]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: ec0dbef)
convert_hf_to_gguf: support split MTP export for HY V3 (#25641)
authorThiago Padilha <redacted>
Tue, 14 Jul 2026 09:43:15 +0000 (06:43 -0300)
committerGitHub <redacted>
Tue, 14 Jul 2026 09:43:15 +0000 (11:43 +0200)
commitcb489bc0fb789c2cb7a9cc9dc44fa71893fe0988
tree6bcbc2afd2365f0c7f840992f3a6afb0260e8a8ftree
parentec0dbef81632bb4233971fe83d91cf0accbfd882commit | diff
convert_hf_to_gguf: support split MTP export for HY V3 (#25641)

- Add a supports_mtp_export capability to ModelBase so architectures can opt
  into --mtp and --no-mtp without extending a central class allowlist.
- Enable the capability for the existing Qwen3.5/3.6 and Step3.5/3.7
  implementations, and for HY V3, whose converter already supports
  filtering the appended MTP layers.
conversion/base.py diff | blob | history
conversion/hunyuan.py diff | blob | history
conversion/qwen.py diff | blob | history
conversion/step3.py diff | blob | history
convert_hf_to_gguf.py diff | blob | history
Packaging of ggml-org/llama.cpp