Installation

fluke is available on PyPI.

To install fluke, you can use pip:

pip install fluke-fl

When installing fluke, the following dependencies are installed:

• numpy: for numerical operations;

• pandas: for handling data;

• scikit-learn: for specific functionalities, like Kernel Density Estimation;

• torch: for managing tensors and neural networks;

• torchmetrics: for metrics computation;

• torchvision: for datasets and models;

• rich: for rich console output;

• typer: for command-line interface;

• wandb: for experiment tracking;

• ‘datasets’: for datasets loading;

• psutil: for system monitoring;

• tensorboard: for experiment tracking;

• clearml: for experiment tracking;

• matplotlib: for plotting;

• seaborn: for plotting;

• opacus: for differential privacy;

• diskcache: for caching;

• cerberus: for data validation;

• hydra-core: for configuration management.

🐳 Docker

You can use also use fluke directly inside a Docker container — no installation needed on your local machine. To do this, you can use the provided Dockerfile. First, clone the repository:

git clone https://github.com/makgyver/fluke.git
cd fluke

Then, build the Docker image:

docker build -t fluke_container .

Then, you can run an interactive session with

docker run --rm fluke_container fluke [ARGS]

where ARGS are the arguments you want to pass to the fluke command as described in the fluke CLI section.