diff --git a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_16.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_16.cu
deleted file mode 100644
index fb26abeb..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_16.cu
+++ /dev/null
@@ -1,5 +0,0 @@
-// 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/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_8.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_8.cu
deleted file mode 100644
index dc168290..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_8.cu
+++ /dev/null
@@ -1,10 +0,0 @@
-// 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/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_1.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_1.cu
deleted file mode 100644
index 9d3cfd8e..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_1.cu
+++ /dev/null
@@ -1,10 +0,0 @@
-// 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/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_2.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_2.cu
deleted file mode 100644
index 2e1883af..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_2.cu
+++ /dev/null
@@ -1,10 +0,0 @@
-// 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/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_4.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_4.cu
deleted file mode 100644
index 2074e954..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_4.cu
+++ /dev/null
@@ -1,10 +0,0 @@
-// 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);
diff --git a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_16.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_16.cu
deleted file mode 100644
index f011a208..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_16.cu
+++ /dev/null
@@ -1,5 +0,0 @@
-// 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/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_4.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_4.cu
deleted file mode 100644
index 24c64cf0..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_4.cu
+++ /dev/null
@@ -1,10 +0,0 @@
-// 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);
diff --git a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_8.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_8.cu
deleted file mode 100644
index 163b1d93..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_8.cu
+++ /dev/null
@@ -1,10 +0,0 @@
-// 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/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_32-ncols2_1.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_32-ncols2_1.cu
deleted file mode 100644
index 0543532e..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_32-ncols2_1.cu
+++ /dev/null
@@ -1,10 +0,0 @@
-// 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/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_32-ncols2_2.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_32-ncols2_2.cu
deleted file mode 100644
index 407b6cf4..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_32-ncols2_2.cu
+++ /dev/null
@@ -1,10 +0,0 @@
-// 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/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_16.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_16.cu
deleted file mode 100644
index f5fd0e23..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_16.cu
+++ /dev/null
@@ -1,5 +0,0 @@
-// 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/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_2.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_2.cu
deleted file mode 100644
index 5e466850..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_2.cu
+++ /dev/null
@@ -1,10 +0,0 @@
-// 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/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_4.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_4.cu
deleted file mode 100644
index 1ada657f..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_4.cu
+++ /dev/null
@@ -1,10 +0,0 @@
-// 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);
diff --git a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_8.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_8.cu
deleted file mode 100644
index bad296b4..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_8.cu
+++ /dev/null
@@ -1,10 +0,0 @@
-// 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/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_64-ncols2_1.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_64-ncols2_1.cu
deleted file mode 100644
index 0d7a9c72..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_64-ncols2_1.cu
+++ /dev/null
@@ -1,10 +0,0 @@
-// 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/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_1.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_1.cu
deleted file mode 100644
index 9d5a9976..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_1.cu
+++ /dev/null
@@ -1,10 +0,0 @@
-// 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/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_2.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_2.cu
deleted file mode 100644
index a6e6f093..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_2.cu
+++ /dev/null
@@ -1,10 +0,0 @@
-// 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/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_4.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_4.cu
deleted file mode 100644
index 86d4ffae..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_4.cu
+++ /dev/null
@@ -1,10 +0,0 @@
-// 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);
diff --git a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_8.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_8.cu
deleted file mode 100644
index 680a13ca..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_8.cu
+++ /dev/null
@@ -1,10 +0,0 @@
-// 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/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqfloat-cpb16.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqfloat-cpb16.cu
deleted file mode 100644
index 2d94e65c..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqfloat-cpb16.cu
+++ /dev/null
@@ -1,10 +0,0 @@
-// This file has been autogenerated by generate_cu_files.py, do not edit manually.
-
-#include "../fattn-wmma-f16.cuh"
-
-DECL_FATTN_WMMA_F16_CASE(64, 16, float);
-DECL_FATTN_WMMA_F16_CASE(80, 16, float);
-DECL_FATTN_WMMA_F16_CASE(96, 16, float);
-DECL_FATTN_WMMA_F16_CASE(112, 16, float);
-DECL_FATTN_WMMA_F16_CASE(128, 16, float);
-DECL_FATTN_WMMA_F16_CASE(256, 16, float);
diff --git a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqfloat-cpb32.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqfloat-cpb32.cu
deleted file mode 100644
index c3d9df3c..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqfloat-cpb32.cu
+++ /dev/null
@@ -1,9 +0,0 @@
-// This file has been autogenerated by generate_cu_files.py, do not edit manually.
-
-#include "../fattn-wmma-f16.cuh"
-
-DECL_FATTN_WMMA_F16_CASE(64, 32, float);
-DECL_FATTN_WMMA_F16_CASE(80, 32, float);
-DECL_FATTN_WMMA_F16_CASE(96, 32, float);
-DECL_FATTN_WMMA_F16_CASE(112, 32, float);
-DECL_FATTN_WMMA_F16_CASE(128, 32, float);
diff --git a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb16.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb16.cu
deleted file mode 100644
index bb680e40..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb16.cu
+++ /dev/null
@@ -1,10 +0,0 @@
-// This file has been autogenerated by generate_cu_files.py, do not edit manually.
-
-#include "../fattn-wmma-f16.cuh"
-
-DECL_FATTN_WMMA_F16_CASE(64, 16, half);
-DECL_FATTN_WMMA_F16_CASE(80, 16, half);
-DECL_FATTN_WMMA_F16_CASE(96, 16, half);
-DECL_FATTN_WMMA_F16_CASE(112, 16, half);
-DECL_FATTN_WMMA_F16_CASE(128, 16, half);
-DECL_FATTN_WMMA_F16_CASE(256, 16, half);
diff --git a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb32.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb32.cu
deleted file mode 100644
index 073f71b1..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb32.cu
+++ /dev/null
@@ -1,10 +0,0 @@
-// This file has been autogenerated by generate_cu_files.py, do not edit manually.
-
-#include "../fattn-wmma-f16.cuh"
-
-DECL_FATTN_WMMA_F16_CASE(64, 32, half);
-DECL_FATTN_WMMA_F16_CASE(80, 32, half);
-DECL_FATTN_WMMA_F16_CASE(96, 32, half);
-DECL_FATTN_WMMA_F16_CASE(112, 32, half);
-DECL_FATTN_WMMA_F16_CASE(128, 32, half);
-DECL_FATTN_WMMA_F16_CASE(256, 32, half);
diff --git a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb8.cu b/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb8.cu
deleted file mode 100644
index d30710c5..00000000
--- a/ml/backend/ggml/ggml/src/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb8.cu
+++ /dev/null
@@ -1,8 +0,0 @@
-// This file has been autogenerated by generate_cu_files.py, do not edit manually.
-
-#include "../fattn-wmma-f16.cuh"
-
-DECL_FATTN_WMMA_F16_CASE(64, 8, half);
-DECL_FATTN_WMMA_F16_CASE(96, 8, half);
-DECL_FATTN_WMMA_F16_CASE(128, 8, half);
-DECL_FATTN_WMMA_F16_CASE(256, 8, half);