]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: 379ac66)
convert : fix conversion for Mistral-Medium-3.5-128B (#24268)
authorDavid Friehs <redacted>
Sun, 7 Jun 2026 19:41:39 +0000 (21:41 +0200)
committerGitHub <redacted>
Sun, 7 Jun 2026 19:41:39 +0000 (21:41 +0200)
commit8a963fc10ee005b3f425d14bcd427eef5c57f157
tree666f2a4acbf827fdf68339f043e0c6b213515ae0tree
parent379ac6673b5cd75c7b4e07d1521c50f1e093878ccommit | diff
convert : fix conversion for Mistral-Medium-3.5-128B (#24268)

Mistral explicitly sets `moe` and `llama_4_scaling` to `null` in
params.json, breaking `key in dict` checks during conversion. Replace
with `dict.get(key) is not None` where this matters.

Fixes `convert-hf-to-gguf.py --mistral-format Mistral-Medium-3.5-128B`
conversion/mistral.py diff | blob | history
convert_hf_to_gguf.py diff | blob | history
Packaging of ggml-org/llama.cpp