]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: 7ea23dd)
mtp: use inp_out_ids for skipping logit computation (#23433)
authorAman Gupta <redacted>
Thu, 21 May 2026 07:23:14 +0000 (15:23 +0800)
committerGitHub <redacted>
Thu, 21 May 2026 07:23:14 +0000 (15:23 +0800)
commit12e5d99078af2feaab0f62f69d230529d66ac623
tree0d92e0ef77343f1d6c0b0d374290e4621ce26bc2tree
parent7ea23ddf7b7cb57e37275774f2d4b718db5c5c26commit | diff
mtp: use inp_out_ids for skipping logit computation (#23433)

when doing a follow-up decode for the draft model, we were always doing the logit computation even though it is not required.
src/models/qwen35.cpp diff | blob | history
src/models/qwen35moe.cpp diff | blob | history
Packaging of ggml-org/llama.cpp