Genji V2 (NovelAI Legacy Model)

Genji V2 is our second Japanese finetune of EleutherAI's GPT-J 6B model. The main difference from the first version of Genji is that the tokenizer was modified to replace rarely used tokens with new Japanese tokens. We have previously released this modified GPT-2 tokenizer on GitHub, but the original version produces different results, depending on whether it is used with Huggingface's slow or fast tokenizer implementation. This release of the model includes a fixed version of the tokenizer files.

Our Genji models never saw too much use, and at this point it won't be long before the model will be retired and Genji will get to enjoy a peaceful retirement.

So, in anticipation of this, for the sake of nostalgia, posterity, and historic preservation, we are releasing the weights of our Genji model publicly on Huggingface Hub under the GPL-3.0 license.

Genji has been converted from our internal format to Hugginface Safetensors for ease of use. Model files in GGUF format are provided as well.

You can find your very own Genji right here where you are!