KNIME, DL4J and CUDA installation failed

[datalytics]

Hi there,

I am new to KNIME. So far doing all my business work in Alteryx.

I have an Lenovo workstation a working copy of Knime 4.4 installed.
'Am trying to enable the CUDA framework to leverage GPU calculation capacity.
I ahve installed the latest nVidia driver and the latest CUDA version 11.x

The KNIME console states the following.

DL4JPluginActivator
No compatible CUDA version seems to be installed on the system! Supported versions are: GPU CUDA Toolkit 7.5 and GPU CUDA Toolkit 8.0

Unfortunately I don’t want to roll back to an outdated CUDA version.

The following article states this to be working for an older KNIME / CUDA version combination.
https://www.knime.com/deeplearning4j

Any idea on what I can do to get Knime 4.4 to work with Cuda 11.x?

Thanks for sharing your insights as well as tips and tricks.

Reinhard
Düsseldorf, Germany

[marvin.kickuth]

Hi Reinhard,

welcome to the forum!

I’m afraid our Deeplearning4J integration only works with the mentioned older CUDA version.
However, we have a Keras integration (KNIME Deep Learning - Keras Integration) that might also fit your needs. That works with a newer CUDA version. Alternatively, at least under Linux, it is possible to install different CUDA versions side by side, so that you can continue working with the current version outside of the DL4J environment.

Kind regards
Marvin

[datalytics]

Hi @marvin.kickuth,

many thanks for your quick turnaround.
Thanks so much for sharing your insights.
Very helpful to know.

I have both options in parallel.
Two SSD’s - one with Win 10 pro and a 2nd one with OSGEOlive / Ubuntu.

So I will play around a bit.

Thanks angain
and looking forward to use this great product.

Reinhard