# Multi-Autoresearch

This repository is organized as three separate projects:

- `pre-training/` contains the existing Autolab pre-training research project.
- `post-training/` contains a compact single-model nano chat post-training
  benchmark.
- `inference/` contains a simpler llama.cpp local inference optimization
  setup focused on getting a GGUF model to run as fast as possible.

Run commands from the relevant project directory so project-local configuration,
scripts, and paths resolve correctly.