Some problems in running attention.py

[ghost]

Hi, I cloned your project and try to use it. Your code is fabulous.
However when I run python attention.py
there's always an error appear:

Caused by op 'pool1', defined at:
File ".\attention.py", line 37, in
ctx = tflib.network.im2latex_cnn(X,NUM_FEATS_START,True)
File "D:\THU\challengeCup\im2latex-tensorflow-master\tflib\network.py", line 77, in im2latex_cnn
X = tflib.ops.max_pool('pool1', X, k=2, s=2)
File "D:\THU\challengeCup\im2latex-tensorflow-master\tflib\ops.py", line 239, in max_pool
padding='SAME', name=name, data_format='NCHW')
File "C:\Users\Ellen\Anaconda3\lib\site-packages\tensorflow\python\ops\nn_ops.py", line 1617, in max_pool
name=name)
File "C:\Users\Ellen\Anaconda3\lib\site-packages\tensorflow\python\ops\gen_nn_ops.py", line 1598, in _max_pool
data_format=data_format, name=name)
File "C:\Users\Ellen\Anaconda3\lib\site-packages\tensorflow\python\framework\op_def_library.py", line 759, in apply_op
op_def=op_def)
File "C:\Users\Ellen\Anaconda3\lib\site-packages\tensorflow\python\framework\ops.py", line 2240, in create_op
original_op=self._default_original_op, op_def=op_def)
File "C:\Users\Ellen\Anaconda3\lib\site-packages\tensorflow\python\framework\ops.py", line 1128, in init
self._traceback = _extract_stack()

InvalidArgumentError (see above for traceback): Default MaxPoolingOp only supports NHWC.
[[Node: pool1 = MaxPoolT=DT_FLOAT, data_format="NCHW", ksize=[1, 1, 2, 2], padding="SAME", strides=[1, 1, 2, 2
], _device="/job:localhost/replica:0/task:0/cpu:0"]]

Does this mean I'm not preprocess the data properly? Or is there any possible reasons for this? How can I correct it?

[ritheshkumar95]

Hello, Thanks for your question! If you notice this issue thread in tensorflow's git ( tensorflow/tensorflow#2660), they seem to point out that NCHW format is supported only in GPU and not CPU. So you're finding this error because I work on NCHW format (comes more naturally for me from my theano roots) and you seem to run your code on CPU based on your stack trace. I'm sorry but I cannot help you more, but point you towards rewriting some of the functions to make it NHCW format. N-> No. of images / batch size H-> Height C-> Channels W-> Width. If I remember correctly, you'd have to change the data format in pool-2d op and conv2d op to NHCW format, and also preprocess the data such that you follow this order. Shouldn't be very hard to do. Hope this helps. Good luck! Best, Rithesh

…

On Fri, Feb 24, 2017, 1:20 PM JoySiTHU ***@***.***> wrote: Hi, I cloned your project and try to use it. Your code is fabulous. However when I run python attention.py there's always an error appear: Caused by op 'pool1', defined at: File ".\attention.py", line 37, in ctx = tflib.network.im2latex_cnn(X,NUM_FEATS_START,True) File "D:\THU\challengeCup\im2latex-tensorflow-master\tflib\network.py", line 77, in im2latex_cnn X = tflib.ops.max_pool('pool1', X, k=2, s=2) File "D:\THU\challengeCup\im2latex-tensorflow-master\tflib\ops.py", line 239, in max_pool padding='SAME', name=name, data_format='NCHW') File "C:\Users\Ellen\Anaconda3\lib\site-packages\tensorflow\python\ops\nn_ops.py", line 1617, in max_pool name=name) File "C:\Users\Ellen\Anaconda3\lib\site-packages\tensorflow\python\ops\gen_nn_ops.py", line 1598, in _max_pool data_format=data_format, name=name) File "C:\Users\Ellen\Anaconda3\lib\site-packages\tensorflow\python\framework\op_def_library.py", line 759, in apply_op op_def=op_def) File "C:\Users\Ellen\Anaconda3\lib\site-packages\tensorflow\python\framework\ops.py", line 2240, in create_op original_op=self._default_original_op, op_def=op_def) File "C:\Users\Ellen\Anaconda3\lib\site-packages\tensorflow\python\framework\ops.py", line 1128, in *init* self._traceback = _extract_stack() InvalidArgumentError (see above for traceback): Default MaxPoolingOp only supports NHWC. [[Node: pool1 = MaxPoolT=DT_FLOAT, data_format="NCHW", ksize=[1, 1, 2, 2], padding="SAME", strides=[1, 1, 2, 2 ], _device="/job:localhost/replica:0/task:0/cpu:0" <http://Relu>]] Does this mean I'm not preprocess the data properly? Or is there any possible reasons for this? How can I correct it? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#5>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AG7s5Yk3FaQ4j0XQUQT2TOsYje2dcaLiks5rfovhgaJpZM4MK6yf> .