deeplearning4j not working

[papaito]

Beginning one or two days ago, nodes in deeplearning4j are not working, throwing different errors.
Same workflows works good in earlier knime installations.

What can I check?
Thanks!

[steffen_KNIME]

Hi,

could you please provide some more information?
When did it work, when not? Which different errors? Did you change the environment?

Regards
Steffen

[papaito]

Thanks steffen!
I have try to increase memory, turn on/off GPU calculation and verbose error logging; no luck.
I’ve removed and reinstalled DL4J.
It seems the CUDA version may cause error.
CUDA 11 production machine, knime AP 4.6.3: no luck (one week ago was working, no changes)
CUDA 8 test machine knime AP 4.6.2: works
I can’t downgrade CUDA in production.

DL4J version on both:
KNIME Deeplearning4J Integration (64bit only) 4.6.0.v202202251624 org.knime.features.ext.dl4j.feature.group KNIME AG, Zurich, Switzerland

Error. Execute failed: The Deeplearning4j library could not be initialized. Maybe there is not enaough memory available for DL4J. Please consider increasing the 'Off Heap Memory Limit0 in the DL4J Preference Page.

[steffen_KNIME]

Hi papaito,

thanks for the information. Could you also provide the knime.log around that error?

Additionally, could you try whether the CPU version works for you instead of GPU?

Thanks
Steffen

@bwilhelm maybe you have another idea when we get the knime.log

[papaito]

Hi Steffen,
today all is working fine.
I don’t know why, I’ve made no changes.

Thank you.

[steffen_KNIME]

Thanks for the update!