Todor Boinovski c01abcece2 hexagon: HMX flash-attention head_dim padding (support DK=DV=72) (llama/26539) ...

Allow HMX flash-attention to run with head_dim not a multiple of 64
(e.g. SigLIP head_dim=72), by operating on DK/DV rounded up to 64 with
zero-filled tail lanes.

2026-09-23 20:46:47 +03:00

..

cmake

CUDA: replace GGML_FA_ALL_QUANTS with GGML_FA_QUANTS, more control over what is compiled (llama/28079)

2026-09-14 20:45:06 +03:00

include

Fix function signature for ggml_backend_sycl_split_buffer_type (llama/28981)

2026-09-23 20:46:47 +03:00

src

hexagon: HMX flash-attention head_dim padding (support DK=DV=72) (llama/26539)

2026-09-23 20:46:47 +03:00

.gitignore

…

CMakeLists.txt

ggml : bump version to 0.24.0 (ggml/1627)

2026-09-14 20:45:06 +03:00