FT-Lab is a clean, reproducible setup for fine-tuning TinyLlama (supports Full FT, LoRA, and QLoRA) and evaluating Retrieval-Augmented Generation (RAG) pipelines using LlamaIndex and LangChain.
Built for small GPUs and designed with controlled experiments and ablation studies in mind.
Feedback and contributions welcome!