sklearn-onnx

Project Introduction: sklearn-onnx

Overview

sklearn-onnx is a powerful tool designed to convert machine learning models built with scikit-learn into the ONNX format. This conversion is crucial for leveraging tools like ONNX Runtime, which can significantly enhance the performance of model scoring. The project is robust, with converters regularly tested using onnxruntime, ensuring compatibility and reliability. Additionally, sklearn-onnx can incorporate external converters to seamlessly convert pipelines that include models or transformers from external libraries.

Documentation

Comprehensive documentation, including tutorials, can be accessed at this site. Here, users can find detailed guides and resources to support their learning and application of the library. Notably, a list of supported scikit-learn models is available here, with the last supported opset being version 21. For issues or queries, the community can browse through existing issues or submit new ones for resolution.

Installation

Users can seamlessly integrate sklearn-onnx into their projects by installing it directly from PyPi using the following command:

pip install skl2onnx

Alternatively, for those who wish to explore the latest updates and features, installing directly from the source is possible:

pip install git+https://github.com/onnx/sklearn-onnx.git

Getting Started

To illustrate how sklearn-onnx can be utilized, here is a quick example:

1. Training a Model:

   First, we train a RandomForestClassifier on the well-known Iris dataset:

   import numpy as np
   from sklearn.datasets import load_iris
   from sklearn.model_selection import train_test_split
   from sklearn.ensemble import RandomForestClassifier
   
   iris = load_iris()
   X, y = iris.data, iris.target
   X = X.astype(np.float32)
   X_train, X_test, y_train, y_test = train_test_split(X, y)
   clr = RandomForestClassifier()
   clr.fit(X_train, y_train)
   

2. Converting to ONNX Format:

   Using the to_onnx function from skl2onnx, the trained model can be converted to ONNX format:

   from skl2onnx import to_onnx
   
   onx = to_onnx(clr, X[:1])
   with open("rf_iris.onnx", "wb") as f:
       f.write(onx.SerializeToString())
   

3. Making Predictions with ONNX Runtime:

   An inference session is established with onnxruntime to make predictions:

   import onnxruntime as rt
   
   sess = rt.InferenceSession("rf_iris.onnx", providers=["CPUExecutionProvider"])
   input_name = sess.get_inputs()[0].name
   label_name = sess.get_outputs()[0].name
   pred_onx = sess.run([label_name], {input_name: X_test.astype(np.float32)})[0]
   

Contributing

Sklearn-onnx encourages contributions from the community in various forms, such as feedback, ideas, or code enhancements. This collaborative approach aims to continuously improve the project and its usability.

License

The project is released under the Apache License v2.0, ensuring open-source accessibility and collaboration.

In summary, sklearn-onnx serves as a bridge between scikit-learn and ONNX, enhancing the usability and performance of machine learning models across different environments.