GGUFs eventually ?

#1
by HMasaki - opened

I like to use ollama and it's easiest to set them up when they're GGUFs ideally Q8 is there a plan to eventually do some GGUFs of this one?

I tried converting it, but something seems to be wrong. I'm trying to solve the problem.

The BPE pre-tokenizer was not recognized!
There are 2 possible reasons for this:

  • the model has not been added to convert_hf_to_gguf_update.py yet
    - the pre-tokenization config has changed upstream
    Check your model files and convert_hf_to_gguf_update.py and update them accordingly.

https://github.com/ggerganov/llama.cpp/pull/6920

Use with ollama

You can use huihui_ai/falcon3-abliterated directly, It's possible for the output to be empty.

ollama run huihui_ai/falcon3-abliterated:10b

That works thank you

Using gguf by bartowski. Indeed empty output (not always). Looks like the problem is q4_0.gguf

It sometimes appears, and it goes away with a try.

Sign up or log in to comment