.decoder(tgt, encoder_output, src_mask, tgt_mask)
def project(self, x):
return self.projection_layer(x)
def build_transformer(src_vocab_size: int, tgt_vocab_size: int, src_seq_len: int, tgt_seq_len: int, d_model: int = 512, num_layers: int = 6, num_heads: int = 8, dropout: float = 0.1, d_ff: int = 2048) -> Transformer:
src_embed = InputEmbeddings(d_model, src_vocab_size)
tgt_embed = InputEmbeddings(d_model, tgt_vocab_size)
src_pos = PositionalEncoding(d_model, src_seq_len, dropout)
tgt_pos = PositionalEncoding(d_model, tgt_seq_len, dropout)
encoder_blocks = []
for _ in range(num_layers):
self_attention = MultiHeadAttentionBlock(d_model, num_heads, dropout)
feed_forward = FeedForwardBlock(d_model, d_ff, dropout)
encoder_block = EncoderBlock(d_model, self_attention, feed_forward, dropout)
encoder_blocks.append(encoder_block)
decoder_blocks = []
for _ in range(num_layers):
self_attention = MultiHeadAttentionBlock(d_model, num_heads, dropout)
cross_attention = MultiHeadAttentionBlock(d_model, num_heads, dropout)
feed_forward = FeedForwardBlock(d_model, d_ff, dropout)
decoder_block = DecoderBlock(d_model, self_attention, cross_attention, feed_forward, dropout)
decoder_blocks.append(decoder_block)
encoder = Encoder(d_model, nn.ModuleList(encoder_blocks))
decoder = Decoder(d_model, nn.ModuleList(decoder_blocks))
projection_layer = ProjectionLayer(d_model, tgt_vocab_size)
transformer = Transformer(encoder, decoder, src_embed, tgt_embed, src_pos, tgt_pos, projection_layer)
for p in transformer.parameters():
if p.dim() > 1:
nn.init.xavier_uniform_(p)
return transformer