metal : add fa-vec tunings for M3 Max, M5 and M5 Pro (#27863)
* metal : add fa-vec tunings for M5 This is a followup contribution to efeda76b948f59ee52ea20db640bc4cf3dfe8ac1 as requested in https://github.com/ggml-org/llama.cpp/discussions/27668 to add support for additional Apple GPUs. I generated this output using the provided instructions: ```sh git clone https://github.com/ggml-org/llama.cpp cd llama.cpp cmake -B build -DGGML_METAL=ON cmake --build build --target ggml-metal-tuning -j ./build/bin/ggml-metal-tuning fa-vec --dtype f16,q8_0 > fa_vec_rows.txt 2> fa_vec_sweep.log ``` This ran on a machine with Apple M5. Assisted-by: pi:llama.cpp/Qwen3.8-27B * metal : add fa-vec tunings for M5 Pro This adds fa_vec_tuned_table records for Apple M5 Pro to ggml-metal-tuning.cpp. Contributed by SerayaEryn in https://github.com/ggml-org/llama.cpp/discussions/27668#discussioncomment-18157544 (F16, Q4_0, Q8_0; M5 Pro, 20 GPU cores). Assisted-by: pi:llama.cpp/Qwen3.8-27B * metal : add fa-vec tunings for M3 Max This adds fa_vec_tuned_table records for Apple M3 Max to ggml-metal-tuning.cpp. Contributed by TeeAaTeeUu in https://github.com/ggml-org/llama.cpp/discussions/27668#discussioncomment-18175220 (F16, Q8_0; M3 Max, MacBook Pro 64GB, low power mode). Assisted-by: pi:llama.cpp/Qwen3.8-27B * cont : whitespaces
This commit is contained in:
@@ -66,6 +66,7 @@ fa_vec_cfg_t fa_vec_baseline_cfg(int dk, int dv) {
|
|||||||
// One row per kept bucket, plus per-(dtype,dk,dv) ne11-collapsed domain defaults
|
// One row per kept bucket, plus per-(dtype,dk,dv) ne11-collapsed domain defaults
|
||||||
// (ne11_b = FA_VEC_NE11_DEFAULT, ne01_b = domain). To retune or add a device, re-run the
|
// (ne11_b = FA_VEC_NE11_DEFAULT, ne01_b = domain). To retune or add a device, re-run the
|
||||||
// sweep and paste its output. See ggml-metal-tuning.h for the row/lookup semantics.
|
// sweep and paste its output. See ggml-metal-tuning.h for the row/lookup semantics.
|
||||||
|
// ref: https://github.com/ggml-org/llama.cpp/pull/27824
|
||||||
constexpr fa_vec_entry_t fa_vec_tuned_table[] = {
|
constexpr fa_vec_entry_t fa_vec_tuned_table[] = {
|
||||||
{ { GGML_METAL_DEVICE_M1_PRO, GGML_TYPE_F16, 32, 32, 3, 3 }, { 2, 4 } },
|
{ { GGML_METAL_DEVICE_M1_PRO, GGML_TYPE_F16, 32, 32, 3, 3 }, { 2, 4 } },
|
||||||
{ { GGML_METAL_DEVICE_M1_PRO, GGML_TYPE_F16, 64, 64, -1, 0 }, { 1, 4 } },
|
{ { GGML_METAL_DEVICE_M1_PRO, GGML_TYPE_F16, 64, 64, -1, 0 }, { 1, 4 } },
|
||||||
@@ -448,6 +449,99 @@ constexpr fa_vec_entry_t fa_vec_tuned_table[] = {
|
|||||||
{ { GGML_METAL_DEVICE_M2_ULTRA, GGML_TYPE_Q8_0, 512, 512, -1, 1 }, { 1, 4 } },
|
{ { GGML_METAL_DEVICE_M2_ULTRA, GGML_TYPE_Q8_0, 512, 512, -1, 1 }, { 1, 4 } },
|
||||||
{ { GGML_METAL_DEVICE_M2_ULTRA, GGML_TYPE_Q8_0, 576, 512, -1, 0 }, { 1, 4 } },
|
{ { GGML_METAL_DEVICE_M2_ULTRA, GGML_TYPE_Q8_0, 576, 512, -1, 0 }, { 1, 4 } },
|
||||||
{ { GGML_METAL_DEVICE_M2_ULTRA, GGML_TYPE_Q8_0, 576, 512, -1, 1 }, { 1, 4 } },
|
{ { GGML_METAL_DEVICE_M2_ULTRA, GGML_TYPE_Q8_0, 576, 512, -1, 1 }, { 1, 4 } },
|
||||||
|
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 32, 32, 1, 3 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 32, 32, 2, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 32, 32, 2, 3 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 32, 32, 2, 4 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 32, 32, 3, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 32, 32, 3, 3 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 64, 64, 2, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 64, 64, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 64, 64, 1, 2 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 64, 64, 2, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 64, 64, 3, 1 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 64, 64, 3, 2 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 96, 96, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 96, 96, 1, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 96, 96, 2, 4 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 96, 96, 3, 1 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 96, 96, 3, 3 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 96, 96, 3, 4 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 128, 128, -1, 1 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 128, 128, 1, 4 }, { 1, 1 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 128, 128, 2, 4 }, { 1, 1 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 192, 192, 1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 192, 192, -1, 1 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 192, 192, 1, 2 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 192, 128, -1, 1 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 192, 128, 1, 3 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 192, 128, 1, 4 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 192, 128, 3, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 256, 256, -1, 0 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 256, 256, -1, 1 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 256, 256, 2, 3 }, { 1, 1 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 320, 256, 1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 320, 256, 3, 0 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 320, 256, -1, 1 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 512, 512, 2, 0 }, { 4, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 512, 512, 3, 0 }, { 4, 1 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 512, 512, 1, 3 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 512, 512, 2, 1 }, { 4, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 512, 512, 2, 3 }, { 4, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 512, 512, 3, 1 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 512, 512, 3, 2 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 576, 512, 2, 0 }, { 4, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 576, 512, 2, 2 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 576, 512, 2, 3 }, { 4, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_F16, 576, 512, 3, 1 }, { 4, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 32, 32, -1, 1 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 32, 32, 1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 32, 