reduce_sum() function does not recognize 2nd argument 'axis'

[kyungyunlee]

There seems to be an issue in reduce_sum() function in lab-06-1-softmax_classifier.py

When the second argument is "axis=1", there is a TypeError saying "unexpected keyword argument."
From my experience with python, this shouldn't be an error, but I am not sure if it is a version issue(I have the most recent one though..) or something else.

In my case, it works when I fix it from
cost = tf.reduce_mean(-tf.reduce_sum(Y * tf.log(hypothesis), axis=1))
to
cost = tf.reduce_mean(-tf.reduce_sum(Y * tf.log(hypothesis), 1))

Also a reference to tensorflow document.
https://www.tensorflow.org/api_docs/python/tf/reduce_sum

[kkweon]

I'm not able to reproduce this error. You should be able to use the axis keyword.
Also, it's not a version issue because axis has been there since the beginning.

The only thing I suspect is that you claim there is an issue in the file lab-06-1-softmax_classifier.py
but your error message shows the error happens in basics-01.py which is not relevant to this file.

It seems like a sementic error to me.

[kyungyunlee]

I am very sorry about the file name and the confusion. I made my own file without cloning directly from git to try out different codes myself, so I was actually running the correct file lab-06-1-softmax_classifier.py and code. I also think that axis keyword should work, but it only works when written without...

Just wondering if it could help...I am using python3 with tensorflow '0.11.0rc2'

Any idea/suggestion/lesson! would be appreciated!!
Thanks

[kkweon]

@kyungyunlee You are using too old version of tensorflow.
The current version is 1.2.1 and yours is 0.11.0 and even it's a rc(release candidate).

Please update your tensorflow first.

[kyungyunlee]

@kkweon
I tried uninstalling and reinstall tensorflow with pip3 multiple times in the past, but it still remained the same, so I was very ignorant about it for a while.
But I just realized that I was using python3.5(default) for running tensorflow but was updating tensorflow for python3.6...
And it solved several problems actually, thank you very very much!

[kkweon]

Great!