TensorFlow Error

Hi,
I’m having troubles using the Keras to Tensorflow Network Converter Node after the Keras Network Learner Node (where the execution was ok). This is the error message:

ERROR Keras to TensorFlow Network Converter 0:93 Execute failed: org/tensorflow/framework/SavedModel

Clearly it depends on the fact that the tensorflow 2 enviroment is not properly installed as you can see in the attached picture.

Any hints? Thanks!
Alfredo

Hello Alfroc,

Are you sure you have configured your DeepLearning libraries in Knime ?
You should check :
In File > Preferences > Knime > Python Deep Learning

If there is no environment in Keras or TensorFlow (2), it’s normal it’s not working for you.
You should have something like this :

Or like this if not configured properly :

Here is the guide to set this up :
KNIME Deep Learning Integration Installation Guide

Br,
Samir

Hi Samir,
unfortunately your suggestions don’t help because I already installed properly the extensions and, as you can see in the picture, the environments. I repeated all the procedure twice after anaconda reinstallation but the problem remains the same.
Best
Alfredo

I found in this link GitHub - fo40225/tensorflow-windows-wheel: Tensorflow prebuilt binary for Windows the following note:

If your CPU didn’t support AVX instructions, you will get ImportError: **DLL load failed: A dynamic link library (DLL) initialization routine failed**. (Win 10) or ImportError: DLL load failed with error code -1073741795 (Win 7) when using tensorflow official release 1.6.0 and up (pip install tensorflow)

You can use pip install <filename.whl> which file download from sse2 folder instead of using official AVX binary.

It might be this the problem? In fact, my 10yo VAIO cpu doesn’t seem to support AVX instructions:

image

So, any hope for me to use tensorflow?

Best,
Alfredo

Hello @alfroc,

I found this “old” post with the same issue as yours. Maybe you can check it out in full (i didn’t read it all :grinning:). This post seems related to your problem :

I let you check this out, but they found a solution if I understood well.

Br,
Samir

1 Like

Hi Samir,
I appreciate your efforts to help me, but I assure you that I have already read all the posts and all installation guides before opening this one.

I belive the problem is what I reported in my previous post:

I have an old CPU that does not support the AVX instructions set.

The solution seems to be to install a compiled version of Tensorflow without AVX, but this is beyond my capabilities.

B,
Alfredo

Hello @alfroc,

I’m sorry to hear that :disappointed_relieved:
Did you try to install SSE2 TensorFlow version from your link above ?
It seems that you can use a “specific” version of TensorFlow for “non-AVX” CPU’s, right ?

You just have to download the TensorFlow version related to your Python version and install it manually with command pip install <path/filename.whl>.
To check your python version, you just have to use command :

  • python --version or python -V
  • py --version or py -V (depending on your installation)

Example : If you have Python 3.8 installed you can install Tensorflow v2.5 from file tensorflow-windows-wheel/2.5.0/py38/CPU+GPU/cuda113cudnn8sse2 at master · fo40225/tensorflow-windows-wheel · GitHub

I hope that you’ll succeed.
Br,
Samir

1 Like

I made a lot of tests without any success and now, not knowing what to do, I must give up.

Hello @alfroc,

You should never give up =). From what I see you have Anaconda installed right ?
Can you try to run these lines on a cmd ? :

conda create --name tf-gpu
conda install -c aaronzs tensorflow-gpu
conda install -c anaconda cudatoolkit
conda install -c anaconda cudnn
conda install keras-gpu

I hope it will work for you.
Br,

Samir

Hi Samir,
as I already said, my CPU is 10yo, it doesn’t have AVX as instruction set, let alone the GPU…!
Again, I really tried everything I could!
B,
Alfredo

In the end I changed my computer, now everything works fine!

2 Likes

@alfroc,

Haha, that’s a solution too. Glad it works for you. Happy Deep learning.

Br,
Samir

1 Like

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.