Best way to deploy R model
Options
Povilas
Registered Posts: 23 ✭✭✭✭
I was wondering what is the best way to deploy custom R model in Dataiku? For Python Dataiku has an option "Add custom Python model" in Visual Analysis. However, I did not noticed anything similar for R. I know there's a possibility to save R model to .RData file and then save this into Dataiku Folder. Is there any better and more convenient way?
Tagged:
Answers
-
Hello Povilas,
I had highlighted the way to deploy custom R models in our last session in Stockholm. You need to create an "R function" endpoint which reads the .RData file in a folder and applies the script. Do you have the API package I used last time? You can use it as a starting base and adapt it to your needs.
Note that there is also a "Custom prediction (R)" endpoint which works similarly to the "R function" endpoint, but adds the ability to enrich the input JSON with external SQL tables. In your case, it should not be needed as JSON files have been enriched beforehand.
There is a visual way to attach models provided you used the visual machine learning interface to develop them. In this case you just need to create a "prediction model" endpoint and select the deployed models you want to use.
Cheers,
Alex -
Hello Alex,
Maybe my question was not accurate enough. I was talking about model creation using Visual Analysis. There you can train and deploy models like this:
https://ibb.co/nFxKs6
As a example here I created Random Forrest models using predefined Dataiku models and created custom Python model using Python code. Those models I can use for prediction on test data or something. Is there any way to create this kind of model using R code instead of Python? I mean to create that green diamond box representing a model using R code (not Python). Or the only way to use any R model is to upload it to Folder and then extract it from there and use it inside R notebook or R recipe? -
Hello Povilas,
We currently support only the Python backend to deploy ML models on an API with a visual interface. For R you will need to write custom code outside of the visual ML interface.
Cheers,
Alex -
Do you record the session please or d you have a blog explaining how to deploy the model?
Thanks.
