Could not build solution.!!

[hy846130226]

Describe the issue

I pull the V1.6.3 code.
and use the following command: .\build.bat --cudnn_home "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8" --cuda_home "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8" --use_tensorrt --tensorrt_home “C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\TensorRT-8.6.1.6” --cmake_generator "Visual Studio 17 2022"

but when I got the solution, I COULD NOT build the nvonnxparser_static project

Urgency

No response

Target platform

C++

Build script

.\build.bat --cudnn_home "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8" --cuda_home "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8" --use_tensorrt --tensorrt_home “C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\TensorRT-8.6.1.6” --cmake_generator "Visual Studio 17 2022"

Error / output

C2664 “bool google::protobuf::TextFormat::Parse(google::protobuf::io::ZeroCopyInputStream *,google::protobuf::Message *)”: could not transfrom “onnx::ModelProto *” into “google::protobuf::Message *” nvonnxparser_static D:\onnxruntime\onnxruntime\build\Windows\Debug_deps\onnx_tensorrt-src\ModelImporter.cpp 358

Visual Studio Version

No response

GCC / Compiler Version

No response

[jywu-msft]

please build RelWithDebInfo or Release flavor

[jywu-msft]

@yf711 can you check if Debug build has new issues?

[hy846130226]

Thanks for your help @jywu-msft !
The release version build successful.
But I also have a question:

I use the nuget package to load the onnxruntime-gpu version 1.16.3, and I could see the folder like:

My question is:
How could I build the 1.16.3 onnxruntime.dll&onnxruntime.lib?

[yf711]

@hy846130226 You can build ORT with --build_shared_lib to generate onnxruntime.dll/lib

Actually, I tried your command to build ORT in debug mode and it passed on my side.
One thing to call out, if you would like to use open-sourced onnx-tensorrt parser, please add --use_tensorrt_oss_parser to your ORT build command. The default behavior without this arg will be using built-in parser from your tensorrt 8.6 binary

[hy846130226]

@hy846130226 You can build ORT with --build_shared_lib to generate onnxruntime.dll/lib

Actually, I tried your command to build ORT in debug mode and it passed on my side. One thing to call out, if you would like to use open-sourced onnx-tensorrt parser, please add --use_tensorrt_oss_parser to your ORT build command. The default behavior without this arg will be using built-in parser from your tensorrt 8.6 binary

Hi @yf711,
thank you very much!

I do not know "You can build ORT with --build_shared_lib to generate onnxruntime.dll/lib", could you please tell me more details or what should I do?

.\build.bat --cudnn_home "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8" --cuda_home "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8" --use_tensorrt --tensorrt_home “C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\TensorRT-8.6.1.6” --cmake_generator "Visual Studio 17 2022" --buildshared_lib

Is this your means?

[hy846130226]

@hy846130226 You can build ORT with --build_shared_lib to generate onnxruntime.dll/lib

Actually, I tried your command to build ORT in debug mode and it passed on my side. One thing to call out, if you would like to use open-sourced onnx-tensorrt parser, please add --use_tensorrt_oss_parser to your ORT build command. The default behavior without this arg will be using built-in parser from your tensorrt 8.6 binary

And my tensorrt version is 8.6.1.6, according to the official onnxruntime website, "The TensorRT execution provider for ONNX Runtime is built and tested with TensorRT 8.6.".

So I don't think I should use the "--use_tensorrt_oss_parser" command

[hy846130226]

Here is the folder snapshot after I build ALL_BUILD:

There is no onnxruntime.dll&onnxruntime.lib?

[hy846130226]

Hi @yf711,

Thanks for your help.
I have resolved this problem according to your words!