Skip to content

Getting Started

Follow these pages in order for a fresh setup:

  1. Installation — clone, set up the Python environment with uv, and install optional extras.
  2. HuggingFace Authentication — log in to the HuggingFace Hub for gated datasets such as FineWeb-2.
  3. Project Structure — orientation to the repository layout: where library code, scripts, configs, and SLURM batch files live.

Once the environment is ready, head to the User Guide to run the data pipeline, train tokenizers, and evaluate models.