From 503983a69a3107a70dc42b69164665743861fb28 Mon Sep 17 00:00:00 2001 From: Georgi Gerganov Date: Wed, 14 Aug 2024 10:29:23 +0300 Subject: [PATCH] cuda : build only necessary templates --- ggml/src/ggml-cuda/fattn.cu | 12 --- ggml/src/ggml-cuda/mmq.cu | 90 +++++++++---------- ggml/src/ggml-cuda/mmq.cuh | 30 +++---- .../fattn-vec-f16-instance-hs128-f16-q4_0.cu | 5 -- .../fattn-vec-f16-instance-hs128-f16-q4_1.cu | 5 -- .../fattn-vec-f16-instance-hs128-f16-q5_0.cu | 5 -- .../fattn-vec-f16-instance-hs128-f16-q5_1.cu | 5 -- .../fattn-vec-f16-instance-hs128-f16-q8_0.cu | 5 -- .../fattn-vec-f16-instance-hs128-q4_0-f16.cu | 5 -- .../fattn-vec-f16-instance-hs128-q4_0-q4_0.cu | 5 -- .../fattn-vec-f16-instance-hs128-q4_0-q4_1.cu | 5 -- .../fattn-vec-f16-instance-hs128-q4_0-q5_0.cu | 5 -- .../fattn-vec-f16-instance-hs128-q4_0-q5_1.cu | 5 -- .../fattn-vec-f16-instance-hs128-q4_0-q8_0.cu | 5 -- .../fattn-vec-f16-instance-hs128-q4_1-f16.cu | 5 -- .../fattn-vec-f16-instance-hs128-q4_1-q4_0.cu | 5 -- .../fattn-vec-f16-instance-hs128-q4_1-q4_1.cu | 5 -- .../fattn-vec-f16-instance-hs128-q4_1-q5_0.cu | 5 -- .../fattn-vec-f16-instance-hs128-q4_1-q5_1.cu | 5 -- .../fattn-vec-f16-instance-hs128-q4_1-q8_0.cu | 5 -- .../fattn-vec-f16-instance-hs128-q5_0-f16.cu | 5 -- .../fattn-vec-f16-instance-hs128-q5_0-q4_0.cu | 5 -- .../fattn-vec-f16-instance-hs128-q5_0-q4_1.cu | 5 -- .../fattn-vec-f16-instance-hs128-q5_0-q5_0.cu | 5 -- .../fattn-vec-f16-instance-hs128-q5_0-q5_1.cu | 5 -- .../fattn-vec-f16-instance-hs128-q5_0-q8_0.cu | 5 -- .../fattn-vec-f16-instance-hs128-q5_1-f16.cu | 5 -- .../fattn-vec-f16-instance-hs128-q5_1-q4_0.cu | 5 -- .../fattn-vec-f16-instance-hs128-q5_1-q4_1.cu | 5 -- .../fattn-vec-f16-instance-hs128-q5_1-q5_0.cu | 5 -- .../fattn-vec-f16-instance-hs128-q5_1-q5_1.cu | 5 -- .../fattn-vec-f16-instance-hs128-q5_1-q8_0.cu | 5 -- .../fattn-vec-f16-instance-hs128-q8_0-f16.cu | 5 -- .../fattn-vec-f16-instance-hs128-q8_0-q4_0.cu | 5 -- .../fattn-vec-f16-instance-hs128-q8_0-q4_1.cu | 5 -- .../fattn-vec-f16-instance-hs128-q8_0-q5_0.cu | 5 -- .../fattn-vec-f16-instance-hs128-q8_0-q5_1.cu | 5 -- .../fattn-vec-f16-instance-hs128-q8_0-q8_0.cu | 5 -- .../fattn-vec-f16-instance-hs256-f16-f16.cu | 5 -- .../fattn-vec-f16-instance-hs64-f16-f16.cu | 5 -- .../fattn-vec-f16-instance-hs64-f16-q4_0.cu | 5 -- .../fattn-vec-f16-instance-hs64-f16-q4_1.cu | 5 -- .../fattn-vec-f16-instance-hs64-f16-q5_0.cu | 5 -- .../fattn-vec-f16-instance-hs64-f16-q5_1.cu | 5 -- .../fattn-vec-f16-instance-hs64-f16-q8_0.cu | 5 -- .../fattn-vec-f32-instance-hs128-f16-q4_0.cu | 5 -- .../fattn-vec-f32-instance-hs128-f16-q4_1.cu | 5 -- .../fattn-vec-f32-instance-hs128-f16-q5_0.cu | 5 -- .../fattn-vec-f32-instance-hs128-f16-q5_1.cu | 5 -- .../fattn-vec-f32-instance-hs128-f16-q8_0.cu | 5 -- .../fattn-vec-f32-instance-hs128-q4_0-f16.cu | 5 -- .../fattn-vec-f32-instance-hs128-q4_0-q4_0.cu | 5 -- .../fattn-vec-f32-instance-hs128-q4_0-q4_1.cu | 5 -- .../fattn-vec-f32-instance-hs128-q4_0-q5_0.cu | 5 -- .../fattn-vec-f32-instance-hs128-q4_0-q5_1.cu | 5 -- .../fattn-vec-f32-instance-hs128-q4_0-q8_0.cu | 5 -- .../fattn-vec-f32-instance-hs128-q4_1-f16.cu | 5 -- .../fattn-vec-f32-instance-hs128-q4_1-q4_0.cu | 5 -- .../fattn-vec-f32-instance-hs128-q4_1-q4_1.cu | 5 -- .../fattn-vec-f32-instance-hs128-q4_1-q5_0.cu | 5 -- .../fattn-vec-f32-instance-hs128-q4_1-q5_1.cu | 5 -- .../fattn-vec-f32-instance-hs128-q4_1-q8_0.cu | 5 -- .../fattn-vec-f32-instance-hs128-q5_0-f16.cu | 5 -- .../fattn-vec-f32-instance-hs128-q5_0-q4_0.cu | 5 -- .../fattn-vec-f32-instance-hs128-q5_0-q4_1.cu | 5 -- .../fattn-vec-f32-instance-hs128-q5_0-q5_0.cu | 5 -- .../fattn-vec-f32-instance-hs128-q5_0-q5_1.cu | 5 -- .../fattn-vec-f32-instance-hs128-q5_0-q8_0.cu | 5 -- .../fattn-vec-f32-instance-hs128-q5_1-f16.cu | 5 -- .../fattn-vec-f32-instance-hs128-q5_1-q4_0.cu | 5 -- .../fattn-vec-f32-instance-hs128-q5_1-q4_1.cu | 5 -- .../fattn-vec-f32-instance-hs128-q5_1-q5_0.cu | 5 -- .../fattn-vec-f32-instance-hs128-q5_1-q5_1.cu | 5 -- .../fattn-vec-f32-instance-hs128-q5_1-q8_0.cu | 5 -- .../fattn-vec-f32-instance-hs128-q8_0-f16.cu | 5 -- .../fattn-vec-f32-instance-hs128-q8_0-q4_0.cu | 5 -- .../fattn-vec-f32-instance-hs128-q8_0-q4_1.cu | 5 -- .../fattn-vec-f32-instance-hs128-q8_0-q5_0.cu | 5 -- .../fattn-vec-f32-instance-hs128-q8_0-q5_1.cu | 5 -- .../fattn-vec-f32-instance-hs128-q8_0-q8_0.cu | 5 -- .../fattn-vec-f32-instance-hs256-f16-f16.cu | 5 -- .../fattn-vec-f32-instance-hs64-f16-f16.cu | 5 -- .../fattn-vec-f32-instance-hs64-f16-q4_0.cu | 5 -- .../fattn-vec-f32-instance-hs64-f16-q4_1.cu | 5 -- .../fattn-vec-f32-instance-hs64-f16-q5_0.cu | 5 -- .../fattn-vec-f32-instance-hs64-f16-q5_1.cu | 5 -- .../fattn-vec-f32-instance-hs64-f16-q8_0.cu | 5 -- .../template-instances/mmq-instance-iq1_s.cu | 5 -- .../template-instances/mmq-instance-iq2_s.cu | 5 -- .../template-instances/mmq-instance-iq2_xs.cu | 5 -- .../mmq-instance-iq2_xxs.cu | 5 -- .../template-instances/mmq-instance-iq3_s.cu | 5 -- .../mmq-instance-iq3_xxs.cu | 5 -- .../template-instances/mmq-instance-iq4_nl.cu | 5 -- .../template-instances/mmq-instance-iq4_xs.cu | 5 -- .../template-instances/mmq-instance-q2_k.cu | 5 -- .../template-instances/mmq-instance-q3_k.cu | 5 -- .../template-instances/mmq-instance-q4_0.cu | 5 -- .../template-instances/mmq-instance-q4_1.cu | 5 -- .../template-instances/mmq-instance-q5_0.cu | 5 -- .../template-instances/mmq-instance-q5_1.cu | 5 -- .../template-instances/mmq-instance-q5_k.cu | 5 -- 102 files changed, 60 insertions(+), 567 deletions(-) delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q8_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-f16.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q8_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-f16.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q8_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-f16.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q8_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-f16.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q8_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-f16.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q8_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs256-f16-f16.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-f16.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q8_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q8_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-f16.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q8_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-f16.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q8_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-f16.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q8_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-f16.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q8_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-f16.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q8_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs256-f16-f16.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-f16.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q8_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/mmq-instance-iq1_s.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_s.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_xs.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_xxs.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/mmq-instance-iq3_s.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/mmq-instance-iq3_xxs.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_nl.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_xs.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/mmq-instance-q2_k.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/mmq-instance-q3_k.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/mmq-instance-q4_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/mmq-instance-q4_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/mmq-instance-q5_0.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/mmq-instance-q5_1.cu delete mode 100644 ggml/src/ggml-cuda/template-instances/mmq-instance-q5_k.cu diff --git a/ggml/src/ggml-cuda/fattn.cu b/ggml/src/ggml-cuda/fattn.cu index 29f608b0f..8e939250b 100644 --- a/ggml/src/ggml-cuda/fattn.cu +++ b/ggml/src/ggml-cuda/fattn.cu @@ -208,13 +208,7 @@ static void ggml_cuda_flash_attn_ext_vec_f16(ggml_backend_cuda_context & ctx, gg FATTN_VEC_F16_CASE(256, GGML_TYPE_F16, GGML_TYPE_F16) #else - FATTN_VEC_F16_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q4_0) - - FATTN_VEC_F16_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q8_0) - - FATTN_VEC_F16_CASE( 64, GGML_TYPE_F16, GGML_TYPE_F16) FATTN_VEC_F16_CASE(128, GGML_TYPE_F16, GGML_TYPE_F16) - FATTN_VEC_F16_CASE(256, GGML_TYPE_F16, GGML_TYPE_F16) #endif // GGML_CUDA_FA_ALL_QUANTS on_no_fattn_vec_case(Q->ne[0]); @@ -283,13 +277,7 @@ static void ggml_cuda_flash_attn_ext_vec_f32(ggml_backend_cuda_context & ctx, gg FATTN_VEC_F32_CASE(256, GGML_TYPE_F16, GGML_TYPE_F16) #else - FATTN_VEC_F32_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q4_0) - - FATTN_VEC_F32_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q8_0) - - FATTN_VEC_F32_CASE( 64, GGML_TYPE_F16, GGML_TYPE_F16) FATTN_VEC_F32_CASE(128, GGML_TYPE_F16, GGML_TYPE_F16) - FATTN_VEC_F32_CASE(256, GGML_TYPE_F16, GGML_TYPE_F16) #endif // GGML_CUDA_FA_ALL_QUANTS on_no_fattn_vec_case(Q->ne[0]); diff --git a/ggml/src/ggml-cuda/mmq.cu b/ggml/src/ggml-cuda/mmq.cu index 78d70cd7a..66036bdd1 100644 --- a/ggml/src/ggml-cuda/mmq.cu +++ b/ggml/src/ggml-cuda/mmq.cu @@ -29,60 +29,60 @@ void ggml_cuda_op_mul_mat_q( const mmq_args args = {src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, stride00, src1_padded_row_size, src1_ncols, ne11, nrows_dst}; switch (src0->type) { - case GGML_TYPE_Q4_0: - mul_mat_q_case(ctx, args, stream); - break; - case GGML_TYPE_Q4_1: - mul_mat_q_case(ctx, args, stream); - break; - case GGML_TYPE_Q5_0: - mul_mat_q_case(ctx, args, stream); - break; - case GGML_TYPE_Q5_1: - mul_mat_q_case(ctx, args, stream); - break; +// case GGML_TYPE_Q4_0: +// mul_mat_q_case(ctx, args, stream); +// break; +// case GGML_TYPE_Q4_1: +// mul_mat_q_case(ctx, args, stream); +// break; +// case GGML_TYPE_Q5_0: +// mul_mat_q_case(ctx, args, stream); +// break; +// case GGML_TYPE_Q5_1: +// mul_mat_q_case(ctx, args, stream); +// break; case GGML_TYPE_Q8_0: mul_mat_q_case(ctx, args, stream); break; - case GGML_TYPE_Q2_K: - mul_mat_q_case(ctx, args, stream); - break; - case GGML_TYPE_Q3_K: - mul_mat_q_case(ctx, args, stream); - break; +// case GGML_TYPE_Q2_K: +// mul_mat_q_case(ctx, args, stream); +// break; +// case GGML_TYPE_Q3_K: +// mul_mat_q_case(ctx, args, stream); +// break; case GGML_TYPE_Q4_K: mul_mat_q_case(ctx, args, stream); break; - case GGML_TYPE_Q5_K: - mul_mat_q_case(ctx, args, stream); - break; +// case GGML_TYPE_Q5_K: +// mul_mat_q_case(ctx, args, stream); +// break; case GGML_TYPE_Q6_K: mul_mat_q_case(ctx, args, stream); break; - case GGML_TYPE_IQ2_XXS: - mul_mat_q_case(ctx, args, stream); - break; - case GGML_TYPE_IQ2_XS: - mul_mat_q_case(ctx, args, stream); - break; - case GGML_TYPE_IQ2_S: - mul_mat_q_case(ctx, args, stream); - break; - case GGML_TYPE_IQ3_XXS: - mul_mat_q_case(ctx, args, stream); - break; - case GGML_TYPE_IQ3_S: - mul_mat_q_case(ctx, args, stream); - break; - case GGML_TYPE_IQ1_S: - mul_mat_q_case(ctx, args, stream); - break; - case GGML_TYPE_IQ4_XS: - mul_mat_q_case(ctx, args, stream); - break; - case GGML_TYPE_IQ4_NL: - mul_mat_q_case(ctx, args, stream); - break; +// case GGML_TYPE_IQ2_XXS: +// mul_mat_q_case(ctx, args, stream); +// break; +// case GGML_TYPE_IQ2_XS: +// mul_mat_q_case(ctx, args, stream); +// break; +// case GGML_TYPE_IQ2_S: +// mul_mat_q_case(ctx, args, stream); +// break; +// case GGML_TYPE_IQ3_XXS: +// mul_mat_q_case(ctx, args, stream); +// break; +// case GGML_TYPE_IQ3_S: +// mul_mat_q_case(ctx, args, stream); +// break; +// case GGML_TYPE_IQ1_S: +// mul_mat_q_case(ctx, args, stream); +// break; +// case GGML_TYPE_IQ4_XS: +// mul_mat_q_case(ctx, args, stream); +// break; +// case GGML_TYPE_IQ4_NL: +// mul_mat_q_case(ctx, args, stream); +// break; default: GGML_ABORT("fatal error"); break; diff --git a/ggml/src/ggml-cuda/mmq.cuh b/ggml/src/ggml-cuda/mmq.cuh index e8a957447..b4a31b982 100644 --- a/ggml/src/ggml-cuda/mmq.cuh +++ b/ggml/src/ggml-cuda/mmq.cuh @@ -2906,24 +2906,24 @@ void mul_mat_q_case(ggml_backend_cuda_context & ctx, const mmq_args & args, cuda #define DECL_MMQ_CASE(type) \ template void mul_mat_q_case(ggml_backend_cuda_context & ctx, const mmq_args & args, cudaStream_t stream) \ -extern DECL_MMQ_CASE(GGML_TYPE_Q4_0); -extern DECL_MMQ_CASE(GGML_TYPE_Q4_1); -extern DECL_MMQ_CASE(GGML_TYPE_Q5_0); -extern DECL_MMQ_CASE(GGML_TYPE_Q5_1); +//extern DECL_MMQ_CASE(GGML_TYPE_Q4_0); +//extern DECL_MMQ_CASE(GGML_TYPE_Q4_1); +//extern DECL_MMQ_CASE(GGML_TYPE_Q5_0); +//extern DECL_MMQ_CASE(GGML_TYPE_Q5_1); extern DECL_MMQ_CASE(GGML_TYPE_Q8_0); -extern DECL_MMQ_CASE(GGML_TYPE_Q2_K); -extern DECL_MMQ_CASE(GGML_TYPE_Q3_K); +//extern DECL_MMQ_CASE(GGML_TYPE_Q2_K); +//extern DECL_MMQ_CASE(GGML_TYPE_Q3_K); extern DECL_MMQ_CASE(GGML_TYPE_Q4_K); -extern DECL_MMQ_CASE(GGML_TYPE_Q5_K); +//extern DECL_MMQ_CASE(GGML_TYPE_Q5_K); extern DECL_MMQ_CASE(GGML_TYPE_Q6_K); -extern DECL_MMQ_CASE(GGML_TYPE_IQ2_XXS); -extern DECL_MMQ_CASE(GGML_TYPE_IQ2_XS); -extern DECL_MMQ_CASE(GGML_TYPE_IQ2_S); -extern DECL_MMQ_CASE(GGML_TYPE_IQ3_XXS); -extern DECL_MMQ_CASE(GGML_TYPE_IQ3_S); -extern DECL_MMQ_CASE(GGML_TYPE_IQ1_S); -extern DECL_MMQ_CASE(GGML_TYPE_IQ4_NL); -extern DECL_MMQ_CASE(GGML_TYPE_IQ4_XS); +//extern DECL_MMQ_CASE(GGML_TYPE_IQ2_XXS); +//extern DECL_MMQ_CASE(GGML_TYPE_IQ2_XS); +//extern DECL_MMQ_CASE(GGML_TYPE_IQ2_S); +//extern DECL_MMQ_CASE(GGML_TYPE_IQ3_XXS); +//extern DECL_MMQ_CASE(GGML_TYPE_IQ3_S); +//extern DECL_MMQ_CASE(GGML_TYPE_IQ1_S); +//extern DECL_MMQ_CASE(GGML_TYPE_IQ4_NL); +//extern DECL_MMQ_CASE(GGML_TYPE_IQ4_XS); // ------------------------------------------------------------------------------------------------------------------------- diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_0.cu deleted file mode 100644 index dd070db28..