Prerequisites¶
Welcome! Before you install the Emmi Framework, let’s make sure your system is ready. This checklist will help make the installation process smooth and successful.
System Requirements¶
Operating System¶
The framework is primarily developed and tested on Linux (Ubuntu 24.04+) and macOS (14.0+).
Warning
Windows is not officially supported at this time. While it may be possible to run the framework, you may encounter issues, and we cannot provide support.
Python¶
A 64-bit installation of Python 3.10 or newer is required. We recommend to use 3.12. You can download it from the official Python website.
Required Tools¶
Git: Used to clone the framework’s source code repository. You can install it from git-scm.com.
uv: Our recommended package manager for installing Python dependencies. Please install it by following the official uv documentation.
Hardware Requirements¶
CPU/RAM: A modern multi-core CPU and at least 16 GB of RAM are recommended for a good experience.
GPU (Highly Recommended): For training models, an NVIDIA GPU with CUDA support is essential. For reference, our internal cluster uses H100 GPUs, but it is possible to train models on a smaller GPUs.
Important
Check Your CUDA Version
Our framework depends on PyTorch, which requires a specific CUDA version.
Check your NVIDIA driver’s CUDA capability by running this command in your terminal:
nvidia-smiNote the CUDA Version shown in the top right.
Compare this to the PyTorch version specified in the
pyproject.tomlfile(s) to ensure compatibility.
Optional¶
GitHub Account: A free account at github.com is useful for reporting issues or communicating with the developers.