]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
model: support youtu-vl model (#18479)
authortt <redacted>
Thu, 1 Jan 2026 18:25:54 +0000 (02:25 +0800)
committerGitHub <redacted>
Thu, 1 Jan 2026 18:25:54 +0000 (19:25 +0100)
commitced765be44ce173c374f295b3c6f4175f8fd109b
tree4f1bcc315586ae08c7a885e5173447fa952d3cd4
parent3ccccc83f79c187e7fd3e19bdc5bf4efe8928697
model: support youtu-vl model (#18479)

* Support Youtu-VL Model

* merge code

* fix bug

* revert qwen2 code & support rsplit in minja.hpp

* update warm info

* fix annotation

* u

* revert minja.hpp

* fix

* Do not write routed_scaling_factor to gguf when routed_scaling_factor is None

* fix expert_weights_scale

* LGTM after whitespace fixes

* fix

* fix

* fix

* layers to layer_index

* enum fix

---------

Co-authored-by: Xuan-Son Nguyen <redacted>
Co-authored-by: Sigbjørn Skjæret <redacted>
17 files changed:
convert_hf_to_gguf.py
convert_hf_to_gguf_update.py
gguf-py/gguf/constants.py
gguf-py/gguf/gguf_writer.py
gguf-py/gguf/tensor_mapping.py
src/llama-model.cpp
src/llama-vocab.cpp
src/llama-vocab.h
src/models/deepseek2.cpp
src/unicode.cpp
tools/mtmd/CMakeLists.txt
tools/mtmd/clip-impl.h
tools/mtmd/clip-model.h
tools/mtmd/clip.cpp
tools/mtmd/models/models.h
tools/mtmd/models/youtuvl.cpp [new file with mode: 0644]
tools/mtmd/mtmd.cpp