GGUF conversion issue

#1
by shigurd - opened

Hi, im having issues with converting this model into GGUF. Im using llama.ccp convert_hf_to_gguf_update.py. Initially got an error due to "- no" in the readme file, but solved it by modifying it (I can see that this has been patched now).

I was able to create the GGUF files, but they end up outputting gibberish (for instance: 122222223555555555555555555541111111111111111111111111111111) for both f32, f16 and Q8_0 when used in ollama. I have tested the llama.ccp script for GGUF conversion of other llama3.2-3B models without any issues like this.

Also tested the model in HF transformers which otherwise works perfectly fine. Anyone have any advice on what im doing wrong?

Sign up or log in to comment