0 votes
Hi dataiku devs and community,

I'm using dss through virtualbox installed on a windows 10 machine. I wanted to know if it is possible to save the trained scikit-learn models through joblib on the local disk (and not in the virtual disk).

This would greatly help my work as I want to use the models locally and not as a REST-API service.

All the best,
Aris
by

1 Answer

0 votes
Hi,

The pickled scikit-learn model can be found in the DSS data directory, in analysis-data/PROJECT/ANALYSIS/MLTASK/sessions/SESSION/ppX/mY/clf.pkl

If you look at the URL when you are within a model in the DSS UI, it will give you all the necessary identifiers to find the proper clf.pkl

You can then use the classical mechanisms of Virtualbox to copy that to your local machine.

An important side note is that DSS performs various preprocessing on the features, so the clf.pkl expects as input the preprocessed data. You would need to apply the exact same preprocessing in order to score with that. DSS does not currently provide an officially supported way of scoring models from using only the clf or files.
by
1,298 questions
1,326 answers
1,507 comments
11,863 users

©Dataiku 2012-2018 - Privacy Policy