keras model / deep learning

Hi Thiru,

quick question: which extension version are you using and how exactly did you find this process ?
I can't seem to find it myself, i.e. I don't have a "Keras Samples" repo.

Mate

hello @Mate
thanks for your reply.  Im using version 9.9.   see the enclosed full screen pic. 
By clicking process - "iris classification' under 'keras samples' folder,  Im getting this process.


thiru

I meant the extension version, not RapidMiner .

+ I don't really understand what you are trying to do here.
Can you please shed some light on why you are trying to do here ?
Btw, what does it have to do with "Keras" ? This is just a normal deep learning training in RapidMiner right ?

deep learning extension version: 0.9.003,  & Keras extension: 1.0.

Im trying to learn the deep learning process , as i'm new to this. 
 I'm following the video   by phillip schlunder.  ->    rapidminer.com/resource/state-deep-learning/

Well, that's a bit older than the one we currently have.
Could you maybe upgrade your extension(s) to the latest version(s) ?

+ we have tutorial processes that can help, amongst them one for the Iris use-case.

Make sure that you have a Python Scripting extension installed and both Python and Keras are configured to point to the correct location of your Python. If you use an Anaconda environment then also specify the environment for Python and for Keras you will need to navigate to the Python in the same environment. Of course TensorFlow and Keras have to be installed (pip or conda). However, Keras extension is no longer supported and works only with TensorFlow 1.xxx (not the current version 2.xxx). So the best is to switch to Deep Learning extension, which is Java native (no need for Python) and continuously supported by RapidMiner.

 I updated the latest version of  , deeplearning, Keras model, Python scripting.  
But still the error persists,  when I'm using the process given in sample repository.   (iris classification in 'Keras sample').
How to resolve this?

When configuring Keras and Python in preferences, have you tried to press the test button in each? Keras installation is non trivial and requires a lot of system bits to satisfy the assumptions of the extension, e. g. it will not work with Python 3.8, Tensorflow 2, needs some libraries, etc. As I said before, the extension has not been updated for a vew years and it is safer to use Deep Learning extension. I have given up on using it a while ago. However, I'll try to reinstall it and will let you know if there are any issues, which I expect to be, perhaps it is no longer compatible with RM 9.9?

OK, I have reinstalled the current version of Keras, configured and tested my Python and Keras extensions in preferences, ensuring that they point to the same version of Python in the same location. I have loaded the Iris example and yes the process shows the "potential" problem, however, when executing it runs fine. What happens is that Keras extension generates a custom (i.e. non-standard RM) model and so RapidMiner does not know if the model type generated and then used when applied are compatible. And yet, RM will pass the model from the output port to the input port and the model will be applied to the test data. All this is assuming that your Python 3.7 / TensorFlow 1.14 / Keras 2.2.5 are correctly installed.

When installing Keras on top of Tensorflow, you will need additional Python packages, i.e. pydot, HDF5, h5py, graphviz, python-graphviz. Note that you need to install graphviz software separately. Depending on your environment, you may need to install some additional software, e.g. cython, so watch the warnings and add those libraries as required. When you get warning on version incompatibilities, e.g. with bleach and htm5lib, you may try ignoring them.

@jacobcybulski
thanks for your clarifications.   ( I got into keras, as it was explained in training video.    rapidminer.com/resource/state-deep-learning )

Anything to be installed if we use just  'deep learning extension'.
does this 'DL extension' has same capacity as 'keras extn' ?

regds
thiru

If you use Deep Learning extension, you will also need to install ND4j extension which manages its back end. If all you need is to run your DL processes on a CPU then you are done. If you have a suitable NVidia GPU then you will also need additional software, such as gou drivers and a GPU development kit. 

@jacobcybulski
thanks again for your inputs.

regdsthiru