Installation

Gunz-ML requires Python 3.11 or higher. It is designed to work seamlessly within the Pekora ecosystem but can be used as a standalone SDK for experiment tracking.

Standard Installation

You can install the library directly from the GitHub repository using pip:

pip install git+https://github.com/sXperfect/gunz-ml.git

Recommended: Mamba/Conda Environment

For development and research, we recommend using a dedicated mamba environment to manage heavy dependencies like PyTorch and MLflow:

# Create the environment
mamba create -n gunz-ml python=3.11

# Activate it
mamba activate gunz-ml

# Install the package in editable mode
pip install -e .

Optional Dependencies

The library uses optional “extras” to keep the base installation lightweight. Depending on your needs, you might want to install:

• [extras]: Full MLflow support with SQL backends.

• [all]: Includes all analysis and visualization tools.

pip install "gunz-ml[extras]"

System Requirements

If you are using the tracking features with a remote database (Juno), ensure you have the necessary system clients for MySQL:

• Ubuntu/Debian: sudo apt-get install libmysqlclient-dev

• macOS: brew install mysql-client