Finetune Llama 3, Mistral & Gemma LLMs 2-5x faster with 80% less memory - View it on GitHub