From dcc09d25961c5d0626bc148e558ee841141748f7 Mon Sep 17 00:00:00 2001 From: Georgi Gerganov Date: Mon, 9 Oct 2023 14:28:27 +0300 Subject: [PATCH] metal : do not use mul_mm kernels when ne00 < 64 (#3542) --- ggml-metal.m | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/ggml-metal.m b/ggml-metal.m index 29cb3c922..7d67db90f 100644 --- a/ggml-metal.m +++ b/ggml-metal.m @@ -1040,7 +1040,7 @@ void ggml_metal_graph_compute( !ggml_is_transposed(src0) && !ggml_is_transposed(src1) && src1t == GGML_TYPE_F32 && - ne00 % 32 == 0 && + ne00 % 32 == 0 && ne00 >= 64 && ne11 > ne11_mm_min) { //printf("matrix: ne00 = %6d, ne01 = %6d, ne02 = %6d, ne11 = %6d, ne12 = %6d\n", ne00, ne01, ne02, ne11, ne12); switch (src0->type) {