This is a minimalist and low-cost practical project for building a large language model (LLM) similar to ChatGPT. Developers only need 8 H100 graphics cards to complete the entire process from data processing, pre-training, fine-tuning, evaluation to inference. It also comes with a ChatGPT-style Web UI for online experience, which is suitable for self-learners to quickly get started and deeply understand the principles of large models.