Hello, I am new to KNIME and working on Deep learning frameworks.

I am try to install keras and tensorflow in my KNIME environment but I am really tired of trying this. I have installed the python required versions of 2.7.0 and latest version of 3.8.0. Instead when ever I am running the keras node it returns kernal error etc. Plz guys help me set this environment. I am really running out of time.
Looking forward for help. Thank YOu

Hi @Faheeemahmed,

Welcome to the KNIME Community!

The recommended way is to setup a new environment using Conda and doing so via the preference page of KNIME Analytics Platform.

You can find the steps to follow here: KNIME Deep Learning Integration Installation Guide

Cheers
Kathrin

4 Likes

Thank You Kathrin for your timely reply.
I have followed the guide and have installed the keras and tensorflow. But the DL python network learner node fails saying, that the keras 2.1.6 and tensorflow 1.8.0 are installed.
can you please explain which python version I should install?
Looking forward to hearing you.
Thank You

Can you please share a screenshot of your Python Deep Learning Preference?
It should look similar to this:

Hello Kathrin, attached is the screenshot.

I would recommend to click on the New environment… button to create a new environment with the correct TensorFlow and Keras version and select the created environment afterwards.

2 Likes

@Kathrin Dear Kathrin, just want to check with you that the Keras extension still works well with Tensorflow 2, right? I have ever found one image to describe the relationships between Knime, Keras, Tensorflow, Python, as showing below. I want to know whether this relationship still applies if I switch to Tensorflow2 and abandon Tensorflow 1 completely? As you know, Keras has been integrated into Tensorflow2 natively, does this affect the Keras extension in Knime?
Thank you very much for your clarification.
knime

Hi @ivanchenph,

the Keras integration uses Tensorflow 1 in the backend. We do have a TensorFlow 2 integration though, which comes with a TensorFlow 2 Reader, Executor and Writer node.

This workflow shows you how you can convert a network which you defined with the Keras nodes into a TensorFlow 2 network Convert a Network created with Keras to TensorFlow 2 – KNIME Hub

Cheers
Kathrin

4 Likes