]> git.djapps.eu Git - pkg/ggml/sources/ggml/commit
overview / pkg / ggml / sources / ggml / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: 336c10a)
Update CUDA graph on scale change plus clear nodes/params (llama/9550)
authoragray3 <redacted>
Sat, 21 Sep 2024 00:41:07 +0000 (01:41 +0100)
committerGeorgi Gerganov <redacted>
Tue, 24 Sep 2024 10:04:37 +0000 (13:04 +0300)
commit78fcac234273b13d8fe45621c2c2f93ee0cc4898
tree7a4b64a26a59b79eae87ef3298fcaaaa59e7ed4ctree
parent336c10a4c3c8ec99af484b25a0cddd397a09cdb2commit | diff
Update CUDA graph on scale change plus clear nodes/params (llama/9550)

* Avoid using saved CUDA graph if scale changes and reset nodes/params on update

Fixes https://github.com/ggerganov/llama.cpp/issues/9451

* clear before resize
src/ggml-cuda.cu diff | blob | history
src/ggml-cuda/common.cuh diff | blob | history
Packaging of ggml-org/ggml