pycalibrate icon indicating copy to clipboard operation
pycalibrate copied to clipboard

pycalibrate is a Python library to visually analyze model calibration in Jupyter Notebooks

Downloads Open In Colab MIT license

pycalibrate

Pycalibrate is a Python package that allows access to the Calibrate tool. Calibrate is a visual analytics tool used to analyze model calibration in Jupyter Notebooks. Below, we show what Calibrate looks like in practice.

Installation

Installing pycalibrate is easy. Simply run:

pip install pycalibrate

You can also use pycalibrate on Colab, by copying our Example Colab Notebook.

You can also pip install directly from the repository by running pip install git+https://github.com/VIDA-NYU/pycalibrate

Usage

One can pycalibrate in just a few lines of code:

from pycalibrate import Calibrate

c = Calibrate(data=dataset) # `dataset` must be a Pandas dataframe

c.add_model(y_preds, y_labels, "ModelName") 
# y_preds is an n x k matrix of predictions
# y_labels is an n x k matrix of one-hot encoded labels

c.visualize() # Voila! 

Calibrate Tool

System screen

Need Help?

Need help? Open up an issue.