But experimenting with Deeplearning I got problems loading some of your public examples - e.g. 09_Simple_Anomaly_Detection_Using_AConvolutional_Net. I'm using KNIME (64bit) with all free extensions on a Windows 10 Prof. PC having 16GB MEM.
The error is related to the 'MNIST Fetcher' node which seems not to be part of the 'all free extension' package. After processing the steps below, I end up with: 'Cannot complete the install because of a conflicting dependency.' shown in the 'Install' window concerning KNIME IMage Processing - Deeplearning4JINtegration'.
Process
Deinstalling all old KNIME versions from my PC
INstallation of KNIME 3.2 via KNIME Full 3.2.0 Installer (64bit)
AFter first Launch, the proposed new Updates, shown in the welcome area, are installed
After Restart, Login to the public Server via the 'EXAMPLES'
Double clicking on '09_Simple_Anomaly_Detection_Using_AConvolutional_Net', found at 04_ANalyses->14_Deep_Learning.
During loading the 'Workflow contains missing nodes' message box appears, indicating that MNIST FEtcher is missing.
The automatic extension finder fails
After activating the 'Stable Community Contributions' in 'Preferences->INstall/Update->Availyable Software Sites' the loading of the workflow (step 5 and 6) brings up the 'Install' window presenting the extension 'with the KNIME IMage Processing - Deeplearning4JINtegration. Pressing'Next' has no effect and clicking on the shown extension brings the 'Details' message: Cannot complete the install because of a conflicting dependency.
there seems to be a version conflict between the very latest KNIME DeepLearning4j Integration (64bit only) extension and the one the KNIME Image Processing - DeepLearning4j Integration references/expects when trying to install.
If you Uninstall the KNIME Deeplearning4J Integration (64bit only) first and then repeat the whole process of fecthing the example from the server, then the installation of the KNIME Image Processing - DeepLearning4j Integration, including the proper dependency, works properly and your workflow will load, configure and run as expected.
Here the full log of the conflict if I ask to upgrade back to the 64bit only version.
Your original request has been modified.
"KNIME Deeplearning4J Integration (64bit only)" is already installed, so an update will be performed instead.
Cannot complete the install because of a conflicting dependency.
Software being installed: KNIME Deeplearning4J Integration (64bit only) 3.2.0.v201607191547 (org.knime.features.ext.dl4j.feature.group 3.2.0.v201607191547)
Software currently installed: KNIME Image Processing - Deeplearning4J Integration 1.0.0.v201607070746 (org.knime.knip.dl4j.feature.feature.group 1.0.0.v201607070746)
Only one of the following can be installed at once:
KNIME Deeplearning4J Integration 3.2.0.v201607070744 (org.knime.features.ext.dl4j.feature.jar 3.2.0.v201607070744)
KNIME Deeplearning4J Integration (64bit only) 3.2.0.v201607191547 (org.knime.features.ext.dl4j.feature.jar 3.2.0.v201607191547)
Cannot satisfy dependency:
From: KNIME Deeplearning4J Integration 3.2.0.v201607070744 (org.knime.features.ext.dl4j.feature.group 3.2.0.v201607070744)
To: org.knime.features.ext.dl4j.feature.jar [3.2.0.v201607070744]
Cannot satisfy dependency:
From: KNIME Deeplearning4J Integration (64bit only) 3.2.0.v201607191547 (org.knime.features.ext.dl4j.feature.group 3.2.0.v201607191547)
To: org.knime.features.ext.dl4j.feature.jar [3.2.0.v201607191547]
Cannot satisfy dependency:
From: KNIME Image Processing - Deeplearning4J Integration 1.0.0.v201607070746 (org.knime.knip.dl4j.feature.feature.group 1.0.0.v201607070746)
To: org.knime.features.ext.dl4j.feature.group [3.2.0.v201607070744]
It turns out it is possible to re-upgrade to the 64bit only version, but then you are back to the problem that the MNIST Fetcher node cannot be found.
Thanks for all the feedback. I commitet a fix which should be available in ~2h on the Stable Community Contributions Update-Site. Let me know if there are still problems!
Sorry for the trouble,
Christian
Edit: Update should be available for knip-deeplearning4j now.