3 ms·Fine-Tuning Language Models with Just Forward Passes3 points by famouswaffles 3y agobrucethemoose2 3y agoThis is huge. Its the breakthrough for finetuning (read: personalizing) LLMs on any machine thats capable of running GPU inference.