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_F16, GGML_TYPE_Q4_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_1.cu deleted file mode 100644 index 54dcde6f5..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_F16, GGML_TYPE_Q4_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_0.cu deleted file mode 100644 index 4ec22f791..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_F16, GGML_TYPE_Q5_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_1.cu deleted file mode 100644 index 3c15bf7f0..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_F16, GGML_TYPE_Q5_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q8_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q8_0.cu deleted file mode 100644 index 7e61b5fdc..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q8_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_F16, GGML_TYPE_Q8_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-f16.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-f16.cu deleted file mode 100644 index fdb15b580..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-f16.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_F16); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_0.cu deleted file mode 100644 index 0f7c417d2..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q4_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_1.cu deleted file mode 100644 index 851f33c43..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q4_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_0.cu deleted file mode 100644 index 763809cbe..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q5_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_1.cu deleted file mode 100644 index f2a276e50..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q5_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q8_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q8_0.cu deleted file mode 100644 index cb227f6f5..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q8_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q8_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-f16.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-f16.cu deleted file mode 100644 index 97ac0520c..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-f16.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_F16); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_0.cu deleted file mode 100644 index c772b4263..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_Q4_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_1.cu deleted file mode 100644 index 5cb743081..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_Q4_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_0.cu deleted file mode 100644 index 98a709d17..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_Q5_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_1.cu deleted file mode 100644 index 4f2f947ae..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_Q5_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q8_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q8_0.cu deleted file mode 100644 index 11f96b6f6..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q8_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_Q8_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-f16.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-f16.cu deleted file mode 100644 index b39bdc061..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-f16.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_F16); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_0.cu deleted file mode 100644 index bbd6a2c7f..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_Q4_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_1.cu deleted file mode 100644 index 9d84ff2b1..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_Q4_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_0.cu deleted file mode 100644 index bc8a5bff6..