is it possible to download a custom python model from DSS once trained?
Furthermore is there a way to add meta data to a model training session from within the custom script? For example if I generate some artifacts and I store them in a managed folder is there a way to store the name of the folder somewhere in the session?
When you say "custom python model", do you mean a custom python model trained from within the Visual ML (see attached image)?
If so, then you may be able to use the following code to extract the trained model object: https://doc.dataiku.com/dss/latest/python-api/saved_models.html?highlight=_clf
import dataiku model = dataiku.Model("5hvOicZn") predictor = model.get_predictor() clf = predictor._clf
Regarding the addition of metadata, my suggestion would be to add this as part of the training session "description" (see the second attached image).
Hopefully this helps, although let me know if you have any other questions.
Thanks a lot Ned,
Yes I meant a custom model trained from a visual recipe. I have already tried the code you posted, but I am able to list models only if they are deployed in the pipeline. Is this the case or I am missing something?
With respect to the metadata, is it possible to set the session description programmatically or through global variables?