>to make the code useful you need data and that data might be free but not open sourced under any license To use something like Meta’s open sourced LLAMA2 mode

>to make the code useful you need data and that data might be free but not open sourced under any license To use something like Meta’s open sourced LLAMA2 model you don’t need the data. The model is self contained. It’s a compressed lossy form of all the data it was trained on. The weights allow you to continue its training with new data of your choosing.