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_Q5_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_1.cu deleted file mode 100644 index a679100c8..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_Q5_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q8_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q8_0.cu deleted file mode 100644 index 8f21bccf7..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q8_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_Q8_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-f16.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-f16.cu deleted file mode 100644 index 858b00fd7..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-f16.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_F16); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_0.cu deleted file mode 100644 index 0fc8011fa..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_Q4_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_1.cu deleted file mode 100644 index 261fdf623..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_Q4_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_0.cu deleted file mode 100644 index 0fb824738..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_Q5_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_1.cu deleted file mode 100644 index a9d9d089b..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_Q5_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q8_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q8_0.cu deleted file mode 100644 index 7d7b27920..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q8_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_Q8_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-f16.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-f16.cu deleted file mode 100644 index a092ee2d5..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-f16.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_F16); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_0.cu deleted file mode 100644 index db55927a1..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q4_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_1.cu deleted file mode 100644 index c3c21cefa..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q4_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_0.cu deleted file mode 100644 index 35dd9f520..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q5_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_1.cu deleted file mode 100644 index 050c22ac7..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q5_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q8_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q8_0.cu deleted file mode 100644 index de4866c5e..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q8_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q8_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs256-f16-f16.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs256-f16-f16.cu deleted file mode 100644 index 57a10bc4b..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs256-f16-f16.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(256, GGML_TYPE_F16, GGML_TYPE_F16); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-f16.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-f16.cu deleted file mode 100644 index e0f08b46a..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-f16.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(64, GGML_TYPE_F16, GGML_TYPE_F16); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_0.cu deleted file mode 100644 index 1c8e8a467..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(64, GGML_TYPE_F16, GGML_TYPE_Q4_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_1.cu deleted file mode 100644 index cefed83fb..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(64, GGML_TYPE_F16, GGML_TYPE_Q4_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_0.cu deleted file mode 100644 index aede6e358..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(64, GGML_TYPE_F16, GGML_TYPE_Q5_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_1.cu deleted file mode 100644 index 1a1a92c78..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(64, GGML_TYPE_F16, GGML_TYPE_Q5_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q8_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q8_0.cu deleted file mode 100644 index ad667473d..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q8_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f16.cuh" - -DECL_FATTN_VEC_F16_CASE(64, GGML_TYPE_F16, GGML_TYPE_Q8_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_0.cu deleted file mode 100644 index 8286ebf37..