ModuleNotFoundError: No module named 'tensorflow.contrib'

[Lawrence-Krukrubo]

I have tf==1.14 installed and python 3.7 I have been awake the whole night trying to sort this. Please please help.
My project is on AIX360 and I can't get it to work. Even after cloning the repo, the above error message keeps appearing whenever I load the CEMExplainer and others

[vijay-arya]

@Lawrence-Krukrubo pls try python 3.6 + tf1.14 for CEM.

[Lawrence-Krukrubo]

I have now installed python=3.6.13 and tf=1.14
I use the pip install aix360 and get the following error:

Failed to build cvxopt ecos scs
ERROR: Could not build wheels for cvxopt, ecos, scs which use PEP 517 and cannot be installed directly

what should I do please?
do you think it will work if I cloned the AIX360 GitHub repo into this environment that has python=3.6.13 and tf=1.14?

[Lawrence-Krukrubo]

Hi @vijay-arya I have read your paper "AI Explainability 360: Impact and Design" and the "AI Explainability 360 Toolkit". This toolkit is the focus of my Master's dissertation which I am working on, at the moment. My focus is on CEM and its unique use of PN and PP to justify and improve decisions made to customers. Please, I need this to work, so I can carry out a few experiments locally. Thanks

[vijay-arya]

@Lawrence-Krukrubo the following should work:

git clone https://github.com/Trusted-AI/AIX360
cd AIX360
pip install -e .[contrastive]

The installation instructions are also available here: https://github.com/Trusted-AI/AIX360/tree/master#installation
The latest version is not in pypi yet, so you will need to clone it from git as above and then install.

[Lawrence-Krukrubo]

I have done every thing step by step and still getting the same error

ModuleNotFoundError: No module named 'tensorflow.contrib

the original installation guide on the page you sent has python=3.10 should I try that version of Python?

[Lawrence-Krukrubo]

I tried to use python=3.10 in a new Conda env and got an incompatibility error for tf=1.14

[Lawrence-Krukrubo]

3.10 won't work as Contrastive expects 3.6, which is what I already have in my env. I have python=3.6.13 and tf=1.14

[Lawrence-Krukrubo]

Is there anything I can do please, is there a way to run the CEMExplainer from any other application like Watson Studio? Please, I am willing to pay for a subscription to get this to work. My entire dissertation hinges on this please

[Lawrence-Krukrubo]

I just tried to run the nn.contrastive example and this runs fine. but the CEMExplainer keeps bringing up the module not found error.

[vijay-arya]

@Lawrence-Krukrubo I hope you are following these steps. Then you should be able to use CEMexplainer.

(base)$ conda create --name aix360 python=3.6
(base)$ conda activate aix360

(aix360)$ git clone https://github.com/Trusted-AI/AIX360
(aix360)$ cd AIX360
(aix360)$ pip install -e .[contrastive]

[Lawrence-Krukrubo]

yes, I did so and it's just a bit frustrating. Let me delete the old env and start all over again, see if it would work.

[Lawrence-Krukrubo]

I just followed your steps judiciously one after another. Exactly how you put it and I am despondent and sad to say the same issue persists...

ModuleNotFoundError: No module named 'tensorflow.contrib'

please isn't there something that can be done, a quick fix or stop-gap just to enable me run my local notebook?

[vijay-arya]

@Lawrence-Krukrubo when do you get this error. Which notebook you are trying to run and the code that gives you the error. Also some info on the system you are using (windows, mac, etc.)

[Lawrence-Krukrubo]

I'm using a Windows system. And I get this error whenever I run any code that has to do with importing CEM or even the Classifier. The moment I run the examples such as the MNIST CEM notebooks. But if I ran the nn.contrastive notebooks, these run fine. I just don't get it as CEM is what I need to do to show the effect of PN and PP values in my dissertation

…

On Mon, 26 Jun 2023, 05:38 Vijay Arya, ***@***.***> wrote: @Lawrence-Krukrubo <https://github.com/Lawrence-Krukrubo> when do you get this error. Which notebook you are trying to run and the code that gives you the error. Also some info on the system you are using (windows, mac, etc.) — Reply to this email directly, view it on GitHub <#182 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AJLJAUAXRPHORRC5HPQXSXLXNEG2TANCNFSM6AAAAAAZS3RYHQ> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[vijay-arya]

@Lawrence-Krukrubo
If you are getting ModuleNotFoundError: No module named 'tensorflow.contrib error, then it means you have a higher version of tensorflow being used somewhere and you are running code that is compatible with TF1.x
https://stackoverflow.com/questions/55311534/how-to-fix-no-module-named-tensorflow-contrib-for-python-project

Since CEM is a white box explainer, the other place to check is if the ML model that you pass to it has been trained using TF 1.14 and not a higher version.

CEM is currently tested on windows, mac, & linux for python 3.6, TF1.14.