From 78458aa83dedd0a9b89e396b052b3a43a1e32ff1 Mon Sep 17 00:00:00 2001 From: Iwan Kawrakow Date: Fri, 25 Apr 2025 19:42:21 +0200 Subject: [PATCH] Fix q4_1 and q5_1 on Arm --- ggml/src/ggml.c | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/ggml/src/ggml.c b/ggml/src/ggml.c index 83a48cb6..ad9393cc 100644 --- a/ggml/src/ggml.c +++ b/ggml/src/ggml.c @@ -741,7 +741,11 @@ static const ggml_type_traits_t type_traits[GGML_TYPE_COUNT] = { .from_float_ref = (ggml_from_float_t) quantize_row_q4_1_ref, .vec_dot = ggml_vec_dot_q4_1_q8_1, #if GGML_USE_IQK_MULMAT +#if defined __AVX2__ .vec_dot_type = GGML_TYPE_Q8_2_X4, +#else + .vec_dot_type = GGML_TYPE_Q8_1_X4, +#endif #else .vec_dot_type = GGML_TYPE_Q8_1, #endif @@ -809,7 +813,11 @@ static const ggml_type_traits_t type_traits[GGML_TYPE_COUNT] = { .from_float_ref = (ggml_from_float_t) quantize_row_q5_1_ref, .vec_dot = ggml_vec_dot_q5_1_q8_1, #if GGML_USE_IQK_MULMAT +#ifdef __AVX2__ .vec_dot_type = GGML_TYPE_Q8_2_X4, +#else + .vec_dot_type = GGML_TYPE_Q8_1_X4, +#endif #else .vec_dot_type = GGML_TYPE_Q8_1, #endif