The openai gpt-oss models, and the z-lab gpt-oss DFlash drafts that reuse their
tokenizer, ship the o200k_harmony BPE tokenizer, which shares the GPT-4o
pre-tokenizer regex already handled by LLAMA_VOCAB_PRE_TYPE_GPT4O. Its checksum
was unrecognized, so get_vocab_base_pre() fell through to the "BPE pre-tokenizer
was not recognized" warning and conversion failed.
Register the gpt-oss-20b tokenizer hash against the existing "gpt-4o" pre-type
in convert_hf_to_gguf_update.py and add the matching entry in
get_vocab_base_pre(), following the dual-location pattern already used for the
z-lab Qwen3.5 DFlash drafts.
Co-authored-by: Joel Farthing <262452229+joelfarthing@users.noreply.github.com>
* convert_hf_to_gguf for Kimi-K2-Instruct
Adapt mainline `PR14653` for tokenizer while maintaining proper MLA
tensors. Tested with this workflow using deepseek fp8_cast_bf16.py and
triton-cpu to upcast the fp8 safetensors to bf16 safetensors then used
this convert_hf_to_gguf.
* Add Kimi-K2 chat template
moonshotai/Kimi-K2-Instruct
https://github.com/ikawrakow/ik_llama.cpp/pull/609#issuecomment-3071259454
* kimi-k2 add ass to template to get response
* Merging mainline - WIP
* Merging mainline - WIP
AVX2 and CUDA appear to work.
CUDA performance seems slightly (~1-2%) lower as it is so often
the case with llama.cpp/ggml after some "improvements" have been made.
* Merging mainline - fix Metal
* Remove check
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Joel Farthing
SamuelOliveirads
firecoperana
Thireus ☠
ubergarm
Fizz~
saood06
Kawrakow