diff options
Diffstat (limited to 'llama.cpp/ggml/src/ggml-cuda/template-instances')
102 files changed, 760 insertions, 0 deletions
diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_16.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_16.cu new file mode 100644 index 0000000..fb26abe --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_16.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(576, 512, 1, 16); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_32.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_32.cu new file mode 100644 index 0000000..1f554d8 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_32.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(576, 512, 1, 32); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_8.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_8.cu new file mode 100644 index 0000000..dc16829 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_8.cu @@ -0,0 +1,10 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(64, 64, 1, 8); +DECL_FATTN_MMA_F16_CASE(80, 80, 1, 8); +DECL_FATTN_MMA_F16_CASE(96, 96, 1, 8); +DECL_FATTN_MMA_F16_CASE(112, 112, 1, 8); +DECL_FATTN_MMA_F16_CASE(128, 128, 1, 8); +DECL_FATTN_MMA_F16_CASE(256, 256, 1, 8); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_1.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_1.cu new file mode 100644 index 0000000..9d3cfd8 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_1.cu @@ -0,0 +1,10 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(64, 64, 16, 1); +DECL_FATTN_MMA_F16_CASE(80, 80, 16, 1); +DECL_FATTN_MMA_F16_CASE(96, 96, 16, 1); +DECL_FATTN_MMA_F16_CASE(112, 112, 16, 1); +DECL_FATTN_MMA_F16_CASE(128, 128, 16, 1); +DECL_FATTN_MMA_F16_CASE(256, 256, 16, 1); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_2.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_2.cu new file mode 100644 index 0000000..2e1883a --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_2.cu @@ -0,0 +1,10 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(64, 64, 16, 2); +DECL_FATTN_MMA_F16_CASE(80, 80, 16, 2); +DECL_FATTN_MMA_F16_CASE(96, 96, 16, 2); +DECL_FATTN_MMA_F16_CASE(112, 112, 16, 2); +DECL_FATTN_MMA_F16_CASE(128, 128, 16, 2); +DECL_FATTN_MMA_F16_CASE(256, 256, 16, 2); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_4.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_4.cu new file mode 100644 index 0000000..517993c --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_4.cu @@ -0,0 +1,11 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(64, 64, 16, 4); +DECL_FATTN_MMA_F16_CASE(80, 80, 16, 4); +DECL_FATTN_MMA_F16_CASE(96, 96, 16, 4); +DECL_FATTN_MMA_F16_CASE(112, 112, 16, 4); +DECL_FATTN_MMA_F16_CASE(128, 128, 16, 4); +DECL_FATTN_MMA_F16_CASE(256, 256, 16, 4); +DECL_FATTN_MMA_F16_CASE(576, 512, 16, 4); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_16.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_16.cu new file mode 100644 index 0000000..f011a20 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_16.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(576, 512, 2, 16); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_32.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_32.cu new file mode 100644 index 0000000..264751d --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_32.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(576, 512, 2, 32); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_4.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_4.cu new file mode 100644 index 0000000..97b19c6 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_4.cu @@ -0,0 +1,11 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(64, 64, 2, 4); +DECL_FATTN_MMA_F16_CASE(80, 80, 2, 4); +DECL_FATTN_MMA_F16_CASE(96, 96, 2, 4); +DECL_FATTN_MMA_F16_CASE(112, 112, 2, 4); +DECL_FATTN_MMA_F16_CASE(128, 128, 2, 4); +DECL_FATTN_MMA_F16_CASE(256, 256, 2, 4); +DECL_FATTN_MMA_F16_CASE(576, 512, 2, 4); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_8.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_8.cu new file mode 100644 index 0000000..163b1d9 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_8.cu @@ -0,0 +1,10 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(64, 64, 2, 8); +DECL_FATTN_MMA_F16_CASE(80, 80, 2, 8); +DECL_FATTN_MMA_F16_CASE(96, 96, 2, 8); +DECL_FATTN_MMA_F16_CASE(112, 112, 2, 8); +DECL_FATTN_MMA_F16_CASE(128, 128, 2, 8); +DECL_FATTN_MMA_F16_CASE(256, 256, 2, 8); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_32-ncols2_1.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_32-ncols2_1.cu new file mode 100644 index 0000000..0543532 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_32-ncols2_1.cu @@ -0,0 +1,10 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(64, 64, 32, 1); +DECL_FATTN_MMA_F16_CASE(80, 80, 32, 1); +DECL_FATTN_MMA_F16_CASE(96, 96, 32, 1); +DECL_FATTN_MMA_F16_CASE(112, 112, 32, 1); +DECL_FATTN_MMA_F16_CASE(128, 128, 32, 1); +DECL_FATTN_MMA_F16_CASE(256, 256, 32, 1); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_32-ncols2_2.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_32-ncols2_2.cu new file mode 100644 index 0000000..407b6cf --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_32-ncols2_2.cu @@ -0,0 +1,10 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(64, 64, 32, 2); +DECL_FATTN_MMA_F16_CASE(80, 80, 32, 2); +DECL_FATTN_MMA_F16_CASE(96, 96, 32, 2); +DECL_FATTN_MMA_F16_CASE(112, 112, 32, 2); +DECL_FATTN_MMA_F16_CASE(128, 128, 32, 2); +DECL_FATTN_MMA_F16_CASE(256, 256, 32, 2); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_16.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_16.cu new file mode 100644 index 0000000..f5fd0e2 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_16.