-
Notifications
You must be signed in to change notification settings - Fork 0
/
model.py
263 lines (206 loc) · 10.1 KB
/
model.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
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
import tensorflow as tf
from tensorflow.models.rnn import rnn_cell
from tensorflow.models.rnn import seq2seq
import numpy as np
class WindowCell():
def __init__(self, input_size, cluster_size, vocab_size, con_size):
self._input_size = input_size
self._cluster_size = cluster_size
self._vocab_size = vocab_size
self._con_size = con_size
@property
def input_size(self):
return self._input_size
@property
def con_size(self):
return self._con_size
@property
def state_size(self):
return self._cluster_size
@property
def output_size(self):
return self._vocab_size
# inputs - tensor of size [batch_size x vocab_size]
# con - tensor of size [batch_size x seq_length x vocab_size]
def __call__(self, inputs, state, con):
with tf.variable_scope(type(self).__name__):
# batch_size x 3(cluster_size)
concat = rnn_cell.linear(inputs, 3 * self._cluster_size, True)
a, b, k = tf.split(1, 3, concat)
ao = tf.exp(a)
bo = tf.exp(b)
ko = state + tf.exp(k) # batch_size x _cluster_size
phi = []
for i in range(self._con_size):
# each phi is [batch_size x 1]
phi.append(tf.reduce_sum(ao * tf.exp(- bo * tf.square(ko - i)), 1, keep_dims=True))
# tf.concat(1, phi) -> [batch_size x seq_length]
# tf.expan_dims(%, 1) -> [batch_size x 1 x seq_length]
# tf.batch_matmul(%, con) -> [batch_size x 1 x vocab_size]
# tf.squeeze(%) -> [batch_size x vocab_size]
wt = tf.squeeze(tf.batch_matmul(tf.expand_dims(tf.concat(1, phi), 1), con), [1])
return wt, ko
class Network():
def __init__(self, cell_fn, input_size, cluster_size, vocab_size, hidden_unit, con_size, num_layers):
self._num_layers = num_layers
self._con_size = con_size
self._vocab_size = vocab_size
self._cluster_size = cluster_size
self.w = WindowCell(input_size, cluster_size, vocab_size, con_size)
self.hn = []
self.hn.append(cell_fn(hidden_unit , 1.0, input_size + self.w.output_size))
for i in range(num_layers - 1):
self.hn.append(cell_fn(hidden_unit , 1.0, input_size + self.hn[i].output_size + self.w.output_size))
self._state_size = self._vocab_size + self.w.state_size + sum(h.state_size for h in self.hn)
#print self._state_size
#for i in range(len(self.hn)):
#print self.hn[i].state_size
#print self._vocab_size
#print self.w.state_size
#exit(0)
def zero_state(self, batch_size, dtype):
zeros = tf.zeros([batch_size, self.state_size], dtype=dtype)
return zeros
def zero_constrain(self, batch_size):
zeros = tf.zeros([batch_size, self._con_size], dtype=tf.float32)
return zeros
@property
def state_size(self):
return self._state_size
# con: [batch_size x seq_length x vocab_size]
def __call__(self, inputs, state, con, scope=None):
cur_state_pos = 0
new_states = []
# state
# wt, h1, w, h2, h3
# vocab_size, 2 x hidden_unit, cluster_size, 2 x hidden_unit, 2 x hidden_unit
with tf.variable_scope(scope or type(self).__name__):
outh = [None] * len(self.hn)
wt_1 = tf.slice(state, [0, 0], [-1, self._vocab_size])
new_states.append(None)
cur_state_pos += self._vocab_size
with tf.variable_scope("hidden1"):
cur_state = tf.slice(state, [0, cur_state_pos], [-1, self.hn[0].state_size])
outh[0], new_state = self.hn[0](tf.concat(1, [inputs, wt_1]), cur_state)
new_states.append(new_state)
cur_state_pos += self.hn[0].state_size
with tf.variable_scope("window"):
cur_state = tf.slice(state, [0, cur_state_pos], [-1, self.w.state_size])
wt, new_state = self.w(outh[0], cur_state, con)
new_states[0] = wt
new_states.append(new_state)
cur_state_pos += self.w.state_size
for i in range(1, len(self.hn)):
with tf.variable_scope("hidden%d" % (i+1)):
cur_state = tf.slice(state, [0, cur_state_pos], [-1, self.hn[i].state_size])
outh[i], new_state = self.hn[i](tf.concat(1, [inputs, outh[i-1], wt]), cur_state)
new_states.append(new_state)
cur_state_pos += self.hn[i].state_size
return tf.concat(1, outh), tf.concat(1, new_states) # TODO add skip connection?
