Discover this year's submissions to the Dataiku Frontrunner Awards and give kudos to your favorite use cases and success stories!READ MORE

Deploy non "Saved Model" for real time scoring

RohitRanga
Level 3
Deploy non "Saved Model" for real time scoring

Hello Dataiku community!

I have a multi-label text classifier trained using scikit learn and stored in a managed folder as a .pkl file. I came across this tutorial for how to create a python model prediction endpoint. 

https://doc.dataiku.com/dss/latest/apinode/endpoint-python-prediction.html#classification

The input to the endpoint needs to be something like this:

{'text': 'I have my graduation ceremony and birthday next week.'}

And the output like this:

{'text': 'I have my graduation ceremony and birthday next week.',
 'predictions': [{'values': ['BIRTHDAY', 'GRADUATION']}]}

 Any thoughts on how this can be done? I would compute embeddings for the text using my custom logic before it is fed into the model. 

0 Kudos
0 Replies