KNIME, Python environments, and Deep Learning settings

KNIME Extensions
,
1

After a bit of confusion and trouble after trying to complete the first deep learning exercise (Keras node kept giving me an error) I was able to successfully configure the Knime settings under the titles “Python (legacy)” and “Python Deep Learning” so that there weren’t any error messages in a yellow font.

It seemed to require 4 separate Python environments in order to account for certain versions being compatible for one thing, but not reusable for the other. And I found I wound up needing to create 4 different environments in Anaconda. My questions are:

  1. …whether that’s normal to require several different environments to achieve these settings without error messages?

  2. Moving forward, if updates are needed for something within each environment… will these occur on their own automatically? Does KNIME notify me? Would it be complicated because of the 4 separate environments?

image
image688×569 14 KB

image
image696×671 17.1 KB

1 Like
2

@jordanm55 welcome to the KNIME forum. It is something of an art to have the right combination of packages for the deep learning nodes to work with python.

If you found a working combination it is best to store, document and keep it. Until knime changes the nodes there should not be any updates but currently you would have to take care of them yourself.

I hope there will be a better integration of DL and a knime python integration like for basic packages.

Sometimes there are workarounds like for macOS and Apple Silicon.

2 Likes