-
Notifications
You must be signed in to change notification settings - Fork 2
/
seq2seq_model.py
384 lines (318 loc) · 15.4 KB
/
seq2seq_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
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
import tensorflow as tf
from tensorflow.contrib import rnn
from tensorflow.python.ops import math_ops
from tensorflow.python.ops import array_ops
from tensorflow.contrib.rnn.python.ops import rnn
from tensorflow.contrib.rnn.python.ops import core_rnn_cell
from six.moves import xrange
import numpy as np
import random
import copy
import data_utils
import seq2seq
class Seq2seq():
def __init__(self,
vocab_size,
buckets,
size,
num_layers,
batch_size,
mode):
self.vocab_size = vocab_size
self.buckets =buckets
# units of rnn cell
self.size = size
# dimension of words
self.num_layers = num_layers
self.batch_size = batch_size
self.learning_rate = tf.Variable(0.5, trainable=False)
self.mode = mode
self.dummy_reply = ["what ?", "yeah .", "you are welcome ! ! ! !"]
# learning rate decay
self.learning_rate_decay = self.learning_rate.assign(self.learning_rate * 0.99)
# input for Reinforcement part
self.loop_or_not = tf.placeholder(tf.bool)
self.reward = tf.placeholder(tf.float32, [None])
batch_reward = tf.stop_gradient(self.reward)
self.RL_index = [None for _ in self.buckets]
# projection function
w_t = tf.get_variable('proj_w', [self.vocab_size, self.size])
w = tf.transpose(w_t)
b = tf.get_variable('proj_b', [self.vocab_size])
output_projection = (w, b)
def sample_loss(labels, inputs):
labels = tf.reshape(labels, [-1, 1])
local_w_t = tf.cast(w_t, tf.float32)
local_b = tf.cast(b, tf.float32)
local_inputs = tf.cast(inputs, tf.float32)
return tf.cast(tf.nn.sampled_softmax_loss(weights = local_w_t,
biases = local_b,
inputs = local_inputs,
labels = labels,
num_sampled = 512,
num_classes = self.vocab_size),
dtype = tf.float32)
softmax_loss_function = sample_loss
#FIXME add RL function
def seq2seq_multi(encoder_inputs, decoder_inputs, mode):
embedding = tf.get_variable("embedding", [self.vocab_size, self.size])
loop_function_RL = None
if mode == 'MLE':
feed_previous = False
elif mode == 'TEST':
feed_previous = True
# need loop_function
elif mode == 'RL':
feed_previous = True
def loop_function_RL(prev, i):
prev = tf.matmul(prev, output_projection[0]) + output_projection[1]
prev_index = tf.multinomial(tf.log(tf.nn.softmax(prev)), 1)
if i == 1:
for index, RL in enumerate(self.RL_index):
if RL is None:
self.RL_index[index] = prev_index
self.index = index
break
else:
self.RL_index[self.index] = tf.concat([self.RL_index[self.index], prev_index], axis = 1)
prev_index = tf.reshape(prev_index, [-1])
# decide which to be the next time step input
sample = tf.nn.embedding_lookup(embedding, prev_index)
from_decoder = tf.nn.embedding_lookup(embedding, decoder_inputs[i])
return tf.where(self.loop_or_not, sample, from_decoder)
return seq2seq.embedding_attention_seq2seq(
encoder_inputs,
decoder_inputs,
cell,
num_encoder_symbols = self.vocab_size,
num_decoder_symbols = self.vocab_size,
embedding_size = self.size,
output_projection = output_projection,
feed_previous = feed_previous,
dtype = tf.float32,
embedding = embedding,
loop = loop_function_RL)
# inputs
self.encoder_inputs = []
self.decoder_inputs = []
self.target_weights = []
for i in xrange(buckets[-1][0]):
self.encoder_inputs.append(tf.placeholder(tf.int32, shape = [None],
name = 'encoder{0}'.format(i)))
for i in xrange(buckets[-1][1] + 1):
self.decoder_inputs.append(tf.placeholder(tf.int32, shape = [None],
name = 'decoder{0}'.format(i)))
self.target_weights.append(tf.placeholder(tf.float32, shape = [None],
name = 'weight{0}'.format(i)))
targets = [self.decoder_inputs[i + 1] for i in xrange(len(self.decoder_inputs) - 1)]
def single_cell():
return tf.contrib.rnn.GRUCell(self.size)
