[Bug Fix] Support isfinite/isnan/isinf for float16/bfloat16 on CUDA/HIP

[youge325]

PR Category

Environment Adaptation

PR Types

Bug fixes

Description

• 在 isfinite_kernel_impl.h 的 GPU 侧 Isfinite/Isnan/Isinf 核函数里，把 “通用浮点” 模板拆成两支：一支只接受标准 float/double，另一支专门匹配 phi::float16 和 phi::bfloat16。这避免了 std::is_floating_point 对这两种自定义半精度类型返回 false 而导致完全没有匹配内核的情况，从而补齐了半精度在 CUDA/HIP 上的 isfinite/isnan/isinf 支持。
• 由于有了独立分支，调用的仍是对应的 isfinite/isnan/isinf 设备实现，逻辑保持一致，但现在 float16/bfloat16 会正确走到实际内核里，不再出现链接缺符号或运行时报 “未注册该数据类型” 的问题。
• 去掉三个模板 IsfiniteKernel/IsinfKernel/IsnanKernel 的 PADDLE_API 修饰，避免在头文件模板定义上做符号导出，引起重复导出或 Windows 下的装饰冲突。
• 编译时的错误日志如下：

[943/2749] Building CUDA object paddle\phi\CMakeFiles\phi.dir\kernels\gpu\isfinite_kernel.cu.obj
FAILED: paddle/phi/CMakeFiles/phi.dir/kernels/gpu/isfinite_kernel.cu.obj
C:\PROGRA~1\NVIDIA~2\CUDA\v13.0\bin\nvcc.exe -forward-unknown-to-host-compiler -DCUDA_TOOLKIT_ROOT_DIR="\"C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0\"" -DCUDA_VERSION_MAJOR=\"13\" -DCUDA_VERSION_MINOR=\"0\" -DCUDNN_MAJOR_VERSION=\"9\" -DEIGEN_STRONG_INLINE=inline -DEIGEN_USE_GPU -DGLOG_NO_ABBREVIATED_SEVERITIES -DGOOGLE_GLOG_DLL_DECL="" -DLAPACK_FOUND -DNOMINMAX -DPADDLE_DISABLE_PROFILER -DPADDLE_DLL_EXPORT -DPADDLE_DLL_INFERENCE -DPADDLE_ON_INFERENCE -DPADDLE_VERSION=0.0.0 -DPADDLE_VERSION_INTEGER=0 -DPADDLE_WITH_AVX -DPADDLE_WITH_CCCL -DPADDLE_WITH_CRYPTO -DPADDLE_WITH_CUDA -DPADDLE_WITH_DNNL -DPADDLE_WITH_MKLML -DPADDLE_WITH_PIP_CUDA_LIBRARIES -DPADDLE_WITH_POCKETFFT -DPADDLE_WITH_SSE3 -DPADDLE_WITH_TENSORRT -DPHI_INNER -DPHI_SHARED -DSTATIC_IR -DTRT_PLUGIN_FP16_AVAILABLE -DUTF8PROC_STATIC -DWIN32_LEAN_AND_MEAN -DYAML_CPP_STATIC_DEFINE -D_XKEYCHECK_H -Dphi_EXPORTS -ID:\Paddle\third_party\cccl\thrust -ID:\Paddle\third_party\cccl\libcudacxx\include -ID:\Paddle\third_party\cccl\cub -ID:\Paddle\build -ID:\Paddle\paddle\fluid\framework\io -ID:\Paddle\patches\thrust -ID:\TensorRT-10.13.3.9\include -ID:\Paddle\build\third_party\install\zlib\include -ID:\Paddle\build\third_party\install -ID:\Paddle\build\third_party\install\gflags\include -ID:\Paddle\build\third_party\install\glog\include -ID:\Paddle\third_party\eigen3 -ID:\Paddle\third_party\threadpool -ID:\Paddle\third_party\dlpack\include -ID:\Paddle\build\third_party\install\xxhash\include -ID:\Paddle\build\third_party\install\warpctc\include -ID:\Paddle\build\third_party\install\warprnnt\include -ID:\Paddle\build\third_party\install\utf8proc\include -ID:\Paddle\build\third_party\install\mklml\include -ID:\Paddle\build\third_party\install\onednn\include -ID:\Paddle\build\third_party\install\protobuf\include -ID:\Paddle\third_party\nlohmann_json\include -ID:\Paddle\build\third_party\install\yaml-cpp\include -ID:\Users\Lenovo\AppData\Local\Programs\Python\Python310\include -ID:\Users\Lenovo\AppData\Local\Programs\Python\Python310\Lib\site-packages\numpy\core\include -ID:\Paddle\build\third_party\pybind\src\extern_pybind\include -ID:\Paddle\build\third_party\install\libuv\include -ID:\Paddle\third_party\cccl -ID:\Paddle\build\third_party\install\cryptopp\include -ID:\Paddle\build\third_party\pocketfft\src -ID:\Paddle\build\third_party\dirent\src\extern_dirent\include -I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include" -ID:\Paddle -ID:\Paddle\paddle\phi\api\include\compat -ID:\Paddle\paddle\phi\api\include\compat\torch\csrc\api\include -ID:\Paddle\build\..\paddle\fluid\framework\io -D_WINDOWS -Xcompiler=" /W0  /GR /EHsc"  -gencode arch=compute_86,code=sm_86 -Xfatbin -compress-all -w --expt-relaxed-constexpr --expt-extended-lambda -Xcompiler "/wd4244 /wd4267 /wd4819 " -Xcompiler /bigobj -std=c++17  -Xcompiler="/arch:AVX" -Xcompiler="-MT -O2 -Ob2" -DNDEBUG -std=c++17 -MD -MT paddle\phi\CMakeFiles\phi.dir\kernels\gpu\isfinite_kernel.cu.obj -MF paddle\phi\CMakeFiles\phi.dir\kernels\gpu\isfinite_kernel.cu.obj.d -x cu -c D:\Paddle\paddle\phi\kernels\gpu\isfinite_kernel.cu -o paddle\phi\CMakeFiles\phi.dir\kernels\gpu\isfinite_kernel.cu.obj -Xcompiler=-Fdpaddle\phi\CMakeFiles\phi.dir\,-FS
isfinite_kernel.cu
tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.cpp
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(141): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi7float16E *,int64_t,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(141): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(141): error C2660: '__cudaGetKernel': function does not take 1 arguments
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt/device_functions.h(2930): note: see declaration of '__cudaGetKernel'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(141): note: while trying to match the argument list '(cudaKernel_t *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(144): error C2912: explicit specialization 'void phi::__wrapper__device_stub_IsinfCUDAKernel<phi::float16,__int64>(const _ZN3phi7float16E *&,int64_t &,bool *&,_ZNSt9enable_ifILb1EvE4typeE *&)' is not a specialization of a function template
