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

Co-authored-by: EC2 Default User <redacted>

diff --git a/ggml/src/ggml-cpu.c b/ggml/src/ggml-cpu.c
index 98c3e21ae3fcff3ae2ab8e132a766cc64f6fdf5d..de1de18ecea7e0f5bcee5e0e7c6ac8b7ec52c68e 100644 (file)
--- 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,