Comment by gliched_robot

Disagree on Nvidia, most folks fine-tune model. Proof: there are about 20k models in huggingface derived from llama 2, all of them trained on Nvidia GPUs.