Issues with TensorRT execution

Currently I create a Simulink model that calls a mexw64 file that was generated for deep learning with TensorRT, I have no issues to generate the mex file but when I try to execute I have the following message:

Invalid MEX-file 'D:\PhD\ITESO\Clases\10._Semestre_7\IDI_X\Simulation\resNet18SemanticSegmentationRT_mex.mexw64': No se puede
encontrar el módulo especificado.

But the file is present on the same folder. Also when I execute the verification of the setup as follows:

gpuEnvObj = coder.gpuEnvConfig;
gpuEnvObj.BasicCodegen = 1;
gpuEnvObj.BasicCodeexec = 1;
gpuEnvObj.DeepLibTarget = 'tensorrt';
gpuEnvObj.DeepCodeexec = 1;
gpuEnvObj.DeepCodegen = 1;
results = coder.checkGpuInstall(gpuEnvObj)

I have the following:

Compatible GPU           : PASSED 
CUDA Environment         : PASSED 
	Runtime   : PASSED 
	cuFFT     : PASSED 
	cuSOLVER  : PASSED 
	cuBLAS    : PASSED 
cuDNN Environment        : PASSED 
TensorRT Environment     : PASSED 
Basic Code Generation    : PASSED 
Basic Code Execution     : PASSED 
Deep Learning (TensorRT) Code Generation: PASSED 
Deep Learning (TensorRT) Code Execution: FAILED (Generated GPU code failed to execute properly with the following error 'Invalid MEX-file 'C:\Users\alis\AppData\Local\Temp\tp22472ee1_cce8_43cd_b782_e5b03357a893\dlEntryPointTest_mex.mexw64': No se puede encontrar el módulo especificado.')

results = 

  struct with fields:

                 gpu: 1
                cuda: 1
               cudnn: 1
            tensorrt: 1
        basiccodegen: 1
       basiccodeexec: 1
         deepcodegen: 1
        deepcodeexec: 0
    tensorrtdatatype: 1
           profiling: 0
           

This is on a Windows 10 machine.

Any idea of waht can be the issue?

Answer 1

[sunho jang]

The same problem happens to me. Have you solved it yet?

Answer 2

[Mikaël LE GRAND]

Hello,

I also encounter the same problem :

gpuEnvObj = coder.gpuEnvConfig;
gpuEnvObj.BasicCodegen = 1;
gpuEnvObj.BasicCodeexec = 1;
gpuEnvObj.DeepLibTarget = 'tensorrt';
gpuEnvObj.DeepCodeexec = 1;
gpuEnvObj.DeepCodegen = 1;
results = coder.checkGpuInstall(gpuEnvObj)
Compatible GPU           : PASSED 
CUDA Environment         : PASSED 
	Runtime   : PASSED 
	cuFFT     : PASSED 
	cuSOLVER  : PASSED 
	cuBLAS    : PASSED 
cuDNN Environment        : PASSED 
TensorRT Environment     : PASSED 
Basic Code Generation    : PASSED 
Basic Code Execution     : PASSED 
Deep Learning (TensorRT) Code Generation: PASSED 
Deep Learning (TensorRT) Code Execution: FAILED (Generated GPU code failed to execute properly with the following error 'Invalid MEX-file 'C:\Users\adm.mle_grand\AppData\Local\Temp\tp05f76285_6ff4_4487_8fca_16965e73b665\dlEntryPointTest_mex.mexw64': Le module spécifié est introuvable.')

results = 

  struct with fields:

                 gpu: 1
                cuda: 1
               cudnn: 1
            tensorrt: 1
        basiccodegen: 1
       basiccodeexec: 1
         deepcodegen: 1
        deepcodeexec: 0
    tensorrtdatatype: 1
           profiling: 0

However the specified function (dlEntryPointTest_mex.mexw64) is present in the directory. Is it a problem of TensorRT release problem ?!? For my part, I use TensorRT-5.1.2.2 release :

I tried other releases of TensorRT but the result is always the same via gpucoderSetup command :

Can someone from the staff of Mathworks help us solve this annoying problem ?

Answer 3

[Mikaël LE GRAND]

Hello,

i finally resolve the problem by reading more accuratly the MATLAB documentation. So :

1/ I have installed CUDA Toolkit 10.1 (R2020a compatible) https://developer.nvidia.com/cuda-10.1-download-archive-update2?target_os=Windows&target_arch=x86_64&target_version=10&target_type=exelocal

2/ I have installed CudNN 7.5.0.56 (R2020a compatible) https://developer.nvidia.com/compute/machine-learning/cudnn/secure/v7.5.0.56/prod/10.1_20190225/cudnn-10.1-windows10-x64-v7.5.0.56.zip

3/ I have installed TensorRT 5.1.2.2 (R2020a compatible) https://developer.nvidia.com/nvidia-tensorrt-5x-download (choose TensorRT 5.1 RC version)

4/ I have setted up the environment variables that MATLAB need :

CUDA_PATH : C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1

CUDA_PATH_V10_1 : C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1

NVIDIA_CUDNN : C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\cuda

NVIDIA_TENSORRT : C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\TensorRT-5.1.2.2

NVTOOLSEXT_PATH : C:\Program Files\NVIDIA Corporation\NvToolsExt\

To finish, I add those path to the PATH variable :

• C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\bin
• C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\libnvvp
• C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\cuda\bin
• C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\TensorRT-5.1.2.2\lib

In MATLAB, it gives :

>> gpuEnvObj = coder.gpuEnvConfig('host');
    gpuEnvObj.BasicCodegen = 1;
    gpuEnvObj.BasicCodeexec = 1;
    gpuEnvObj.DeepLibTarget = 'tensorrt';
    gpuEnvObj.DeepCodeexec = 1;
    gpuEnvObj.DeepCodegen = 1;
    gpuEnvObj.Profiling = 1;
    coder.checkGpuInstall(gpuEnvObj)
Compatible GPU           : PASSED 
CUDA Environment         : PASSED 
	Runtime   : PASSED 
	cuFFT     : PASSED 
	cuSOLVER  : PASSED 
	cuBLAS    : PASSED 
cuDNN Environment        : PASSED 
TensorRT Environment     : PASSED 
Profiling Environment    : PASSED 
Basic Code Generation    : PASSED 
Basic Code Execution     : PASSED 
Deep Learning (TensorRT) Code Generation: PASSED 
Deep Learning (TensorRT) Code Execution: PASSED 

ans = 

  struct with fields:

                 gpu: 1
                cuda: 1
               cudnn: 1
            tensorrt: 1
        basiccodegen: 1
       basiccodeexec: 1
         deepcodegen: 1
        deepcodeexec: 1
    tensorrtdatatype: 1
           profiling: 1

>> 

I hope this could help...