Takeaways from hundreds of LLM finetuning experiments with LoRA

Takeaways from hundreds of LLM finetuning experiments with LoRA. 39 comments, 258 points on Hacker News.