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(576, 512, 4, 16); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_2.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_2.cu new file mode 100644 index 0000000..5e46685 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_2.cu @@ -0,0 +1,10 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(64, 64, 4, 2); +DECL_FATTN_MMA_F16_CASE(80, 80, 4, 2); +DECL_FATTN_MMA_F16_CASE(96, 96, 4, 2); +DECL_FATTN_MMA_F16_CASE(112, 112, 4, 2); +DECL_FATTN_MMA_F16_CASE(128, 128, 4, 2); +DECL_FATTN_MMA_F16_CASE(256, 256, 4, 2); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_4.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_4.cu new file mode 100644 index 0000000..989626d --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_4.cu @@ -0,0 +1,11 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(64, 64, 4, 4); +DECL_FATTN_MMA_F16_CASE(80, 80, 4, 4); +DECL_FATTN_MMA_F16_CASE(96, 96, 4, 4); +DECL_FATTN_MMA_F16_CASE(112, 112, 4, 4); +DECL_FATTN_MMA_F16_CASE(128, 128, 4, 4); +DECL_FATTN_MMA_F16_CASE(256, 256, 4, 4); +DECL_FATTN_MMA_F16_CASE(576, 512, 4, 4); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_8.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_8.cu new file mode 100644 index 0000000..bad296b --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_8.cu @@ -0,0 +1,10 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(64, 64, 4, 8); +DECL_FATTN_MMA_F16_CASE(80, 80, 4, 8); +DECL_FATTN_MMA_F16_CASE(96, 96, 4, 8); +DECL_FATTN_MMA_F16_CASE(112, 112, 4, 8); +DECL_FATTN_MMA_F16_CASE(128, 128, 4, 8); +DECL_FATTN_MMA_F16_CASE(256, 256, 4, 8); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_64-ncols2_1.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_64-ncols2_1.cu new file mode 100644 index 0000000..0d7a9c7 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_64-ncols2_1.cu @@ -0,0 +1,10 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(64, 64, 64, 1); +DECL_FATTN_MMA_F16_CASE(80, 80, 64, 1); +DECL_FATTN_MMA_F16_CASE(96, 96, 64, 1); +DECL_FATTN_MMA_F16_CASE(112, 112, 64, 1); +DECL_FATTN_MMA_F16_CASE(128, 128, 64, 1); +DECL_FATTN_MMA_F16_CASE(256, 256, 64, 1); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_1.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_1.cu new file mode 100644 index 0000000..9d5a997 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_1.cu @@ -0,0 +1,10 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(64, 64, 8, 1); +DECL_FATTN_MMA_F16_CASE(80, 80, 8, 1); +DECL_FATTN_MMA_F16_CASE(96, 96, 8, 1); +DECL_FATTN_MMA_F16_CASE(112, 112, 8, 1); +DECL_FATTN_MMA_F16_CASE(128, 128, 8, 1); +DECL_FATTN_MMA_F16_CASE(256, 256, 8, 1); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_2.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_2.cu new file mode 100644 index 0000000..a6e6f09 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_2.cu @@ -0,0 +1,10 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(64, 64, 8, 2); +DECL_FATTN_MMA_F16_CASE(80, 80, 8, 2); +DECL_FATTN_MMA_F16_CASE(96, 96, 8, 2); +DECL_FATTN_MMA_F16_CASE(112, 112, 8, 2); +DECL_FATTN_MMA_F16_CASE(128, 128, 8, 2); +DECL_FATTN_MMA_F16_CASE(256, 256, 8, 2); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_4.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_4.cu new file mode 100644 index 0000000..173de7a --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_4.cu @@ -0,0 +1,11 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(64, 64, 8, 4); +DECL_FATTN_MMA_F16_CASE(80, 80, 8, 4); +DECL_FATTN_MMA_F16_CASE(96, 96, 8, 4); +DECL_FATTN_MMA_F16_CASE(112, 112, 8, 4); +DECL_FATTN_MMA_F16_CASE(128, 128, 8, 4); +DECL_FATTN_MMA_F16_CASE(256, 256, 8, 4); +DECL_FATTN_MMA_F16_CASE(576, 512, 8, 4); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_8.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_8.cu new file mode 100644 index 0000000..680a13c --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_8.cu @@ -0,0 +1,10 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +DECL_FATTN_MMA_F16_CASE(64, 64, 8, 8); +DECL_FATTN_MMA_F16_CASE(80, 80, 8, 8); +DECL_FATTN_MMA_F16_CASE(96, 96, 8, 8); +DECL_FATTN_MMA_F16_CASE(112, 112, 8, 8); +DECL_FATTN_MMA_F16_CASE(128, 128, 8, 8); +DECL_FATTN_MMA_F16_CASE(256, 256, 8, 8); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq112-dv112.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq112-dv112.cu new file mode 100644 index 0000000..a8b15ad --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq112-dv112.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-tile.cuh" + +DECL_FATTN_TILE_CASE(112, 112); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq128-dv128.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq128-dv128.cu new file mode 100644 index 0000000..1da1810 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq128-dv128.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-tile.cuh" + +DECL_FATTN_TILE_CASE(128, 128); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq256-dv256.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq256-dv256.cu new file mode 100644 index 0000000..bc65c72 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq256-dv256.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-tile.cuh" + +DECL_FATTN_TILE_CASE(256, 256); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq40-dv40.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq40-dv40.cu new file mode 100644 index 0000000..10b330f --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq40-dv40.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-tile.cuh" + +DECL_FATTN_TILE_CASE(40, 40); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq576-dv512.