Adding NVMe SSDs to Enable and Accelerate 100B Model Fine-Tuning on a Single GPU

I also submitted a link to the post on Hugging Face: https://news.ycombinator.com/item?id=39691238