File size: 7,485 Bytes
238ab50
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
# code taken from https://towardsdatascience.com/how-to-code-the-transformer-in-pytorch-24db27c8f9ec
# and https://pytorch.org/tutorials/beginner/transformer_tutorial.html

import torch
import math
import copy


class Embedder(torch.nn.Module):
    def __init__(self, vocab_size, d_model):
        super().__init__()
        self.embed = torch.nn.Embedding(vocab_size, d_model)

    def forward(self, x):
        return self.embed(x)


class PositionalEncoder(torch.nn.Module):
    def __init__(self, d_model, dropout=0.1, max_seq_len=80):
        super().__init__()
        self.dropout = torch.nn.Dropout(p=dropout)

        position = torch.arange(max_seq_len).unsqueeze(1)
        div_term = torch.exp(torch.arange(0, d_model, 2) * (-math.log(10000.0) / d_model))
        pe = torch.zeros(max_seq_len, 1, d_model)
        pe[:, 0, 0::2] = torch.sin(position * div_term)
        pe[:, 0, 1::2] = torch.cos(position * div_term)
        self.register_buffer('pe',
                             pe)  # notifies PyTorch that this value should be saved like a model parameter but should not have gradients

    def forward(self, x):
        x = x + self.pe[:x.size(0)]
        return self.dropout(x)


class MultiHeadAttention(torch.nn.Module):
    def __init__(self, heads, d_model, dropout=0.1):
        super().__init__()

        self.d_model = d_model
        self.d_k = d_model // heads
        self.h = heads

        self.q_linear = torch.nn.Linear(d_model, d_model)
        self.v_linear = torch.nn.Linear(d_model, d_model)
        self.k_linear = torch.nn.Linear(d_model, d_model)
        self.dropout = torch.nn.Dropout(dropout)
        self.out = torch.nn.Linear(d_model, d_model)

    def forward(self, q, k, v, mask=None):
        bs = q.size(0)

        # perform linear operation and split into h heads

        k = self.k_linear(k).view(bs, -1, self.h, self.d_k)
        q = self.q_linear(q).view(bs, -1, self.h, self.d_k)
        v = self.v_linear(v).view(bs, -1, self.h, self.d_k)

        # transpose to get dimensions bs * h * sl * d_model

        k = k.transpose(1, 2)
        q = q.transpose(1, 2)
        v = v.transpose(1, 2)

        # calculate attention using function we will define next
        scores = attention(q, k, v, self.d_k, mask, self.dropout)

        # concatenate heads and put through final linear layer
        concat = scores.transpose(1, 2).contiguous().view(bs, -1, self.d_model)

        output = self.out(concat)

        return output


def attention(q, k, v, d_k, mask=None, dropout=None):
    scores = torch.matmul(q, k.transpose(-2, -1)) / math.sqrt(d_k)
    if mask is not None:
        mask = mask.unsqueeze(1)
        scores = scores.masked_fill(mask == 0, -1e9)
    scores = torch.nn.functional.softmax(scores, dim=-1)

    if dropout is not None:
        scores = dropout(scores)

    output = torch.matmul(scores, v)
    return output


class FeedForward(torch.nn.Module):
    def __init__(self, d_model, d_ff=2048, dropout=0.1):
        super().__init__()
        # We set d_ff as a default to 2048
        self.linear_1 = torch.nn.Linear(d_model, d_ff)
        self.dropout = torch.nn.Dropout(dropout)
        self.linear_2 = torch.nn.Linear(d_ff, d_model)

    def forward(self, x):
        x = self.dropout(torch.nn.functional.relu(self.linear_1(x)))
        x = self.linear_2(x)
        return x


class Norm(torch.nn.Module):
    def __init__(self, d_model, eps=1e-6):
        super().__init__()

        self.size = d_model
        # create two learnable parameters to calibrate normalization
        self.alpha = torch.nn.Parameter(torch.ones(self.size))
        self.bias = torch.nn.Parameter(torch.zeros(self.size))
        self.eps = eps

    def forward(self, x):
        norm = self.alpha * (x - x.mean(dim=-1, keepdim=True)) / (x.std(dim=-1, keepdim=True) + self.eps) + self.bias
        return norm