cell = single_cell()
if self.num_layers > 1:
cell = tf.contrib.rnn.MultiRNNCell([single_cell() for _ in range(self.num_layers)])
if self.mode == 'MLE':
self.outputs, self.losses = seq2seq.model_with_buckets(
self.encoder_inputs, self.decoder_inputs, targets,
self.target_weights, self.buckets, lambda x, y: seq2seq_multi(x, y, self.mode),
softmax_loss_function = softmax_loss_function)
for b in xrange(len(self.buckets)):
self.outputs[b] = [tf.matmul(output, output_projection[0]) + output_projection[1]
for output in self.outputs[b]]
self.update = []
optimizer = tf.train.GradientDescentOptimizer(self.learning_rate)
for b in xrange(len(self.buckets)):
gradients = tf.gradients(self.losses[b], tf.trainable_variables())
clipped_gradients, _ = tf.clip_by_global_norm(gradients, 5.0)
self.update.append(optimizer.apply_gradients(zip(clipped_gradients, tf.trainable_variables())))
elif self.mode == 'TEST':
self.outputs, self.losses = seq2seq.model_with_buckets(
self.encoder_inputs, self.decoder_inputs, targets,
self.target_weights, self.buckets, lambda x, y: seq2seq_multi(x, y, self.mode),
softmax_loss_function = softmax_loss_function)
for b in xrange(len(self.buckets)):
self.outputs[b] = [tf.matmul(output, output_projection[0]) + output_projection[1]
for output in self.outputs[b]]
elif self.mode == 'RL':
self.outputs, self.losses = seq2seq.model_with_buckets(
self.encoder_inputs, self.decoder_inputs, targets,
self.target_weights, self.buckets, lambda x, y: seq2seq_multi(x, y, self.mode),
softmax_loss_function = softmax_loss_function, per_example_loss = True)
for b in xrange(len(self.buckets)):
self.outputs[b] = [tf.matmul(output, output_projection[0]) + output_projection[1]
for output in self.outputs[b]]
for i, b in enumerate(self.outputs):
prev_index = tf.multinomial(tf.log(tf.nn.softmax(b[self.buckets[i][1] - 1])), 1)
self.RL_index[i] = tf.concat([self.RL_index[i], prev_index], axis = 1)
self.update = []
optimizer = tf.train.GradientDescentOptimizer(0.01)
#optimizer = tf.train.GradientDescentOptimizer(self.learning_rate)
for b in xrange(len(self.buckets)):
scaled_loss = tf.multiply(self.losses[b], batch_reward)
self.losses[b] = tf.reduce_mean(scaled_loss)
gradients = tf.gradients(self.losses[b], tf.trainable_variables())
clipped_gradients, _ = tf.clip_by_global_norm(gradients, 5.0)
self.update.append(optimizer.apply_gradients(zip(clipped_gradients, tf.trainable_variables())))
# specify saver
self.saver = tf.train.Saver(max_to_keep = 2)
# token_vector: list [batch_size, vocab_size] of length max_length
# return: list of length batch_size, each contain the list of the decoded sentence
def token2word(self, token_vector):
sentence_list = [[] for _ in xrange(self.batch_size)]
for logit in token_vector:
outputs = np.argmax(logit, axis = 1)
for i in xrange(self.batch_size):
sentence_list[i].append(outputs[i])
for i in xrange(self.batch_size):
if data_utils.EOS_ID in sentence_list[i]:
sentence_list[i] = sentence_list[i][:sentence_list[i].index(data_utils.EOS_ID)]
if data_utils.PAD_ID in sentence_list[i]:
sentence_list[i] = sentence_list[i][:sentence_list[i].index(data_utils.PAD_ID)]
sentence_temp = [tf.compat.as_str(self.vocab_list[output]) for output in sentence_list[i]]
sentence_list[i] = " ".join(word for word in sentence_temp)
return sentence_list
# decoding function for reinforcement learning sampling output
def token2word_RL(self, token_vector):
sentence_list = [[] for _ in xrange(self.batch_size)]
for i in xrange(self.batch_size):
token_list = list(token_vector[i])
if data_utils.EOS_ID in token_list:
sentence_list[i] = token_list[:token_list.index(data_utils.EOS_ID)]
sentence_temp = [tf.compat.as_str(self.vocab_list[output]) for output in sentence_list[i]]
sentence_list[i] = " ".join(word for word in sentence_temp)
