TerraGPU can be installe and used via anaconda environments and containers. A Docker container is provided and this same container can be converted to a Singularity container without loosing any functionalities. A demo cpu based conda environment is provided with limited capabilities for CPU-only users.
CPU support is limited and the author does not provide any guarantee of usability. Limited functionalities have been adapted for CPU in order to provide demo support or local testing before migrating to production environments testing.
The container is built on top of NGC NVIDIA RAPIDS containers. It provides a built-in anaconda environment called rapids which includes all other dependencies for software development of Earth Science applications on GPUS and AI/ML frameworks.
This application is powered by PyTorch and PyTorch Lighning AI/ML backends, including CUML for the development of machine learning models support by NVIDIA RAPIDS.
A Dockerhub public extension will be provided as soon as permissions are granted from the software release process.
module load singularity
singularity build --sandbox terragpu docker://gitlab.nccs.nasa.gov:5050/nccs-lcluc/terragpu/terragpu:latestgit clone https://github.com/nasa-cisto-ai/terragpu.git
cd terragpu; conda env create -f requirements/environment_gpu.yml;
conda activate terragpu
python setup.py installAs an example, you can shell into the container:
singularity shell --nv terragpuThen activate the Anaconda environment:
source active rapids
And start developing your Python applications:
python --version