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq576-dv512.cu new file mode 100644 index 0000000..254b7d2 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq576-dv512.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-tile.cuh" + +DECL_FATTN_TILE_CASE(576, 512); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq64-dv64.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq64-dv64.cu new file mode 100644 index 0000000..5caffac --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq64-dv64.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-tile.cuh" + +DECL_FATTN_TILE_CASE(64, 64); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq72-dv72.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq72-dv72.cu new file mode 100644 index 0000000..8f9d531 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq72-dv72.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-tile.cuh" + +DECL_FATTN_TILE_CASE(72, 72); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq80-dv80.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq80-dv80.cu new file mode 100644 index 0000000..90abb3b --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq80-dv80.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-tile.cuh" + +DECL_FATTN_TILE_CASE(80, 80); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq96-dv96.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq96-dv96.cu new file mode 100644 index 0000000..7292c0a --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq96-dv96.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-tile.cuh" + +DECL_FATTN_TILE_CASE(96, 96); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-f16.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-f16.cu new file mode 100644 index 0000000..c357abd --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-f16.cu @@ -0,0 +1,7 @@ +// 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_F16, GGML_TYPE_F16); +DECL_FATTN_VEC_CASE(128, GGML_TYPE_F16, GGML_TYPE_F16); +DECL_FATTN_VEC_CASE(256, GGML_TYPE_F16, GGML_TYPE_F16); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q4_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q4_0.cu new file mode 100644 index 0000000..4b14865 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q4_0.cu @@ -0,0 +1,7 @@ +// 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_F16, GGML_TYPE_Q4_0); +DECL_FATTN_VEC_CASE(128, GGML_TYPE_F16, GGML_TYPE_Q4_0); +DECL_FATTN_VEC_CASE(256, GGML_TYPE_F16, GGML_TYPE_Q4_0); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q4_1.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q4_1.cu new file mode 100644 index 0000000..ef77157 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q4_1.cu @@ -0,0 +1,7 @@ +// 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_F16, GGML_TYPE_Q4_1); +DECL_FATTN_VEC_CASE(128, GGML_TYPE_F16, GGML_TYPE_Q4_1); +DECL_FATTN_VEC_CASE(256, GGML_TYPE_F16, GGML_TYPE_Q4_1); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q5_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q5_0.cu new file mode 100644 index 0000000..9ae11cc --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q5_0.cu @@ -0,0 +1,7 @@ +// 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_F16, GGML_TYPE_Q5_0); +DECL_FATTN_VEC_CASE(128, GGML_TYPE_F16, GGML_TYPE_Q5_0); +DECL_FATTN_VEC_CASE(256, GGML_TYPE_F16, GGML_TYPE_Q5_0); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q5_1.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q5_1.cu new file mode 100644 index 0000000..10ed48a --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q5_1.cu @@ -0,0 +1,7 @@ +// 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_F16, GGML_TYPE_Q5_1); +DECL_FATTN_VEC_CASE(128, GGML_TYPE_F16, GGML_TYPE_Q5_1); +DECL_FATTN_VEC_CASE(256, GGML_TYPE_F16, GGML_TYPE_Q5_1); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q8_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q8_0.cu new file mode 100644 index 0000000..4fcc3f3 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q8_0.cu @@ -0,0 +1,7 @@ +// 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_F16, GGML_TYPE_Q8_0); +DECL_FATTN_VEC_CASE(128, GGML_TYPE_F16, GGML_TYPE_Q8_0); +DECL_FATTN_VEC_CASE(256, GGML_TYPE_F16, GGML_TYPE_Q8_0); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-f16.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-f16.cu new file mode 100644 index 0000000..7ca5053 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-f16.cu @@ -0,0 +1,7 @@ +// 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_Q4_0, GGML_TYPE_F16); +DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_F16); +DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_0, GGML_TYPE_F16); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q4_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q4_0.cu new file mode 100644 index 0000000..6ef1a48 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q4_0.cu @@ -0,0 +1,7 @@ +// 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_Q4_0, GGML_TYPE_Q4_0); +DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q4_0); +DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_0, GGML_TYPE_Q4_0); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q4_1.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q4_1.cu new file mode 100644 index 0000000..4c0532c --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q4_1.cu @@ -0,0 +1,7 @@ +// 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_Q4_0, GGML_TYPE_Q4_1); +DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q4_1); +DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_0, GGML_TYPE_Q4_1); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q5_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q5_0.cu new file mode 100644 index 0000000..ed3d7ba --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q5_0.cu @@ -0,0 +1,7 @@ +// 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_Q4_0, GGML_TYPE_Q5_0); +DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q5_0); +DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_0, GGML_TYPE_Q5_0); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q5_1.