rwitz

/

go-bruins-v2

like 30

Thank you! What parameters are you using?

I also have the same problem, how to replace the output of <0x0A> with carriage return and line feed?

I'm not sure, but i guess they use GGUF version as I've seen that issue before in GGUF version of another model.

I used TheBlock's quantized GGUF version https://huggingface.co/TheBloke/go-bruins-v2-GGUF and LoneStriker's exl2 quantized version https://huggingface.co/LoneStriker/go-bruins-v2-8.0bpw-h8-exl2-2 has the same problem. It outputs <0x0A> instead of carriage return and line feed.

This issue should now be fixed in the GGUF. It was caused by lack of tokenizer.model, but I've sourced the tokenizer.model from Mistral 7B and re-generated the GGUFs

@rwitz please include tokenizer.model in future uploads, thanks