Suppose you want to restore the trained/saved model and test new inputs on. The way the model is written makes it inflexible to the number of inputs passed in, you must always pass in a number of examples equal to the batchsize the model was trained on. You can't for example run the model on just one test image, which would be nice.

Hi,
Why do not you add bias term for the conv_transpose which is in the line 51 in ops.py?

convt = tf.nn.conv2d_transpose(x, w, output_shape=outputShape, strides=[1,2,2,1])

I think it should be:
convt = tf.nn.conv2d_transpose(x, w, output_shape=outputShape, strides=[1,2,2,1]) + b

I'm having trouble running your code and am getting the following error.
Any thoughts of what might be wrong?
Thanks.

$python main.py
....
Extracting MNIST_data/train-images-idx3-ubyte.gz
Traceback (most recent call last):
  File "main.py", line 79, in <module>
    model = LatentAttention()
  File "main.py", line 12, in __init__
    self.mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
  File "/home/medgar/models/autoencoders/variational-autoencoder-kvfrans/input_data.py", line 148, in read_data_sets
    train_images = extract_images(local_file)
  File "/home/medgar/models/autoencoders/variational-autoencoder-kvfrans/input_data.py", line 38, in extract_images
    buf = bytestream.read(rows * cols * num_images)
  File "/usr/lib/python2.7/gzip.py", line 275, in read
    chunk = self.extrabuf[offset: offset + size]
TypeError: only integer scalar arrays can be converted to a scalar index

Let me know if you want the changes pushed.

Hi,when i run this code ,i got error!

/root/miniconda3/bin/python /root/PycharmProjects/variational-autoencoder/main.py
Traceback (most recent call last):
File "/root/PycharmProjects/variational-autoencoder/main.py", line 6, in
from scipy.misc import imsave as ims
File "/root/miniconda3/lib/python3.6/site-packages/scipy/misc/init.py", line 68, in
from scipy.interpolate._pade import pade as _pade
File "/root/miniconda3/lib/python3.6/site-packages/scipy/interpolate/init.py", line 187, in
from .ndgriddata import *
File "/root/miniconda3/lib/python3.6/site-packages/scipy/interpolate/ndgriddata.py", line 10, in
from .interpnd import LinearNDInterpolator, NDInterpolatorBase,
ImportError: cannot import name 'LinearNDInterpolator'

How can i fix this?

Anyone else have this problem?

epoch 427: genloss 74090.617188 latloss 126.223068
epoch 428: genloss 77959.625000 latloss 83.587059
epoch 429: genloss 76822.828125 latloss 191.646500
epoch 430: genloss 166910.765625 latloss 58514.851562
epoch 431: genloss 131478.968750 latloss 59600.101562
epoch 432: genloss nan latloss nan
epoch 433: genloss nan latloss nan
epoch 434: genloss nan latloss nan
epoch 435: genloss nan latloss nan
epoch 436: genloss nan latloss nan
epoch 437: genloss nan latloss nan

The images decoded then go "dark" right after that.