fix q4_0_8_8 format for corrupted tokens issue (llama/10198)

Co-authored-by: EC2 Default User <ec2-user@ip-172-31-62-167.us-west-2.compute.internal>
2024-12-20 05:07:52 +00:00 · 2024-11-07 02:02:08 -06:00 · 2024-11-07 02:02:08 -06:00 · 807f848c2f
commit 807f848c2f
parent 42398f13b0
1 changed files with 2 additions and 0 deletions
--- 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,