Hi, I'am exploring your code recently, I've encountered a problem when I try to run python Omniglot.py.
System: Ubuntu 16.04
tensroflow version: 1.1.0-rc2
cuda 8.0
python: 2.7

The error information is as following, could you please help me to find out the solution?

python Omniglot.py
Traceback (most recent call last):
  File "Omniglot.py", line 112, in <module>
    omniglot()
  File "Omniglot.py", line 31, in omniglot
    output_var, output_var_flatten, params = memory_augmented_neural_network(input_ph, target_ph, batch_size=batch_size, nb_class=nb_class, memory_shape=memory_shape, controller_size=controller_size, input_size=input_size, nb_reads=nb_reads)
  File "/data/yangyu/workspace/MANN/MANN/Model.py", line 33, in memory_augmented_neural_network
    W_key = tf.get_variable('W_key', shape=shape,initializer=tf.random_uniform_initializer(-1*high, high))
  File "/data/yangyu/anaconda2/lib/python2.7/site-packages/tensorflow/python/ops/variable_scope.py", line 1065, in get_variable
    use_resource=use_resource, custom_getter=custom_getter)
  File "/data/yangyu/anaconda2/lib/python2.7/site-packages/tensorflow/python/ops/variable_scope.py", line 962, in get_variable
    use_resource=use_resource, custom_getter=custom_getter)
  File "/data/yangyu/anaconda2/lib/python2.7/site-packages/tensorflow/python/ops/variable_scope.py", line 367, in get_variable
    validate_shape=validate_shape, use_resource=use_resource)
  File "/data/yangyu/anaconda2/lib/python2.7/site-packages/tensorflow/python/ops/variable_scope.py", line 303, in _true_getter
    is_scalar = shape is not None and not shape

ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

mldl@mldlUB1604:/ub16_prj/NTM-One-Shot-TF$ python Omniglot.py
Shapes Recieved in Update: V, dim, val are ==> [64, 128] [64] [64]
Shapes Recieved in Body of Update: v, d2, chg are ==> [128] [] []
wlu_tm1 : [16, 4]
Shapes Recieved in Update: V, dim, val are ==> [16, 128, 40] [16] [16, 40]
Shapes Recieved in Body of Update: v, d2, chg are ==> [128, 40] [] [40]
Compiling the Model
Output, Target shapes: [16, 50, 5] [16, 50, 5]
Done
Training the model
Traceback (most recent call last):
File "Omniglot.py", line 112, in
omniglot()
File "Omniglot.py", line 85, in omniglot
for i, (batch_input, batch_output) in generator:
File "/home/mldl/ub16_prj/NTM-One-Shot-TF/MANN/Utils/Generator.py", line 36, in next
return (self.num_iter - 1), self.sample(self.nb_samples)
File "/home/mldl/ub16_prj/NTM-One-Shot-TF/MANN/Utils/Generator.py", line 41, in sample
sampled_character_folders = random.sample(self.character_folders, nb_samples)
File "/usr/lib/python2.7/random.py", line 323, in sample
raise ValueError("sample larger than population")
ValueError: sample larger than population
mldl@mldlUB1604:/ub16_prj/NTM-One-Shot-TF$ ll data/
total 18412
drwxrwxr-x 5 mldl mldl 4096 8月 4 18:24 ./
drwxrwxr-x 7 mldl mldl 4096 8月 4 18:16 ../
drwxr-xr-x 32 mldl mldl 4096 10月 21 2015 images_background/
-rw-rw-r-- 1 mldl mldl 1319022 8月 4 18:23 images_background_small1.zip
-rw-rw-r-- 1 mldl mldl 1580917 8月 4 18:23 images_background_small2.zip
-rw-rw-r-- 1 mldl mldl 9464212 8月 4 18:23 images_background.zip
-rw-rw-r-- 1 mldl mldl 6462886 8月 4 18:23 images_evaluation.zip
drwxrwxr-x 2 mldl mldl 4096 8月 3 17:26 omniglot/
drwxrwxr-x 2 mldl mldl 4096 8月 4 18:23 one-shot-classification/
mldl@mldlUB1604:~/ub16_prj/NTM-One-Shot-TF$ ll data/images_background

Hi
Nice work with the code
Just wanted to document how your code differs from the paper.
For example, you have an extra a_t formed using W_add, b_add and h_t
The paper simply adds the key k_t in place of a_t
Also, it's not clear how they actually got k_t.

