sperate examples

example/15_grouped_gemm/CMakeLists.txt

 add_example_executable(example_grouped_gemm_xdl_fp16 grouped_gemm_xdl_fp16.cpp)
-add_example_executable(example_grouped_gemm_bias_xdl_fp16 grouped_gemm_bias_xdl_fp16.cpp)

example/28_grouped_gemm_bias/CMakeLists.txt

+add_example_executable(example_grouped_gemm_bias_xdl_fp16 grouped_gemm_bias_xdl_fp16.cpp)

example/CMakeLists.txt

@@ -46,3 +46,4 @@ add_subdirectory(24_batched_gemm_c_permute)
 add_subdirectory(25_gemm_bias_c_permute)
 add_subdirectory(26_contraction)
 add_subdirectory(27_layernorm)
+add_subdirectory(28_grouped_gemm_bias)

profiler/include/profile_grouped_gemm_impl.hpp

@@ -7,7 +7,7 @@
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
-#include "ck/tensor_operation/gpu/device/device_grouped_gemm_xdl.hpp"
+#include "ck/tensor_operation/gpu/device/device_grouped_gemm.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/tensor_operation_instance/gpu/grouped_gemm.hpp"