-
Notifications
You must be signed in to change notification settings - Fork 0
/
train_model.py
44 lines (35 loc) · 1.54 KB
/
train_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
import tensorflow as tf
import numpy as np
# from tensorflow.examples.tutorials.mnist import input_data
import emnist_input_data as input_data #input_data
import tensorflow_autoencoder as model
# Paths
model_save_path = './trained_models_emnist/autoencoder_model' # './trained_models/autoencoder_model'
mnist_data_path = './EMNIST_data/' # './MNIST_data/'
# Parameters
mnist_width = 28
corruption_level = 0.3
n_visible = mnist_width * mnist_width
code_layer_size = 50
layers = [n_visible, 500, code_layer_size]
# Hyperparameters
alpha = 0.01
# Load MNIST data
mnist = input_data.read_data_sets(mnist_data_path, one_hot=True)
trX, trY, teX, teY = mnist.train.images, mnist.train.labels, mnist.test.images, mnist.test.labels
saver = tf.train.Saver()
# Train on training data, every epoch evaluate with same evaluation data
def train_model(sess):
for i in range(100):
for start, end in zip(range(0, len(trX), 128), range(128, len(trX), 128)):
input_ = trX[start:end]
noise_mask = np.random.binomial(1, 1 - corruption_level, input_.shape)
sess.run(model.train_op, feed_dict={model.original: input_, model.corrupted: noise_mask * input_})
noise_mask = np.random.binomial(1, 1 - corruption_level, teX.shape)
print('Cost: ', sess.run(model.cost, feed_dict={model.original: teX, model.corrupted: noise_mask * teX}))
saver.save(sess, model_save_path)
# Run training / viewing
with tf.Session() as sess:
tf.initialize_all_variables().run()
# saver.restore(sess, model_save_path)
train_model(sess)