-
Notifications
You must be signed in to change notification settings - Fork 1
/
myconv.py
188 lines (170 loc) · 9.68 KB
/
myconv.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
"""Torch modules for graph attention networks(GAT)."""
# pylint: disable= no-member, arguments-differ, invalid-name
import torch as th
from torch import nn
from dgl import function as fn
from dgl.nn.pytorch import edge_softmax
from dgl._ffi.base import DGLError
from dgl.nn.pytorch.utils import Identity
from dgl.utils import expand_as_pair
# pylint: enable=W0235
class myGATConv(nn.Module):
"""
Adapted from
https://docs.dgl.ai/_modules/dgl/nn/pytorch/conv/gatconv.html#GATConv
"""
def __init__(self,
edge_feats,
num_etypes,
num_ntypes,
in_dims,
in_feats,
out_feats,
num_heads,
feat_drop=0.,
attn_drop=0.,
negative_slope=0.2,
residual=False,
activation=None,
allow_zero_in_degree=False,
bias=False,
alpha=0.):
super(myGATConv, self).__init__()
self._edge_feats = edge_feats
self._num_heads = num_heads
self._in_src_feats, self._in_dst_feats = expand_as_pair(in_feats)
self._out_feats = out_feats
self._allow_zero_in_degree = allow_zero_in_degree
self._num_ntypes = num_ntypes
self._num_etypes = num_etypes
# self.edge_emb = nn.Embedding(num_etypes, edge_feats)
self.edge_emb = nn.ParameterList([nn.Parameter(th.FloatTensor(size=(1, edge_feats))) for _ in range(num_etypes)])
if isinstance(in_feats, tuple):
self.fc_src = nn.Linear(
self._in_src_feats, out_feats * num_heads, bias=False)
self.fc_dst = nn.Linear(
self._in_dst_feats, out_feats * num_heads, bias=False)
else:
# self.fc_list = nn.ModuleList([nn.Linear(self._in_src_feats, out_feats * num_heads, bias=False) for in_dim in in_dims])
self.fc = nn.Linear(
self._in_src_feats, out_feats * num_heads, bias=False)
self.fc_e = nn.Linear(edge_feats, edge_feats*num_heads, bias=False)
# self.attn_l = nn.Parameter(th.FloatTensor(size=(1, num_heads, out_feats)))
# self.attn_r = nn.Parameter(th.FloatTensor(size=(1, num_heads, out_feats)))
# self.attn_e = nn.Parameter(th.FloatTensor(size=(1, num_heads, edge_feats)))
# self.attn_l = nn.Embedding(num_ntypes, 1 * num_heads * out_feats)
# self.attn_r = nn.Embedding(num_ntypes, 1 * num_heads * out_feats)
# self.attn_e = nn.Embedding(num_etypes, 1 * num_heads * edge_feats)
self.attn_l = nn.ParameterList([nn.Parameter(th.FloatTensor(size=(1, num_heads, out_feats))) for _ in range(num_ntypes)])
self.attn_r = nn.ParameterList([nn.Parameter(th.FloatTensor(size=(1, num_heads, out_feats))) for _ in range(num_ntypes)])
self.attn_e = nn.ParameterList([nn.Parameter(th.FloatTensor(size=(1, num_heads, edge_feats))) for _ in range(num_etypes)])
self.feat_drop = nn.Dropout(feat_drop)
self.attn_drop = nn.Dropout(attn_drop)
self.leaky_relu = nn.LeakyReLU(negative_slope)
if residual:
if self._in_dst_feats != out_feats:
self.res_fc = nn.Linear(
self._in_dst_feats, num_heads * out_feats, bias=False)
else:
self.res_fc = Identity()
else:
self.register_buffer('res_fc', None)
self.reset_parameters()
self.activation = activation
self.bias = bias
if bias:
self.bias_param = nn.Parameter(th.zeros((1, num_heads, out_feats)))
self.alpha = alpha
def reset_parameters(self):
gain = nn.init.calculate_gain('relu')
if hasattr(self, 'fc'):
nn.init.xavier_normal_(self.fc.weight, gain=gain)
else:
nn.init.xavier_normal_(self.fc_src.weight, gain=gain)
nn.init.xavier_normal_(self.fc_dst.weight, gain=gain)
for i in range(self._num_ntypes):
nn.init.xavier_normal_(self.attn_l[i], gain=gain)
nn.init.xavier_normal_(self.attn_r[i], gain=gain)
for i in range(self._num_etypes):
nn.init.xavier_normal_(self.attn_e[i], gain=gain)
nn.init.xavier_normal_(self.edge_emb[i], gain=gain)
if isinstance(self.res_fc, nn.Linear):
nn.init.xavier_normal_(self.res_fc.weight, gain=gain)
nn.init.xavier_normal_(self.fc_e.weight, gain=gain)
# not sure
# self.edge_emb.reset_parameters()
# self.attn_l.reset_parameters()
# self.attn_r.reset_parameters()
# self.attn_e.reset_parameters()
def set_allow_zero_in_degree(self, set_value):
self._allow_zero_in_degree = set_value
def forward(self, graph, feat, e_feat, e_count, n_count, res_attn=None):
with graph.local_scope():
if not self._allow_zero_in_degree:
if (graph.in_degrees() == 0).any():
raise DGLError('There are 0-in-degree nodes in the graph, '
'output for those nodes will be invalid. '
'This is harmful for some applications, '
'causing silent performance regression. '
'Adding self-loop on the input graph by '
'calling `g = dgl.add_self_loop(g)` will resolve '
'the issue. Setting ``allow_zero_in_degree`` '
'to be `True` when constructing this module will '
'suppress the check and let the code run.')