32, 1, 4 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 32, 32, 2, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 32, 32, 2, 4 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 32, 32, 3, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 64, 64, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 64, 64, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 64, 64, 2, 3 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 64, 64, 3, 2 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 96, 96, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 96, 96, 1, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 96, 96, 1, 4 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 96, 96, 2, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 96, 96, 2, 4 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 96, 96, 3, 4 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 128, 128, 1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 128, 128, 3, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 128, 128, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 128, 128, 1, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 128, 128, 2, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 128, 128, 3, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 192, 192, 3, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 192, 192, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 192, 192, 1, 1 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 192, 192, 1, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 192, 192, 1, 4 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 192, 192, 2, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 192, 192, 3, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 192, 128, 1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 192, 128, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 192, 128, 1, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 192, 128, 2, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 192, 128, 2, 4 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 192, 128, 3, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 256, 256, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 256, 256, -1, 1 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 320, 256, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 320, 256, -1, 1 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 512, 512, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 512, 512, -1, 1 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 576, 512, 2, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 576, 512, 3, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M3_MAX, GGML_TYPE_Q8_0, 576, 512, -1, 1 }, { 1, 4 } },
|
||||||
|
{ { 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_PRO, GGML_TYPE_F16, 32, 32, 1, 3 }, { 2, 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, 1 }, { 2, 4 } },
|
||||||
{ { GGML_METAL_DEVICE_M4_PRO, GGML_TYPE_F16, 32, 32, 2, 3 }, { 2, 4 } },
|
{ { GGML_METAL_DEVICE_M4_PRO, GGML_TYPE_F16, 32, 32, 2, 3 }, { 2, 4 } },
|
||||||
@@ -508,6 +602,7 @@ constexpr fa_vec_entry_t fa_vec_tuned_table[] = {
|
|||||||
{ { GGML_METAL_DEVICE_M4_PRO, GGML_TYPE_Q8_0, 576, 512, 1, 1 }, { 1, 2 } },
|
{ { GGML_METAL_DEVICE_M4_PRO, GGML_TYPE_Q8_0, 576, 512, 1, 1 }, { 1, 2 } },
|
||||||
{ { GGML_METAL_DEVICE_M4_PRO, GGML_TYPE_Q8_0, 576, 512, 1, 3 }, { 1, 2 } },
|
{ { GGML_METAL_DEVICE_M4_PRO, GGML_TYPE_Q8_0, 576, 512, 1, 3 }, { 1, 2 } },
|
||||||
{ { GGML_METAL_DEVICE_M4_PRO, GGML_TYPE_Q8_0, 576, 512, 1, 4 }, { 1, 2 } },
|
{ { GGML_METAL_DEVICE_M4_PRO, GGML_TYPE_Q8_0, 576, 512, 1, 4 }, { 1, 2 } },
|
||||||
|
|
||||||
{ { GGML_METAL_DEVICE_M4_MAX, GGML_TYPE_F16, 32, 32, 2, 1 }, { 2, 4 } },
|
{ { GGML_METAL_DEVICE_M4_MAX, GGML_TYPE_F16, 32, 32, 2, 1 }, { 2, 4 } },
|
||||||
{ { GGML_METAL_DEVICE_M4_MAX, GGML_TYPE_F16, 32, 32, 2, 3 }, { 2, 4 } },
|
{ { GGML_METAL_DEVICE_M4_MAX, GGML_TYPE_F16, 32, 32, 2, 3 }, { 2, 4 } },
|
||||||
{ { GGML_METAL_DEVICE_M4_MAX, GGML_TYPE_F16, 32, 32, 3, 1 }, { 2, 4 } },
|
{ { GGML_METAL_DEVICE_M4_MAX, GGML_TYPE_F16, 32, 32, 3, 1 }, { 2, 4 } },
|
||||||
@@ -699,7 +794,215 @@ constexpr fa_vec_entry_t fa_vec_tuned_table[] = {
|
|||||||
{ { GGML_METAL_DEVICE_M4_MAX, GGML_TYPE_Q8_0, 576, 512, -1, 0 }, { 1, 4 } },
|
{ { GGML_METAL_DEVICE_M4_MAX, GGML_TYPE_Q8_0, 576, 512, -1, 0 }, { 1, 4 } },
|
||||||
{ { GGML_METAL_DEVICE_M4_MAX, GGML_TYPE_Q8_0, 576, 512, -1, 1 }, { 1, 4 } },
|
{ { GGML_METAL_DEVICE_M4_MAX, GGML_TYPE_Q8_0, 576, 512, -1, 1 }, { 1, 4 } },
|
||||||
|
|
||||||
{ { GGML_METAL_DEVICE_M5_MAX, GGML_TYPE_F16, 32, 32, -1, 1 }, { 2, 4 } },
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 32, 32, 1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 32, 32, 1, 3 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 32, 32, 2, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 32, 32, 2, 3 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 32, 32, 3, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 32, 32, 3, 2 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 32, 32, 3, 3 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 32, 32, 3, 4 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 64, 64, 1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 64, 64, 2, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 64, 64, -1, 1 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 64, 64, 1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 64, 64, 1, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 64, 64, 1, 4 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 64, 64, 2, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 64, 64, 2, 4 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 64, 64, 3, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 96, 96, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 96, 96, 1, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 96, 96, 1, 4 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 96, 96, 2, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 96, 96, 3, 3 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 128, 128, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 128, 128, 3, 0 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 128, 128, -1, 1 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 128, 128, 1, 2 }, { 1, 1 