]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
CUDA: use mma FA kernel for gqa > 4 on RTX 4000 (#15035)
authorJohannes Gäßler <redacted>
Sat, 2 Aug 2025 14:37:08 +0000 (16:37 +0200)
committerGitHub <redacted>
Sat, 2 Aug 2025 14:37:08 +0000 (16:37 +0200)
commit03d46982180c2fb624bd2a233e46426ab22be5d1
tree152219d415af8c1b7e9dcd8106e793ec0db2f1f7
parent3303c19b1691088275ee864a823697177c94a15d
CUDA: use mma FA kernel for gqa > 4 on RTX 4000 (#15035)
ggml/src/ggml-cuda/fattn.cu