llama-graph: fix for MLA with FA causing extra overhead for small batches#14198
Closed
jukofyork wants to merge 1 commit intoggml-org:masterfrom
jukofyork:mla-fa-performance-jumps--fix
Closed
llama-graph: fix for MLA with FA causing extra overhead for small batches#14198jukofyork wants to merge 1 commit intoggml-org:masterfrom jukofyork:mla-fa-performance-jumps--fix
jukofyork wants to merge 1 commit intoggml-org:masterfrom
jukofyork:mla-fa-performance-jumps--fix