mlflavors 0.1.0rc1

MLflavors: A collection of custom MLflow flavors.

The MLflavors package adds MLflow support for some popular machine learning frameworks currently not considered for inclusion as MLflow built-in flavors. You can use this package just like MLflow built-in flavors to save and load your models. Some of the key benefits are listed below:

• Save your trained model as an MLflow artifact

• Load your model from MLflow for batch inference tasks

• Serve your model for real-time inference to an endpoint in the cloud (e.g. Databricks, Azure ML, AWS Sagemaker, etc.) using MLflow built-in deployment tools

• Get inspiration for creating your own MLflow custom flavor

The following open-source libraries are currently supported:

Framework	Tutorials	Category
Orbit	MLflow-Orbit	Time Series Forecasting
Sktime	MLflow-Sktime	Time Series Forecasting
StatsForecast	MLflow-StatsForecast	Time Series Forecasting
PyOD	MLflow-PyOD	Anomaly Detection
SDV	MLflow-SDV	Synthetic Data Generation

The MLflow interface for the supported frameworks closely follows the design of built-in flavors.

Documentation

Usage examples for all flavors and the API reference can be found in the package documenation.

Installation

Installing from PyPI:

$ pip install mlflavors

Quickstart

Save an Orbit ETS model as an artifact to MLflow:

import mlflavors

from orbit.models import ETS
from orbit.utils.dataset import load_iclaims

df = load_iclaims()

test_size = 52
train_df = df[:-test_size]
test_df = df[-test_size:]

ets = ETS(
    response_col="claims",
    date_col="week",
    seasonality=52,
    seed=8888,
)
ets.fit(df=train_df)

mlflavors.orbit.save_model(
    orbit_model=ets,
    path="model",
)

Make a prediction loading the model from MLflow in native format:

loaded_model = mlflavors.orbit.load_model("model")
loaded_model.predict(test_df, decompose=True, store_prediction_array=True, seed=2023)

Make a prediction loading the model from MLflow in pyfunc format:

# Convert test data to 2D numpy array so it can be passed to pyfunc predict using
# a single-row Pandas DataFrame configuration argument
X_test_array = test_df.to_numpy()

# Create configuration DataFrame
predict_conf = pd.DataFrame(
    [
        {
            "X": X_test_array,
            "X_cols": test_df.columns,
            "X_dtypes": list(test_df.dtypes),
            "decompose": True,
            "store_prediction_array": True,
            "seed": 2023,
        }
    ]
)

loaded_pyfunc = mlflavors.orbit.pyfunc.load_model("model")
loaded_pyfunc.predict(predict_conf)

Contributing

Contributions from the community are welcome, I will be happy to support the inclusion and development of new features and flavors. To open an issue or request a new feature, please open a GitHub issue.

Versioning

Versions and changes are documented in the changelog .

Development

To set up your local development environment, create a virtual environment, such as:

$ conda create -n mlflavors-dev python=3.9
$ source activate mlflavors-dev

Install project locally:

$ python -m pip install --upgrade pip
$ pip install -e ".[dev,docs]"

Install pre-commit hooks:

$ pre-commit install

Run tests:

$ pytest tests/unit --cov

Build Sphinx docs:

$ cd docs
$ make html