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_F16, GGML_TYPE_Q4_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_1.cu deleted file mode 100644 index 458786882..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_F16, GGML_TYPE_Q4_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_0.cu deleted file mode 100644 index d89103ce0..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_F16, GGML_TYPE_Q5_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_1.cu deleted file mode 100644 index bb75fd42f..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_F16, GGML_TYPE_Q5_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q8_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q8_0.cu deleted file mode 100644 index b1629817e..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q8_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_F16, GGML_TYPE_Q8_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-f16.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-f16.cu deleted file mode 100644 index d8657604d..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-f16.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_F16); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_0.cu deleted file mode 100644 index 2e5bd2f1a..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q4_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_1.cu deleted file mode 100644 index be5f302d9..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q4_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_0.cu deleted file mode 100644 index 8dd91cd72..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q5_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_1.cu deleted file mode 100644 index 4cb791502..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q5_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q8_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q8_0.cu deleted file mode 100644 index 09dea4267..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q8_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q8_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-f16.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-f16.cu deleted file mode 100644 index 0fbb60769..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-f16.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_F16); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_0.cu deleted file mode 100644 index 2aeab83b2..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_Q4_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_1.cu deleted file mode 100644 index 599415b49..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_Q4_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_0.cu deleted file mode 100644 index e4f8e3083..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_Q5_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_1.cu deleted file mode 100644 index 34d166527..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_Q5_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q8_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q8_0.cu deleted file mode 100644 index 4bebef45a..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q8_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_Q8_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-f16.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-f16.cu deleted file mode 100644 index 326468da2..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-f16.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_F16); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_0.cu deleted file mode 100644 index 511b58f4e..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_Q4_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_1.cu deleted file mode 100644 index d9906d142..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_Q4_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_0.cu deleted file mode 100644 index f61c183ab..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_Q5_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_1.cu deleted file mode 100644 index c10450fd2..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_Q5_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q8_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q8_0.cu deleted file mode 100644 index 2d5cb195c..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q8_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_Q8_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-f16.