RAPIDS.AI Install on Ubuntu
There are different software packages available for using NVIDIA CUDA/OPENCL architectures.
In addition to Google-supported Tensorflow and other packages, RAPIDS.AI is a useful open GPU environment. Installation has become more straight forward (after you setup your CUDA NDVIDIA drivers) RAPIDS has useful and well maintained documentation and Jupyter Notebooks (for example, see Jupyter Notebooks for Machine Learning).
Installation
Follow conda package selection for a pure rapids installation.
In order to install an environment useful for point cloud processing, use one of the following.
CUDA 10.2 PointCloud Processing (Ubuntu 18.04)
conda create -y -n PC_cudf -c rapidsai-nightly -c nvidia -c anaconda -c conda-forge -c defaults \
ipython spyder python=3.8 rapids=0.16 cudatoolkit=10.1 cuspatial gdal=3 numpy scipy dask h5py pandas \
pytables hdf5 cython matplotlib tabulate scikit-learn pyflann cyflann scikit-image opencv ipywidgets \
scikit-learn laszip liblas
Activate the environment conda activate PC_cudf
and install additional packages with pip install laspy
and tables pip install tables
CUDA 11.0 PointCloud Processing (Ubuntu 20.04)
conda create -y -n PC_cudf -c rapidsai-nightly -c nvidia -c anaconda -c conda-forge -c defaults \
ipython spyder python=3.8 rapids=0.16 cudatoolkit=11.0 cuspatial gdal=3 numpy scipy dask h5py pandas \
pytables hdf5 cython matplotlib tabulate scikit-learn pyflann cyflann scikit-image opencv ipywidgets \
scikit-learn laszip liblas
Activate the environment conda activate PC_cudf
and install additional packages with pip install laspy
and tables pip install tables