if isinstance(feat, tuple):
h_src = self.feat_drop(feat[0])
h_dst = self.feat_drop(feat[1])
if not hasattr(self, 'fc_src'):
self.fc_src, self.fc_dst = self.fc, self.fc
feat_src = self.fc_src(h_src).view(-1, self._num_heads, self._out_feats)
feat_dst = self.fc_dst(h_dst).view(-1, self._num_heads, self._out_feats)
else:
h_src = h_dst = self.feat_drop(feat)
feat_src = feat_dst = self.fc(h_src).view(
-1, self._num_heads, self._out_feats)
if graph.is_block:
feat_dst = feat_src[:graph.number_of_dst_nodes()]
device = th.device('cuda:0' if th.cuda.is_available() else 'cpu')
e_feat.sort(key=lambda x: x[0])
all_index = [x[1] for x in e_feat]
e_feat = [x[0] for x in e_feat]
e_feat = th.tensor(e_feat, dtype=th.long).to(device)
all_index = th.tensor(all_index, dtype=th.long).to(device)
_, all_index2 = th.sort(all_index)
# embed_e_feat = self.edge_emb(e_feat)
embed_e_feat = th.cat([e_feat[sum(e_count[0:i]):sum(e_count[0:i+1])].unsqueeze(-1) * self.edge_emb[e_feat[sum(e_count[0:i])]] for i in range(len(e_count))], 0)
embed_e_feat = self.fc_e(embed_e_feat).view(-1, self._num_heads, self._edge_feats)
# ee = (embed_e_feat * self.attn_e).sum(dim=-1).unsqueeze(-1)
# el = (feat_src * self.attn_l).sum(dim=-1).unsqueeze(-1)
# er = (feat_dst * self.attn_r).sum(dim=-1).unsqueeze(-1)
# ee = th.cat([embed_e_feat[sum(e_count[0:i]):sum(e_count[0:i+1])] * self.attn_e(e_feat[sum(e_count[0:i])]).view(-1, self._num_heads, self._edge_feats) for i in range(len(e_count))], 0).sum(dim=-1).unsqueeze(-1)
# el = th.cat([feat_src[sum(n_count[0:i]):sum(n_count[0:i+1])] * self.attn_l(th.tensor([i], dtype=th.long).to(device)[0]).view(-1, self._num_heads, self._out_feats) for i in range(len(n_count))], 0).sum(dim=-1).unsqueeze(-1)
# er = th.cat([feat_dst[sum(n_count[0:i]):sum(n_count[0:i+1])] * self.attn_r(th.tensor([i], dtype=th.long).to(device)[0]).view(-1, self._num_heads, self._out_feats) for i in range(len(n_count))], 0).sum(dim=-1).unsqueeze(-1)
ee = th.cat([embed_e_feat[sum(e_count[0:i]):sum(e_count[0:i+1])] * self.attn_e[e_feat[sum(e_count[0:i])]] for i in range(len(e_count))], 0).sum(dim=-1).unsqueeze(-1)
el = th.cat([feat_src[sum(n_count[0:i]):sum(n_count[0:i+1])] * self.attn_l[th.tensor([i], dtype=th.long).to(device)[0]] for i in range(len(n_count))], 0).sum(dim=-1).unsqueeze(-1)
er = th.cat([feat_dst[sum(n_count[0:i]):sum(n_count[0:i+1])] * self.attn_r[th.tensor([i], dtype=th.long).to(device)[0]] for i in range(len(n_count))], 0).sum(dim=-1).unsqueeze(-1)
ee = ee.index_select(0, all_index2)
graph.srcdata.update({'ft': feat_src, 'el': el})
graph.dstdata.update({'er': er})
graph.edata.update({'ee': ee})
graph.apply_edges(fn.u_add_v('el', 'er', 'e'))
e = self.leaky_relu(graph.edata.pop('e')+graph.edata.pop('ee'))
# compute softmax
graph.edata['a'] = self.attn_drop(edge_softmax(graph, e))
if res_attn is not None:
graph.edata['a'] = graph.edata['a'] * (1-self.alpha) + res_attn * self.alpha
# message passing
graph.update_all(fn.u_mul_e('ft', 'a', 'm'),
fn.sum('m', 'ft'))
rst = graph.dstdata['ft']
# residual
if self.res_fc is not None:
resval = self.res_fc(h_dst).view(h_dst.shape[0], -1, self._out_feats)
rst = rst + resval
# bias
if self.bias:
rst = rst + self.bias_param
# activation
if self.activation:
rst = self.activation(rst)
return rst, graph.edata.pop('a').detach()