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-f16.cu deleted file mode 100644 index b384f34d7..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-f16.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_F16); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_0.cu deleted file mode 100644 index 446e293b1..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_Q4_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_1.cu deleted file mode 100644 index 6f4302988..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_Q4_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_0.cu deleted file mode 100644 index 1cd8ba88f..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_Q5_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_1.cu deleted file mode 100644 index 1ee2eab65..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_Q5_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q8_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q8_0.cu deleted file mode 100644 index 2bc77816a..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q8_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_Q8_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-f16.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-f16.cu deleted file mode 100644 index d55ced08b..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-f16.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_F16); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_0.cu deleted file mode 100644 index 8361e99c4..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q4_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_1.cu deleted file mode 100644 index 7507a67c4..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q4_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_0.cu deleted file mode 100644 index 61f050b23..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q5_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_1.cu deleted file mode 100644 index d4a49d9c9..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q5_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q8_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q8_0.cu deleted file mode 100644 index d14627897..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q8_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q8_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs256-f16-f16.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs256-f16-f16.cu deleted file mode 100644 index e73f917a1..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs256-f16-f16.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(256, GGML_TYPE_F16, GGML_TYPE_F16); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-f16.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-f16.cu deleted file mode 100644 index d40825dfc..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-f16.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(64, GGML_TYPE_F16, GGML_TYPE_F16); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_0.cu deleted file mode 100644 index b5c6869f4..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(64, GGML_TYPE_F16, GGML_TYPE_Q4_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_1.cu deleted file mode 100644 index 4e21b0cca..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(64, GGML_TYPE_F16, GGML_TYPE_Q4_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_0.cu deleted file mode 100644 index 2eac321b3..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(64, GGML_TYPE_F16, GGML_TYPE_Q5_0); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_1.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_1.cu deleted file mode 100644 index f7d2c3b4e..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(64, GGML_TYPE_F16, GGML_TYPE_Q5_1); diff --git a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q8_0.cu b/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q8_0.cu deleted file mode 100644 index a013f400b..000000000 --- a/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q8_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../fattn-vec-f32.cuh" - -DECL_FATTN_VEC_F32_CASE(64, GGML_TYPE_F16, GGML_TYPE_Q8_0); diff --git a/ggml/src/ggml-cuda/template-instances/mmq-instance-iq1_s.cu b/ggml/src/ggml-cuda/template-instances/mmq-instance-iq1_s.cu deleted file mode 100644 index 84ec85029..000000000 --- a/ggml/src/ggml-cuda/template-instances/mmq-instance-iq1_s.cu +++ /dev/null @@ -1,5 +0,0 @@ -// 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/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_s.cu