LLMScratchs
PublicThis repository contains a comprehensive implementation of the full lifecycle of Large Language Model (LLM大模型) development and application, covering Pretraining (预训练), Supervised Fine-Tuning (SFT全参), Direct Preference Optimization (DPO偏好优化), Inference(推理) and Deployment. It supports frameworks such as PyTorch, HuggingFace, and DeepSpeed.