return sentence_list
# calculate logP(b|a)
# a and b are both list of token ids. ex:[1,2,3,4,5...]
def prob(self, a, b, X, bucket_id):
# define softmax
def softmax(x):
e_x = np.exp(x)
return e_x / e_x.sum()
# function X, not trainable, batch = 1
temp = self.batch_size
self.batch_size = 1
encoder_input, decoder_input, weight = self.get_batch({bucket_id: [(a, b)]}, bucket_id)
self.batch_size = temp
outputs = X(encoder_input, decoder_input, weight, bucket_id)
r = 0.0
for logit, i in zip(outputs, b):
r += np.log10(softmax(logit[0])[i])
return r
# this function is specify for training of Reinforcement Learning case
def RL_readmap(self, map_path):
self.vocab_dict, self.vocab_list = data_utils.read_map(map_path)
def run(self, sess, encoder_inputs, decoder_inputs, target_weights,
bucket_id, forward_only = False, X = None, Y = None):
encoder_size, decoder_size = self.buckets[bucket_id]
input_feed = {}
for l in xrange(encoder_size):
input_feed[self.encoder_inputs[l].name] = encoder_inputs[l]
for l in xrange(decoder_size):
input_feed[self.decoder_inputs[l].name] = decoder_inputs[l]
input_feed[self.target_weights[l].name] = target_weights[l]
last_target = self.decoder_inputs[decoder_size].name
input_feed[last_target] = np.zeros([self.batch_size], dtype = np.int32)
if self.mode == 'MLE':
if forward_only:
output_feed = [self.losses[bucket_id], self.outputs[bucket_id]]
outputs = sess.run(output_feed, input_feed)
return outputs[0], outputs[1]
else:
output_feed = [self.losses[bucket_id], self.update[bucket_id]]
outputs = sess.run(output_feed, input_feed)
return outputs[0], outputs[1]
elif self.mode == 'TEST':
output_feed = [self.outputs[bucket_id]]
outputs = sess.run(output_feed, input_feed)
return outputs[0]
elif self.mode == 'RL':
# check mode: sample or from decoder input
# True for sample and False for from decoder input
input_feed[self.loop_or_not] = True
# step 1: get seq2seq sampled output
output_feed = [self.RL_index[bucket_id]]
outputs = sess.run(output_feed, input_feed)
# sentence_rl is a batch sized list of sampled decoded natural sentence
#sentence_rl = self.token2word_RL(outputs[0])
#for a in sentence_rl:
# print(a)
# step 2: get rewards according to some rules
reward = np.ones((self.batch_size), dtype = np.float32)
new_data = []
for i in xrange(self.batch_size):
token_ids = list(outputs[0][i])
if data_utils.EOS_ID in token_ids:
token_ids = token_ids[:token_ids.index(data_utils.EOS_ID)]
new_data.append(([], token_ids + [data_utils.EOS_ID]))
#print(token_ids)
# in this case, X is language model score
# reward 1: ease of answering
'''
temp_reward = [self.prob(token_ids, data_utils.convert_to_token(tf.compat.as_bytes(sen), self.vocab_dict,
False) + [data_utils.EOS_ID], X, bucket_id)/float(len(sen)) for sen in self.dummy_reply]
r1 = -np.mean(temp_reward)
'''
# reward 2: semantic coherence
r_input = list(reversed([o[i] for o in encoder_inputs]))
