Skip to content

matheus695p/explainability-ml-models

Repository files navigation

Machine learning model explainer

Overview

Welcome to the ml_explainer project, a tool for generating explainability reports for machine learning models. This Kedro project was inspired by the spaceflights tutorial and was generated using Kedro 0.18.14.

Project Objective

The primary goal of this project is to create an ML package that can automatically generate comprehensive explanability reports for a wide range of machine learning models. These reports follow a specific template and utilize GenAI to provide detailed insights into the decision-making processes of the models.

Key Features

  • Explainability Reports: Generate detailed and informative reports that explain the inner workings of machine learning models.

  • Template-Based: Follow a predefined template to ensure consistency and clarity in the generated reports.

  • GenAI Integration: Leverage GenAI to enhance the explanations by providing insights and context from a vast knowledge base.

Example of Explanability report generated using GenAI

Interpretability Report Output Example

Getting Started

To get started with this project, follow these steps:

  1. Clone the repository to your local machine.
  2. Create a .env file in the project's root directory.
  3. Add your OpenAI API keys to the .env file as follows:
OPENAI_API_BASE="YOUR_OPENAI_API_BASE"
OPENAI_API_KEY="YOUR_OPENAI_API_KEY_HERE"

How to Install Dependencies

Declare project dependencies in the following files:

  • src/requirements.txt: For pip installation.
  • src/environment.yml: For conda installation.

To install dependencies, follow the instructions below:

Using pip

Navigate to the project directory and run the following command:

pip install -r src/requirements.txt

Generate a GenAI interpretability report

kedro run

Specifically to the report:

kedro run --pipeline explainer --nodes=explainability_report

About

Machine learning explainability reports with GenAI

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages