TensorRT 4.0.16 sampleUFFMNIST input dimension error

AI & Data Science Deep Learning (Training & Inference) TensorRT
1

Hello,
I have a Host PC with Ubuntu x86_64 16.04.4 LTS.
I’m trying to execute the c++ sampleUFFMNIST sample provided as part of the TensorRT package.
I succeeded to do it with TensorRT 3.0.4 after I created an application with Nsight Eclipse 9.0.
But when i upgraded the TensorRT version to 4.0.16 I started to get the following runtime error:

"ERROR: Parameter check failed at: Network.cpp::addInput::281, condition: isIndexedCHW(dims) && volume(dims) < MAX_TENSOR_SIZE
"

I didn’t change anything in the original source code just prepared a new Nsight Eclipse project with the required settings that weren’t changed at all when I moved from Tensor 3.0.4 to TensorRT 4.0.16.

Thanks,

2

It seems that I succeeded to bypass the problem:
I replaced the usage of Dim3 with DimCHW as an input to the registerInput API as was done in the sampleUFFMNIST copy of the TensorRT 3.0.4.
With this change the error was disappeared.

It is very strange because the usage of Dim3 is part of the installed TensorRT 4.0.16 sampleUFFMNIST code…

Any kind of explanation will be great.