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q5_1.cu new file mode 100644 index 0000000..687f254 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q5_1.cu @@ -0,0 +1,7 @@ +// 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_Q4_0, GGML_TYPE_Q5_1); +DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q5_1); +DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_0, GGML_TYPE_Q5_1); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q8_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q8_0.cu new file mode 100644 index 0000000..41107c4 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q8_0.cu @@ -0,0 +1,7 @@ +// 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_Q4_0, GGML_TYPE_Q8_0); +DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q8_0); +DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_0, GGML_TYPE_Q8_0); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-f16.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-f16.cu new file mode 100644 index 0000000..d523ce0 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-f16.cu @@ -0,0 +1,7 @@ +// 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_Q4_1, GGML_TYPE_F16); +DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_F16); +DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_1, GGML_TYPE_F16); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q4_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q4_0.cu new file mode 100644 index 0000000..8b9ed35 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q4_0.cu @@ -0,0 +1,7 @@ +// 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_Q4_1, GGML_TYPE_Q4_0); +DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_Q4_0); +DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_1, GGML_TYPE_Q4_0); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q4_1.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q4_1.cu new file mode 100644 index 0000000..0553e46 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q4_1.cu @@ -0,0 +1,7 @@ +// 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_Q4_1, GGML_TYPE_Q4_1); +DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_Q4_1); +DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_1, GGML_TYPE_Q4_1); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q5_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q5_0.cu new file mode 100644 index 0000000..8390eaf --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q5_0.cu @@ -0,0 +1,7 @@ +// 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_Q4_1, GGML_TYPE_Q5_0); +DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_Q5_0); +DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_1, GGML_TYPE_Q5_0); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q5_1.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q5_1.cu new file mode 100644 index 0000000..f61e19d --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q5_1.cu @@ -0,0 +1,7 @@ +// 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_Q4_1, GGML_TYPE_Q5_1); +DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_Q5_1); +DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_1, GGML_TYPE_Q5_1); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q8_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q8_0.cu new file mode 100644 index 0000000..86a1882 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q8_0.cu @@ -0,0 +1,7 @@ +// 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_Q4_1, GGML_TYPE_Q8_0); +DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_Q8_0); +DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_1, GGML_TYPE_Q8_0); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-f16.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-f16.cu new file mode 100644 index 0000000..1d7af47 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-f16.cu @@ -0,0 +1,7 @@ +// 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); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q4_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q4_0.cu new file mode 100644 index 0000000..837224d --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q4_0.cu @@ -0,0 +1,7 @@ +// 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); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q4_1.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q4_1.cu new file mode 100644 index 0000000..0dd7dd6 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q4_1.cu @@ -0,0 +1,7 @@ +// 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); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q5_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q5_0.cu new file mode 100644 index 0000000..41b859f --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q5_0.cu @@ -0,0 +1,7 @@ +// 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); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q5_1.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q5_1.cu new file mode 100644 index 0000000..d2e5ffd --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q5_1.cu @@ -0,0 +1,7 @@ +// 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); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q8_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q8_0.cu new file mode 100644 index 0000000..81ff740 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q8_0.cu @@ -0,0 +1,7 @@ +// 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); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-f16.