LNK2001:unresolved external symbol OrtSessionOptionsAppendExecutionProvider_Tensorrt

[chwjshlwel]

Describe the bug I run the onnxruntime c++ code as follows:

#include "stdafx.h"
#include <windows.h>
#include <windowsx.h>
#include <onnxruntime_cxx_api.h>
#include <cuda_provider_factory.h>
#include <onnxruntime_c_api.h>
#include <tensorrt_provider_factory.h>

Ort::Env env{ ORT_LOGGING_LEVEL_WARNING, "test" };
Ort::SessionOptions sf;
int device_id = 0;
Ort::ThrowOnError(OrtSessionOptionsAppendExecutionProvider_Tensorrt(sf, device_id));
Ort::ThrowOnError(OrtSessionOptionsAppendExecutionProvider_CUDA(sf, device_id));
Ort::Session session(env, L"mobilenet.onnx", sf);

System information

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): windows 10
• ONNX Runtime installed from (source or binary): build from github
• ONNX Runtime version: newest
• Python version:3.7
• Visual Studio version (if applicable): vs2019
• GCC/Compiler version (if compiling from source):
• CUDA/cuDNN version: cuda 11.3/cudnn8.2, TensorRT-8.0.1.6
• GPU model and memory:RTX 3080(10G)

To Reproduce when I build the solution, I got the error LNK2001:unresolved external symbol OrtSessionOptionsAppendExecutionProvider_Tensorrt

But I linked both onnxruntime.lib, onnxruntime_provider_cuda.lib,onnxruntime_provider_tensorrt.lib,nvinfer.lib i also put all the dll with the same path with the exe Please help me to solve this problem. Thanks!

[yuslepukhin]

How did you create a solution? This may be helpful.

Does onnxruntime.dll exports the symbol in question? For example, for the CPU only build it is exporting OrtSessionOptionsAppendExecutionProvider_CPU

d:\dev\ort_trans>dumpbin /EXPORTS d:\dev\ort_trans\cpu_build\Debug\Debug\onnxruntime.dll Microsoft (R) COFF/PE Dumper Version 14.29.30040.0 Copyright (C) Microsoft Corporation. All rights reserved.

Dump of file d:\dev\ort_trans\cpu_build\Debug\Debug\onnxruntime.dll

File Type: DLL

Section contains the following exports for onnxruntime.dll [...]

ordinal hint RVA      name

      1    0 000441A7 OrtGetApiBase = @ILT+274850(OrtGetApiBase)
      2    1 0005B7D0 OrtSessionOptionsAppendExecutionProvider_CPU = @ILT+370635(OrtSessionOptionsAppendExecutionProvider_CPU)

[chwjshlwel]

for OrtSessionOptionsAppendExecutionProvider_CPU and OrtSessionOptionsAppendExecutionProvider_CUDA, there are no problem. The problem is just for TensorRT. I check the version, and maybe it is the version problem. I will try the version onnxruntime 1.8.2, cuda 11.1, cudnn 8.0.5, tensorrt 7.2.2.3, and see if there is problem. But when I use CUDA provider, there has some error. If I used the onnxruntime_providers_cuda.dll built by mine, it has errors. While if I used the onnxruntime_providers_cuda.dll downloaded from https://github.com/microsoft/onnxruntime/releases/download/v1.8.1/onnxruntime-win-gpu-x64-1.8.1.zip, it has no error. I think the dlls have problems. It shows build successed while it has error when test_all but I use --skip_test and it is built successed. [Debug.zip](https://github.com/microsoft/onnxruntime/files/6972168/Debug.zip) I attached my onnxruntime_providers_cuda.dll and onnxruntime_providers_tensorrt.dll

[chwjshlwel]

Is any link that I can download onnxruntime_providers_tensorrt.dll. I think I build failed even it shows I build successed.

[yuslepukhin]

Is any link that I can download onnxruntime_providers_tensorrt.dll. I think I build failed even it shows I build successed.

This is DLL is build depended, it must be built together with other DLLs. So you must build something locally, it must come from your build tree.

[chwjshlwel]

Thanks. I upload the onnxruntime_providers_tensorrt.lib and onnxruntime_providers_tensorrt.dll. Please help me to see if these files are right. Debug (2).zip

[yuslepukhin]

Thanks. I upload the onnxruntime_providers_tensorrt.lib and onnxruntime_providers_tensorrt.dll. Please help me to see if these files are right. Debug (2).zip

I can not tell if these DLLs are correct for you build. They are build defendant. Have you followed the instructions to run the build?

[RyanUnderhill]

Just to reaffirm, you need to follow these steps to have TensorRT available:

https://onnxruntime.ai/docs/how-to/build/eps.html#tensorrt

The only library you need to link with in your code is just onnxruntime.lib (or access the onnxruntime.dll through GetProcAddress), onnxruntime.dll will dynamically load the other libraries itself.

[stale[bot]]

This issue has been automatically marked as stale due to inactivity and will be closed in 7 days if no further activity occurs. If further support is needed, please provide an update and/or more details.