]> git.djapps.eu Git - pkg/ggml/sources/whisper.cpp/commit
CUDA: MoE helper in device code, better tile sizes (llama/15525)
authorJohannes Gäßler <redacted>
Mon, 25 Aug 2025 15:23:40 +0000 (17:23 +0200)
committerGeorgi Gerganov <redacted>
Sat, 20 Sep 2025 10:42:41 +0000 (13:42 +0300)
commit1e856b2919f9e30013359bed7feb2eaf330017b7
tree1e31b6c030b6581eeadebd335d540bda4e6aa883
parent54be54f4cef4b4e77128763499e628c8bf1f6a1e
CUDA: MoE helper in device code, better tile sizes (llama/15525)

* CUDA: MoE helper in device code, better tile sizes

* reduce superfluous CUDA blocks
ggml/src/ggml-cuda/common.cuh
ggml/src/ggml-cuda/mmq.cu
ggml/src/ggml-cuda/mmq.cuh
ggml/src/ggml-cuda/vendors/hip.h