datawhalechina/self-llm

Provides comprehensive tutorials covering 50+ open-source LLMs with model-specific environment configurations, deployment strategies (CLI, web demos, LangChain integration), and multiple fine-tuning approaches including distributed full-parameter training, LoRA, and P-tuning. Targets resource-constrained users by emphasizing local deployment and private domain adaptation without API dependencies, supporting both NVIDIA GPUs and specialized accelerators like AMD and Huawei Ascend NPUs.

28,927 stars. Actively maintained with 3 commits in the last 30 days.