File size: 4,748 Bytes
8896a5f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from __future__ import print_function,division

import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.nn.utils.rnn import PackedSequence

class LastHundredEmbed(nn.Module):
    
    def forward(self, x):
        return x[:,:,-100:]

class IdentityEmbed(nn.Module):
    
    def forward(self, x):
        return x

class FullyConnectedEmbed(nn.Module):
    def __init__(self, nin, nout, dropout=0.5, activation=nn.ReLU()):
        super(FullyConnectedEmbed, self).__init__()
        self.nin = nin
        self.nout = nout
        self.dropout_p = dropout
        
        self.transform = nn.Linear(nin, nout)
        self.drop = nn.Dropout(p = self.dropout_p)
        self.activation = activation
        
    def forward(self, x):
        t = self.transform(x)
        t = self.activation(t)
        t = self.drop(t)
        return t

class LMEmbed(nn.Module):
    def __init__(self, nin, nout, lm, padding_idx=-1, transform=nn.ReLU()
                , sparse=False):
        super(LMEmbed, self).__init__()

        if padding_idx == -1:
            padding_idx = nin-1

        self.lm = lm
        self.embed = nn.Embedding(nin, nout, padding_idx=padding_idx, sparse=sparse)
        self.proj = nn.Linear(lm.hidden_size(), nout)
        self.transform = transform
        self.nout = nout

    def forward(self, x):
        packed = type(x) is PackedSequence
        h_lm = self.lm.encode(x)

        # embed and unpack if packed
        if packed:
            h = self.embed(x.data)
            h_lm = h_lm.data
        else:
            h = self.embed(x)

        # project
        h_lm = self.proj(h_lm)
        h = self.transform(h + h_lm)

        # repack if needed
        if packed:
            h = PackedSequence(h, x.batch_sizes)

        return h


class Linear(nn.Module):
    def __init__(self, nin, nhidden, nout, padding_idx=-1,
                 sparse=False, lm=None):
        super(Linear, self).__init__()
        
        if padding_idx == -1:
            padding_idx = nin-1
        
        if lm is not None:
            self.embed = LMEmbed(nin, nhidden, lm, padding_idx=padding_idx, sparse=sparse)
            self.proj = nn.Linear(self.embed.nout, nout)
            self.lm = True
        else:
            self.proj = nn.Embedding(nin, nout, padding_idx=padding_idx, sparse=sparse)
            self.lm = False

        self.nout = nout
        
        
    def forward(self, x):
        
        if self.lm:
            h = self.embed(x)
            if type(h) is PackedSequence:
                h = h.data
                z = self.proj(h)
                z = PackedSequence(z, x.batch_sizes)
            else:
                h = h.view(-1, h.size(2))
                z = self.proj(h)
                z = z.view(x.size(0), x.size(1), -1)
        else:
            if type(x) is PackedSequence:
                z = self.embed(x.data)
                z = PackedSequence(z, x.batch_sizes)
            else:
                z = self.embed(x)
    
        return z


class StackedRNN(nn.Module):
    def __init__(self, nin, nembed, nunits, nout, nlayers=2, padding_idx=-1, dropout=0,
                 rnn_type='lstm', sparse=False, lm=None):
        super(StackedRNN, self).__init__()
        
        if padding_idx == -1:
            padding_idx = nin-1
        
        if lm is not None:
            self.embed = LMEmbed(nin, nembed, lm, padding_idx=padding_idx, sparse=sparse)
            nembed = self.embed.nout
            self.lm = True
        else:
            self.embed = nn.Embedding(nin, nembed, padding_idx=padding_idx, sparse=sparse)
            self.lm = False

        if rnn_type == 'lstm':
            RNN = nn.LSTM
        elif rnn_type == 'gru':
            RNN = nn.GRU

        self.dropout = nn.Dropout(p=dropout)
        if nlayers == 1:
            dropout = 0

        self.rnn = RNN(nembed, nunits, nlayers, batch_first=True
                      , bidirectional=True, dropout=dropout)
        self.proj = nn.Linear(2*nunits, nout)
        self.nout = nout

        
        
    def forward(self, x):
        
        if self.lm:
            h = self.embed(x)
        else:
            if type(x) is PackedSequence:
                h = self.embed(x.data)
                h = PackedSequence(h, x.batch_sizes)
            else:
                h = self.embed(x)
            
        h,_ = self.rnn(h)
        
        if type(h) is PackedSequence:
            h = h.data
            h = self.dropout(h)
            z = self.proj(h)
            z = PackedSequence(z, x.batch_sizes)
        else:
            h = h.view(-1, h.size(2))
            h = self.dropout(h)
            z = self.proj(h)
            z = z.view(x.size(0), x.size(1), -1)
    
        return z