b/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_s.cu deleted file mode 100644 index 583c4e5a5..000000000 --- a/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_s.cu +++ /dev/null @@ -1,5 +0,0 @@ -// 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/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_xs.cu b/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_xs.cu deleted file mode 100644 index edaf1560d..000000000 --- a/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_xs.cu +++ /dev/null @@ -1,5 +0,0 @@ -// 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/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_xxs.cu b/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_xxs.cu deleted file mode 100644 index 233d9342c..000000000 --- a/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_xxs.cu +++ /dev/null @@ -1,5 +0,0 @@ -// 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/ggml/src/ggml-cuda/template-instances/mmq-instance-iq3_s.cu b/ggml/src/ggml-cuda/template-instances/mmq-instance-iq3_s.cu deleted file mode 100644 index 6092dc713..000000000 --- a/ggml/src/ggml-cuda/template-instances/mmq-instance-iq3_s.cu +++ /dev/null @@ -1,5 +0,0 @@ -// 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/ggml/src/ggml-cuda/template-instances/mmq-instance-iq3_xxs.cu b/ggml/src/ggml-cuda/template-instances/mmq-instance-iq3_xxs.cu deleted file mode 100644 index 1d5bd201f..000000000 --- a/ggml/src/ggml-cuda/template-instances/mmq-instance-iq3_xxs.cu +++ /dev/null @@ -1,5 +0,0 @@ -// 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/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_nl.cu b/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_nl.cu deleted file mode 100644 index eb02fab00..000000000 --- a/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_nl.cu +++ /dev/null @@ -1,5 +0,0 @@ -// 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/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_xs.cu b/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_xs.cu deleted file mode 100644 index 1eb3b7430..000000000 --- a/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_xs.cu +++ /dev/null @@ -1,5 +0,0 @@ -// 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/ggml/src/ggml-cuda/template-instances/mmq-instance-q2_k.cu b/ggml/src/ggml-cuda/template-instances/mmq-instance-q2_k.cu deleted file mode 100644 index 6415369dc..000000000 --- a/ggml/src/ggml-cuda/template-instances/mmq-instance-q2_k.cu +++ /dev/null @@ -1,5 +0,0 @@ -// 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/ggml/src/ggml-cuda/template-instances/mmq-instance-q3_k.cu b/ggml/src/ggml-cuda/template-instances/mmq-instance-q3_k.cu deleted file mode 100644 index ffb6213af..000000000 --- a/ggml/src/ggml-cuda/template-instances/mmq-instance-q3_k.cu +++ /dev/null @@ -1,5 +0,0 @@ -// 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/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_0.cu b/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_0.cu deleted file mode 100644 index 0c0b0c8a8..000000000 --- a/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// 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/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_1.cu b/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_1.cu deleted file mode 100644 index ee67f6942..000000000 --- a/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// 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/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_0.cu b/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_0.cu deleted file mode 100644 index cc57fb975..000000000 --- a/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_0.cu +++ /dev/null @@ -1,5 +0,0 @@ -// 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/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_1.cu b/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_1.cu deleted file mode 100644 index 721ac790c..000000000 --- a/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_1.cu +++ /dev/null @@ -1,5 +0,0 @@ -// 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/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_k.cu b/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_k.cu deleted file mode 100644 index a2e90ffd5..000000000 --- a/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_k.cu +++ /dev/null @@ -1,5 +0,0 @@ -// This file has been autogenerated by generate_cu_files.py, do not edit manually. - -#include "../mmq.cuh" - -DECL_MMQ_CASE(GGML_TYPE_Q5_K);