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-f16.cu new file mode 100644 index 0000000..a38dae1 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-f16.cu @@ -0,0 +1,7 @@ +// 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); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q4_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q4_0.cu new file mode 100644 index 0000000..2304571 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q4_0.cu @@ -0,0 +1,7 @@ +// 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); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q4_1.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q4_1.cu new file mode 100644 index 0000000..84b83e5 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q4_1.cu @@ -0,0 +1,7 @@ +// 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); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q5_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q5_0.cu new file mode 100644 index 0000000..39f80e2 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q5_0.cu @@ -0,0 +1,7 @@ +// 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); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q5_1.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q5_1.cu new file mode 100644 index 0000000..cf4e661 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q5_1.cu @@ -0,0 +1,7 @@ +// 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); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q8_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q8_0.cu new file mode 100644 index 0000000..6565418 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q8_0.cu @@ -0,0 +1,7 @@ +// 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); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-f16.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-f16.cu new file mode 100644 index 0000000..a1bc3f5 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-f16.cu @@ -0,0 +1,7 @@ +// 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); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q4_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q4_0.cu new file mode 100644 index 0000000..4b76a9b --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q4_0.cu @@ -0,0 +1,7 @@ +// 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); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q4_1.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q4_1.cu new file mode 100644 index 0000000..77d0412 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q4_1.cu @@ -0,0 +1,7 @@ +// 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); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q5_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q5_0.cu new file mode 100644 index 0000000..6e170fe --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q5_0.cu @@ -0,0 +1,7 @@ +// 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); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q5_1.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q5_1.cu new file mode 100644 index 0000000..b617cd7 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q5_1.cu @@ -0,0 +1,7 @@ +// 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); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q8_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q8_0.cu new file mode 100644 index 0000000..a5b768b --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q8_0.cu @@ -0,0 +1,7 @@ +// 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); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/generate_cu_files.py b/llama.cpp/ggml/src/ggml-cuda/template-instances/generate_cu_files.py new file mode 100755 index 0000000..e382df1 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/generate_cu_files.py @@ -0,0 +1,99 @@ +#!/usr/bin/env python3 + +from glob import glob +import os + +HEAD_SIZES_KQ = [40, 64, 72, 80, 96, 112, 128, 256, 576] + +TYPES_KV = ["GGML_TYPE_F16", "GGML_TYPE_Q4_0", "GGML_TYPE_Q4_1", "GGML_TYPE_Q5_0", "GGML_TYPE_Q5_1", "GGML_TYPE_Q8_0"] + +SOURCE_FATTN_TILE = """// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-tile.cuh" + +DECL_FATTN_TILE_CASE({head_size_kq}, {head_size_v}); +""" + +SOURCE_FATTN_VEC = """// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-vec.cuh" + +DECL_FATTN_VEC_CASE( 64, {type_k}, {type_v}); +DECL_FATTN_VEC_CASE(128, {type_k}, {type_v}); +DECL_FATTN_VEC_CASE(256, {type_k}, {type_v}); +""" + +SOURCE_FATTN_MMA_START = """// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../fattn-mma-f16.cuh" + +""" + +SOURCE_FATTN_MMA_CASE = "DECL_FATTN_MMA_F16_CASE({head_size_kq}, {head_size_v}, {ncols1}, {ncols2});\n" + +TYPES_MMQ = [ + "GGML_TYPE_Q4_0", "GGML_TYPE_Q4_1", "GGML_TYPE_Q5_0", "GGML_TYPE_Q5_1", "GGML_TYPE_Q8_0", + "GGML_TYPE_Q2_K", "GGML_TYPE_Q3_K", "GGML_TYPE_Q4_K", "GGML_TYPE_Q5_K", "GGML_TYPE_Q6_K", + "GGML_TYPE_IQ2_XXS", "GGML_TYPE_IQ2_XS", "GGML_TYPE_IQ2_S", "GGML_TYPE_IQ3_XXS", "GGML_TYPE_IQ3_S", + "GGML_TYPE_IQ1_S", "GGML_TYPE_IQ4_NL", "GGML_TYPE_IQ4_XS", "GGML_TYPE_MXFP4" +] + +SOURCE_MMQ = """// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE({type}); +""" + +SOURCE_MMF = """// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmf.cuh" + +DECL_MMF_CASE({type}); +""" + + +def get_short_name(long_quant_name): + return long_quant_name.replace("GGML_TYPE_", "").lower() + + +for filename in glob("*.cu"): + os.remove(filename) + +for head_size_kq in HEAD_SIZES_KQ: + head_size_v = head_size_kq if head_size_kq != 576 else 512 + with open(f"fattn-tile-instance-dkq{head_size_kq}-dv{head_size_v}.cu", "w") as f: + f.write(SOURCE_FATTN_TILE.format(head_size_kq=head_size_kq, head_size_v=head_size_v)) + +for type_k in TYPES_KV: + for type_v in TYPES_KV: + with open(f"fattn-vec-instance-{get_short_name(type_k)}-{get_short_name(type_v)}.cu", "w") as f: + f.write(SOURCE_FATTN_VEC.format(type_k=type_k, type_v=type_v)) + +for ncols in [8, 16, 32, 64]: + for ncols2 in [1, 2, 4, 8, 16, 32]: + if ncols2 > ncols: + continue + ncols1 = ncols // ncols2 + with open(f"fattn-mma-f16-instance-ncols1_{ncols1}-ncols2_{ncols2}.cu", "w") as f: + f.write(SOURCE_FATTN_MMA_START) + + for head_size_kq in HEAD_SIZES_KQ: + if head_size_kq == 40: + continue + if head_size_kq == 72: + continue + if head_size_kq != 576 and ncols2 in (16, 32): + continue + if head_size_kq == 576 and ncols2 not in (4, 16, 32): + continue + head_size_v = head_size_kq if head_size_kq != 576 else 512 + f.write(SOURCE_FATTN_MMA_CASE.format(ncols1=ncols1, ncols2=ncols2, head_size_kq=head_size_kq, head_size_v=head_size_v)) + +for type in TYPES_MMQ: + with open(f"mmq-instance-{get_short_name(type)}.cu", "w") as f: + f.write(SOURCE_MMQ.format(type=type)) + +for type in range(1, 17): + with open(f"mmf-instance-ncols_{type}.cu", "w") as f: + f.write(SOURCE_MMF.format(type=type)) diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_1.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_1.cu new file mode 100644 index 0000000..f594d5d --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_1.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmf.cuh" + +DECL_MMF_CASE(1); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_10.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_10.cu new file mode 100644 index 0000000..9cc6772 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_10.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmf.cuh" + +DECL_MMF_CASE(10); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_11.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_11.cu new file mode 100644 index 0000000..317f487 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_11.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmf.cuh" + +DECL_MMF_CASE(11); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_12.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_12.cu new file mode 100644 index 0000000..dc00332 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_12.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmf.cuh" + +DECL_MMF_CASE(12); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_13.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_13.cu new file mode 100644 index 0000000..0782101 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_13.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmf.cuh" + +DECL_MMF_CASE(13); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_14.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_14.cu new file mode 100644 index 0000000..a23ad6a --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_14.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmf.cuh" + +DECL_MMF_CASE(14); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_15.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_15.cu new file mode 100644 index 0000000..0fe3f78 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_15.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmf.cuh" + +DECL_MMF_CASE(15); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_16.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_16.cu new file mode 100644 index 0000000..5440863 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_16.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmf.cuh" + +DECL_MMF_CASE(16); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_2.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_2.cu new file mode 100644 index 0000000..3b90179 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_2.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmf.cuh" + +DECL_MMF_CASE(2); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_3.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_3.cu new file mode 100644 index 0000000..56e940b --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_3.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmf.cuh" + +DECL_MMF_CASE(3); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_4.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_4.cu new file mode 100644 index 0000000..a7665d4 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_4.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmf.cuh" + +DECL_MMF_CASE(4); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_5.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_5.cu new file mode 100644 index 0000000..3a1dff2 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_5.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmf.cuh" + +DECL_MMF_CASE(5); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_6.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_6.cu new file mode 100644 index 0000000..400fb7c --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_6.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmf.cuh" + +DECL_MMF_CASE(6); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_7.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_7.cu new file mode 100644 index 0000000..954a1c7 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_7.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmf.cuh" + +DECL_MMF_CASE(7); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_8.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_8.cu new file mode 100644 index 0000000..f1bd09c --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_8.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmf.cuh" + +DECL_MMF_CASE(8); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_9.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_9.cu new file mode 100644 index 0000000..1255ac2 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_9.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmf.cuh" + +DECL_MMF_CASE(9); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq1_s.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq1_s.cu new file mode 100644 index 0000000..84ec850 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq1_s.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_IQ1_S); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_s.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_s.cu new file mode 100644 index 0000000..583c4e5 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_s.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_IQ2_S); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_xs.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_xs.cu new file mode 100644 index 0000000..edaf156 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_xs.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_IQ2_XS); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_xxs.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_xxs.cu new file mode 100644 index 0000000..233d934 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_xxs.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_IQ2_XXS); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq3_s.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq3_s.cu new file mode 100644 index 0000000..6092dc7 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq3_s.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_IQ3_S); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq3_xxs.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq3_xxs.cu new file mode 100644 index 0000000..1d5bd20 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq3_xxs.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_IQ3_XXS); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_nl.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_nl.cu new file mode 100644 index 0000000..eb02fab --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_nl.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_IQ4_NL); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_xs.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_xs.cu new file mode 100644 index 0000000..1eb3b74 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_xs.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_IQ4_XS); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-mxfp4.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-mxfp4.cu new file mode 100644 index 0000000..c14624c --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-mxfp4.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_MXFP4); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q2_k.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q2_k.cu new file mode 100644 index 0000000..6415369 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q2_k.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_Q2_K); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q3_k.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q3_k.cu new file mode 100644 index 0000000..ffb6213 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q3_k.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_Q3_K); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_0.cu new file mode 100644 index 0000000..0c0b0c8 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_0.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_Q4_0); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_1.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_1.cu new file mode 100644 index 0000000..ee67f69 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_1.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_Q4_1); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_k.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_k.cu new file mode 100644 index 0000000..9eeb3cd --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_k.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_Q4_K); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_0.cu new file mode 100644 index 0000000..cc57fb9 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_0.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_Q5_0); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_1.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_1.cu new file mode 100644 index 0000000..721ac79 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_1.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_Q5_1); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_k.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_k.cu new file mode 100644 index 0000000..a2e90ff --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_k.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_Q5_K); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q6_k.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q6_k.cu new file mode 100644 index 0000000..470938f --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q6_k.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_Q6_K); diff --git a/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q8_0.cu b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q8_0.cu new file mode 100644 index 0000000..974477b --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/template-instances/mmq-instance-q8_0.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_Q8_0); |