if data_utils.PAD_ID in r_input:
r_input = r_input[:r_input.index(data_utils.PAD_ID)]
r2 = self.prob(r_input, token_ids, X, bucket_id) / float(len(token_ids)) if len(token_ids) != 0 else 0
# reward 3: sentiment analysis score
word_token = [self.vocab_list[token].decode('utf-8') for token in token_ids]
r3 = Y(word_token, np.array([len(token_ids)], dtype = np.int32))
'''
print('r1: %s' % r1)
print('r2: %s' % r2)
print('r3: %s' % r3)
'''
#reward[i] = 0.7 * r1 + 0.7 * r2 + r3
reward[i] = 0 * r2 + r3
#print(reward)
# advantage
reward = reward - np.mean(reward)
_, decoder_inputs, target_weights = self.get_batch({bucket_id: new_data}, bucket_id, order = True)
# step 3: update seq2seq model
for l in xrange(decoder_size):
input_feed[self.decoder_inputs[l].name] = decoder_inputs[l]
input_feed[self.target_weights[l].name] = target_weights[l]
input_feed[self.reward] = reward
input_feed[self.loop_or_not] = False
output_feed = [self.losses[bucket_id], self.update[bucket_id]]
#output_feed = [self.losses[bucket_id]]
outputs = sess.run(output_feed, input_feed)
return outputs[0]
def get_batch(self, data, bucket_id, rand = True, order = False):
# data should be [whole_data_length x (source, target)]
# decoder_input should contain "GO" symbol and target should contain "EOS" symbol
encoder_size, decoder_size = self.buckets[bucket_id]
encoder_inputs, decoder_inputs = [], []
encoder_input, decoder_input = random.choice(data[bucket_id])
c = 0
for i in xrange(self.batch_size):
if rand:
encoder_input, decoder_input = random.choice(data[bucket_id])
if order:
encoder_input, decoder_input = data[bucket_id][i]
c += 1
encoder_pad = [data_utils.PAD_ID] * (encoder_size - len(encoder_input))
encoder_inputs.append(list(reversed(encoder_input + encoder_pad)))
decoder_pad = [data_utils.PAD_ID] * (decoder_size - len(decoder_input) - 1)
decoder_inputs.append([data_utils.GO_ID] + decoder_input + decoder_pad)
batch_encoder_inputs, batch_decoder_inputs, batch_weights = [], [], []
for length_idx in xrange(encoder_size):
batch_encoder_inputs.append(np.array([encoder_inputs[batch_idx][length_idx]
for batch_idx in xrange(self.batch_size)], dtype = np.int32))
for length_idx in xrange(decoder_size):
batch_decoder_inputs.append(np.array([decoder_inputs[batch_idx][length_idx]
for batch_idx in xrange(self.batch_size)], dtype = np.int32))
batch_weight = np.ones(self.batch_size, dtype = np.float32)
for batch_idx in xrange(self.batch_size):
# We set weight to 0 if the corresponding target is a PAD symbol.
# The corresponding target is decoder_input shifted by 1 forward.
if length_idx < decoder_size - 1:
target = decoder_inputs[batch_idx][length_idx + 1]
if length_idx == decoder_size - 1 or target == data_utils.PAD_ID:
batch_weight[batch_idx] = 0.0
batch_weights.append(batch_weight)
return batch_encoder_inputs, batch_decoder_inputs, batch_weights
if __name__ == '__main__':
test = Seq2seq(50, 100, 200, 300, 1, 128)