Get started with using machine learning libraries on Nvidia Jetson Nano

Note: It will take more time than usual (e.g. ~30 mins for pandas) to install packages using pip in jetson nano as most of them are built from source (Ref: https://forums.developer.nvidia.com/t/jetson-nano-for-data-science/112805)

Warning!!! - your casual pip3 install torch torchvision torchaudio won't work gave an error: package directory 'torch/cuda' does not exist. helper link - https://forums.developer.nvidia.com/t/pytorch-for-jetson-version-1-8-0-now-available/72048#5331080

#install pytorch 
wget [https://nvidia.box.com/shared/static/p57jwntv436lfrd78inwl7iml6p13fzh.whl](https://nvidia.box.com/shared/static/p57jwntv436lfrd78inwl7iml6p13fzh.whl) -O torch-1.8.0-cp36-cp36m-linux_aarch64.whl
sudo apt-get install python3-pip libopenblas-base libopenmpi-dev
pip3 install Cython
pip3 install numpy torch-1.8.0-cp36-cp36m-linux_aarch64.whl

Numpy install issue requires - sudo apt-get install python3-dev

sudo apt-get install git cmake
sudo apt-get install libatlas-base-dev gfortran
sudo apt-get install libhdf5-serial-dev hdf5-tools
sudo apt-get install python3-dev
pip3 install scipy

Refs

• https://learninone209186366.wordpress.com/2019/07/24/how-to-install-the-librosa-library-in-jetson-nano-or-aarch64-module/
• https://stackoverflow.com/questions/61475274/runtimeerror-path-failed-executing-please-point-llvm-config-to-the-path-for/63501475#63501475

sudo pip3 install --upgrade setuptools
sudo pip3 install cython
#Install LLVM 9 or above to install llvmlite (dependecy for librosa)

sudo apt install llvm-10
cd /usr/bin
#create a simi-link 
sudo ln -s llvm-config-10 llvm-config
#You might get an error llvm-config already exist while creating simi-link
# In that case remove existing first - rm llvm-config
pip3 install llvmlite

sudo apt-get install libblas-dev liblapack-dev libatlas-base-dev gfortran
pip3 install sounfile
pip3 install librosa

cat /usr/local/cuda-10.2/version.txt

pip3 install torchaudio wont work Alternate bulding from source will also throw an error - A workaround has been mentioned here - pytorch/audio#658

https://github.com/rbonghi/jetson_stats

https://docs.nvidia.com/deeplearning/tensorrt/install-guide/index.html

https://github.com/NVIDIA-AI-IOT/torch2trt Issue - Need to check cuda installation corrosposing to tensorrt version ?

# https://github.com/NVIDIA-AI-IOT/torch2trt
pip install nvidia-pyindex
pip install nvidia-tensorrt

git clone https://github.com/NVIDIA-AI-IOT/torch2trt
cd torch2trt
sudo python setup.py install --plugins

https://github.com/debauchee/barrier

https://www.youtube.com/watch?v=-G9IIauHOhA

Note - You will initially need a keyboard and mouse to setup your device jetson device as client. Alternatively if you donot have a keyboard enable "On screen keyboard" from System Settings - Universal Access - Typing Once you are diconnect from the network you would need to connect the external mouse and keybard again to activate the barrier client.

SSL issues: Please generate your own Barrier.pem key as described in - debauchee/barrier#231

**Use macbook pro keyboard layout ** Keyboard setting -> Text settings -> Add English (UK, intl, Macintosh)