Hello,
Our team received the following error message:
AttributeError: 'module' object has no attribute 'control_flow_ops'
We have googled it and found that it might be caused by the newer version of TensorFlow, which is 0.11.0rc2 on the server. If we can roll back to version 0.10.0rc0 or 0.9.0, the problem may be resolved. I was wondering if you could show us how to do it? Thank you very much in advance!
Best,
Yang
Created by Yang Yu yangyu Hello David,
Thank you very much for your help!
Currently I am using the "FROM tschaffter/tensorflow-gpu" command in my docker file. What do you suggest I should change this line in order to use one of those repos you mentioned? Or should I add extra lines in the docker file?
Thanks,
Yang @yangyu - if you want a different version of TensorFlow installed you need to change your Dockerfile.
If you are building off of a TensorFlow container from their repo, just use an older version:
[All TensorFlow Docker tags](https://hub.docker.com/r/tensorflow/tensorflow/tags/)
Or if TensorFlow is installed on a later line of your Dockerfile, you just need to change the rc to whichever number you want (although you may find other conflicts in your code).
Either way, the answer will be in the Dockerfile you use to build your container. Dear Yang,
I redirect the question to @darvinyi.
Thanks,
Thomas