caliber 0.1.16

Model-agnostic calibration and performance enhancement.

Caliber is a toolkit designed to enhance the calibration and overall performance of your machine learning models. It operates independently of your specific models, offering a flexible and agnostic approach. Caliber’s methods are lightweight and generally fast, making them easy to integrate as add-ons within your existing pipeline with minimal effort.

Installation

You can install caliber via

pip install caliber

The following extras are available:

• torch: install PyTorch dependencies.

• docs: install Sphinx dependencies.

Alternatively, you can build the package using Poetry.

License

This project is licensed under the Apache-2.0 License.