From f5e85d43a048f3d5adefb4c5e29867d8077fba62 Mon Sep 17 00:00:00 2001 From: Strongtut Date: Fri, 28 Aug 2026 05:37:37 -0700 Subject: [PATCH] metal : add fa-vec tunings for M4 (#27875) This adds fa_vec_tuned_table records for Apple M4 to ggml-metal-tuning.cpp. Includes F16, Q4_0, Q4_1, Q5_0, Q5_1, and Q8_0. (M4, 10 GPU Cores) Co-authored-by: Strongtut <8432058+Strongtut@users.noreply.github.com> --- ggml/src/ggml-metal/ggml-metal-tuning.cpp | 226 ++++++++++++++++++++++ 1 file changed, 226 insertions(+) diff --git a/ggml/src/ggml-metal/ggml-metal-tuning.cpp b/ggml/src/ggml-metal/ggml-metal-tuning.cpp index c2139fe20..285590d16 100644 --- a/ggml/src/ggml-metal/ggml-metal-tuning.cpp +++ b/ggml/src/ggml-metal/ggml-metal-tuning.cpp @@ -542,6 +542,232 @@ constexpr fa_vec_entry_t fa_vec_tuned_table[] = { { { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 576, 512, 1, 1 }, { 1, 2 } }, { { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 576, 512, 1, 4 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 32, 32, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 32, 32, 1, 3 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 32, 32, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 32, 32, 2, 3 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 32, 32, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 32, 32, 3, 3 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 64, 64, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 64, 64, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 64, 64, 1, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 64, 64, 2, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 96, 96, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 96, 96, 1, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 96, 96, 1, 4 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 96, 96, 2, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 96, 96, 2, 4 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 128, 128, -1, 1 }, { 2, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 128, 128, 1, 2 }, { 1, 1 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 128, 128, 1, 4 }, { 1, 1 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 128, 128, 2, 2 }, { 1, 1 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 128, 128, 2, 4 }, { 1, 1 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 192, 192, 3, 0 }, { 2, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 192, 192, -1, 1 }, { 2, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 192, 192, 1, 2 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 192, 128, -1, 1 }, { 2, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 192, 128, 1, 2 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 192, 128, 1, 4 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 256, 256, -1, 0 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 256, 256, 3, 0 }, { 2, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 256, 256, -1, 1 }, { 2, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 320, 256, 3, 0 }, { 2, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 320, 256, -1, 1 }, { 2, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 512, 512, 3, 0 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 512, 512, 3, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 512, 512, 3, 2 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_F16, 512, 512, 3, 3 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 32, 32, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 32, 32, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 32, 32, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 32, 32, 2, 4 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 32, 32, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 64, 64, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 64, 64, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 64, 64, 1, 2 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 64, 64, 2, 2 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 64, 64, 2, 3 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 64, 64, 3, 2 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 64, 64, 3, 3 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 96, 96, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 96, 96, 1, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 128, 128, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 128, 128, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 128, 128, 1, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 128, 128, 2, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 128, 128, 3, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 192, 192, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 192, 192, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 192, 128, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 192, 128, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 192, 128, 1, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 192, 128, 2, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 192, 128, 3, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 256, 256, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 256, 256, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 320, 256, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 320, 256, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 512, 512, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 512, 512, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 576, 512, 2, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 576, 512, 3, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 576, 512, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 576, 512, 1, 1 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 576, 512, 1, 2 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 576, 512, 1, 3 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_0, 576, 512, 1, 4 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 32, 32, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 32, 32, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 32, 32, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 32, 32, 2, 4 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 32, 32, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 64, 64, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 64, 64, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 64, 64, 1, 2 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 64, 64, 2, 2 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 64, 64, 2, 3 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 64, 64, 3, 2 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 64, 64, 3, 3 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 96, 96, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 96, 96, 1, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 96, 96, 2, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 128, 128, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 128, 128, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 128, 128, 1, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 128, 128, 2, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 128, 128, 3, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 192, 192, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 192, 192, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 192, 128, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 192, 128, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 192, 128, 1, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 192, 128, 2, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 192, 128, 3, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 256, 256, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 256, 256, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 320, 256, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 320, 256, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 576, 512, 2, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 576, 512, 3, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 576, 512, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 576, 512, 1, 1 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 576, 512, 1, 2 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 576, 512, 1, 3 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q4_1, 576, 512, 1, 4 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 32, 32, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 32, 32, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 32, 32, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 32, 32, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 64, 64, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 64, 64, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 64, 64, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 64, 64, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 64, 64, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 96, 96, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 96, 96, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 96, 96, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 96, 96, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 128, 128, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 128, 128, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 128, 128, 1, 2 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 128, 128, 1, 3 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 128, 128, 2, 2 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 128, 128, 2, 3 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 128, 128, 3, 2 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 128, 128, 3, 3 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 192, 192, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 192, 192, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 192, 128, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 192, 128, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 256, 256, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 256, 256, -1, 1 }, { 2, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 256, 256, 1, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 256, 256, 3, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 320, 256, -1, 1 }, { 2, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 320, 256, 1, 2 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 320, 256, 2, 2 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 320, 256, 3, 2 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 512, 512, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 512, 512, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 576, 512, 2, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 576, 512, 2, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 576, 512, 2, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 576, 512, 2, 3 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_0, 576, 512, 2, 4 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 32, 32, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 32, 32, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 32, 32, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 32, 32, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 64, 64, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 64, 64, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 64, 64, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 64, 64, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 64, 64, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 96, 96, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 96, 96, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 96, 96, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 96, 96, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 128, 128, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 128, 128, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 128, 128, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 128, 128, 1, 4 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 128, 128, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 128, 128, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 128, 128, 3, 4 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 192, 192, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 192, 192, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 192, 128, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 192, 128, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 256, 256, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 256, 256, -1, 1 }, { 2, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 256, 256, 1, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 256, 256, 3, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 320, 256, -1, 1 }, { 2, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 320, 256, 1, 2 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 320, 256, 1, 4 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 320, 256, 2, 2 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 320, 256, 3, 2 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 512, 512, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 512, 512, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 576, 512, 2, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 576, 512, 2, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 576, 512, 2, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 576, 512, 2, 3 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q5_1, 576, 512, 2, 4 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 32, 32, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 32, 32, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 32, 32, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 32, 32, 2, 4 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 32, 32, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 64, 64, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 64, 64, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 64, 64, 1, 2 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 64, 64, 2, 2 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 64, 64, 3, 2 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 64, 64, 3, 3 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 96, 96, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 128, 128, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 128, 128, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 128, 128, 1, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 128, 128, 2, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 192, 192, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 192, 192, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 192, 192, 1, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 192, 192, 2, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 192, 192, 2, 4 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 192, 192, 3, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 192, 128, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 192, 128, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 256, 256, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 256, 256, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 320, 256, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 320, 256, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 512, 512, -1, 0 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 512, 512, -1, 1 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 576, 512, 2, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 576, 512, 3, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4, GGML_TYPE_Q8_0, 576, 512, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M4_PRO, GGML_TYPE_F16, 32, 32, 1, 3 }, { 2, 4 } }, { { GGML_METAL_DEVICE_M4_PRO, GGML_TYPE_F16, 32, 32, 2, 1 }, { 2, 4 } }, { { GGML_METAL_DEVICE_M4_PRO, GGML_TYPE_F16, 32, 32, 2, 3 }, { 2, 4 } },