From 427291b5b34cd914a31b3fd3b61a68f6184f4b9f Mon Sep 17 00:00:00 2001 From: Isaac <34376531+init-22@users.noreply.github.com> Date: Sat, 5 Sep 2026 00:08:34 +0530 Subject: [PATCH] metal : add remaining fa-vec tunings for M3 (#28396) * addition of m3 in fa_vec_tuned_table * adding q4_0,q4_1,q5_0,q5_1 in ggml-metal-tuning * Fix formatting in ggml-metal-tuning.cpp --- ggml/src/ggml-metal/ggml-metal-tuning.cpp | 172 ++++++++++++++++++++++ 1 file changed, 172 insertions(+) diff --git a/ggml/src/ggml-metal/ggml-metal-tuning.cpp b/ggml/src/ggml-metal/ggml-metal-tuning.cpp index 7de01fac1..8cdc55a0a 100644 --- a/ggml/src/ggml-metal/ggml-metal-tuning.cpp +++ b/ggml/src/ggml-metal/ggml-metal-tuning.cpp @@ -1525,6 +1525,178 @@ constexpr fa_vec_entry_t fa_vec_tuned_table[] = { { { GGML_METAL_DEVICE_M3, GGML_TYPE_F16, 576, 512, 2, 1 }, { 4, 2 } }, { { GGML_METAL_DEVICE_M3, GGML_TYPE_F16, 576, 512, 2, 2 }, { 4, 2 } }, { { GGML_METAL_DEVICE_M3, GGML_TYPE_F16, 576, 512, 3, 1 }, { 4, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 32, 32, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 32, 32, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 32, 32, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 32, 32, 2, 4 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 32, 32, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 64, 64, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 64, 64, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 64, 64, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 64, 64, 1, 4 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 64, 64, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 64, 64, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 64, 64, 3, 4 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 96, 96, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 96, 96, 1, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 96, 96, 2, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 128, 128, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 128, 128, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 128, 128, 1, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 128, 128, 2, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 128, 128, 3, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 192, 192, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 192, 192, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 192, 192, 1, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 192, 192, 2, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 192, 192, 3, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 192, 192, 3, 2 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 192, 192, 3, 3 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 192, 128, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 192, 128, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 192, 128, 1, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 192, 128, 2, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 256, 256, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 256, 256, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 320, 256, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 320, 256, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 320, 256, 3, 4 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 512, 512, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 512, 512, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 576, 512, 2, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 576, 512, 3, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 576, 512, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_0, 576, 512, 1, 4 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 32, 32, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 32, 32, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 32, 32, 1, 4 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 32, 32, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 32, 32, 2, 4 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 32, 32, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 64, 64, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 64, 64, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 64, 64, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 64, 64, 1, 4 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 64, 64, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 64, 64, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 96, 96, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 96, 96, 1, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 96, 96, 1, 3 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 96, 96, 2, 2 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 128, 128, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 128, 128, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 128, 128, 1, 2 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 128, 128, 2, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 128, 128, 3, 1 }, { 2, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 128, 128, 3, 2 }, { 2, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 128, 128, 3, 3 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 192, 192, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 192, 192, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 192, 192, 1, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 192, 192, 2, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 192, 192, 3, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 192, 128, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 192, 128, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 256, 256, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 256, 256, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 320, 256, 2, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 320, 256, 3, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 320, 256, 1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 320, 256, 2, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 320, 256, 3, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 320, 256, 3, 3 }, { 2, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 512, 512, 2, 0 }, { 4, 1 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 512, 512, 2, 4 }, { 4, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 512, 512, 3, 1 }, { 4, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 512, 512, 3, 2 }, { 4, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 576, 512, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 576, 512, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q4_1, 576, 512, 1, 4 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 32, 32, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 32, 32, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 32, 32, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 32, 32, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 64, 64, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 64, 64, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 64, 64, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 64, 64, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 64, 64, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 96, 96, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 96, 96, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 96, 96, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 96, 96, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 128, 128, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 128, 128, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 128, 128, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 128, 128, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 128, 128, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 192, 192, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 192, 192, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 192, 128, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 192, 128, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 192, 128, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 192, 128, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 192, 128, 2, 4 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 192, 128, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 256, 256, -1, 0 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 256, 256, -1, 1 }, { 2, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 256, 256, 1, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 256, 256, 2, 2 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 256, 256, 3, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 320, 256, 1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 320, 256, -1, 1 }, { 2, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 320, 256, 2, 2 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 320, 256, 3, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 512, 512, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 512, 512, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 512, 512, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 576, 512, 2, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 576, 512, 3, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 576, 512, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 576, 512, 1, 1 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_0, 576, 512, 1, 4 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 32, 32, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 32, 32, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 32, 32, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 32, 32, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 64, 64, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 64, 64, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 64, 64, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 64, 64, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 64, 64, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 96, 96, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 96, 96, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 96, 96, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 96, 96, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 128, 128, 3, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 128, 128, -1, 1 }, { 4, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 128, 128, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 128, 128, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 128, 128, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 192, 192, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 192, 192, -1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 192, 128, 1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 192, 128, 2, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 192, 128, -1, 1 }, { 4, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 192, 128, 1, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 192, 128, 1, 4 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 192, 128, 2, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 192, 128, 3, 1 }, { 2, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 256, 256, -1, 0 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 256, 256, -1, 1 }, { 2, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 256, 256, 3, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 320, 256, 3, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 320, 256, -1, 1 }, { 2, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 320, 256, 1, 2 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 320, 256, 2, 2 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 320, 256, 3, 2 }, { 1, 2 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 512, 512, -1, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 512, 512, -1, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 576, 512, 2, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 576, 512, 3, 0 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 576, 512, 2, 4 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 576, 512, 3, 1 }, { 1, 4 } }, + { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q5_1, 576, 512, 3, 3 }, { 1, 4 } }, { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q8_0, 32, 32, -1, 1 }, { 4, 4 } }, { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q8_0, 32, 32, 1, 1 }, { 2, 4 } }, { { GGML_METAL_DEVICE_M3, GGML_TYPE_Q8_0, 32, 32, 2, 1 }, { 2, 4 } },