tensorflow:AttributeError: 'module' object has no attribute 'mul'

Question

I have used tensorflow for ONE day,but there comes some troubles,when I import tensorflow, there would be AttributeError: 'module' object has no attribute 'XXXXXX'

Environment

I use ubuntu14.04, python2.7, CUDA toolkit 8.0 and CuDNN v5. And versions of my six and protobuf are: Name: six Version: 1.10.0 Location: /usr/local/lib/python2.7/dist-packages Requires: Name: protobuf Version: 3.2.0 Location: /usr/local/lib/python2.7/dist-packages Requires: six, setuptools

here is my test code:

 import tensorflow as tf a = tf.placeholder(tf.int16) b = tf.placeholder(tf.int16) add = tf.add(a, b) mul = tf.mul(a, b) with tf.Session() as sess: # Run every operation with variable input print "Addition with variables: %i" % sess.run(add, feed_dict={a: 2, b: 3}) print "Multiplication with variables: %i" % sess.run(mul, feed_dict={a: 2, b: 3})

I get this output:

Is there any problem with the tensorflow installation? or any other problems?

Answer 1

According to thetensorflow 1.0.0 release notes ,

tf.mul , tf.sub and tf.neg are deprecated in favor of tf.multiply , tf.subtract and tf.negative .

You'll need to replace tf.mul with tf.multiply .

Answer 2

This operation was previously available in 0.x versions. With the release of TF 1.0 they introduced breaking changes to the API . In addition to

tf.mul , tf.sub and tf.neg are deprecated in favor of tf.multiply , tf.subtract and tf.negative

many other functions were renamed and changed with the following justification:

Several python API calls have been changed to resemble NumPy more closely.

So a lot of the scripts that you already found on the web or from the books will not work. Good thing is that majority of them can be fixed with their migration script . It can be run with tf_upgrade.py --infile foo.py --outfile foo-upgraded.py . It will not be able to solve everything (limitations are listed here ), but will save you a lot of work.

Answer 3

2.0 Compatible Answer :

The Commands for tf.multiply , if we want to migrate from Tensorflow 1.x to 2.x are shown below:

tf.compat.v1.math.multiply, tf.compat.v1.multiply, tf.compat.v2.math.multiply, tf.compat.v2.multiply

The Commands for tf.subtract , if we want to migrate from Tensorflow 1.x to 2.x are shown below:

tf.compat.v1.math.subtract, tf.compat.v1.subtract, tf.compat.v2.math.subtract, tf.compat.v2.subtract

The Commands for tf.negative , if we want to migrate from Tensorflow 1.x to 2.x are shown below:

tf.compat.v1.math.negative, tf.compat.v1.negative, tf.compat.v2.math.negative, 
tf.compat.v2.negative

See this Tensorflow Migration Guide for more details.

Answer 4

在 python-3 中使用tf.multiply而不是tf.mul 。