📄

Delta Machines integration

Learning and inference

Coming Soon

We’re finalizing practical documentation on our Delta Machines technology - a machine learning library designed to deliver cutting-edge predictive performance to businesses in a transparent and productive way.
Check back soon for full details.

Learning

Delta Machines can be fitted and manipulated in Python as any Scikit-Learn estimator, with the standard fit method.

Both methods accept these 5 array types:

  1. JAX arrays
  2. Numpy arrays
  3. Pandas dataframes and series
  4. Polars dataframes and series
  5. Pyarrow tables and arrays

Inference

A Delta Machines model can be deployed in 4 ways:

  1. As a Python Scikit-Learn estimator
  2. As a shell command
  3. As a web service
  4. As a code library

Python Scikit-Learn estimator

The Scikit-Learn estimator method requires to use the standard predict and predict_proba methods. The estimator itself can be saved and loaded from a JSON file using the save and load methods. These method accept either a file name in argument, or a typing.TextIOstream. 


                # Model training (in a production batch)
                model = delta_machines.ClassifierModel()
                model.fit(X, y)
                model.save("my_model.json")
            

                # Model pre-loading (in a production batch or realtime service)
                model = delta_machines.ClassifierModel()
                model.load("my_model.json")
            

                # Model use (in the same process)
                predictions = model.predict(X)

Shell command

To be completed.

Web service

To be completed.

Code library

To be completed.

⬅ Return to documentation homepage