fsc_torch_jetson_cpp

A tutorial and sample code for running a pytorch nerual network using c++ on Nvidia jetson nano computer.

Author

Dr. Longhao Qian

How to use

• Read this tutorial about how to correctly install pytorch c++ lib on jetson nano.

• Run the python/verify_pytorch.py to verify whether pytorch c++ lib function correctly.

• run to train a simple neural network and save the trained network to .pt weights file.

• run to compile the src/pytorchTest/pytorch_load_test.cpp and src/pytorchTest/pytorch_load_test_gpu.cpp to get the execution files.

• run to build_release/pytorchTest/pytorch_load_test_gpu and build_release/pytorchTest/pytorch_load_test and observe the results.

CMake settings

• read this tutorial about cmake commands to incooperate pytorch lib in compilation.

Tensor definition in c++

• read this tutorial about tensor definition in c++

CPU/GPU execution

• read this tutorial about CPU/GPU nerual network execution.

Libtorch documentation

• using libtorch to load pytorch model is in this link.