-
Notifications
You must be signed in to change notification settings - Fork 17
/
EasyDGL.py
198 lines (156 loc) · 8.86 KB
/
EasyDGL.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
"""
@version: 1.0
@author: Chao Chen
@contact: chao.chen@sjtu.edu.cn
"""
import pickle
import tensorflow.compat.v1 as tf
import module.coding as C
import module.temporal as T
from model.Base import Sequential, FeedForward, layernorm
def clip_by_value(input_tensor):
return tf.clip_by_value(input_tensor, clip_value_min=0., clip_value_max=100.)
def gelu(input_tensor):
"""Gaussian Error Linear Unit.
This is a smoother version of the RELU.
Original paper: https://arxiv.org/abs/1606.08415
Args:
input_tensor: float Tensor to perform activation.
Returns:
`input_tensor` with the GELU activation applied.
"""
cdf = 0.5 * (1.0 + tf.erf(input_tensor / tf.sqrt(2.0)))
return input_tensor * cdf
class EasyDGL(Sequential):
def __init__(self, num_items, FLAGS):
super().__init__(num_items, FLAGS)
self.mask = num_items
self.seqslen += 1
self.num_items += 1
self.masklen = FLAGS.masklen
self.time_scale = FLAGS.time_scale
self.mark_lookup_table = pickle.load(open(FLAGS.mark, 'rb')).toarray()
self.num_events = self.mark_lookup_table.shape[-1]
self.ct_reg = FLAGS.ct_reg
with tf.variable_scope("CSTMA"):
self.item_embs = C.Embedding(self.num_items, self.num_units, self.l2_reg,
zero_pad=True, scale=True, scope="item_embs")
self.mark_embs = C.Embedding(self.num_events, self.num_units, self.l2_reg,
zero_pad=True, scale=False, scope="mark_embs")
self.pcoding = C.PositionCoding(self.seqslen, self.num_units, self.l2_reg, scope="spatial_embs")
self.tcoding = C.TimeSinusoidCoding(self.num_units)
self.output_bias = self.output_bias(inf_pad=True)
self.list_attention = list()
self.list_dense = list()
for i in range(FLAGS.num_blocks):
with tf.variable_scope("num_blocks_%d" % i):
attention = T.BiMAU(self.num_units, self.num_heads,
self.num_events, self.attention_probs_dropout_rate)
fforward = FeedForward([self.num_units, self.num_units], self.hidden_dropout_rate)
self.list_attention.append(attention)
self.list_dense.append(fforward)
def __call__(self, features, is_training):
seqs_ids = features['seqs_i']
seqs_ts = features['seqs_t'] / self.time_scale
seqs_spans = clip_by_value(seqs_ts[:, 1:] - seqs_ts[:, :-1])
seqs_spans = tf.concat([seqs_spans[:, :1], seqs_spans], axis=-1)
seqs_marks = tf.where(tf.equal(seqs_ids, self.mask), tf.zeros_like(seqs_ids), seqs_ids)
seqs_marks = tf.nn.embedding_lookup(self.mark_lookup_table, seqs_marks)
# temporal encoding
seqs_tcodes = self.tcoding.code(seqs_ts)
# positional encoding
seqs_units = self.item_embs(seqs_ids) + seqs_tcodes
posn_codes = self.pcoding.code(seqs_units)
# event mark encoding
marks_codes = tf.nn.embedding_lookup(self.mark_embs.lookup_table, seqs_marks)
marks_codes = tf.reduce_sum(marks_codes, axis=2)
seqs_units = tf.concat([seqs_units, posn_codes, marks_codes], axis=-1)
# Dropout
seqs_units = tf.layers.dropout(seqs_units, rate=self.hidden_dropout_rate,
training=tf.convert_to_tensor(is_training))
seqs_masks = tf.expand_dims(tf.to_float(tf.not_equal(seqs_ids, 0)), 1)
seqs_masks = tf.tile(seqs_masks, [self.num_heads, tf.shape(seqs_ids)[1], 1])
