PaddleOCR on Jetson ORIN Nano super(8gb)

[harrryy007]

Hi , i want to deploy my custom PaddleOCR model on my jetson orin nano super (8gb) .
But when i tried to use ur docker or ur pip i saw that there is no version for cuda 12.6 which is latest version ..

Can you please guide me to deploy my model?( i have converted this model for number plate ... to onnx as well as trt for optimization ) .

when i try to install paddleocr it says no version found .

[WhaleAmbassador]

I decided to build from source entirely for C++ . Cloning Paddle
git clone https://github.com/PaddlePaddle/Paddle.git
and then CMake with these cd build
cmake ..
-DCMAKE_CXX_FLAGS=""
-DCMAKE_C_FLAGS=""
-DWITH_NCCL=OFF
-DWITH_CONTRIB=OFF
-DWITH_MKL=OFF
-DWITH_MKLDNN=OFF
-DWITH_TESTING=OFF
-DCMAKE_BUILD_TYPE=Release
-DON_INFER=ON
-DWITH_PYTHON=OFF
-DWITH_GPU=ON

then a did a make -2j for 2 core processing. You need to create some Swap Space or else the Orin will crash because some of those make files are massive when compiled. I have not been successful with getting paddle to run on the Orin, but this seems like the best way to go if you have a different form of Jetpack than 12.3.

I am running into an issue where most of my CMake flags are getting errors where they are being written to -64m which is for x86_64 architecture and the Orin uses ARM architecture.

I apologize for not being more specific, but I am incredibly new to this and I was just curious if you've maybe come up with a solution, besides just re flashing the Orin to meet the specifications of Paddle.