ML Experiment Migration Tool

This tool automates the migration of machine learning (ML) experiment data from Weights & Biases (W&B) to DagsHub, leveraging MLflow for a seamless and unified experiment tracking experience. It enables researchers and ML engineers to efficiently transfer run data, including metrics, parameters, and artifacts, ensuring collaborative, transparent, and scalable project management.

Features

• Automatic Migration: Migrate runs, including their metrics, parameters, and artifacts, from W&B to DagsHub.
• Unified Tracking: Use MLflow to log experiments consistently across platforms.
• Collaboration and Version Control: Facilitates project sharing, collaboration, and version control.

Requirements

Before running the project, ensure you have Python 3.8 or later installed. You'll also need to install the necessary Python packages:

pip install -r requirements.txt

This will install MLflow, DagsHub, Weights & Biases, tqdm, and other dependencies required for the script to run.

How to Run

1. Set up your environment: Make sure you have access tokens or credentials set up for both Weights & Biases and DagsHub.

2. Prepare your command: The script accepts several command-line arguments to specify the source (Weights & Biases) and destination (DagsHub) for the migration:

   • wb_owner: Owner of the W&B project.
   • wb_project: Name of the W&B project.
   • dh_owner: Owner of the DagsHub repository.
   • dh_repo: Name of the DagsHub repository.
   • --run_id (optional): Specific run ID from W&B to migrate. If not provided, all runs from the specified W&B project will be migrated.

3. Execute the script: Run the script from your command line with the necessary arguments. Here's an example command:

   python import_from_wb.py <wb_owner> <wb_project> <dh_owner> <dh_repo> --run_id <optional_run_id>
   

   Replace <wb_owner>, <wb_project>, <dh_owner>, <dh_repo>, and <optional_run_id> with your specific details.

Contributing

We welcome contributions to this project! If you have suggestions for improvements or encounter any issues, please open an issue or submit a pull request.