]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
mtmd, model : merge HunyuanOCR into HunyuanVL and fix OCR vision precision (#23329)
authorwendadawen <redacted>
Wed, 20 May 2026 22:35:37 +0000 (06:35 +0800)
committerGitHub <redacted>
Wed, 20 May 2026 22:35:37 +0000 (00:35 +0200)
commit6a257d44633d4a752183ed778b88d2924d0a6b9d
treea5beb5d6663d90063f27c2c0eacb09090c9b5cc2
parent3a479c9132072815cb70a443b4efa45bb66b3f59
mtmd, model : merge HunyuanOCR into HunyuanVL and fix OCR vision precision (#23329)

- HunyuanOCR shares the same HF arch and vision layout as HunyuanVL butwas split into a separate path that skipped the +0.1 bilinear sampler used by the HF reference.
- Collapse OCR into the HUNYUANVL projector + HUNYUAN_VL text arch
16 files changed:
conversion/hunyuan.py
gguf-py/gguf/constants.py
gguf-py/gguf/tensor_mapping.py
src/llama-chat.cpp
src/llama-chat.h
tools/cli/README.md
tools/completion/README.md
tools/mtmd/CMakeLists.txt
tools/mtmd/clip-impl.h
tools/mtmd/clip-model.h
tools/mtmd/clip.cpp
tools/mtmd/models/hunyuanocr.cpp [deleted file]
tools/mtmd/models/hunyuanvl.cpp [new file with mode: 0644]
tools/mtmd/models/models.h
tools/mtmd/mtmd.cpp
tools/server/README.md