mira-score 0.1.7

MIRA: A Score for Conditional Distribution Accuracy and Model Comparison

MIRA: A Score for Conditional Distribution Accuracy and Model Comparison

This repository contains the MIRA score implementation.

Overview

MIRA is a score for evaluating conditional distribution accuracy and comparing models. This package provides tools to compute MIRA scores and includes a demonstration notebook showing baseline comparisons.

Installation Instructions

Follow these steps to set up the environment and run the examples:

1. Create and Activate a Conda Environment

conda create -n mira_env python=3.8
conda activate mira_env

2. Install the Package

pip install mira_score

Or, to install from source in editable mode:

git clone https://github.com/SammyS15/mira-score.git
cd mira-score
pip install -e .

This will install the mira_score package along with all required dependencies.

4. Verify Installation

Test that the package is correctly installed:

python3 -c "from mira_score import mira, mira_bootstrap, get_device; print('MIRA successfully installed!')"