File size: 7,196 Bytes
3ad8be1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import dgl
import dgl.function as fn
import torch
import torch.nn as nn
import torch.nn.functional as F
from dgl.nn.pytorch import edge_softmax
from dgl import softmax_edges

class FC(nn.Module):
    def __init__(self, d_graph_layer, fc_hidden_dim, dropout, n_tasks):
        super(FC, self).__init__()

        self.predict = nn.ModuleList()
        for index,dim in enumerate(fc_hidden_dim):
            self.predict.append(nn.Linear(d_graph_layer, dim))
            self.predict.append(nn.Dropout(dropout))
            self.predict.append(nn.LeakyReLU())
            self.predict.append(nn.BatchNorm1d(dim))
            d_graph_layer = dim
        self.predict.append(nn.Linear(d_graph_layer, n_tasks))

    def forward(self, h):
        for layer in self.predict:
            h = layer(h)
        # return torch.sigmoid(h)
        return h

class EdgeWeightAndSum(nn.Module):
    """
    for normal use, please delete the 'temporary version' line and meanwhile recover the 'normal version'
    """
    def __init__(self, in_feats):
        super(EdgeWeightAndSum, self).__init__()
        self.in_feats = in_feats
        self.atom_weighting = nn.Sequential(
            nn.Linear(in_feats, 1),
            nn.Tanh()
        )

    def forward(self, g, edge_feats):
        with g.local_scope():
            g.edata['e'] = edge_feats
            g.edata['w'] = self.atom_weighting(g.edata['e'])
            weights = g.edata['w']  # temporary version
            h_g_sum = dgl.sum_edges(g, 'e', 'w')
        # return h_g_sum, g.edata['w']  # normal version
        return h_g_sum, weights  # temporary version

class MultiHeadAttention(nn.Module):
    def __init__(self, in_feats, num_head, merge):
        super(MultiHeadAttention, self).__init__()
        self.heads = nn.ModuleList()
        for i in range(num_head):
            self.heads.append(EdgeWeightAndSum(in_feats))
        self.merge = merge

    def forward(self, g, edge_feats):
        h_g_heads, weight_heads = [], []
        for attn_head in self.heads:
            h_g_head, weigh = attn_head(g, edge_feats)
            h_g_heads.append(h_g_head)
            weight_heads.append(weigh)

        if self.merge == 'concat':
            return torch.cat(h_g_heads, dim=1), torch.cat(weight_heads, dim=1)
        else:
            return torch.mean(torch.stack(h_g_heads)), torch.mean(torch.stack(weight_heads))

class EdgeWeightAndSum_v2(nn.Module):
    """
    for normal use, please delete the 'temporary version' line and meanwhile recover the 'normal version'
    """
    def __init__(self, in_feats):
        super(EdgeWeightAndSum_v2, self).__init__()
        self.in_feats = in_feats
        self.atom_weighting = nn.Sequential(
            nn.Linear(in_feats, 1),
            nn.LeakyReLU()
        )

    def forward(self, g, edge_feats):
        with g.local_scope():
            g.edata['e'] = edge_feats
            g.edata['w'] = edge_softmax(g, self.atom_weighting(g.edata['e']))
            weights = g.edata['w']  # temporary version
            h_g_sum = dgl.sum_edges(g, 'e', 'w')
        # return h_g_sum, g.edata['w']  # normal version
        return h_g_sum, weights  # temporary version

class MultiHeadAttention_v2(nn.Module):
    def __init__(self, in_feats, num_head, merge):
        super(MultiHeadAttention_v2, self).__init__()
        self.heads = nn.ModuleList()
        for i in range(num_head):
            self.heads.append(EdgeWeightAndSum_v2(in_feats))
        self.merge = merge

    def forward(self, g, edge_feats):
        h_g_heads, weight_heads = [], []
        for attn_head in self.heads:
            h_g_head, weigh = attn_head(g, edge_feats)
            h_g_heads.append(h_g_head)
            weight_heads.append(weigh)

        if self.merge == 'concat':
            return torch.cat(h_g_heads, dim=1), torch.cat(weight_heads, dim=1)
        else:
            return torch.mean(torch.stack(h_g_heads)), torch.mean(torch.stack(weight_heads))

class EdgeWeightAndSum_v3(nn.Module):
    """
    for normal use, please delete the 'temporary version' line and meanwhile recover the 'normal version'
    """
    def __init__(self, in_feats):
        super(EdgeWeightAndSum_v3, self).__init__()
        self.in_feats = in_feats
        self.atom_weighting = nn.Sequential(
            nn.Linear(in_feats, 1),
            nn.LeakyReLU()
        )

    def forward(self, g, edge_feats):
        with g.local_scope():
            g.edata['e'] = edge_feats
            g.edata['e2'] = self.atom_weighting(g.edata['e'])
            g.edata['w'] = softmax_edges(g, 'e2')
            weights = g.edata['w']  # temporary version
            h_g_sum = dgl.sum_edges(g, 'e', 'w')
        # return h_g_sum, g.edata['w']  # normal version
        return h_g_sum, weights  # temporary version

class MultiHeadAttention_v3(nn.Module):
    def __init__(self, in_feats, num_head, merge):
        super(MultiHeadAttention_v3, self).__init__()
        self.heads = nn.ModuleList()
        for i in range(num_head):
            self.heads.append(EdgeWeightAndSum_v3(in_feats))
        self.merge = merge

    def forward(self, g, edge_feats):
        h_g_heads, weight_heads = [], []
        for attn_head in self.heads:
            h_g_head, weigh = attn_head(g, edge_feats)
            h_g_heads.append(h_g_head)
            weight_heads.append(weigh)

        if self.merge == 'concat':
            return torch.cat(h_g_heads, dim=1), torch.cat(weight_heads, dim=1)
        else:
            return torch.mean(torch.stack(h_g_heads)), torch.mean(torch.stack(weight_heads))

class ReadsOutLayer(nn.Module):
    """
    for normal use, please delete the 'temporary version' line and meanwhile recover the 'normal version'
    """
    def __init__(self, in_feats, pooling, num_head=None, attn_merge=None):
        super(ReadsOutLayer, self).__init__()
        self.pooling = pooling

        if self.pooling == 'w_sum':
            self.weight_and_sum = EdgeWeightAndSum(in_feats)
        elif self.pooling == 'multi_head':
            self.weight_and_sum = MultiHeadAttention(in_feats, num_head, attn_merge)
        elif self.pooling == 'w_sum_v2':
            self.weight_and_sum = EdgeWeightAndSum_v2(in_feats)
        elif self.pooling == 'multi_head_v2':
            self.weight_and_sum = MultiHeadAttention_v2(in_feats, num_head, attn_merge)
        elif self.pooling == 'w_sum_v3':
            self.weight_and_sum = EdgeWeightAndSum_v3(in_feats)
        elif self.pooling == 'multi_head_v3':
            self.weight_and_sum = MultiHeadAttention_v3(in_feats, num_head, attn_merge)


    def forward(self, bg, edge_feats):

        # h_g_sum, weights = self.weight_and_sum(bg, edge_feats)  # temporary version
        with bg.local_scope():
            bg.edata['e'] = edge_feats
            h_g_max = dgl.max_edges(bg, 'e')
            if self.pooling == 'mean':
                h_p = dgl.mean_edges(bg, 'e')
            elif self.pooling == 'sum':
                h_p = dgl.sum_edges(bg,'e')
            else:
                h_p, weights = self.weight_and_sum(bg, edge_feats)  # normal version
        bg.edata['weights'] = weights
        return torch.cat([h_p, h_g_max], dim=1)  # normal version