Where can I find a compatible draft model for Llama-3_1-Nemotron-51B-Instruct?

[mashdragon]

I'm struggling to find a compatible model for this quantized model family.

Original model: https://huggingface.co/nvidia/Llama-3_1-Nemotron-51B-Instruct

Draft models I've tried:

• https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-GGUF
• https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF
• https://huggingface.co/bartowski/Meta-Llama-3.1-8B-Instruct-GGUF

Mismatch error from llama.cpp:

common_speculative_are_compatible: draft model special tokens must match target model to use speculation
common_speculative_are_compatible: tgt: bos = 128000 (1), eos = 128001 (0)
common_speculative_are_compatible: dft: bos = 128000 (1), eos = 128009 (0)
srv    load_model: the draft model 'Llama-3.2-3B-Instruct-uncensored-Q5_K_M.gguf' is not compatible with the target model 'Llama-3_1-Nemotron-51B-Instruct-Q5_K_M.gguf'

A pointer to a compatible smaller model, or some instructions on how to distill the Nemotron one to create my own compatible model, would be much appreciated.

[mashdragon]

This was a GGUF config issue that appears to be fixed with #11008