Lots of Error when compiling paddle-inference.dll3.0.0 with vs2022

[SpaceView]

identifier "seed_tensor_ptr" is undefined
identifier "seed_tensor_ptr" is undefined
命令“"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.3\bin\nvcc.exe" --use-local-env -ccbin "C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Tools\MSVC\14.38.33130\bin\HostX64\x64" -x cu -ID:\aiOCR\devPaddle -ID:\aiOCR\Paddle\paddle\fluid\framework\io -ID:\aiOCR\Paddle\patches\thrust -ID:\aiOCR\devPaddle\third_party\install\zlib\include -ID:\aiOCR\devPaddle\third_party\install -ID:\aiOCR\devPaddle\third_party\install\gflags\include -ID:\aiOCR\devPaddle\third_party\install\glog\include -ID:\aiOCR\Paddle\third_party\eigen3 -ID:\aiOCR\Paddle\third_party\threadpool -ID:\aiOCR\Paddle\third_party\dlpack\include -ID:\aiOCR\devPaddle\third_party\install\xxhash\include -ID:\aiOCR\devPaddle\third_party\install\warpctc\include -ID:\aiOCR\devPaddle\third_party\install\warprnnt\include -ID:\aiOCR\devPaddle\third_party\install\utf8proc\include -ID:\aiOCR\devPaddle\third_party\install\mklml\include -ID:\aiOCR\devPaddle\third_party\install\onednn\include -ID:\aiOCR\devPaddle\third_party\install\protobuf\include -ID:\aiOCR\Paddle\third_party\nlohmann_json\include -I"D:\aiOCR\devPaddle\third_party\install\yaml-cpp\include" -IC:\anaconda3\include -I"C:\anaconda3\Lib\site-packages\numpy\core\include" -ID:\aiOCR\devPaddle\third_party\pybind\src\extern_pybind\include -ID:\aiOCR\devPaddle\third_party\install\cryptopp\include -ID:\aiOCR\devPaddle\third_party\pocketfft\src -ID:\aiOCR\devPaddle\third_party\dirent\src\extern_dirent\include -I"C:\Programs\cudnn-windows-x86_64-9.0.0.312\include" -I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.3\include" -ID:\aiOCR\Paddle -ID:\aiOCR\devPaddle..\paddle\fluid\framework\io -I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.3\include" --keep-dir x64\Debug -maxrregcount=0 --machine 64 --compile -cudart static -Wno-deprecated-gpu-targets -gencode arch=compute_89,code=sm_89 -w --expt-relaxed-constexpr --expt-extended-lambda -std=c++17 -std=c++17 -Xcompiler="/EHsc /wd4244 /wd4267 /wd4819 /bigobj /arch:AVX -Zi -Ob0" -g -D_WINDOWS -D_MWAITXINTRIN_H_INCLUDED -D__STRICT_ANSI__ -DPHI_INNER -DNOMINMAX -DWIN32_LEAN_AND_MEAN -DTRT_PLUGIN_FP16_AVAILABLE -D"CUDA_VERSION_MAJOR="12"" -D"CUDA_VERSION_MINOR="3"" -D"CUDA_TOOLKIT_ROOT_DIR="C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.3"" -D"CUDNN_MAJOR_VERSION="9"" -DGLOG_NO_ABBREVIATED_SEVERITIES -DGOOGLE_GLOG_DLL_DECL= -DEIGEN_STRONG_INLINE=inline -DUTF8PROC_STATIC -DPADDLE_WITH_MKLML -DLAPACK_FOUND -DPADDLE_WITH_DNNL -DPADDLE_WITH_CRYPTO -DPADDLE_WITH_POCKETFFT -DPADDLE_VERSION=3.0.0 -DPADDLE_VERSION_INTEGER=3000000 -DPADDLE_DISABLE_PROFILER -DPADDLE_WITH_AVX -DPADDLE_WITH_SSE3 -D_XKEYCHECK_H -DPADDLE_DLL_INFERENCE -DPADDLE_WITH_CUDA -DEIGEN_USE_GPU -DPADDLE_ON_INFERENCE -DPADDLE_DLL_EXPORT -DSTATIC_IR -DYAML_CPP_STATIC_DEFINE -D"CMAKE_INTDIR="Debug"" -D_MBCS -DWIN32 -D_WINDOWS -DPHI_INNER -DNOMINMAX -DWIN32_LEAN_AND_MEAN -DTRT_PLUGIN_FP16_AVAILABLE -D"CUDA_VERSION_MAJOR="12"" -D"CUDA_VERSION_MINOR="3"" -D"CUDA_TOOLKIT_ROOT_DIR="C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.3"" -D"CUDNN_MAJOR_VERSION="9"" -DGLOG_NO_ABBREVIATED_SEVERITIES -DGOOGLE_GLOG_DLL_DECL= -DEIGEN_STRONG_INLINE=inline -DUTF8PROC_STATIC -DPADDLE_WITH_MKLML -DLAPACK_FOUND -DPADDLE_WITH_DNNL -DPADDLE_WITH_CRYPTO -DPADDLE_WITH_POCKETFFT -DPADDLE_VERSION=3.0.0 -DPADDLE_VERSION_INTEGER=3000000 -DPADDLE_DISABLE_PROFILER -DPADDLE_WITH_AVX -DPADDLE_WITH_SSE3 -D_XKEYCHECK_H -DPADDLE_DLL_INFERENCE -DPADDLE_WITH_CUDA -DEIGEN_USE_GPU -DPADDLE_ON_INFERENCE -DPADDLE_DLL_EXPORT -DSTATIC_IR -DYAML_CPP_STATIC_DEFINE -D"CMAKE_INTDIR="Debug"" -Xcompiler "/EHsc /W0 /nologo /Od /FS /Zi /RTC1 /MTd /GR" -Xcompiler "/FdD:\aiOCR\devPaddle\paddle\phi\Debug\libphi_core.pdb" -o phi_core.dir\Debug/kernels/fusion/gpu/fused_attention_grad_kernel.cu.obj "D:\aiOCR\Paddle\paddle\phi\kernels\fusion\gpu\fused_attention_grad_kernel.cu"”已退出，返回代码为 2。

//-------------------------------------
where seed_tensor_ptr error seems coming from here,
#ifndef PADDLE_WITH_HIP
assert(seed_tensor_ptr == nullptr);
#endif

[wangguan1995]

我也遇到了同类问题 #72379

[gongshaotian]

@wangguan1995 注释后问题解决了吗