This repo goes over how to use the GPT-J-6B model weights with huggingface. Originally the model was not officially supported by huggingface but the weights could be still used with huggingface through other methods. The model is now officially supported, much easier to use and the notebook has been changed to reflect that.
See original_youtube branch for old method
This repos old method goes along with the Youtube video seen here
This repo is forked from paulcjh and has a few tweaks and removals to make the process a bit more clear and straightforward.