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(156): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi7float16E *,unsigned int,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(156): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(156): error C2660: '__cudaGetKernel': function does not take 1 arguments
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt/device_functions.h(2930): note: see declaration of '__cudaGetKernel'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(156): note: while trying to match the argument list '(cudaKernel_t *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(159): error C2912: explicit specialization 'void phi::__wrapper__device_stub_IsinfCUDAKernel<phi::float16,unsigned int>(const _ZN3phi7float16E *&,unsigned int &,bool *&,_ZNSt9enable_ifILb1EvE4typeE *&)' is not a specialization of a function template
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(171): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi8bfloat16E *,int64_t,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(171): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(171): error C2660: '__cudaGetKernel': function does not take 1 arguments
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt/device_functions.h(2930): note: see declaration of '__cudaGetKernel'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(171): note: while trying to match the argument list '(cudaKernel_t *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(174): error C2912: explicit specialization 'void phi::__wrapper__device_stub_IsinfCUDAKernel<phi::bfloat16,__int64>(const _ZN3phi8bfloat16E *&,int64_t &,bool *&,_ZNSt9enable_ifILb1EvE4typeE *&)' is not a specialization of a function template
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(186): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi8bfloat16E *,unsigned int,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(186): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(186): error C2660: '__cudaGetKernel': function does not take 1 arguments
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt/device_functions.h(2930): note: see declaration of '__cudaGetKernel'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(186): note: while trying to match the argument list '(cudaKernel_t *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(189): error C2912: explicit specialization 'void phi::__wrapper__device_stub_IsinfCUDAKernel<phi::bfloat16,unsigned int>(const _ZN3phi8bfloat16E *&,unsigned int &,bool *&,_ZNSt9enable_ifILb1EvE4typeE *&)' is not a specialization of a function template
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(471): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi7float16E *,int64_t,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(471): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(471): error C2660: '__cudaGetKernel': function does not take 1 arguments
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt/device_functions.h(2930): note: see declaration of '__cudaGetKernel'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(471): note: while trying to match the argument list '(cudaKernel_t *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(474): error C2912: explicit specialization 'void phi::__wrapper__device_stub_IsnanCUDAKernel<phi::float16,__int64>(const _ZN3phi7float16E *&,int64_t &,bool *&,_ZNSt9enable_ifILb1EvE4typeE *&)' is not a specialization of a function template
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(486): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi7float16E *,unsigned int,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(486): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(486): error C2660: '__cudaGetKernel': function does not take 1 arguments
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt/device_functions.h(2930): note: see declaration of '__cudaGetKernel'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(486): note: while trying to match the argument list '(cudaKernel_t *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(489): error C2912: explicit specialization 'void phi::__wrapper__device_stub_IsnanCUDAKernel<phi::float16,unsigned int>(const _ZN3phi7float16E *&,unsigned int &,bool *&,_ZNSt9enable_ifILb1EvE4typeE *&)' is not a specialization of a function template
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(501): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi8bfloat16E *,int64_t,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(501): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(501): error C2660: '__cudaGetKernel': function does not take 1 arguments
