Due to recent updates, all users are required to create an Altair One account to login to the RapidMiner community. Click the Register button to create your account using the same email that you have previously used to login to the RapidMiner community. This will ensure that any previously created content will be synced to your Altair One account. Once you login, you will be asked to provide a username that identifies you to other Community users. Email us at Community with questions.

Keras Extension Dependencies

ZAMZAM Member Posts: 21 Maven
edited April 2020 in Help
Hello,

For some reason, my Keras extension cant find "Keras" and "Scikit-learn" libraries on my computer although they are in the same folder as the other required libraries such as "Tensorflow".

Anyone encountered a similar problem or can help, i would appreciate that.

Thanks a lot.
Tagged:

Best Answer

Answers

  • SGolbertSGolbert RapidMiner Certified Analyst, Member Posts: 344 Unicorn
    edited April 2019
    Hi,

    I also wanted to ask what are the future plans for the Keras extension ( @pschlunder ). I think that as long as DL4J has similar capabilities, it makes more sense to use that, so long the implementation from RapidMiner is done efficiently. To illustrate the last point, a lot of the computing time when using Keras from Python comes from Python-TensorFlow communication. If that could be avoided in RapidMiner+DL4J, we have the potential of having an excellent performance.

    Regards,
    Sebastian
  • sgenzersgenzer Administrator, Moderator, Employee, RapidMiner Certified Analyst, Community Manager, Member, University Professor, PM Moderator Posts: 2,959 Community Manager
    edited April 2019
    I will defer to @pschlunder for the details but, in short, the DL4J extension is going to be supported moving forward in lieu of the Keras one - for all the reasons stated above. In the RM ecosystem I would strongly recommend staying with DL4J.

    [that said, a little birdie has told me that there will be new RM integration with Jupyter notebooks in the very near future, so you could likely avoid some of the Keras issues but simply taking advantage of that...stay tuned :wink: )

  • pschlunderpschlunder Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, RMResearcher, Member Posts: 96 RM Research
    Hi,

    thanks for using our Deep Learning functionality. Sorry to hear, that the keras extension is causing you problems. When saying, that tensorflow can be found, is it with the python scripting operator? If so please ensure, that the python path for the Keras extension is set to the same folder in the general RapidMiner settings in the Keras tab. Unfortunately you need to set the path for both the Python and the Keras extension.

    As mentioned by others maybe you could try the currently under development Deep Learning extension. Many of the previously available functionality from the Keras extension is already there. As Sebastian mentioned, it's not relying on Python and hence works more performant since data can be kept in Java.

    For the future expect Deep Learning capabilities from the new extension, if you need to apply models created in Keras (e.g. from colleagues or previous projects), please try out the "Read Keras Model" operator from Deep Learning extension. It enables you to load models trained in python and apply them inside RapidMiner.

    Hope this helps,

    Philipp
Sign In or Register to comment.