This is a collection of exercises to learn libtorch, the C++ backend of Pytorch. Most of us learned Pytorch in an interactive environment like Jupyter notebook. Such interactivity made learning tensor API a breeze! Now, we can do the same with C++.
All notebooks should be run on C++ kernel e.g. https://github.com/jupyter-xeus/xeus-cling
mamba makes it easier to install necessary dependencies into an isolated environment.
mamba create -n cling
mamba activate cling
mamba install xeus-cling jupyterlab -c conda-forgemamba is a C++ implementation of conda. To install mamba,
name=Miniforge3-$(uname)-$(uname -m).sh
wget "https://github.com/conda-forge/miniforge/releases/latest/download/$name"
bash $name -b -p ~/mambaforge && ~/mambaforge/bin/mamba init bash && rm $nameDownload libTorch:
The following script takes care of downloading libtorch and extracting under libs directory
bash ./libs/setup.sh
mamba activate cling
jupyter-lab