Fast.ai course: https://course.fast.ai/
Based on: https://github.com/brtknr/fastai-jetson-nano
Setup dependencies is a bit of a hassle as for ARM devices there is not anaconda available.. so we need to install all dependencies manually
EDIT: 24/04/2020 , pytorch v1.5.0 available on jetsons https://forums.developer.nvidia.com/t/pytorch-for-jetson-nano-version-1-5-0-now-available/72048
sudo apt-get -y update
sudo apt-get -y upgrade
sudo apt-get -y purge \
python3-numpy python3-scipy \
python3-matplotlib python3-pandas python3-pil
sudo apt-get -y install \
python3-pip build-essential python-dev git \
cython python-requests python-typing \
g++ cmake python-dev \
nodejs npm freetype2-demos \
zlib1g-dev zip libjpeg8-dev libhdf5-dev \
libhdf5-serial-dev hdf5-tools \
pkg-config libfreetype6-dev libpng-dev \
libblas3 liblapack3 liblapack-dev libblas-dev \
gfortran htop
download requirements.txt file from this repo: https://github.com/tdurand/fastai-jetsonxavier/blob/master/requirements.txt
pip3 install -r requirements.txt
Links for wheel files: https://github.com/pytorch/pytorch#nvidia-jetson-platforms
Go to the link and get the latest .whl file
pip3 install torch-1.4.0-cp36-cp36m-linux_aarch64.whlreference guide: https://forums.developer.nvidia.com/t/pytorch-for-jetson-nano-version-1-4-0-now-available
sudo apt-get install libjpeg-dev zlib1g-dev
git clone --branch v0.5.0 https://github.com/pytorch/vision torchvision
cd torchvision
python3 setup.py installpip3 install fastai --no-depsecho "done with part1 - now logout, login again and run setup_jupyter.sh"
jupyter labextension install @jupyter-widgets/jupyterlab-manager
jupyter labextension install @jupyterlab/statusbar
jupyter lab --generate-config
cp jupyter_notebook_config.py .jupyter/jupyter_notebook_config.py
jupyter notebook passwordgit clone https://github.com/fastai/course-v3.git
cd course-v3
jupyter notebookif you get, couldn't find this dependency, just restart the terminal and type the command again