AttributeError: 'module' object has no attribute 'pack'

[haozheji]

Training images using keras with tf backend.The error shows that tf (version==0.12.head) has no attribute "pack".

Traceback (most recent call last):
  File "binary_convnets.py", line 30, in <module>
    model.add(Flatten())
  File "/usr/local/lib/python2.7/dist-packages/keras/models.py", line 324, in add
    output_tensor = layer(self.outputs[0])
  File "/usr/local/lib/python2.7/dist-packages/keras/engine/topology.py", line 517, in __call__
    self.add_inbound_node(inbound_layers, node_indices, tensor_indices)
  File "/usr/local/lib/python2.7/dist-packages/keras/engine/topology.py", line 571, in add_inbound_node
    Node.create_node(self, inbound_layers, node_indices, tensor_indices)
  File "/usr/local/lib/python2.7/dist-packages/keras/engine/topology.py", line 155, in create_node
    output_tensors = to_list(outbound_layer.call(input_tensors[0], mask=input_masks[0]))
  File "/usr/local/lib/python2.7/dist-packages/keras/layers/core.py", line 440, in call
    return K.batch_flatten(x)
  File "/usr/local/lib/python2.7/dist-packages/keras/backend/tensorflow_backend.py", line 862, in batch_flatten
    x = tf.reshape(x, tf.pack([-1, prod(shape(x)[1:])]))
AttributeError: 'module' object has no attribute 'pack'

[zhaopku]

As far as I know, tf.pack has been renamed as tf.stack.

[aselle]

Yes, the keras needs to be updated since this is renamed to stack in tf 1.0.
@fchollet, what's the best way to handle this?

[hahnsangkim]

I have the same issue. I've upgraded to tensorflow version 1.0 and then tried to run a python program that uses keras like this
from keras.models import Sequential
from keras.layers import Flatten, Dense

Update: It was actually something wrong in my code. It turns out that tf.pack is OK. I upgraded keras = 1.2.0 which still remains the same.

[aselle]

I apologize, but I'm confused now. If the issue is resolved please close it. If not, please re-state exactly what is still wrong and what you would like fixed. Thanks so much!

[asif-shakeel]

I am having this problem as well. I pip installed tensorflow in anaconda with python3 environment. Can we revert anaconda to an earlier version of tensorflow to at least temporarily get around this?

[aselle]

@tawantinsuyu, unfortunately no, we won't downgrade the default version, because that would break users who already updated. To use an older version, an easy way is to use docker. You can also find the previous .whl from our web site and install that, but we don't recommend that. Good luck. Thanks!

[aselle]

Automatically closing due to lack of recent activity. Please update the issue when new information becomes available, and we will reopen the issue. Thanks!

[xiezhongzhao]

I have the same problem, my version of tensorflow is 1.0.1, I changed the 'pack' into 'stack', but it did't work. I hope you can help me @zhaopku

[navoj]

In page 105 of the book https://www.scribd.com/read/365181872/Building-Machine-Learning-Projects-with-TensorFlow# it also has a call to 'pack' that does not work with the latest version of Tensorflow.

[renelikestacos]

I had the same issue with TensorFlow. When i tried it in PyCharm it worked. Firstly I thought uninstall and install keras solved it. Still have the same error at Jupyter Notebook (Python 3.6)

[Somayeh12]

Any solution to this problem? I am having the same problem. tensorflow 1.13, python 3.6.
I replaced tf.pack with tf.stack, but the output of tf.stack is different than tf.pack.
Any solution?

[krishna0890227]

We can replace tf.pack() with tf.stack in Jupyter Notebook as well.

[ShauryaUppal-1Mg]

Any solution to this problem? I am having the same problem. tensorflow 1.13, python 3.6.
I replaced tf.pack with tf.stack, but the output of tf.stack is different than tf.pack.
Any solution?

How did you solve this issue? I am facing same issue.