-
Notifications
You must be signed in to change notification settings - Fork 8
/
lstm.py
278 lines (219 loc) · 9.88 KB
/
lstm.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
"""
This is a modified version of https://github.com/tensorflow/models/tree/master/tutorials/rnn/ptb
author: Sebastian Gehrmann
This file trains a simple language model and extracts the states from the training set.
The extracted states can be used for LSTMVis (https://github.com/HendrikStrobelt/LSTMVis)
For a description how to use the code, please look at
https://github.com/sebastianGehrmann/tensorflow-statereader
"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import time
import h5py
import numpy as np
import tensorflow as tf
import reader
flags = tf.flags
logging = tf.logging
flags.DEFINE_string("data_path", None,
"Where the training/test data is stored.")
flags.DEFINE_string("save_path", None,
"Model output directory.")
flags.DEFINE_string("load_path", None,
"Checkpoint directory.")
flags.DEFINE_bool("use_fp16", False,
"Train using 16-bit floats instead of 32bit floats")
# Hyperparameters
flags.DEFINE_float("init_scale", 0.1, "TBD")
flags.DEFINE_float("learning_rate", 1.0, "initial learning rate")
flags.DEFINE_integer("max_grad_norm", 5, "Max norm of the gradient")
flags.DEFINE_integer("num_layers", 2, "Layers of the LSTM")
flags.DEFINE_integer("num_steps", 30, "Steps to unroll the LSTM")
flags.DEFINE_integer("hidden_size", 200, "Cell states")
flags.DEFINE_integer("max_epoch", 4, "How many epochs with max LR")
flags.DEFINE_integer("max_max_epoch", 10, "How long to train for")
flags.DEFINE_float("dropout", 1.0, "Dropout")
flags.DEFINE_float("lr_decay", 0.5, "Multiplier of LR")
flags.DEFINE_integer("batch_size", 20, "Batchsize")
flags.DEFINE_integer("vocab_size", 6500, "Size of Vocabulary")
FLAGS = flags.FLAGS
def data_type():
return tf.float16 if FLAGS.use_fp16 else tf.float32
class PTBInput(object):
"""The input data."""
def __init__(self, data, name=None):
self.batch_size = batch_size = FLAGS.batch_size
self.num_steps = num_steps = FLAGS.num_steps
self.epoch_size = ((len(data) // batch_size) - 1) // num_steps
self.input_data, self.targets = reader.ptb_producer(
data, batch_size, num_steps, name=name)
class PTBModel(object):
"""The PTB model."""
def __init__(self, is_training, input_):
self._input = input_
batch_size = input_.batch_size
num_steps = input_.num_steps
self._size = FLAGS.hidden_size
vocab_size = FLAGS.vocab_size
lstm_cell = tf.nn.rnn_cell.BasicLSTMCell(
self._size, forget_bias=0.0, state_is_tuple=True)
if is_training and FLAGS.dropout < 1:
lstm_cell = tf.nn.rnn_cell.DropoutWrapper(
lstm_cell, output_keep_prob=FLAGS.dropout)
cell = tf.nn.rnn_cell.MultiRNNCell(
[lstm_cell] * FLAGS.num_layers, state_is_tuple=True)
self._initial_state = cell.zero_state(batch_size, data_type())
with tf.device("/cpu:0"):
embedding = tf.get_variable(
"embedding", [vocab_size, self._size], dtype=data_type())
inputs = tf.nn.embedding_lookup(embedding, input_.input_data)
if is_training and FLAGS.dropout < 1:
inputs = tf.nn.dropout(inputs, FLAGS.dropout)
outputs = []
self._states = []
state = self._initial_state
with tf.variable_scope("RNN"):
for time_step in range(num_steps):
if time_step > 0: tf.get_variable_scope().reuse_variables()
(cell_output, state) = cell(inputs[:, time_step, :], state)
outputs.append(cell_output)
self._states.append(state)
output = tf.reshape(tf.concat_v2(outputs, 1), [-1, self._size])
softmax_w = tf.get_variable(
"softmax_w", [self._size, vocab_size], dtype=data_type())
# print(softmax_w, "SOFTMAX W")
softmax_b = tf.get_variable("softmax_b", [vocab_size], dtype=data_type())
# print(softmax_b, "SOFTMAX B")
logits = tf.matmul(output, softmax_w) + softmax_b # 400 x 10k
# print(logits, "LOGITS")
# print(input_.targets, "TARGET")
loss = tf.nn.seq2seq.sequence_loss_by_example(
[logits],
[tf.reshape(input_.targets, [-1])],
[tf.ones([batch_size * num_steps], dtype=data_type())])
self._cost = cost = tf.reduce_sum(loss) / batch_size
self._final_state = state
if not is_training:
return
self._lr = tf.Variable(0.0, trainable=False)
tvars = tf.trainable_variables()
grads, _ = tf.clip_by_global_norm(tf.gradients(cost, tvars),
FLAGS.max_grad_norm)
optimizer = tf.train.GradientDescentOptimizer(self._lr)
self._train_op = optimizer.apply_gradients(
zip(grads, tvars),
global_step=tf.contrib.framework.get_or_create_global_step())
self._new_lr = tf.placeholder(
tf.float32, shape=[], name="new_learning_rate")
self._lr_update = tf.assign(self._lr, self._new_lr)
def assign_lr(self, session, lr_value):
session.run(self._lr_update, feed_dict={self._new_lr: lr_value})
@property
def input(self):
return self._input
@property
def initial_state(self):
return self._initial_state
@property
def cost(self):
return self._cost
@property
def states(self):
return self._states
@property
def final_state(self):
return self._final_state
@property
def lr(self):
return self._lr
@property
def train_op(self):
return self._train_op
@property
def size(self):
return self._size
def run_epoch(session, model, eval_op=None, verbose=False):
"""Runs the model on the given data."""
