]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
llama: Add option to merge gate and exp weights (#19139)
authorAman Gupta <redacted>
Thu, 26 Feb 2026 13:01:08 +0000 (21:01 +0800)
committerGitHub <redacted>
Thu, 26 Feb 2026 13:01:08 +0000 (21:01 +0800)
commitb68d75165ad37ba1256cc45a43ec4f51cf813c3e
tree3c7541bf60e9579ba1af1aa60d6961c1091b9a1a
parentffaafde16ffebd2853467f0dd833625a726ce08e
llama: Add option to merge gate and exp weights (#19139)

* llama: Add option to merge gate and exp weights

* Update convert_hf_to_gguf.py

Co-authored-by: Sigbjørn Skjæret <redacted>
* Update convert_hf_to_gguf.py

Co-authored-by: Sigbjørn Skjæret <redacted>
* update constants.py

* add gate_up for the all MoE models

* convert: simplify merge tensor condition

* update constants.py

* reduce number of models, add create_tensor_gate_up helper

---------

Co-authored-by: Sigbjørn Skjæret <redacted>
12 files changed:
convert_hf_to_gguf.py
gguf-py/gguf/constants.py
gguf-py/gguf/tensor_mapping.py
src/llama-arch.cpp
src/llama-arch.h
src/llama-graph.cpp
src/llama-graph.h
src/llama-model.cpp
src/llama-model.h
src/models/deepseek2.cpp
src/models/qwen35moe.cpp
src/models/qwen3next.cpp