Harshvardhan Chandirasekar

try this: in https://github.com/ceccocats/tkDNN/blob/38106a9495438465118f8e36021126f5fbcbb57d/CMakeLists.txt#L5 remove "-Wno-deprecated-declarations -Wno-unused-variable" they are unsupported command line options for MSVC, and you will have to comment out and probably replace "unistd.h" with "windows.h" in the...

there is support for tensorrt8 ,checkout the tensorrt8 branch https://github.com/ceccocats/tkDNN/tree/tensorrt8..if you are asking for python support for tensorrt8 in tkDNNi am not sure about it

Regarding the int8 model conversion ,have you downloaded the dataset and followed the steps mentioned in the Readme? Cause it looks there is a problem with your dataFilePath for image...

Unfortunately this repo doesn't support TorchRt models .You would need to manually export the weights from the pytorch model to layer wise weights and then write the model again using...

It should , I was able to run it on both windows and linux with cudnn 8.3.3

I tested it with cuda 11.4 , but I don't think using cuda 10.2 should be an issue

Yup it should also work with tensorrt 8.4 ea

are you cross compiling ?if so then you can try this guide https://forums.developer.nvidia.com/t/solved-libcudart-so-error-adding-symbols-file-in-wrong-format/71084/3

can you share more details on your issue like a screenshot of the wrong predictions ,with what models are you experiencing this issue and your setup (windows or linux) ?

have you removed the comment from this line https://github.com/ceccocats/tkDNN/blob/a992c9feb5fb5c7a64da59f6c5a6f0c1c1a6cf2d/include/tkDNN/DetectionNN.h#L20 to use opencv contrib ?