Update GGML to b6646 (#12245)

Notable EOLs with this change:
- MacOS v12 and v13 are no longer supported (v14+ required)
- AMD gfx900 and gfx906 are no longer supported

ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-f16.cu

+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../fattn-vec.cuh"
+
+DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_0, GGML_TYPE_F16);
+DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_F16);
+DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_0, GGML_TYPE_F16);

ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q4_0.cu

+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../fattn-vec.cuh"
+
+DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_0, GGML_TYPE_Q4_0);
+DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_Q4_0);
+DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_0, GGML_TYPE_Q4_0);

ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q4_1.cu

+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../fattn-vec.cuh"
+
+DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_0, GGML_TYPE_Q4_1);
+DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_Q4_1);
+DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_0, GGML_TYPE_Q4_1);

ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q5_0.cu

+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../fattn-vec.cuh"
+
+DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_0, GGML_TYPE_Q5_0);
+DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_Q5_0);
+DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_0, GGML_TYPE_Q5_0);

ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q5_1.cu

+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../fattn-vec.cuh"
+
+DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_0, GGML_TYPE_Q5_1);
+DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_Q5_1);
+DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_0, GGML_TYPE_Q5_1);

ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q8_0.cu

+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../fattn-vec.cuh"
+
+DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_0, GGML_TYPE_Q8_0);
+DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_Q8_0);
+DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_0, GGML_TYPE_Q8_0);

ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-f16.cu

+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../fattn-vec.cuh"
+
+DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_1, GGML_TYPE_F16);
+DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_F16);
+DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_1, GGML_TYPE_F16);

ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q4_0.cu

+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../fattn-vec.cuh"
+
+DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_1, GGML_TYPE_Q4_0);
+DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_Q4_0);
+DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_1, GGML_TYPE_Q4_0);

ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q4_1.cu

+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../fattn-vec.cuh"
+
+DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_1, GGML_TYPE_Q4_1);
+DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_Q4_1);
+DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_1, GGML_TYPE_Q4_1);

ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q5_0.cu

+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../fattn-vec.cuh"
+
+DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_1, GGML_TYPE_Q5_0);
+DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_Q5_0);
+DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_1, GGML_TYPE_Q5_0);

ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q5_1.cu

+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../fattn-vec.cuh"
+
+DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_1, GGML_TYPE_Q5_1);
+DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_Q5_1);
+DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_1, GGML_TYPE_Q5_1);

ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q8_0.cu

+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../fattn-vec.cuh"
+
+DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_1, GGML_TYPE_Q8_0);
+DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_Q8_0);
+DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_1, GGML_TYPE_Q8_0);

ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-f16.cu

+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../fattn-vec.cuh"
+
+DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q8_0, GGML_TYPE_F16);
+DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_F16);
+DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q8_0, GGML_TYPE_F16);

ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q4_0.cu

+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../fattn-vec.cuh"
+
+DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q8_0, GGML_TYPE_Q4_0);
+DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q4_0);
+DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q8_0, GGML_TYPE_Q4_0);

ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q4_1.cu

+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../fattn-vec.cuh"
+
+DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q8_0, GGML_TYPE_Q4_1);
+DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q4_1);
+DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q8_0, GGML_TYPE_Q4_1);

ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q5_0.cu

+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../fattn-vec.cuh"
+
+DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q8_0, GGML_TYPE_Q5_0);
+DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q5_0);
+DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q8_0, GGML_TYPE_Q5_0);

ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q5_1.cu

+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../fattn-vec.cuh"
+
+DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q8_0, GGML_TYPE_Q5_1);
+DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q5_1);
+DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q8_0, GGML_TYPE_Q5_1);

ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q8_0.cu

+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../fattn-vec.cuh"
+
+DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q8_0, GGML_TYPE_Q8_0);
+DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q8_0);
+DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q8_0, GGML_TYPE_Q8_0);

ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-f16.cu → ml/backend/ggml/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_1.cu

 // This file has been autogenerated by generate_cu_files.py, do not edit manually.
 
-#include "../fattn-vec-f32.cuh"
+#include "../mmf.cuh"
 
-DECL_FATTN_VEC_F32_CASE(64, GGML_TYPE_F16, GGML_TYPE_F16);
+DECL_MMF_CASE(1);

ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-f16.cu → ml/backend/ggml/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_10.cu

 // This file has been autogenerated by generate_cu_files.py, do not edit manually.
 
-#include "../fattn-vec-f16.cuh"
+#include "../mmf.cuh"
 
-DECL_FATTN_VEC_F16_CASE(64, GGML_TYPE_F16, GGML_TYPE_F16);
+DECL_MMF_CASE(10);