In [17]:

    

import edward as ed
import tensorflow as tf
from edward.models import Bernoulli, Normal
from scipy.misc import imsave
from tensorflow.contrib import slim
from tensorflow.examples.tutorials.mnist import input_data
from keras.layers import Dense

M = 1 # batch size during training
d = 2 # latent variable dimension

# Probability model (subgraph)
z = Normal(loc=tf.zeros([M, d]), scale=tf.ones([M, d]))
h = Dense(256, activation='relu')(z)
x = Bernoulli(logits=Dense(28 * 28, activation=None)(h))

# Variational model (subgraph)
x_ph = tf.placeholder(tf.float32, [M, 28 * 28])
qh = Dense(256, activation='relu')(x_ph)
qz = Normal(loc=Dense(d, activation=None)(qh),
scale=Dense(d, activation='softplus')(qh))

# Bind p(x, z) and q(z | x) to the same TensorFlow placeholder for x.
mnist = input_data.read_data_sets("data/mnist", one_hot=True)
data = {x: x_ph}

inference = ed.KLqp({z: qz}, data)
optimizer = tf.train.RMSPropOptimizer(0.01, epsilon=1.0)
inference.initialize(optimizer=optimizer)

tf.initialize_all_variables().run()

n_epoch = 100
n_iter_per_epoch = 1000
for _ in range(n_epoch):
    for _ in range(n_iter_per_epoch):
        x_train, _ = mnist.train.next_batch(M)
        info_dict = inference.update(feed_dict={x_ph: x_train})

# Generate images.
imgs = x.value().eval()
for m in range(M):
    imsave("img/%d.png" % m, imgs[m].reshape(28, 28))


    

    




Extracting data/mnist\train-images-idx3-ubyte.gz
Extracting data/mnist\train-labels-idx1-ubyte.gz
Extracting data/mnist\t10k-images-idx3-ubyte.gz
Extracting data/mnist\t10k-labels-idx1-ubyte.gz






    




---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-17-c641b3737c0e> in <module>()
     25 data = {x: x_ph}
     26 
---> 27 inference = ed.KLqp({z: qz}, data)
     28 optimizer = tf.train.RMSPropOptimizer(0.01, epsilon=1.0)
     29 inference.initialize(optimizer=optimizer)

c:\programdata\miniconda3\envs\tensorflow\lib\site-packages\edward\inferences\klqp.py in __init__(self, *args, **kwargs)
     57   """
     58   def __init__(self, *args, **kwargs):
---> 59     super(KLqp, self).__init__(*args, **kwargs)
     60 
     61   def initialize(self, n_samples=1, kl_scaling=None, *args, **kwargs):

c:\programdata\miniconda3\envs\tensorflow\lib\site-packages\edward\inferences\variational_inference.py in __init__(self, *args, **kwargs)
     30   """
     31   def __init__(self, *args, **kwargs):
---> 32     super(VariationalInference, self).__init__(*args, **kwargs)
     33 
     34   def initialize(self, optimizer=None, var_list=None, use_prettytensor=False,

c:\programdata\miniconda3\envs\tensorflow\lib\site-packages\edward\inferences\inference.py in __init__(self, latent_vars, data)
     69     self.latent_vars = latent_vars
     70 
---> 71     check_data(data)
     72     self.data = {}
     73     for key, value in six.iteritems(data):

c:\programdata\miniconda3\envs\tensorflow\lib\site-packages\edward\util\random_variables.py in check_data(data)
     40         elif key.dtype != value.dtype:
     41           raise TypeError("Key-value pair in data does not have same "
---> 42                           "dtype: {}, {}".format(key.dtype, value.dtype))
     43       elif isinstance(value, (float, list, int, np.ndarray, np.number, str)):
     44         if not key.shape.is_compatible_with(np.shape(value)):

TypeError: Key-value pair in data does not have same dtype: <dtype: 'int32'>, <dtype: 'float32'>

In [ ]:

    




    

In [ ]:

    




    

In [16]:

    

x


    

    
Out[16]:





<ed.RandomVariable 'Bernoulli_5/' shape=(100, 784) dtype=int32>

In [15]:

    

x_ph


    

    
Out[15]:





<tf.Tensor 'Placeholder_1:0' shape=(100, 784) dtype=float32>

In [ ]:

    




    

In [ ]:

    




    

In [ ]: