Public release of LoTR, a novel approach for parameter-efficient fine-tuning of LLMs which represents a gradient update to parameters in a form of tensor decomposition.
arXiv:2402.01376
- Python Published by daskol over 2 years ago
