You can finetune a 7B Llama 2 model on a single GPU A100 for example, so you could even use Google Colab. But in general, most universities and companies are supportive of open source these days.

→ View original post on X — @rasbt,