def decoder(inputs, initial_state, network, con, loop_function=None, scope=None):
with tf.variable_scope(scope or "rnn_decoder"):
state = initial_state
outputs = []
prev = None
for i, inp in enumerate(inputs):
if loop_function is not None and prev is not None:
with variable_scope.variable_scope("loop_function", reuse=True):
inp = loop_function(prev, i)
if i > 0:
tf.get_variable_scope().reuse_variables()
output, state = network(inp, state, con)
#output, state = network(inp, initial_state, con)
outputs.append(output)
if loop_function is not None:
prev = output
return outputs, state
class ConstrainedModel():
def __init__(self, args, infer=False):
self.args = args
if infer:
args.batch_size = 1
args.seq_length = 1
if args.model == 'rnn':
cell_fn = rnn_cell.BasicRNNCell
elif args.model == 'gru':
cell_fn = rnn_cell.GRUCell
elif args.model == 'lstm':
cell_fn = rnn_cell.BasicLSTMCell
else:
raise Exception("model type not supported: {}".format(args.model))
#cell = cell_fn(args.rnn_size)
con_size = 50 #args.seq_length
#self.cell = cell = rnn_cell.MultiRNNCell([cell] * args.num_layers)
self.network = Network(cell_fn, args.vocab_size, 20, args.vocab_size, args.rnn_size, con_size, args.num_layers)
self.input_data = tf.placeholder(tf.int32, [args.batch_size, args.seq_length])
self.targets = tf.placeholder(tf.int32, [args.batch_size, args.seq_length])
self.con_data = tf.placeholder(tf.int32, [args.batch_size, con_size])
self.initial_state = self.network.zero_state(args.batch_size, tf.float32)
with tf.variable_scope('rnnlm'):
softmax_w = tf.get_variable("softmax_w", [args.rnn_size * args.num_layers, args.vocab_size])
softmax_b = tf.get_variable("softmax_b", [args.vocab_size])
#embedding = tf.get_variable("embedding", [args.vocab_size, args.rnn_size])
embedding = tf.constant(np.identity(args.vocab_size, dtype=np.float32))
inputs = tf.split(1, args.seq_length, tf.nn.embedding_lookup(embedding, self.input_data))
inputs = [tf.squeeze(input_, [1]) for input_ in inputs]
# [(batch_size * seq_length) x vocab_size]
con = tf.nn.embedding_lookup(embedding, self.con_data)
def loop(prev, _):
prev = tf.nn.xw_plus_b(prev, softmax_w, softmax_b)
prev_symbol = tf.stop_gradient(tf.argmax(prev, 1))
return tf.nn.embedding_lookup(embedding, prev_symbol)
#outputs, states = seq2seq.rnn_decoder(inputs, self.initial_state, cell, loop_function=loop if infer else None, scope='rnnlm')
outputs, states = decoder(inputs, self.initial_state, self.network, con, loop_function=loop if infer else None, scope='rnnlm')
# turn a list of output into row matrix where each row is output
output = tf.reshape(tf.concat(1, outputs), [-1, args.rnn_size * args.num_layers])
self.logits = tf.nn.xw_plus_b(output, softmax_w, softmax_b)
self.probs = tf.nn.softmax(self.logits)
loss = seq2seq.sequence_loss_by_example([self.logits],
[tf.reshape(self.targets, [-1])],
[tf.ones([args.batch_size * args.seq_length])],
args.vocab_size)
self.cost = tf.reduce_sum(loss) / args.batch_size / args.seq_length
print states
self.final_state = states
self.lr = tf.Variable(0.0, trainable=False)
tvars = tf.trainable_variables()
grads, _ = tf.clip_by_global_norm(tf.gradients(self.cost, tvars),
args.grad_clip)
optimizer = tf.train.AdamOptimizer(self.lr)
self.train_op = optimizer.apply_gradients(zip(grads, tvars))
def sample(self, sess, chars, vocab, num=200, prime=''):
state = self.network.zero_state(1, tf.float32).eval()
#con = self.network.zero_constrain(1).eval()
con_text = prime + 'the american auto industry'
con_text = con_text.upper() + ' ' * (50 - len(con_text))
con = np.expand_dims(map(vocab.get, con_text), 0)
print con
print con_text
for char in prime[:-1]:
x = np.zeros((1, 1))
x[0, 0] = vocab[char]
feed = {self.input_data: x, self.initial_state:state, self.con_data: con}
[state] = sess.run([self.final_state], feed)
#print "max = "
#print np.argmax(state[0][:self.network._vocab_size])
#print state[0][668:668+20]
pos = self.network._vocab_size + self.network.hn[0].state_size
print state[0][pos:pos+self.network._cluster_size]
def weighted_pick(weights):
t = np.cumsum(weights)
s = np.sum(weights)
return(int(np.searchsorted(t, np.random.rand(1)*s)))
ret = prime
char = prime[-1]
for n in xrange(num):
x = np.zeros((1, 1))
x[0, 0] = vocab[char]
feed = {self.input_data: x, self.initial_state:state, self.con_data: con}
[probs, state] = sess.run([self.probs, self.final_state], feed)
p = probs[0]
# sample = int(np.random.choice(len(p), p=p))
sample = weighted_pick(p)
pred = chars[sample]
ret += pred
char = pred
return ret