# Run the stacked transformer.
# `sequence_output` shape = [batch_size, seq_length, hidden_size].
prev_outputs = seqs_units
for i, attention in enumerate(self.list_attention):
with tf.variable_scope("layer_%d" % i):
layer_inputs = prev_outputs
with tf.variable_scope("attention"):
with tf.variable_scope("self"):
# sequential-temporal representations
attention_outs, seqs_intny = attention(layer_inputs, layer_inputs, seqs_masks,
seqs_spans, seqs_marks, is_training)
# Run a linear projection of `hidden_size` then add a residual
# with `layer_input`.
with tf.variable_scope("output"):
attention_outs = tf.layers.dense(attention_outs, self.num_units)
attention_outs = tf.layers.dropout(attention_outs, rate=self.hidden_dropout_rate,
training=is_training)
attention_outs = layernorm(attention_outs + layer_inputs[:, :, :self.num_units])
# The activation is only applied to the "intermediate" hidden layer.
with tf.variable_scope("intermediate"):
intermediate_outputs = tf.layers.dense(attention_outs, 2 * self.num_units,
activation=gelu)
# Down-project back to `hidden_size` then add the residual.
with tf.variable_scope("output"):
layer_outputs = tf.layers.dense(intermediate_outputs, self.num_units)
layer_outputs = tf.layers.dropout(layer_outputs, rate=self.hidden_dropout_rate,
training=is_training)
layer_outputs = layernorm(layer_outputs + attention_outs)
prev_outputs = layer_outputs
# likelihood for point process
if is_training:
tf.add_to_collection("LLE_PP", seqs_intny)
seqs_outs = prev_outputs
with tf.variable_scope("cls/predictions"):
with tf.variable_scope("transform"):
seqs_outs = tf.layers.dense(seqs_outs, self.num_units, activation=gelu)
seqs_outs = layernorm(seqs_outs)
if is_training:
seqs_outs = tf.batch_gather(seqs_outs, features['masked_positions'])
seqs_outs = tf.reshape(seqs_outs, [tf.shape(seqs_ids)[0] * self.masklen, self.num_units])
else:
# only using the latest representations for making predictions
seqs_outs = tf.reshape(seqs_outs[:, -1], [tf.shape(seqs_ids)[0], self.num_units])
# compute logits
logits = tf.matmul(seqs_outs, self.item_embs.lookup_table, transpose_b=True)
logits = tf.nn.bias_add(logits, self.output_bias)
return logits
def train(self, features, labels):
logits = self.__call__(features, is_training=True)
log_probs = tf.log(tf.nn.softmax(logits, -1) + 1e-5) # (bs*seqsLen, num_items)
# continuous-time regularization
regularizer = tf.losses.get_regularization_loss()
if self.ct_reg != 0.:
masked_positions = features['masked_positions']
seqs_spans = clip_by_value(features['seqs_t'][:, 1:] - features['seqs_t'][:, :-1])
seqs_spans = tf.concat([seqs_spans[:, :1], seqs_spans], axis=-1)
seqs_spans = tf.batch_gather(seqs_spans, masked_positions)
next_mark_onehot = tf.to_float(tf.nn.embedding_lookup(self.mark_lookup_table, labels))
if self.num_heads != 1:
seqs_spans = tf.tile(seqs_spans, [self.num_heads, 1])
next_mark_onehot = tf.tile(next_mark_onehot, [self.num_heads, 1, 1])
masked_positions = tf.tile(masked_positions, [self.num_heads, 1])
for seqs_intny in tf.get_collection("LLE_PP"):
seqs_intny = tf.batch_gather(seqs_intny, masked_positions)
ct_regularizer = T.MAU.biased_likelihood(
seqs_intny, next_mark_onehot, seqs_spans)
regularizer += self.ct_reg * ct_regularizer / self.num_heads
# softmax based loss without negative sampling
labels = tf.reshape(labels, [-1])
label_ids = tf.one_hot(labels, depth=self.num_items, dtype=tf.float32)
label_weights = tf.to_float(tf.not_equal(labels, 0))
per_example_loss = -tf.reduce_sum(log_probs * label_ids, axis=[-1])
numerator = tf.reduce_sum(label_weights * per_example_loss)
denominator = tf.reduce_sum(label_weights) + 1e-5
loss = numerator / denominator
# perform gradient-based optimization
loss = loss + regularizer
train_op = self.trainOp(loss)
with tf.variable_scope("Sequential/TRAIN"):
_, loss_op = tf.metrics.mean(loss, name='loss')
loss_vars = tf.get_collection(tf.GraphKeys.LOCAL_VARIABLES, scope="main/Sequential/TRAIN")
assert len(loss_vars) > 0, "(train)metric local variables should not be None."
loss_init_op = tf.variables_initializer(loss_vars)
return train_op, loss_op, loss_init_op