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 128, 128, 1, 4 }, { 1, 1 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 128, 128, 2, 4 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 128, 128, 3, 3 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 128, 128, 3, 4 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 192, 192, -1, 1 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 192, 192, 1, 2 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 192, 192, 1, 4 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 192, 128, 1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 192, 128, 2, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 192, 128, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 192, 128, 1, 2 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 192, 128, 1, 4 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 192, 128, 3, 2 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 192, 128, 3, 3 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 256, 256, -1, 0 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 256, 256, 1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 256, 256, -1, 1 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 256, 256, 1, 4 }, { 1, 1 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 320, 256, 2, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 320, 256, 3, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 320, 256, -1, 1 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 320, 256, 1, 2 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 320, 256, 1, 4 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 512, 512, 2, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 512, 512, 1, 1 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 512, 512, 2, 1 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 512, 512, 2, 3 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 512, 512, 3, 1 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 512, 512, 3, 2 }, { 4, 1 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 512, 512, 3, 3 }, { 4, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 512, 512, 3, 4 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 576, 512, 2, 0 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 576, 512, -1, 1 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 576, 512, 1, 2 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 576, 512, 1, 4 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_F16, 576, 512, 2, 2 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 32, 32, -1, 1 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 32, 32, 1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 32, 32, 2, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 32, 32, 2, 4 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 32, 32, 3, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 32, 32, 3, 4 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 64, 64, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 64, 64, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 64, 64, 1, 2 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 64, 64, 1, 3 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 64, 64, 2, 2 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 64, 64, 3, 2 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 64, 64, 3, 3 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 96, 96, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 96, 96, 1, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 96, 96, 2, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 96, 96, 2, 3 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 96, 96, 3, 4 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 128, 128, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 128, 128, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 128, 128, 2, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 192, 192, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 192, 192, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 192, 192, 1, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 192, 192, 2, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 192, 192, 3, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 192, 128, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 192, 128, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 192, 128, 2, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 192, 128, 2, 3 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 192, 128, 3, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 256, 256, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 256, 256, -1, 1 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 320, 256, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 320, 256, -1, 1 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 320, 256, 2, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 512, 512, 3, 0 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 512, 512, 2, 1 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 512, 512, 3, 1 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 576, 512, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5, GGML_TYPE_Q8_0, 576, 512, -1, 1 }, { 1, 4 } },
|
||||||
|
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 32, 32, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 32, 32, 1, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 32, 32, 1, 4 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 32, 32, 2, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 32, 32, 2, 4 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 32, 32, 3, 2 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 32, 32, 3, 3 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 64, 64, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 64, 64, 1, 2 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 64, 64, 3, 2 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 64, 64, 3, 3 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 64, 64, 3, 4 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 96, 96, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 128, 128, -1, 1 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 128, 128, 1, 4 }, { 1, 1 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 192, 192, -1, 1 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 192, 192, 1, 2 