start_time = time.time()
costs = 0.0
iters = 0
state = session.run(model.initial_state)
fetches = {
"cost": model.cost,
"final_state": model.final_state,
}
if eval_op is not None:
fetches["eval_op"] = eval_op
all_states = []
for step in range(model.input.epoch_size):
feed_dict = {}
for i, (c, h) in enumerate(model.initial_state):
feed_dict[c] = state[i].c
feed_dict[h] = state[i].h
vals, stat = session.run([fetches, model.states], feed_dict)
curr_states = np.array([s[0][0] for s in stat])
if len(all_states) == 0:
all_states = curr_states
else:
all_states = np.vstack((all_states, curr_states))
cost = vals["cost"]
state = vals["final_state"]
costs += cost
iters += model.input.num_steps
if verbose and step % (model.input.epoch_size // 10) == 10:
print("%.3f perplexity: %.3f speed: %.0f wps" %
(step * 1.0 / model.input.epoch_size, np.exp(costs / iters),
iters * model.input.batch_size / (time.time() - start_time)))
return np.exp(costs / iters), all_states
def main(_):
if not FLAGS.data_path:
raise ValueError("Must set --data_path to PTB data directory")
raw_data = reader.ptb_raw_data(FLAGS.data_path, True)
train_data, valid_data, _ = raw_data
with tf.Graph().as_default():
initializer = tf.random_uniform_initializer(-FLAGS.init_scale,
FLAGS.init_scale)
with tf.name_scope("Train"):
train_input = PTBInput(data=train_data, name="TrainInput")
with tf.variable_scope("Model", reuse=None, initializer=initializer):
m = PTBModel(is_training=True, input_=train_input)
tf.summary.scalar("Training Loss", m.cost)
tf.summary.scalar("Learning Rate", m.lr)
with tf.name_scope("Train_states"):
train_input = PTBInput(data=train_data, name="TrainInput")
with tf.variable_scope("Model", reuse=True, initializer=initializer):
mstates = PTBModel(is_training=False, input_=train_input)
tf.summary.scalar("Training Loss", mstates.cost)
with tf.name_scope("Valid"):
valid_input = PTBInput(data=valid_data, name="ValidInput")
with tf.variable_scope("Model", reuse=True, initializer=initializer):
mvalid = PTBModel(is_training=False, input_=valid_input)
tf.summary.scalar("Validation Loss", mvalid.cost)
sv = tf.train.Supervisor(logdir=FLAGS.save_path)
with sv.managed_session() as session:
if FLAGS.load_path:
sv.saver.restore(session, tf.train.latest_checkpoint(FLAGS.load_path))
else:
for i in range(FLAGS.max_max_epoch):
lr_decay = FLAGS.lr_decay ** max(i + 1 - FLAGS.max_epoch, 0.0)
m.assign_lr(session, FLAGS.learning_rate * lr_decay)
print("Epoch: %d Learning rate: %.3f" % (i + 1, session.run(m.lr)))
train_perplexity, stat = run_epoch(session, m, eval_op=m.train_op,
verbose=True)
print(stat.shape)
print("Epoch: %d Train Perplexity: %.3f" % (i + 1, train_perplexity))
valid_perplexity, stat = run_epoch(session, mvalid)
print("Epoch: %d Valid Perplexity: %.3f" % (i + 1, valid_perplexity))
# run and store the states on training set
train_perplexity, stat = run_epoch(session, mstates, eval_op=m.train_op,
verbose=True)
f = h5py.File("states.h5", "w")
stat = np.reshape(stat, (-1, mstates.size))
f["states1"] = stat
f.close()
if FLAGS.save_path:
print("Saving model to %s." % FLAGS.save_path)
sv.saver.save(session, FLAGS.save_path, global_step=sv.global_step)
if __name__ == "__main__":
tf.app.run()