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt/device_functions.h(2930): note: see declaration of '__cudaGetKernel'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(501): note: while trying to match the argument list '(cudaKernel_t *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(504): error C2912: explicit specialization 'void phi::__wrapper__device_stub_IsnanCUDAKernel<phi::bfloat16,__int64>(const _ZN3phi8bfloat16E *&,int64_t &,bool *&,_ZNSt9enable_ifILb1EvE4typeE *&)' is not a specialization of a function template
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(516): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi8bfloat16E *,unsigned int,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(516): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(516): error C2660: '__cudaGetKernel': function does not take 1 arguments
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt/device_functions.h(2930): note: see declaration of '__cudaGetKernel'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(516): note: while trying to match the argument list '(cudaKernel_t *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(519): error C2912: explicit specialization 'void phi::__wrapper__device_stub_IsnanCUDAKernel<phi::bfloat16,unsigned int>(const _ZN3phi8bfloat16E *&,unsigned int &,bool *&,_ZNSt9enable_ifILb1EvE4typeE *&)' is not a specialization of a function template
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(711): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi7float16E *,int64_t,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(711): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(711): error C2660: '__cudaGetKernel': function does not take 1 arguments
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt/device_functions.h(2930): note: see declaration of '__cudaGetKernel'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(711): note: while trying to match the argument list '(cudaKernel_t *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(714): error C2912: explicit specialization 'void phi::__wrapper__device_stub_IsfiniteCUDAKernel<phi::float16,__int64>(const _ZN3phi7float16E *&,int64_t &,bool *&,_ZNSt9enable_ifILb1EvE4typeE *&)' is not a specialization of a function template
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(726): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi7float16E *,unsigned int,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(726): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(726): error C2660: '__cudaGetKernel': function does not take 1 arguments
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt/device_functions.h(2930): note: see declaration of '__cudaGetKernel'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(726): note: while trying to match the argument list '(cudaKernel_t *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(729): error C2912: explicit specialization 'void phi::__wrapper__device_stub_IsfiniteCUDAKernel<phi::float16,unsigned int>(const _ZN3phi7float16E *&,unsigned int &,bool *&,_ZNSt9enable_ifILb1EvE4typeE *&)' is not a specialization of a function template
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(741): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi8bfloat16E *,int64_t,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(741): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(741): error C2660: '__cudaGetKernel': function does not take 1 arguments
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt/device_functions.h(2930): note: see declaration of '__cudaGetKernel'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(741): note: while trying to match the argument list '(cudaKernel_t *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(744): error C2912: explicit specialization 'void phi::__wrapper__device_stub_IsfiniteCUDAKernel<phi::bfloat16,__int64>(const _ZN3phi8bfloat16E *&,int64_t &,bool *&,_ZNSt9enable_ifILb1EvE4typeE *&)' is not a specialization of a function template
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(756): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi8bfloat16E *,unsigned int,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(756): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(756): error C2660: '__cudaGetKernel': function does not take 1 arguments
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt/device_functions.h(2930): note: see declaration of '__cudaGetKernel'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(756): note: while trying to match the argument list '(cudaKernel_t *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(759): error C2912: explicit specialization 'void phi::__wrapper__device_stub_IsfiniteCUDAKernel<phi::bfloat16,unsigned int>(const _ZN3phi8bfloat16E *&,unsigned int &,bool *&,_ZNSt9enable_ifILb1EvE4typeE *&)' is not a specialization of a function template