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 192, 128, 2, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 192, 128, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 192, 128, 3, 1 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 256, 256, -1, 0 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 256, 256, -1, 1 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 256, 256, 1, 4 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 320, 256, 3, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 320, 256, -1, 1 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 320, 256, 1, 2 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 512, 512, 2, 1 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 512, 512, 2, 3 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 512, 512, 3, 3 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 576, 512, -1, 1 }, { 2, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 576, 512, 1, 1 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 576, 512, 1, 2 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 576, 512, 1, 3 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 576, 512, 1, 4 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_F16, 576, 512, 2, 2 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 32, 32, -1, 1 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 32, 32, 1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 32, 32, 1, 4 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 32, 32, 2, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 32, 32, 3, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 64, 64, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 64, 64, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 64, 64, 1, 2 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 96, 96, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 96, 96, 1, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 96, 96, 2, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 96, 96, 3, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 128, 128, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 128, 128, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 128, 128, 1, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 192, 192, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 192, 192, -1, 1 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 192, 192, 1, 3 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 192, 192, 2, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 192, 192, 2, 3 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 192, 192, 2, 4 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 192, 192, 3, 4 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 192, 128, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 192, 128, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 192, 128, 1, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 192, 128, 2, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 256, 256, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 256, 256, -1, 1 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 320, 256, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 320, 256, -1, 1 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 512, 512, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 512, 512, 1, 0 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 512, 512, -1, 1 }, { 1, 2 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 512, 512, 2, 1 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q4_0, 512, 512, 2, 3 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 32, 32, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 32, 32, 1, 2 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 32, 32, 1, 3 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 32, 32, 2, 2 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 32, 32, 2, 3 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 32, 32, 3, 2 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 32, 32, 3, 3 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 64, 64, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 64, 64, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 64, 64, 3, 2 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 64, 64, 3, 3 }, { 4, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 96, 96, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 96, 96, 1, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 96, 96, 2, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 128, 128, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 128, 128, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 128, 128, 1, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 128, 128, 2, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 192, 192, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 192, 192, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 192, 192, 1, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 192, 192, 2, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 192, 192, 3, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 192, 128, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 192, 128, -1, 1 }, { 2, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 192, 128, 1, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 192, 128, 3, 2 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 256, 256, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 256, 256, -1, 1 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 320, 256, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 320, 256, -1, 1 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 576, 512, -1, 0 }, { 1, 4 } },
|
||||||
|
{ { GGML_METAL_DEVICE_M5_PRO, GGML_TYPE_Q8_0, 576, 512, -1, 1 }, { 1, 4 } },
|
||||||
|
|
||||||
|
{ { GGML_METAL_DEVICE_M5_MAX, GGML_TYPE_F16, 32, 32, -1, 1 }, { 2, 4 } },
|
||||||
{ { GGML_METAL_DEVICE_M5_MAX, GGML_TYPE_F16, 32, 32, 1, 2 }, { 1, 4 } },
|
{ { GGML_METAL_DEVICE_M5_MAX, GGML_TYPE_F16, 32, 32, 1, 2 }, { 1, 4 } },
|
||||||
{ { GGML_METAL_DEVICE_M5_MAX, GGML_TYPE_F16, 32, 32, 1, 4 }, { 1, 4 } },
|
{ { GGML_METAL_DEVICE_M5_MAX, GGML_TYPE_F16, 32, 32, 1, 4 }, { 1, 4 } },
|
||||||
{ { GGML_METAL_DEVICE_M5_MAX, GGML_TYPE_F16, 32, 32, 2, 2 }, { 4, 4 } },
|
{ { GGML_METAL_DEVICE_M5_MAX, GGML_TYPE_F16, 32, 32, 2, 2 }, { 4, 4 } },
|
||||||
|
|||||||
Reference in New Issue
Block a user