Tensorflow, Keras version numbers

I am trying to set python deep learning up, but cannot deal with the information on the different versions. The message on the console tells me to “Please make sure that Keras 2.1.6 and TensorFlow 1.8.0 are installed in your Python environment.”
I did, but then the Python DL preferences complains that this tensorflow version is too old.
I have been playing around with the different versions of both Keras and Tensorflow, but couldn’t find a combination compatible with KNIME 4.3.1.

Could you advise?


Despite the error messages, Keras seems to work.

you could share your workflow to see what it looks like. There are TF 2 nodes in KNIME if I remember correctly.

Hi there,

The screenshot shows the inconsistency in the instructions: Preferences v. actual error message below. The issue is highlighted in the error message.

Hi @ribizli,

The Warning on the preference page is only relevant if you want to use the TensorFlow 2 integration. Note that there are two sections on the preference page. One for “Keras” and one for “TensorFlow 2”.
The section for “Keras” is not showing an error/warning => You can use the Keras nodes (which use TensorFlow 1 internally).
The section for “TensorFlow 2” is not showing a warning => You cannot use the TensorFlow 2 nodes.

Therefore, your setup is fine for using the Keras nodes and you can ignore the warning. To also use the TensorFlow 2 integration you could select different conda environments for “Keras” and “TensorFlow 2” to satisfy all requirements

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