Hi all,
I am recently working on writing an extension for predicting POJO (Plain
Old Java Object) models created by H2O.ai, which is complex to load in OSGi
environment due to lack of proper packaging. When it comes to external
users, they have to do more preprocessing to load it into the CEP.
In this extension, users only need to give the path to a particular model.
Before that user has to place the h2o-genmodel.jar in
<CEP_HOME>/repository/components/lib directory.
eg.
from input_stream#h2opojo:predict('path_to_model_folder')
For python H2O API users, I have also implemented a simple python method to
output the model into a specific folder named by model class name.
If the user is not using python environment, then the user has to compile
the model externally and need to rename the folder which contains the
compiled classes using the model_name. (model_name.java)
h2opjo extension : https://github.com/roshanmadhushanka/h2opojo
model generation python :
https://github.com/roshanmadhushanka/FeatureProcessor
--
Roshan Alwis
*Trainee Software Engineer*
*WSO2*
m: 0715894672
a: No 3/56, Aluthgama Rd, Elpitiya.
e: [email protected]
<https://web.facebook.com/alwisroshan> <https://twitter.com/rm_alwis>
<https://lk.linkedin.com/in/roshanalwis>
_______________________________________________
Architecture mailing list
[email protected]
https://mail.wso2.org/cgi-bin/mailman/listinfo/architecture