Keras installation, problem to install graphviz package
I want to start with the Keras extension. So I followed the install script at https://community.rapidminer.com/t5/RapidMiner-Studio-Knowledge-Base/Keras-Deep-Learning-extension/ta-p/40839
I'm using a new MacBook. So I followed the MacOS/OSX Annaconda install script. The results are:
conda create –n keras (successfully executed)
source activate keras (successfully executed)
conda install pandas (successfully executed)
conda install scikit-learn (successfully executed)
conda install -c conda-forge keras (successfully executed)
conda install –c anaconda graphviz (unsuccessful but successful using Anaconda Navigator)
conda install –c conda-forge pydotplus (unsuccessful but successful using Anaconda Navigator)
The last install step states "In RapidMiner Studio Keras and Python Scripting panels in preferences, specify the path to your new conda environment Python executable." I set it to: /Users/luc/anaconda3/envs/keras/bin/python
Still RapidMiner Preferences is complaining about graphviz:
I tried to change my path in .bash_profile to: export PATH="/Users/luc/anaconda3/envs/keras/bin:$PATH"
Such does not bring a solution. What could be the problem?
Best Answer
-
luc_bartkowski Member Posts: 46 Maven
By installing graphviz using 'pip install graphviz' I solved my problem.
2
Answers
tagging @pschlunder and @jpuente
Scott
I had the same issue with graphviz and pydot. I followed your recommendation and used pip to install both. Then indeed the keras test in RapiMiner's preferences returned an Ok installation. Yet, now when running a keras sample example (SP 500 regression), I get an error message: "Pydot failed to call GraphViz. Please install GraphViz and ensure that its executablesare in $PATH. (Script line 197)". Anyone a clue how to fix this?
Hi @Gottfried,
please check if the PATH set for the Python extension and the one for the Keras extension match. Therefore open the RapidMiner settings and open the respective tabs. My guess is, that your Python extension points to another path, than Keras. The Keras extension relies on the python extension for somethings, hence the need to set both path alike.
Hope this helps.
Philipp
Hi @Gottfried,
could you share your process xml please? And if possible a reduced data set for testing?
Regards,
Philipp
I just ran any of the examples coming with the Keras extension setup, e.g. the SP500 regression example. I did no change to neither the proposed process nor the dataset. Do I really need to paste a copy here?
I encountered the same problem. Has anyone solved it yet?
Did you ever get a resolution to this problem?