README.01-configure.md

Configure NVIDIA Jetson TX2 Board

Link to required NVIDIA packages

[HOST] Check for available SSH connection to Jetson TX2

# to connect Jetson board
$ ssh nvidia@JETSON_IP_ADDRESS
# to copy files to Jetson board
$ scp setup.txt nvidia@JETSON_IP_ADDRESS:/home/nvidia

Using JetPack to install CUDA, OpenCV4Tegra and cuDNN is highly recommended, manual installation may be problematic!

[HOST] Copy CUDA/TensorFlow packages to Jetson TX2

$ scp $HOME/nvidia/jetpack_download/cuda-repo-l4t-8-0-local_8.0.84-1_arm64.deb nvidia@JETSON_IP_ADDRESS:/home/nvidia
$ scp $HOME/nvidia/jetpack_download/libcudnn6_6.0.21-1+cuda8.0_arm64.deb nvidia@JETSON_IP_ADDRESS:/home/nvidia
$ scp $HOME/nvidia/jetpack_download/libcudnn6-dev_6.0.21-1+cuda8.0_arm64.deb nvidia@JETSON_IP_ADDRESS:/home/nvidia
$ scp $HOME/nvidia/jetpack_download/libcudnn6-doc_6.0.21-1+cuda8.0_arm64.deb nvidia@JETSON_IP_ADDRESS:/home/nvidia
$ scp $HOME/nvidia/jetpack_download/nv-gie-repo-ubuntu1604-ga-cuda8.0-trt2.1-20170614_1-1_arm64.deb nvidia@JETSON_IP_ADDRESS:/home/nvidia
$ scp $HOME/nvidia/jetpack_download/libopencv4tegra-repo_2.4.13-17-g5317135_arm64_l4t-r26.deb nvidia@JETSON_IP_ADDRESS:/home/nvidia

[TARGET] Update Ubuntu packages

$ sudo apt update
$ sudo apt upgrade
$ sudo apt autoremove

[TARGET] Install CUDA and TensorFlow

$ sudo dpkg -i /home/nvidia/cuda-repo-l4t-8-0-local_8.0.84-1_arm64.deb
$ sudo dpkg -i /home/nvidia/libcudnn6_6.0.21-1+cuda8.0_arm64.deb
$ sudo dpkg -i /home/nvidia/libcudnn6-dev_6.0.21-1+cuda8.0_arm64.deb
$ sudo dpkg -i /home/nvidia/libcudnn6-doc_6.0.21-1+cuda8.0_arm64.deb
$ sudo dpkg -i /home/nvidia/nv-gie-repo-ubuntu1604-ga-cuda8.0-trt2.1-20170614_1-1_arm64.deb
$ sudo dpkg -i /home/nvidia/libopencv4tegra-repo_2.4.13-17-g5317135_arm64_l4t-r26.deb
$ sudo apt update
$ sudo apt install cuda-toolkit-8.0
$ sudo apt install tensorrt-2.1.2

Add these commands at the end of $HOME.profile

export PATH=/usr/local/cuda-8.0/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64:$LD_LIBRARY_PATH