NVIDIA containers run your code faster on OSCAR
You know the drill if you want to setup your environment to run AI/ML code. Specialized AI/ML code has many requirements before you can run the it. Some users will use Anaconda, virtualenv, or pray that all the packages, drivers and modules are already installed on your GPU system.
NVIDIA has a resource where you can browse and download pre-built containers with optimized software modules installed. Check it out here — https://ngc.nvidia.com/catalog/containers
OK, now what? How do we take advantage of these containers on Brown’s HPC Cluster — OSCAR. Docker is not available for users on OSCAR because of security concerns. Good news. OSCAR users do have the ability to run containers with Singularity.
Check out these steps to run NVIDIA containers on OSCAR — https://docs.ccv.brown.edu/oscar/gpu-computing/ampere-architecture-gpus#to-use-ngc-containers-via-singularity
Let us know if your code execution performance improves.