]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: 03c0946)
ggml : group all experts in a single ggml_mul_mat_id (#6505)
authorslaren <redacted>
Thu, 18 Apr 2024 13:18:48 +0000 (15:18 +0200)
committerGitHub <redacted>
Thu, 18 Apr 2024 13:18:48 +0000 (15:18 +0200)
commit0d56246f4b9764158525d894b96606f6163c53a8
tree43e57dfbbde67b701020fc3e2ac885e846925d26tree
parent03c0946d73c63ea73e1d85015b7088298443d438commit | diff
ggml : group all experts in a single ggml_mul_mat_id (#6505)

* ggml : group all experts in a single ggml_mul_mat_id
cuda : improve mmid row copy

* cuda : fix bin bcast with non-cont src0

* test-backend-ops : only run all mul mat tests for base types

* llama : disable moe offloading with SYCL

---------

Co-authored-by: Georgi Gerganov <redacted>
12 files changed:
examples/imatrix/imatrix.cpp diff | blob | history
ggml-cuda.cu diff | blob | history
ggml-cuda/binbcast.cu diff | blob | history
ggml-cuda/convert.cu diff | blob | history
ggml-metal.m diff | blob | history
ggml-metal.metal diff | blob | history
ggml-sycl.cpp diff | blob | history
ggml.c diff | blob | history
ggml.h diff | blob | history
llama.cpp diff | blob | history
scripts/compare-commits.sh diff | blob | history
tests/test-backend-ops.cpp diff | blob | history
Packaging of ggml-org/llama.cpp