# build an encoder layer with one multi-head attention layer and one # feed-forward layer
class EncoderLayer(torch.nn.Module):
    def __init__(self, d_model, heads, dropout=0.1):
        super().__init__()
        self.norm_1 = Norm(d_model)
        self.norm_2 = Norm(d_model)
        self.attn = MultiHeadAttention(heads, d_model)
        self.ff = FeedForward(d_model)
        self.dropout_1 = torch.nn.Dropout(dropout)
        self.dropout_2 = torch.nn.Dropout(dropout)

    def forward(self, x, mask):
        x2 = self.norm_1(x)
        x = x + self.dropout_1(self.attn(x2, x2, x2, mask))
        x2 = self.norm_2(x)
        x = x + self.dropout_2(self.ff(x2))
        return x


# build a decoder layer with two multi-head attention layers and
# one feed-forward layer
class DecoderLayer(torch.nn.Module):
    def __init__(self, d_model, heads, dropout=0.1):
        super().__init__()
        self.norm_1 = Norm(d_model)
        self.norm_2 = Norm(d_model)
        self.norm_3 = Norm(d_model)

        self.dropout_1 = torch.nn.Dropout(dropout)
        self.dropout_2 = torch.nn.Dropout(dropout)
        self.dropout_3 = torch.nn.Dropout(dropout)

        self.attn_1 = MultiHeadAttention(heads, d_model)
        self.attn_2 = MultiHeadAttention(heads, d_model)
        self.ff = FeedForward(d_model)

    def forward(self, x, e_outputs, src_mask, trg_mask):
        x2 = self.norm_1(x)
        x = x + self.dropout_1(self.attn_1(x2, x2, x2, trg_mask))
        x2 = self.norm_2(x)
        x = x + self.dropout_2(self.attn_2(x2, e_outputs, e_outputs,
                                           src_mask))
        x2 = self.norm_3(x)
        x = x + self.dropout_3(self.ff(x2))
        return x


# We can then build a convenient cloning function that can generate multiple layers:
def get_clones(module, N):
    return torch.nn.ModuleList([copy.deepcopy(module) for i in range(N)])


class Encoder(torch.nn.Module):
    def __init__(self, vocab_size, d_model, N, heads):
        super().__init__()
        self.N = N
        self.embed = Embedder(vocab_size, d_model)
        self.pe = PositionalEncoder(d_model)
        self.layers = get_clones(EncoderLayer(d_model, heads), N)
        self.norm = Norm(d_model)

    def forward(self, src, mask):
        x = self.embed(src)
        x = self.pe(x)
        for i in range(self.N):
            x = self.layers[i](x, mask)
        return self.norm(x)


class Decoder(torch.nn.Module):
    def __init__(self, vocab_size, d_model, N, heads):
        super().__init__()
        self.N = N
        self.embed = Embedder(vocab_size, d_model)
        self.pe = PositionalEncoder(d_model)
        self.layers = get_clones(DecoderLayer(d_model, heads), N)
        self.norm = Norm(d_model)

    def forward(self, trg, e_outputs, src_mask, trg_mask):
        x = self.embed(trg)
        x = self.pe(x)
        for i in range(self.N):
            x = self.layers[i](x, e_outputs, src_mask, trg_mask)
        return self.norm(x)


class Transformer(torch.nn.Module):
    def __init__(self, src_vocab, trg_vocab, d_model, N, heads):
        super().__init__()
        self.encoder = Encoder(src_vocab, d_model, N, heads)
        self.decoder = Decoder(trg_vocab, d_model, N, heads)
        self.out = torch.nn.Linear(d_model, trg_vocab)

    def forward(self, src, trg, src_mask, trg_mask):
        e_outputs = self.encoder(src, src_mask)
        d_output = self.decoder(trg, e_outputs, src_mask, trg_mask)
        output = self.out(d_output)
        return output


# we don't perform softmax on the output as this will be handled
# automatically by our loss function