GGUF draft in BF16, Q8_0 and Q4_K_M formats for speculative decoding with Bielik-11B-v3.1-Instruct. This draft is intended to run alongside the target model.
Use it with the Bielik-11B-v3.1-Instruct GGUF in llama.cpp:
llama-server -m Bielik-11B-v3.1-Instruct.Q4_K_M.gguf \
-md Bielik-11B-v3.1-DFlash.Q4_K_M.gguf \
--spec-type draft-dflash --spec-draft-n-max 15 --jinja
Available files: Q4_K_M, Q8_0, and bf16.
See the source draft model for architecture and training details.