You need trusted, GPU-optimized containers for AI frameworks. Where do they come from?
Why D is the answer
The NVIDIA NGC catalog
NGC hosts GPU-optimized containers, models and SDKs.
In detail
- The NGC catalog holds trusted, GPU-optimized containers.
- It also hosts models and SDKs.
Use case
An illustrative scenario, not a real incident.
A team needs a PyTorch container tuned for GPUs, and pulls a trusted one from NGC.
See it in different ways
flowchart LR A["NGC catalog"] --> B["API key if needed"] B --> C["Pull from nvcr.io"] C --> D["Run with GPUs"]
sequenceDiagram participant U as Operator participant R as nvcr.io participant H as GPU host U->>R: Signs in with an NGC API key R-->>H: GPU-optimized image H->>H: Runs with GPUs
flowchart LR S0["Browse NGC"] --> S1["Pull image"] --> S2["Run"]
Key point: NGC hosts GPU-optimized containers.
Why the other options are wrong
- A. A local build from a public Dockerfile
- A local build is not NVIDIA's tested image.
- B. A shared drive
- A shared drive is not a registry.
- C. An unofficial community mirror
- An unofficial mirror may not be trusted.