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(893): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi8bfloat16E *,unsigned int,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(893): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(893): error C2660: '__cudaRegisterFunction': function does not take 9 arguments
C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include\crt/host_runtime.h(246): note: see declaration of '__cudaRegisterFunction'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(893): note: while trying to match the argument list '(void **, char *, const char [123], int, uint3 *, uint3 *, dim3 *, dim3 *, int *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(894): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi8bfloat16E *,int64_t,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(894): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(894): error C2660: '__cudaRegisterFunction': function does not take 9 arguments
C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include\crt/host_runtime.h(246): note: see declaration of '__cudaRegisterFunction'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(894): note: while trying to match the argument list '(void **, char *, const char [123], int, uint3 *, uint3 *, dim3 *, dim3 *, int *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(895): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi7float16E *,unsigned int,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(895): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(895): error C2660: '__cudaRegisterFunction': function does not take 9 arguments
C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include\crt/host_runtime.h(246): note: see declaration of '__cudaRegisterFunction'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(895): note: while trying to match the argument list '(void **, char *, const char [122], int, uint3 *, uint3 *, dim3 *, dim3 *, int *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(896): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi7float16E *,int64_t,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(896): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(896): error C2660: '__cudaRegisterFunction': function does not take 9 arguments
C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include\crt/host_runtime.h(246): note: see declaration of '__cudaRegisterFunction'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(896): note: while trying to match the argument list '(void **, char *, const char [122], int, uint3 *, uint3 *, dim3 *, dim3 *, int *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(909): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi8bfloat16E *,unsigned int,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(909): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(909): error C2660: '__cudaRegisterFunction': function does not take 9 arguments
C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include\crt/host_runtime.h(246): note: see declaration of '__cudaRegisterFunction'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(909): note: while trying to match the argument list '(void **, char *, const char [120], int, uint3 *, uint3 *, dim3 *, dim3 *, int *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(910): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi8bfloat16E *,int64_t,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(910): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(910): error C2660: '__cudaRegisterFunction': function does not take 9 arguments
C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include\crt/host_runtime.h(246): note: see declaration of '__cudaRegisterFunction'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(910): note: while trying to match the argument list '(void **, char *, const char [120], int, uint3 *, uint3 *, dim3 *, dim3 *, int *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(911): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi7float16E *,unsigned int,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(911): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(911): error C2660: '__cudaRegisterFunction': function does not take 9 arguments
C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include\crt/host_runtime.h(246): note: see declaration of '__cudaRegisterFunction'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(911): note: while trying to match the argument list '(void **, char *, const char [119], int, uint3 *, uint3 *, dim3 *, dim3 *, int *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(912): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi7float16E *,int64_t,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(912): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(912): error C2660: '__cudaRegisterFunction': function does not take 9 arguments
