]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: c749cb0)
server: accept null sampling params (#25538)
authorPascal <redacted>
Fri, 10 Jul 2026 20:07:29 +0000 (22:07 +0200)
committerGitHub <redacted>
Fri, 10 Jul 2026 20:07:29 +0000 (22:07 +0200)
commit4f37f519722aa3242eecb7649466b4a4a2d6d6da
treec14a7da5e9e7e6ddad3707576cdfd9d68f65657ctree
parentc749cb041706647f460bb918cccc9d91995205abcommit | diff
server: accept null sampling params (#25538)

* server: accept null sampling params

Extend the schema validation to treat a null value as absent, so
clients can send null on nullable params (temperature, top_p, ...)
to request the server default. This matches the OpenAI spec and the
json_value convention used elsewhere.

Add has_field() to skip null in the field eval guards.

* has_field -> has_value​
tools/server/server-schema.cpp diff | blob | history
Packaging of ggml-org/llama.cpp