Introduction to NGC Catalog

NVIDIA’s NGC Catalog is designed to provide a comprehensive repository of GPU-optimized software specifically curated for AI, HPC (High-Performance Computing), and data science workloads. NGC stands for NVIDIA GPU Cloud. It is not a cloud service in itself but rather a catalog of tools and content optimized for NVIDIA GPUs, available via web access and CLI tools.

The NGC Catalog is a curated hub by NVIDIA that hosts GPU-accelerated containers, pre-trained AI models, software development kits (SDKs), Helm charts, and other valuable resources. It is designed to simplify the deployment of complex AI workflows, helping teams move from idea to production faster.

Features of NGC Catalog

Some important NGC Catalog features are as follows:

• Pre-trained Models: Ready-to-use models for computer vision, NLP, speech, and more.
• GPU-Optimized Containers: Pre-built containers for TensorFlow, PyTorch, RAPIDS, and other frameworks.
• SDKs: Access to NVIDIA’s specialized SDKs like DeepStream, Riva, Clara, and others.
• Helm Charts: Kubernetes-ready charts to deploy AI solutions on the cloud or on-premises.
• Collections: Thematic groupings of resources for specific domains such as healthcare, robotics, and finance.
• Security and Reliability: Content is scanned for vulnerabilities and validated for performance on NVIDIA GPUs.

 

 

NGC Console

The NGC Console is the web-based graphical user interface for interacting with the NGC Catalog. It allows users to explore, organize, and deploy resources with ease. It also supports enterprise features like team collaboration, model versioning, and private registries.

How to Access NGC Console?

• Visit the official NGC website: https://ngc.nvidia.com
• Create or log in to your NVIDIA Developer account.
• Browse the catalog for models, containers, SDKs, or Helm charts.
• Use the web interface or CLI to pull or deploy resources to your local or cloud infrastructure.

Containers, Collections, Models, and Helm Charts

Containers

NGC containers are Docker-compatible environments that package applications, libraries, and dependencies needed to run AI or HPC tasks. These containers are tested and optimized for NVIDIA GPUs to ensure high performance and scalability.

Collections

Collections are curated sets of containers, models, and Helm charts grouped by application domain. They make it easier for users to find everything needed for a specific field, like autonomous vehicles or genomics, in one place.

Models

These are pre-trained deep learning models that can be used as-is or fine-tuned for specific tasks. NGC models save time and computation effort, allowing developers to focus on innovation rather than training from scratch.

Helm Charts

Helm charts in NGC simplify the deployment of applications on Kubernetes clusters. These charts define, install, and upgrade complex applications, ensuring consistency and repeatability in deployments.