Also I assumed they had used CNNs to get features before going to the LSTM controller, since it was an image input

When I run this code.The terminal appears following:
/usr/bin/python2.7 /home/cr/PycharmProjects/NTM-One-Shot-TF-master/Omniglot.py
Shapes Recieved in Update: V, dim, val are ==> [64, 128] [64] [64]
Shapes Recieved in Body of Update: v, d2, chg are ==> [128] [] []
wlu_tm1 : [16, 4]
Shapes Recieved in Update: V, dim, val are ==> [16, 128, 40] [16] [16, 40]
Shapes Recieved in Body of Update: v, d2, chg are ==> [128, 40] [] [40]
Compiling the Model
Output, Target shapes: [16, 50, 5] [16, 50, 5]
Done
Traceback (most recent call last):
File "/home/cr/PycharmProjects/NTM-One-Shot-TF-master/Omniglot.py", line 113, in
omniglot()
File "/home/cr/PycharmProjects/NTM-One-Shot-TF-master/Omniglot.py", line 87, in omniglot
for i, (batch_input, batch_output) in generator:
File "/home/cr/PycharmProjects/NTM-One-Shot-TF-master/MANN/Utils/Generator.py", line 36, in next
return (self.num_iter - 1), self.sample(self.nb_samples)
File "/home/cr/PycharmProjects/NTM-One-Shot-TF-master/MANN/Utils/Generator.py", line 56, in sample
for (filename, angle, shift) in zip(image_files, angles, shifts)], dtype=np.float32)
File "/home/cr/PycharmProjects/NTM-One-Shot-TF-master/MANN/Utils/Images.py", line 29, in load_transform
original = imread(image_path, flatten=True)
File "/usr/local/lib/python2.7/dist-packages/scipy-0.19.1-py2.7-linux-x86_64.egg/scipy/misc/pilutil.py", line 156, in imread
im = Image.open(name)
File "/usr/lib/python2.7/dist-packages/PIL/Image.py", line 1996, in open
fp = builtins.open(fp, "rb")
IOError: [Errno 21] Is a directory: './data/omniglot/images_background/Blackfoot_(Canadian_Aboriginal_Syllabics)/character11'
Training the model

Process finished with exit code 1

How to handle this problem？

Do you train the model use GPU?I tried to run your project to train it,but very slowly.So can you tell me your hardware and how much time you used to finish the whole training process?
thanks

Is this a special tech to initialize weights?

Hi, I get a little confused about the memory module.
The shape of the memory in your code is Batchsize x Memory_size x Slotsize(or vector_dim). So actually when the training process is finished, there are Batchsize memories generated, not just one memory.
That means, once the value of batchsize is changed, the memory changed, which seems make no sense.
During test process, the value of Batchsize should be the same with that in the training process. So what if I got a batch that contains N samples (N<Batchsize), which memory should each sample use?(when N = Batchsize, each sample use one memory).

The key problem, I think, is why there should be more than 1 memory?

Thank you.

Hi, thanks for your code, I use these to do sequence prediction task, I find there is a problem that the vector in different memory slots tend to be same. So do you know how to fix this problem.
And for MANN, I find w_write(t) = w_read(t-1) + w_lt(t), why use t-1 timestep read weight rather than t timestep read weight, I think w_read(t) is more related to w_write(t), is there some consideration ?
thanks

Hi,

Thanks for creating this, it works well. However, I am having trouble understanding the output from the testing script. Could you explain what I am looking at?

Warm Regards

Describe the bug
A clear and concise description of what the bug is.

To Reproduce
Steps to reproduce the behavior:

1. Go to '...'
2. Click on '....'
3. Scroll down to '....'
4. See error

Expected behavior
A clear and concise description of what you expected to happen.

Screenshots
If applicable, add screenshots to help explain your problem.

Desktop (please complete the following information):

• OS: [e.g. iOS]
• Browser [e.g. chrome, safari]
• Version [e.g. 22]

Smartphone (please complete the following information):

• Device: [e.g. iPhone6]
• OS: [e.g. iOS8.1]
• Browser [e.g. stock browser, safari]
• Version [e.g. 22]

Additional context
Add any other context about the problem here.

I receive such error while running

Traceback (most recent call last):
File "/root/tensorflow/anomaly-detector/NTM-One-Shot-TF/Omniglot.py", line 112, in
omniglot()
File "/root/tensorflow/anomaly-detector/NTM-One-Shot-TF/Omniglot.py", line 85, in omniglot
for i, (batch_input, batch_output) in generator:
File "/root/tensorflow/anomaly-detector/NTM-One-Shot-TF/MANN/Utils/Generator.py", line 36, in next
return (self.num_iter - 1), self.sample(self.nb_samples)
File "/root/tensorflow/anomaly-detector/NTM-One-Shot-TF/MANN/Utils/Generator.py", line 56, in sample
for (filename, angle, shift) in zip(image_files, angles, shifts)], dtype=np.float32)
File "/root/tensorflow/anomaly-detector/NTM-One-Shot-TF/MANN/Utils/Images.py", line 29, in load_transform
original = imread(image_path, flatten=True)
File "/usr/local/lib/python2.7/dist-packages/scipy/misc/pilutil.py", line 156, in imread
im = Image.open(name)
File "/usr/lib/python2.7/dist-packages/PIL/Image.py", line 2251, in open
fp = builtins.open(fp, "rb")
IOError: [Errno 21] Is a directory: './data/omniglot/images_background/Japanese_(katakana)/character45'

Last two dirs in path may vary