ImportError for graph_util from tensorflow.python.framework

[mbaytas]

Importing graph_util from tensorflow.python.framework fails with an ImportError.

I discovered this while I was trying to go through the TensorFlow for Poets Codelab, an the the re-training failed.

I can reproduce the same error in my Python shell. I am able to verify that TensorFlow is installed correctly by importing tensorflow, tensor_shape from tensorflow.python.framework, and gfile from tensorflow.python.platform (as in the retraining script); but trying to import graph_util results in the same ImportError:

>>> import tensorflow
# Success
>>> from tensorflow.python.framework import tensor_shape
# Success
>>> from tensorflow.python.platform import gfile
# Success
>>> from tensorflow.python.framework import graph_util
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ImportError: cannot import name graph_util

I have also posted this as a question on Stack Overflow, and haven't received any answers so far. I am not sure if this is a bug or a configuration problem etc. Feels like a bug, since I can import tensor_shape from tensorflow.python.framework no problem.

Environment info

Operating System: OS X El Capitan (v10.11.5)
Python: v2.7.11 via Homebrew

TensorFlow 0.9.0 installed from binary pip package for Mac OS X, CPU only, Python 2.7 (TF_BINARY_URL=https://storage.googleapis.com/tensorflow/mac/tensorflow-0.9.0-py2-none-any.whl).

Steps to reproduce

In Python shell:

>>> from tensorflow.python.framework import graph_util
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ImportError: cannot import name graph_util

What have I tried?

• Looked to see if the file graph_util is there inside tensorflow/python/framework. It is.
• Looked to see if the current source code I am examining has undergone any changes since v0.9.0 that would fix this issue that appears in the release that I am using. It doesn't appear so to me.

[Nadja-K]

I'm having the same problem and can't seem to find a solution with the master branch.
While this is no real solution for the problem itself you could try pulling the r0.9 version from git.
This seems to work for me at the very least.

[aselle]

Assigning @danmane since he is on sync rotation next week to see the patch through.

[vrv]

I suspect the patch went through, though don't know which commit fixed this :(

[mbaytas]

The issue persists.

[vrv]

Ah, graph_util.py was moved from python/client to python/framework after r0.9. You should either check out the code from the r0.9 branch, or use the nightly PIPs linked from github.com/tensorflow/tensorflow to get access to more recent installs that match the latest source code in master.

[mbaytas]

Solved my problem by checking out the r0.9 branch.

[doganai]

Are newer versions deprecated?

[hudeqiWH]

I also encountered the same problem as yours, and I used the following two steps to solve the problem：
1.Activate TensorFlow，and stay active during following steps as well as running retrain.py：
source ~/tensorflow/bin/activate # bash, sh, ksh, or zsh
2.Upgrade TensorFlow version by those steps:
(tensorflow)$ easy_install -U pip
(tensorflow)$ pip install --upgrade tensorflow # for Python 2.7
Reference tensorflow official link:https://www.tensorflow.org/install/install_linux