This project serves as a template to learn how to integrate GitHub Actions and GitHub API using ghapi, along with the Weights & Biases (WandB) API, to create a CI/CD workflow for machine learning projects.

The project focuses on creating a streamlined process for running experiments and tracking results using GitHub Actions and WandB. This setup automates the workflow by enabling experiment tracking directly from pull requests, simplifying collaboration and monitoring for machine learning models.

This project is the final submission for the CI/CD for Machine Learning (GitOps) course by Weights & Biases.

• ChatOps Integration: Listens for the comment "/wandb <run id>" on a pull request to trigger a comparison workflow.
• Baseline Fetching: Fetches the run tagged as “baseline” from your WandB project using the WandB API.
• Comparison Report: Automatically generates a report comparing the baseline run with the provided run ID.
• PR Commenting: Posts a comment on the pull request with the URL to the comparison report, making it easy to access.

First, install the required dependencies by running the following command:

pip install -r requirements.txt

You need to set the following WandB environment variables to integrate the WandB API:

• WANDB_ENTITY
• WANDB_PROJECT
• WANDB_API_KEY

python main.py

To integrate the workflow with GitHub Actions, first, create the following repository secrets on GitHub for your project:

• WANDB_ENTITY
• WANDB_PROJECT
• WANDB_API_KEY

These secrets will be used to communicate with WandB during the workflow execution.

When you create a pull request, you can trigger an experiment run on WandB by commenting on the PR with the following command:

/wandb experiment-run-id

This will trigger the workflow, and you’ll see a workspace on your WandB dashboard where you can monitor the experiment's progress.

This project is licensed under the MIT License.