C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include\crt/host_runtime.h(246): note: see declaration of '__cudaRegisterFunction'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(912): note: while trying to match the argument list '(void **, char *, const char [119], int, uint3 *, uint3 *, dim3 *, dim3 *, int *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(931): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi8bfloat16E *,unsigned int,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(931): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(931): error C2660: '__cudaRegisterFunction': function does not take 9 arguments
C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include\crt/host_runtime.h(246): note: see declaration of '__cudaRegisterFunction'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(931): note: while trying to match the argument list '(void **, char *, const char [120], int, uint3 *, uint3 *, dim3 *, dim3 *, int *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(932): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi8bfloat16E *,int64_t,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(932): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(932): error C2660: '__cudaRegisterFunction': function does not take 9 arguments
C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include\crt/host_runtime.h(246): note: see declaration of '__cudaRegisterFunction'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(932): note: while trying to match the argument list '(void **, char *, const char [120], int, uint3 *, uint3 *, dim3 *, dim3 *, int *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(933): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi7float16E *,unsigned int,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(933): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(933): error C2660: '__cudaRegisterFunction': function does not take 9 arguments
C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include\crt/host_runtime.h(246): note: see declaration of '__cudaRegisterFunction'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(933): note: while trying to match the argument list '(void **, char *, const char [119], int, uint3 *, uint3 *, dim3 *, dim3 *, int *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(934): error C2440: 'type cast': cannot convert from 'overloaded-function' to 'void (__cdecl *)(const _ZN3phi7float16E *,int64_t,bool *,_ZNSt9enable_ifILb1EvE4typeE *)'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(934): note: None of the functions with this name in scope match the target type
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(934): error C2660: '__cudaRegisterFunction': function does not take 9 arguments
C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include\crt/host_runtime.h(246): note: see declaration of '__cudaRegisterFunction'
C:\Users\Lenovo\AppData\Local\Temp\tmpxft_00002620_00000000-7_isfinite_kernel.cudafe1.stub.c(934): note: while trying to match the argument list '(void **, char *, const char [119], int, uint3 *, uint3 *, dim3 *, dim3 *, int *)'
[944/2749] Building CUDA object paddle\phi\CMakeFiles\phi.dir\kernels\gpu\inverse_grad_kernel.cu.obj
inverse_grad_kernel.cu
tmpxft_00008b7c_00000000-7_inverse_grad_kernel.cudafe1.cpp
[945/2749] Building CUDA object paddle\phi\CMakeFiles\phi.dir\kernels\gpu\kldiv_loss_grad_kernel.cu.obj
kldiv_loss_grad_kernel.cu
tmpxft_00005cc4_00000000-7_kldiv_loss_grad_kernel.cudafe1.cpp
[946/2749] Building CUDA object paddle\phi\CMakeFiles\phi.dir\kernels\gpu\kldiv_loss_kernel.cu.obj
kldiv_loss_kernel.cu
tmpxft_00000b18_00000000-7_kldiv_loss_kernel.cudafe1.cpp
[947/2749] Building CUDA object paddle\phi\CMakeFiles\phi.dir\kernels\gpu\kron_grad_kernel.cu.obj
kron_grad_kernel.cu
tmpxft_00003e30_00000000-7_kron_grad_kernel.cudafe1.cpp
[948/2749] Building CUDA object paddle\phi\CMakeFiles\phi.dir\kernels\gpu\kthvalue_grad_kernel.cu.obj
kthvalue_grad_kernel.cu
tmpxft_00006d80_00000000-7_kthvalue_grad_kernel.cudafe1.cpp
[949/2749] Building CUDA object paddle\phi\CMakeFiles\phi.dir\kernels\gpu\kron_kernel.cu.obj
kron_kernel.cu
tmpxft_000050d8_00000000-7_kron_kernel.cudafe1.cpp
[950/2749] Building CUDA object paddle\phi\CMakeFiles\phi.dir\kernels\gpu\elementwise_grad_kernel.cu.obj
elementwise_grad_kernel.cu
tmpxft_00006924_00000000-7_elementwise_grad_kernel.cudafe1.cpp
ninja: build stopped: subcommand failed.

• issue: paddle.utils.run_check()报错 #75848
• 拆分自 fix bugs when compiling with VS2022 #75611

[paddle-bot]

你的PR提交成功，感谢你对开源项目的贡献!
请关注后续CI自动化测试结果，详情请参考Paddle-CI手册。
Your PR has been submitted. Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

[youge325]

/re-run all-failed

[youge325]

/re-run all-failed

[youge325]

/re-run all-failed

[risemeup1]

LGTM