GitHub: https://github.com/pfekin/LARA I've been working on LARA (Lightweight Additive Residual Adaptation), a research project on making post-training modular for frozen language models. I've also developed a small PyTorch library that implements it. The main idea is to train a low-rank residual ad
