IntelLabs/RAG-FiT
A library for fine-tuning LLMs on retrieval-augmented generation datasets using parameter-efficient techniques.
Velocity · 7d
+1.1
★ / day
Trend
→steady
star history
RAG-FiT enhances LLMs to better leverage external information by fine-tuning on specially created RAG-augmented datasets. The framework provides modular workflows for dataset creation, training via parameter-efficient fine-tuning (PEFT), inference, and evaluation using RAG-specific metrics. Workflows are customizable through configuration files and support integration with tools like Haystack and DeepEval.