]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: edb18b6)
Force FP32 compute in GLM4 FFN Down (#13101)
authorCity <redacted>
Fri, 25 Apr 2025 12:38:34 +0000 (14:38 +0200)
committerGitHub <redacted>
Fri, 25 Apr 2025 12:38:34 +0000 (14:38 +0200)
commit558a764713468f26f5a163d25a22100c9a04a48f
tree2f2d9fe9e55f931044fae069efb17e83c30dd55atree
parentedb18b6e8f5ea6509ad43057f8bb98fc557dbc4ecommit | diff
Force FP32 compute in GLM4 FFN Down (#13101)

* Force FP32 compute in cuBLAS GEMM

* Revert "Force FP32 compute in cuBLAS GEMM"

This reverts commit 6efd872732159ab88ee7b3c1d77ba5ebc83079bd.

* Force F32 compute in GLM4 ffn down

* Edit comment to clarify issue

Co-authored-by: Johannes Gäßler <redacted>
---------

Co-authored-by: Johannes Gäßler <redacted>
src/llama-graph.cpp diff | blob | history
Packaging of ggml-org/llama.cpp