Flegyas / latentis

A Python package for analyzing and transforming neural latent spaces.

🚧🚧🚧 Developers At Work. The library will be ready next month. Feel free to contact us directly for any specific request! 🚧🚧🚧

---

Latentis: Your Gateway to Latent Space Communication

Welcome to Latentis, the first-of-its-kind Python library dedicated to the innovative field of Latent Space Communication. Latentis is designed to empower researchers, data scientists, and enthusiasts to unlock new insights by providing a comprehensive suite of tools where latent spaces are the core ingredient.

Core Features

Latentis offers a structured suite of tools designed for efficiency and ease of use:

• Data Download & Processing: streamline the acquisition and preparation of complex datasets (via HuggingFace Datasets).
• Advanced Encoding: either employ pre-trained models or bring your own to encode anything.
• Benchmarking Tools: standard and customizable benchmarking tools, allowing for thorough evaluation and refinement of methods.

Getting Started

Ease into your next research project with:

pip install latentis

Development installation

Setup the development environment:

git clone [e-mail]:flegyas/latentis.git
cd latentis
conda env create -f env.yaml
conda activate latentis
pre-commit install

Run the tests:

pre-commit run --all-files
pytest -v

Update the dependencies

Re-install the project in edit mode:

pip install -e .[dev]