/tag/singularity
NVIDIA DGX BasePOD™
Introducing the NVIDIA DGX BasePOD™ As artificial intelligence (AI) and machine learning (ML) continue to change how academic research is conducted, the NVIDIA DGX BasePOD, or BasePOD, brings new AI and ML functionality to Rivanna, UVA’s High-Performance Computing (HPC) system. The BasePOD is a cluster of high-performance GPUs that allows large deep-learning models to be created and utilized at UVA.
The NVIDIA DGX BasePOD™ on Rivanna, hereafter referred to as the POD, is comprised of:
10 DGX A100 nodes 8 A100 GPU devices and 2 TB local node memory (per node) 80 GB GPU memory (per GPU device) Compared to the regular GPU nodes, the POD contains advanced features such as:
Software Containers
Overview Containers bundle an application, the libraries and other executables it may need, and even the data used with the application into portable, self-contained files called images. Containers simplify installation and management of software with complex dependencies and can also be used to package workflows. Singularity is a container application targeted to multi-user, high-performance computing systems. It interoperates well with Slurm and with the Lmod modules system. Singularity can be used to create and run its own containers, or it can import Docker containers.
Creating Singularity Containers To create your own image from scratch, you must have root privileges on some computer running Linux (any version).
Launching RStudio Server from a Singularity Container
Rocker provides many software containers for R. Due to the default permission settings of our file system, launching an RStudio Server session is not straightforward. If you are interested in using their containers on Rivanna, please follow these steps.
Pull container Use Singularity to pull the container. We will use geospatial in this example.
singularity pull docker://rocker/geospatial You should see geospatial_latest.sif in your current directory.
One-time setup The commands in this section are to be executed as a one-time setup on the frontend. You may need to repeat the steps here when running a new rocker container.
Create a directory where you have write permissions, e.
Rivanna HPC Software
Overview Research Computing at UVA offers a variety of standard software packages for all Rivanna users. We also install requested software based on the needs of the high-performance computing (HPC) community as a whole. Software used by a single group should be installed by that group’s members, ideally on leased storage controlled by the group. Departments with a set of widely-used software packages may install them to the lsp_apps space. The Research Computing group also provides limited assistance for individual installations.
For help installing research software on your PC, please contact Research Software Support at res-consult@virginia.edu.
Software Modules and Containers Software on Rivanna is accessed via environment modules or containers.
Custom Jupyter Kernels
You can create custom kernels from an Anaconda environment or a Singularity container. In both cases you’ll need to install the ipykernel package.
Jupyter kernel based on Anaconda environment To create a custom kernel of the Anaconda environment myenv that uses Python 3.7:
module load anaconda conda create -n myenv python=3.7 ipykernel <other_packages> source activate myenv python -m ipykernel install –user –name myenv –display-name "My Env" Note:
You can customize the display name for your kernel. It is shown when you hover over a tile in JupyterLab. If you do not specify a display name, the default Python [conda env:<ENV_NAME>] will be shown.
Keras on Rivanna
Description Keras is a high-level neural networks application programming interface (API), written in Python and capable of running on top of TensorFlow, CNTK, or Theano. On Rivanna, we provide TensorFlow containers that include the Keras API. Since version 1.12.0, TensorFlow contains its own Keras API implementation as described on the TensorFlow website.
Using Keras with TensorFlow containers Like TensorFlow itself, Python code that utlizes the Keras package can be run interactively as Jupyter Notebooks, in interactive shell jobs, or non-interctively as Slurm batch jobs. Rivanna provides several nodes with graphics processing units (GPUs) that should be used when running Keras code.
Rivanna HPC Software
Overview Research Computing at UVA offers a variety of standard software packages for all Rivanna users. We also install requested software based on the needs of the HPC community as a whole. Software used by a single group should be installed by that group’s members, ideally on leased storage controlled by the group. Departments with a set of widely-used software packages may install them to the lsp_apps space. The Research Computing group also provides limited assistance for individual installations.
For help installing research software on your PC, please contact Research Software Support at res-consult@virginia.edu.
Software Modules and Containers Software on Rivanna is provided via environment modules or as containers.