From 2319126a70b541f8670225a04a38202bbdccbedb Mon Sep 17 00:00:00 2001 From: snadampal <87143774+snadampal@users.noreply.github.com> Date: Thu, 7 Nov 2024 02:02:08 -0600 Subject: [PATCH] fix q4_0_8_8 format for corrupted tokens issue (#10198) Co-authored-by: EC2 Default User --- ggml/src/ggml-cpu.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/ggml/src/ggml-cpu.c b/ggml/src/ggml-cpu.c index 98c3e21ae..de1de18ec 100644 --- a/ggml/src/ggml-cpu.c +++ b/ggml/src/ggml-cpu.c @@ -409,6 +409,8 @@ static const struct ggml_type_traits_cpu type_traits_cpu[GGML_TYPE_COUNT] = { .gemm = ggml_gemm_q4_0_4x8_q8_0, }, [GGML_TYPE_Q4_0_8_8] = { + .vec_dot = NULL, + .vec_dot_type = GGML_TYPE_Q8_0, .nrows = 1, .ncols = 8, .gemv = ggml_gemv_q4_0_8x8_q8_0,