]> git.djapps.eu Git - pkg/ggml/sources/ggml/commit
overview / pkg / ggml / sources / ggml / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: c2b13bb)
Fix WARP_SIZE=16 bug of Intel GPU (llama/8266)
authorluoyu-intel <redacted>
Fri, 5 Jul 2024 05:06:13 +0000 (05:06 +0000)
committerGeorgi Gerganov <redacted>
Mon, 8 Jul 2024 10:03:28 +0000 (13:03 +0300)
commitea98f39dedcd0391c75ca4f1c5a35d5becea8dbc
treebac4c2ee2c413c255a81e21931c0f5aee037a3c1tree
parentc2b13bbef9ab16b96f8a0e609b28e5b67b547202commit | diff
Fix WARP_SIZE=16 bug of Intel GPU (llama/8266)

* fix group_norm ut

* split softmax

* fix softmax

* add concat support condition

* revert debug code

* move QK_WARP_SIZE to presets.hpp
src/CMakeLists.txt diff | blob | history
src/ggml-sycl.cpp diff | blob | history
src/ggml-sycl/backend.hpp diff | blob | history
src/ggml-sycl/dmmv.cpp diff | blob | history
src/ggml-sycl/norm.cpp diff | blob | history
src/ggml-sycl/presets.hpp diff | blob | history
src/ggml-sycl/softmax.cpp [new file with mode: 0644] blob
src/ggml-sycl/softmax.hpp [new file with mode: 0644] blob
Packaging of ggml-org/ggml