Is it possible to extract the trained model object in dataiku and use it in python recipe?

[minliang1215]

In sklearn, we can dump the trained model object to a pickle object, and use it in other parts of the code after loading it. Is there an equivalent function in Dataiku? The only available actions I can see when I click on a trained model are predicting, scoring, evaluation and convert the training process to python code.

Thanks!

[UserBird]

Hi,

Using the Dataiku API, you can retrieve the model and use it inside a python recipe.

Something like this will work:

import dataiku

m = dataiku.Model(my_model_id)

my_predictor = m.get_predictor()

The  predictor is a DSS object that allows you to apply the same pipeline as the visual model (preprocessing + scoring). If you want to retrieve the scikit-learn model itself, just add one more line:

my_clf = my_predictor.clf

You can find more information here: https://doc.dataiku.com/dss/latest/api/python/saved_models.html

Cheers, 

Du