Confusion about the models

[1337sml]

Thanks for this great project! I have a few issues with the models though.

It is rather confusing that you get an error message regarding a missing OpenAI token if you try the setup guide. For me the problem was that I was not logged into huggingface and that triggered the except part which tries to use OpenAI. Maybe mentioning huggingface as the modelprovider would be good.

The hardware requirements guide lists several models https://github.com/hitomi-team/sukima/wiki/Hardware-Requirements but does not mention the actual model names.

[pw45000]

For usage, check out my guide. The API key error is a red herring, which can be telling of a multitude of issues. Suitable models would be Diablo-GPT-Large or using Hourai. So, for an example, you can load DiabloGPT model by switching the value of the "model" field as "microsoft/DialoGPT-large".